British Journal of Surgery

doi: 10.1002/bjs.10038pmid: N/A

Article PDF first page preview Close This content is only available as a PDF. © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd

Glynne-Jones, R; Hall, M

doi: 10.1002/bjs.9930pmid: 26458069

Radiotherapy has been an accepted treatment for locally advanced (T3/4 N1/2) rectal cancer (LARC) since the 1990s, and preoperative chemoradiotherapy (CRT) is the standard today. Questions are emerging, however, regarding the blanket use of a multimodal approach that involves neoadjuvant CRT, total mesorectal excision (TME) and adjuvant chemotherapy in all patients without selection. In the past decade, significant advances in the management of rectal cancer have been achieved, reflecting improvements in surgical technique, MRI and histopathology reporting. The main aims have been to minimize the risk of local and distant recurrence, and preserve the sphincter with good long-term function. The introduction of systemic chemotherapeutic agents (oxaliplatin, irinotecan) and targeted biological therapies offers multiple possible modifications to the standard preoperative CRT multimodal approach. Treatment decisions are often based on MRI in most developed healthcare systems. Although MRI-based risk stratification strategies have been incorporated into clinical practice guidelines1 and eligibility criteria in clinical trials (NCT01558921, ISRCTN09351447)2,3, the evidence for MRI on the basis of prospective observational studies, such as MERCURY4, has yet to be validated in randomized phase III trials. There is variability in interpretation. Some radiologists do not define macroscopic extramural vascular invasion. Many of the clinical data on which these treatment decisions are founded predate good-quality TME, the use of MRI and modern histopathology. In this previous era, radiotherapy may well have simply compensated for poor surgery and these data may not be relevant in 2015. High-quality TME is now performed in 70–80 per cent of patients5, reducing local recurrence rates to approximately 2 per cent after radiotherapy or CRT, and only 6 per cent without. The rate of metastatic disease, however, remains consistently in the region of 30–35 per cent. The population at risk is also changing, as a result of screening programmes that have led to earlier diagnosis, creating further controversy between radical surgery and organ-sparing approaches for T1/2 cancers. Difficulties in performing trials in rectal cancer reflect poor choice of primary endpoint, inadequate imaging/staging, insufficient quality assurance, variations in the quality of pathological assessment, the impact of non-operative ‘watch-and-wait’ strategies and the use of postoperative chemotherapy. For many of these reasons, progress in the non-surgical aspects of treatment of LARC seems to have stalled. Various strategies have been explored in small phase II studies: integrating novel chemotherapy and/or biological agents into CRT schedules, integrating induction or consolidation chemotherapy into standard or short-course preoperative radiotherapy (SCPRT) schedules, as well as alternating chemotherapy and split-course radiotherapy. Most represent strategies to increase pathological complete response rates. Few of these non-surgical adaptive strategies seem to have resulted in significant advances in understanding the disease or improving outcomes. Ramping up the neoadjuvant chemotherapy component, in numerous iterative attempts to add chemotherapy before CRT, seems attractive theoretically but does not appear to have resulted in tangible advantages6,7. In contrast, additional courses of FOLFOX (folinic acid, 5-fluorouracil, oxaliplatin) as consolidation after CRT and before TME have the potential to increase the pathological complete response rate; theoretically this could broaden the options for patients in terms of less invasive treatment strategies8, although long-term oncological outcomes are lacking. The European Organization for Research and Treatment of Cancer 22921 trial9 has indicated that postoperative adjuvant chemotherapy after preoperative CRT does not improve disease-free or overall survival. These approaches hypothesize that better results might be obtained by adding more chemotherapy to existing strategies. The question that seems to have been largely ignored is whether radiotherapy is needed at all, or whether neoadjuvant chemotherapy alone could produce the same results in the current TME era. Omitting CRT has the advantages of improved wound healing, less frequent anastomotic leaks, avoidance of long-term radiation toxicity and a smaller risk of second malignancy. There has been little drive to replace SCPRT/CRT with neoadjuvant chemotherapy alone for resectable cT3 cancers, although this strategy forms the basis of the ongoing PROSPECT trial (NCT01515787). Preliminary results of the Chinese FOWARC study10 also suggest that neoadjuvant FOLFOX alone and CRT achieve similar pathological complete response rates, tumour regression grades, and downstaging and curative resection rates, but with reduced surgical morbidity for FOLFOX alone. Future studies need defined efficacy endpoints and clear stopping rules that should be relevant to currently achieved outcomes. Local recurrence is no longer a major problem for LARC. Improvements in survival, numbers requiring a permanent stoma and quality of life should take greater precedence than the complete elimination of local recurrence. Neoadjuvant chemotherapy has been introduced as the primary management of many cancers11 despite a lack of evidence of superiority in terms of overall survival. The FOxTROT (Fluoropyrimidine, Oxaliplatin and Targeted Receptor pre-Operative Therapy for colon cancer) trial12 of neoadjuvant chemotherapy in colonic cancer, which demonstrated significant downstaging, is also beginning to influence colorectal clinicians. Even if this approach does not improve survival, there may be many other potential advantages in terms of facilitating sphincter- or organ-sparing surgery, reducing surgical morbidity compared with CRT and creating opportunities for translational research in LARC. Treatment selection for LARC now needs to take account of factors other than staging and performance status. Mucinous/signet ring tumours, representing about 20 per cent of rectal cancers, respond poorly to fluoropyrimidine or doublet chemotherapy13,14 and may fare better with other agents. These tumours also respond poorly to conventional CRT. For this non-responsive subset, intensive neoadjuvant chemotherapy can facilitate the development of novel chemotherapy scheduling and the exploration of novel agents, and enhance the opportunity for translational research. The role of CRT in these patients needs to be clarified so that recognized adverse effects in terms of poor sexual, urinary and bowel function can be avoided. A comparison of preoperative CRT versus neoadjuvant FOLFIRINOX (folinic acid, 5-fluorouracil, irinotecan, oxaliplatin) followed by preoperative CRT is currently under way in a randomized phase III study in LARC (NCT01804790). Yet another trial integrating two to four courses of induction chemotherapy before CRT or SCPRT and randomized against CRT alone or SCPRT may not be the best way forward. There are potential advantages to neoadjuvant chemotherapy alone without CRT, including greater levels of treatment compliance to chemotherapy, earlier delivery of systemic therapy, and better wound healing with no long-term radiation effects. This concept is explored in the Cancer Research UK BACCHUS (Bevacizumab And Combination Chemotherapy in Rectal Cancer Until Surgery) trial (NCT01650428). Trials randomizing chemotherapy alone against CRT might prove difficult to perform, but if chemotherapy alone is as effective as CRT they may ultimately offer major gains in terms of quality of life. Future studies that assess gene expression and mutations may identify patients who are more likely to respond to CRT and who can then avoid radical surgery. Alternatively, others may avoid radiotherapy with all its drawbacks and instead receive more aggressive chemotherapy and/or immunological or biological options. Acknowledgements R.G.-J. is principal investigator of the randomized phase II BACCHUS trial. The trial is funded by Cancer Research UK and Roche, who are providing the bevacizumab and funding for PET. R.G.-J. has received honoraria from Roche, Sanofi-Aventis and Merck KGaA, and research funding from Merck KGaA, Roche and Sanofi. M.H. has sat on advisory boards for Roche, Merck Sharp & Dohme, Boehringer Ingelheim, GSK and Astra Zeneca. Disclosure: The authors declare no other conflict of interest. References 1 Glimelius B , Tiret E, Cervantes A, Arnold D; ESMO Guidelines Working Group . Rectal cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up . Ann Oncol 2013 ; 24 ( Suppl 6 ): vi81 – vi88 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Nilsson PJ , van Etten B, Hospers GA, Påhlman L, van de Velde CJ, Beets-Tan RG et al. Short-course radiotherapy followed by neo-adjuvant chemotherapy in locally advanced rectal cancer – the RAPIDO trial . BMC Cancer 2013 ; 13 : 279 . Google Scholar Crossref Search ADS PubMed WorldCat 3 ARISTOTLE: a phase III trial comparing standard versus novel chemoradiation treatment (CRT) as pre-operative treatment for magnetic resonance imaging (MRI)-defined locally advanced rectal cancer . (ISRCTN number ISRCTN09351447, EudraCT number 2008-005782-59.) http://www.isrctn.com/ISRCTN09351447 [accessed 9 August 2015 ]. 4 Taylor FG , Quirke P, Heald RJ, Moran BJ, Blomqvist L, Swift IR et al. ; Magnetic Resonance Imaging in Rectal Cancer European Equivalence Study Group . Preoperative magnetic resonance imaging assessment of circumferential resection margin predicts disease-free survival and local recurrence: 5-year follow-up results of the MERCURY study . J Clin Oncol 2014 ; 32 : 34 – 43 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Rödel C , Liersch T, Becker H, Fietkau R, Hohenberger W, Hothorn T et al. ; German Rectal Cancer Study Group . Preoperative chemoradiotherapy and postoperative chemotherapy with fluorouracil and oxaliplatin versus fluorouracil alone in locally advanced rectal cancer: initial results of the German CAO/ARO/AIO-04 randomised phase 3 trial . Lancet Oncol 2012 ; 13 : 679 – 687 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Dewdney A , Cunningham D, Tabernero J, Capdevila J, Glimelius B, Cervantes A et al. Multicenter randomized phase II clinical trial comparing neoadjuvant oxaliplatin, capecitabine, and preoperative radiotherapy with or without cetuximab followed by total mesorectal excision in patients with high-risk rectal cancer (EXPERT-C) . J Clin Oncol 2012 ; 30 : 1620 – 1627 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Fernandez-Martos C , Garcia-Albeniz X, Pericay C, Maurel J, Aparicio J, Montagut C et al. Chemoradiation, surgery and adjuvant chemotherapy versus induction chemotherapy followed by chemoradiation and surgery: long-term results of the Spanish GCR-3 phase II randomized trial . Ann Oncol 2015 ; 26 : 1722 – 1728 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Garcia-Aguilar J , Chow OS, Smith DD, Marcet JE, Cataldo PA, Varma MG et al. ; Timing of Rectal Cancer Response to Chemoradiation Consortium. Effect of adding mFOLFOX6 after neoadjuvant chemoradiation in locally advanced rectal cancer: a multicentre, phase 2 trial . Lancet Oncol 2015 ; 16 : 957 – 966 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Bosset JF , Calais G, Mineur L, Maingon P, Stojanovic-Rundic S, Bensadoun RJ et al. Fluorouracil-based adjuvant chemotherapy after preoperative chemoradiotherapy in rectal cancer: long-term results of the EORTC 22921 randomised study . Lancet Oncol 2014 ; 15 : 184 – 190 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Deng Y , Chi P, Lan P, Wang L, Cui L, Chen D et al. A multi-center randomized controlled trial of mFOLFOX6 with or without radiation in neoadjuvant treatment of local advanced rectal cancer (FOWARC study): preliminary results . J Clin Oncol 2015 ; 33 ( Suppl ): Abstract 3500. Google Scholar OpenURL Placeholder Text WorldCat 11 Kehoe S , Hook J, Nankivell M, Jayson G, Kitchener H, Lopes T et al. Primary chemotherapy versus primary surgery for newly diagnosed advanced ovarian cancer (CHORUS): and open-label, randomised, controlled, non-inferiority study . Lancet 2015 ; 386 : 249 – 257 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Foxtrot Collaborative Group . Feasibility of preoperative chemotherapy for locally advanced, operable colon cancer: the pilot phase of a randomised controlled trial . Lancet Oncol 2012 ; 13 : 1152 – 1160 . Crossref Search ADS PubMed WorldCat 13 Negri FV , Wotherspoon A, Cunningham D, Norman AR, Chong G, Ross PJ. Mucinous histology predicts for reduced fluorouracil responsiveness and survival in advanced colorectal cancer . Ann Oncol 2005 ; 16 : 1305 – 1310 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Catalano V , Loupakis F, Graziano F, Torresi U, Bisonni R, Mari D et al. Mucinous histology predicts for poor response rate and overall survival of patients with colorectal cancer and treated with first-line oxaliplatin- and/or irinotecan-based chemotherapy . Br J Cancer 2009 ; 100 : 881 – 887 . Google Scholar Crossref Search ADS PubMed WorldCat © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd

Petter-Puchner, A H

doi: 10.1002/bjs.9932pmid: 26356133

Midline closure is a seemingly simple procedure in open surgery, but represents a field where technical progress and research in abdominal wall repair intersects. A variety of new closure techniques (such as small bites, controlled traction) as well as sophisticated devices (elastic sutures, mesh sutures, mesh prophylaxis) are available and reflect the quest to reduce disturbingly high rates of incisional herniation. Results are mixed, however, with hernia rates ranging from 2 to 30 per cent in colorectal procedures, with equally high figures in open bariatric surgery1–4. Despite this frequency, there is little evidence of significant improvement or alteration in surgical practice over the past three decades5. Although a growing number of multicentre randomized clinical trials and recent guidelines from the European Hernia Society (EHS)6 provide sound recommendations, this evidence seems to have been ignored by many surgeons who adhere to techniques associated with a high probability of the development of wound complications and incisional hernias7. There are three crucial aspects to midline closure that merit emphasis: suture technique, preservation of the anatomy of the abdominal wall and prophylactic mesh placement. There is increasing evidence that non-resorbable, or at least slowly resorbable, running sutures should be preferred over single sutures or resorbable materials3. It is important to remember that the application of too much traction can cause ischaemia of entrapped muscle as well as trauma to the fascia, no matter what material has been chosen. To allow an equal distribution of tension, self-locking knots should be considered and a suture to wound ratio of 4 : 1, as recommended by the EHS6,8. It is still not clear, however, whether single running sutures are superior to slings and whether small bites are superior to large ones. These aspects of closure are the subjects of ongoing studies, such as the ESTOIH (Effect of Stitch Technique on the Occurrence of Incisional Hernia After Abdominal Wall Closure) trial9. This study uses an elastic suture, indicating a trend towards thermoplastic materials adaptive to local demands. Although thermoplastic sutures have been tested only in animal studies10, other novel approaches are also close to translation to clinical practice, such as mesh suture that has an outer porous coat into which native tissue can grow9. The mesh structure formed a broad band leading to fast integration and performed better than polypropylene in a large animal model11. Preserving the integrity of fascial edges along with sparing of the rectus muscle is a key element of successful midline closure, whether performed with running sutures, slings, meshes or, perhaps in future, mesh sutures. To preserve fascial edges it is essential to identify and create a clean margin zone of some 2–4 cm along the linea alba. Blind stitches through subcutaneous fat or attempts to grasp recessed fascial edges with the needle inevitably lead to weak spots along the suture line. Preparation of the fascia is not time-consuming and should be seen as an essential step in all closure techniques. This is particularly so with prophylactic mesh placement, as described in detail in the EHS guidelines6. There is good evidence that prophylactic mesh placement is beneficial in high-risk patients7. Patients at high risk of incisional hernia include those having open surgery for abdominal aortic aneurysm and those with morbid obesity12. A recent study13 derived from the Danish Hernia Database reported a cumulative incidence of about 10 per cent after transabdominal aortic reconstructive surgery, implying that, in future, prophylactic mesh should be seen as the standard of care. In obese patients, new findings underline the importance of distinguishing between visceral obesity and body mass index14. The latter alone seems a poor predictor of the development of an incisional hernia. Although prophylactic mesh placement has also been found beneficial in a recent study of open colorectal surgery3, it is still not fully understood how much overlap is needed to prevent hernia development and whether resorbable biological mesh materials elicit sufficient ingrowth to provide a durable repair in the long term. Guidelines on midline closure have been designed for all surgeons who use this approach to the abdominal cavity, and not exclusively for abdominal wall and hernia experts. Only wide acceptance will pave the way for a significant reduction in wound complications that contribute to the current incidence of incisional herniation. The wider use of prophylactic mesh in the cohorts of patients outlined above seems appropriate. Training courses related to suture materials and techniques should be integrated into all surgical training programmes. Even experienced surgeons should consider participating in such courses fully to appreciate the possibilities and limitations of new closure concepts. Disclosure The author declares no conflict of interest. References 1 Song IH , Ha HK, Choi SG, Jeon BG, Kim MJ, Park KJ. Analysis of risk factors for the development of incisional and parastomal hernias in patients after colorectal surgery . J Korean Soc Coloproctol 2012 ; 28 : 299 – 303 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Mishra A , Keeler BD, Maxwell-Armstrong C, Simpson JA, Acheson AG. The influence of laparoscopy on incisional hernia rates: a retrospective analysis of 1057 colorectal cancer resections . Colorectal Dis 2014 ; 16 : 815 – 821 . Google Scholar Crossref Search ADS PubMed WorldCat 3 García-Ureña MÁ , López-Monclús J, Hernando LA, Montes DM, Valle de Lersundi AR, Pavón CC et al. Randomized controlled trial of the use of a large-pore polypropylene mesh to prevent incisional hernia in colorectal surgery . Ann Surg 2015 ; 261 : 876 – 881 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Abo-Ryia MH , El-Khadrawy OH, Abd-Allah HS. Prophylactic preperitoneal mesh placement in open bariatric surgery: a guard against incisional hernia development . Obes Surg 2013 ; 23 : 1571 – 1574 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Cameron AE , Gray RC, Talbot RW, Wyatt AP. Abdominal wound closure: a trial of Prolene and Dexon . Br J Surg 1980 ; 67 : 487 – 488 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Muysoms FE , Antoniou SA, Bury K, Campanelli G, Conze J, Cuccurullo D et al. ; European Hernia Society. European Hernia Society guidelines on the closure of abdominal wall incisions . Hernia 2015 ; 19 : 1 – 24 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Meijer EJ , Timmermans L, Jeekel J, Lange JF, Muysoms FE. The principles of abdominal wound closure . Acta Chir Belg 2013 ; 113 : 239 – 244 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Israelsson LA , Millbourn D. Prevention of incisional hernias: how to close a midline incision . Surg Clin North Am 2013 ; 93 : 1027 – 1040 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Fortelny RH , Baumann P, Thasler WE, Albertsmeier M, Riedl S, Steurer W et al. Effect of suture technique on the occurrence of incisional hernia after elective midline abdominal wall closure: study protocol for a randomized controlled trial . Trials 2015 ; 16 : 52 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Lambertz A , Vogels RR, Busch D, Schuster P, Jockenhövel S, Neumann UP et al. Laparotomy closure using an elastic suture: a promising approach . J Biomed Mater Res B Appl Biomater 2015 ; 103 : 417 – 423 . Google Scholar Crossref Search ADS PubMed WorldCat 11 Dumanian GA , Tulaimat A, Dumanian ZP. Experimental study of the characteristics of a novel mesh suture . Br J Surg 2015 ; 102 : 1285 – 1292 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Nieuwenhuizen J , Eker HH, Timmermans L, Hop WC, Kleinrensink GJ, Jeekel J et al. ; PRIMA Trialist Group . A double blind randomized controlled trial comparing primary suture closure with mesh augmented closure to reduce incisional hernia incidence . BMC Surg 2013 ; 13 : 48 . Google Scholar Crossref Search ADS PubMed WorldCat 13 Henriksen NA , Helgstrand F, Vogt KC, Jorgensen LN, Bisgaard T; Database Danish Hernia; Danish Vascular Registry . Risk factors for incisional hernia repair after aortic reconstructive surgery in a nationwide study . J Vasc Surg 2013 ; 57 : 1524 – 1530 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Aquina CT , Rickles AS, Probst CP, Kelly KN, Deeb AP, Monson JR et al. ; Muscle and Adiposity Research Consortium (MARC) . Visceral obesity, not elevated BMI, is strongly associated with incisional hernia after colorectal surgery . Dis Colon Rectum 2015 ; 58 : 220 – 227 . Google Scholar Crossref Search ADS PubMed WorldCat © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd

Levolger, S; van Vugt, J L A; de Bruin, R W F; IJzermans, J N M

doi: 10.1002/bjs.9893pmid: 26375617

Abstract Background Preoperative risk assessment in cancer surgery is of importance to improve treatment and outcome. The aim of this study was to assess the impact of CT-assessed sarcopenia on short- and long-term outcomes in patients undergoing surgical resection of gastrointestinal and hepatopancreatobiliary malignancies. Methods A systematic search of Embase, PubMed and Web of Science was performed to identify relevant studies published before 30 September 2014. PRISMA guidelines for systematic reviews were followed. Screening for inclusion, checking the validity of included studies and data extraction were carried out independently by two investigators. Results After screening 692 records, 13 observational studies with a total of 2884 patients were included in the analysis. There was wide variation in the reported prevalence of sarcopenia (17·0–79 per cent). Sarcopenia was independently associated with reduced overall survival in seven of ten studies, irrespective of tumour site. Hazard ratios (HRs) of up to 3·19 (hepatic cancer), 1·63 (pancreatic cancer), 1·85 (colorectal cancer) and 2·69 (colorectal liver metastases, CLM) were reported. For oesophageal cancer, the HR was 0·31 for increasing muscle mass. In patients with colorectal cancer and CLM, sarcopenia was independently associated with postoperative mortality (colorectal cancer: odds ratio (OR) 43·3), complications (colorectal cancer: OR 0·96 for increasing muscle mass; CLM: OR 2·22) and severe complications (CLM: OR 3·12). Conclusion Sarcopenia identified before surgery by single-slice CT is associated with impaired overall survival in gastrointestinal and hepatopancreatobiliary malignancies, and increased postoperative morbidity in patients with colorectal cancer with or without hepatic metastases. Introduction Advanced surgical techniques, developments in perioperative care and the introduction of enhanced recovery programmes have improved surgical outcomes1–5. Nevertheless, risk assessment before major abdominal surgery remains of paramount importance to improve outcomes further after cancer surgery. Known factors that are predictive of short-term outcome include albumin levels, American Society of Anesthesiologists (ASA) classification and emergency surgery, whereas advanced age and disseminated disease determine long-term outcome6–8. Outcomes of patients with similar age, tumour stage and ASA classification may be very different in clinical practice. Therefore, the risk factors commonly used to predict outcome after cancer surgery may reflect the patient's general health status and physiological reserves insufficiently. An important risk factor for worse outcome is frailty, which is poorly reflected by the traditional determinants of outcome9–13. Frailty is defined as a biological syndrome characterized by decreased reserve and resilience to stress factors across multiple physiological systems, and has been shown to be associated with adverse health outcomes14,15. A hallmark sign of frailty is sarcopenia, the involuntary loss of skeletal muscle mass16–18. The prevalence of sarcopenia in healthy individuals increases with advanced age, ranging from 9 per cent at 45 years and up to 64 per cent in individuals aged over 85 years19. Sarcopenia is characterized by a loss of skeletal muscle mass, skeletal muscle strength and physical performance20. It has been shown to impair physical performance and survival in geriatric, non-cancer populations21,22, and to impair survival in a variety of clinical conditions, such as cancer23. Up to 80 per cent of patients with advanced cancer are affected by cancer-induced cachexia, a clinical condition that also results in skeletal muscle wasting with or without loss of body fat24–26. Cachectic patients are more prone to a reduced effect of therapy and increased chemotherapy toxicity27–29. It has been estimated that as many as 30 per cent of cancer-related deaths result from cachexia30–33. One study23 showed that sarcopenia was associated with decreased survival in obese patients with cancer by using CT to assess reduced skeletal muscle mass34 (Fig. 1). Fig. 1 Open in new tabDownload slide Transverse CT image at the level of L3 showing a cross-sectional area of skeletal muscle mass highlighted in red, including the psoas, paraspinal, transverse abdominal, external oblique, internal oblique and rectus abdominis muscles A systematic review was undertaken to investigate the influence of low skeletal muscle mass or skeletal muscle density assessed by CT on short- and long-term outcomes in patients undergoing surgery for gastrointestinal and hepatopancreatobiliary malignancies. Methods Eligibility criteria were established a priori. A systematic search was performed to identify all original articles on patients undergoing surgical resection of malignancies of the gastrointestinal tract or hepatopancreatobiliary system, in which preoperative abdominal CT was used to assess skeletal muscle mass. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines35 were followed. Included in the analysis were studies that reported on the prevalence of sarcopenia and at least one of the following outcomes: postoperative mortality, postoperative complications, length of intensive care (ICU) stay, length of hospital stay, disease-free survival and overall survival. The search was limited to papers in English with a publication date from January 2000 to September 2014. Three search strings with corresponding search terms were constructed (Table S1, supporting information). The same search strings were used to develop queries in the Embase, PubMed and Web of Science databases. The Embase database search was performed using the following query: (‘sarcopenia’:de,ab,ti OR ‘analytic morphomics’:de,ab,ti OR ‘body composition’:de,ab,ti OR ‘muscle depletion’:de,ab,ti OR ‘muscle mass’:de,ab,ti OR ‘psoas area’:de,ab,ti OR ‘myopenia’:de,ab,ti OR ‘core muscle’:de,ab,ti OR ‘lean body mass’:de,ab,ti OR ‘muscular atrophy’:de,ab,ti) AND (‘cancer’:de,ab,ti OR ‘neoplasms’:de,ab,ti OR ‘malignancy’:de,ab,ti) AND (‘surgery’:de,ab,ti OR ‘resection’:de,ab,ti OR ‘esophagectomy’:de,ab,ti OR ‘gastrectomy’:de,ab,ti OR ‘hepatectomy’:de,ab,ti OR ‘colectomy’:de,ab,ti OR ‘pancreatectomy’:de,ab,ti or ‘cholecystectomy’:de,ab,ti). Similar queries were constructed for PubMed and Web of Science. Duplicate records were removed and abstracts were screened independently by two investigators to determine which records were eligible for further analysis. Abstracts were included for initial analysis if sarcopenia in patients undergoing surgical treatment for gastrointestinal or hepatopancreatobiliary malignancies was described. Abstracts that described sarcopenia determined by means other than abdominal CT or patients undergoing non-surgical treatment were excluded from further analysis. Records without abstracts, case reports, review articles, opinion articles and experimental studies were excluded. Eligibility of studies and assessment of methodological quality Full-text articles of the remaining records were subsequently retrieved and screened independently by two investigators. All original articles that met the inclusion criteria were included. Additional relevant references were sought in the included full-text articles. Two investigators independently assessed the methodological quality of the included studies using the Newcastle–Ottawa quality assessment scale for cohort studies36 for each a priori defined outcome measure. Data extraction Data regarding study design and results were extracted independently by two investigators for each eligible study. Extracted data included age, sex distribution, patient selection, prevalence of sarcopenia, postoperative mortality, postoperative complications, length of ICU stay, length of hospital stay, disease-free survival and overall survival. If univariable and multivariable analyses had been performed to adjust for known risk factors, the latter was used for interpretation of the results. Statistical analysis Outcomes are reported as originally shown. The prevalence of sarcopenia described in this review applies to the total population of each study. Therefore, rates could not be provided for subgroups (such as by cancer stage) separately. No meta-analysis was performed because there was great heterogeneity between studies. Results The literature search was performed on 30 September 2014 and identified an initial 692 records, of which 27 were found to be potentially relevant (Fig. 2). From these 27 records, seven full-text articles were excluded as sarcopenia was assessed by means other than abdominal CT, four articles did not report relevant outcome data, and three articles reported on a population that received non-surgical treatment for the studied tumours. The remaining 13 studies matched the inclusion criteria37–49. Cross-referencing yielded no additional results. The included studies provided data on patients with oesophageal, gastric, pancreatic, primary liver and colorectal cancer, and resectable hepatic colorectal metastases (Table 1). No studies reported on patients with bile duct or gallbladder cancer. Fig. 2 Open in new tabDownload slide PRISMA flow chart showing selection of articles for review Table 1 Studies of the effects of sarcopenia in patients operated on for gastrointestinal and hepatopancreatobiliary malignancies Reference . Malignancy, patient selection . Disease stage (%) . n (men) . Age(years) . BMI(kg/m2) . Muscle(s) measured(level), cut-offs . Quality points by outcome* . . . . . . . . Short-term morbidity or mortality . DFS . OS . Awad et al.37 Oesophageal and gastric cancer, WHO performance status 0–2 Locally advanced 47 (34) 63† Before NACRT: 24·6† Before resection: 23·8† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 5 n.r. 4 Sheetz et al.46 Oesophageal cancer, all consecutive patients IS: 9·6 I: 24·3 II: 33·5 III: 27·4 IV: 5·2 230 (202) 62† Overall: 28·6† TPA, PMD (L4) – 7 6 7 Yip et al.38 Oesophageal cancer IS: 6 I: 3 II: 51 III: 40 35 (30) 63† Before NACRT: 26·7† Before resection: 25·8† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 5 5 5 Harimoto et al.41 Hepatocellular cancer, all consecutive patients I: 15·6 II: 51·1 III: 26·3 IV: 7·0 186 (145) – Sarcopenia: 20·5† No sarcopenia: 24·0† CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 7 6 6 Itoh et al.40 Hepatocellular cancer, all patients without simultaneous procedures n.a. 190 (146) – < 18·5: 7·9% ≥ 18·5 to < 25: 68·4% ≥ 25 to < 30: 21·6% ≥ 30: 2·1% CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 n.r. 6 6 Voron et al.39 Hepatocellular cancer, all consecutive patients n.a. 109 (92) 62† Overall: 24·6† Sarcopenia: 25·6† No sarcopenia: 26·9† CSAMM/m2 (L3) F 38·9 cm2 M 52·4 cm2 7 6 6 Peng et al.47 Pancreatic cancer, all consecutive patients IS: 0·2 I: 5·9 II: 16·9 III: 71·5 IV: 4·0 n.a.: 1·6 557 (296) 66† ≥ 30: 20·1% TPA (L3) F 362 mm2/m2 M 492 mm2/m2 6 n.r. 5 Jung et al.49 Colorectal cancer All stage III receiving adjuvant chemotherapy 229 (134) 61‡ Sarcopenia: 22·2† (< 30: 87·8%) No sarcopenia: 23·6† (< 30: 71·1%) TPA/m2 (L4) – n.r. 7 7 Lieffers et al.42 Colorectal cancer II: 31·6 III: 35·5 IV: 32·9 234 (135) 63† Overall: 28·5† Sarcopenia: 26·1† No sarcopenia: 30·0† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 6 n.r. n.r. Reisinger et al.43 Colorectal cancer, all consecutive patients I–II: 46·7 III–IV: 53·3 310 (155) 69† > 25: 58·7% CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 7 n.r. n.r. Sabel et al.48 Colorectal cancer, all consecutive patients I: 24 II: 33 II: 30 IV: 11 n.a.: 2 302 (157) 68† Overall: 28·7† PMD (L4) – 7 7 7 Peng et al.45 Colorectal liver metastases, all consecutive patients All stage IV 259 (155) 58‡ ≥ 30: 26·0% TPA/m2 (L3) 500 mm2 6 5 5 van Vledder et al.44 Colorectal liver metastases, all consecutive patients All stage IV 196 (120) 65‡ Sarcopenia: 23·7† No sarcopenia: 26·7† CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 n.r. 7 7 Reference . Malignancy, patient selection . Disease stage (%) . n (men) . Age(years) . BMI(kg/m2) . Muscle(s) measured(level), cut-offs . Quality points by outcome* . . . . . . . . Short-term morbidity or mortality . DFS . OS . Awad et al.37 Oesophageal and gastric cancer, WHO performance status 0–2 Locally advanced 47 (34) 63† Before NACRT: 24·6† Before resection: 23·8† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 5 n.r. 4 Sheetz et al.46 Oesophageal cancer, all consecutive patients IS: 9·6 I: 24·3 II: 33·5 III: 27·4 IV: 5·2 230 (202) 62† Overall: 28·6† TPA, PMD (L4) – 7 6 7 Yip et al.38 Oesophageal cancer IS: 6 I: 3 II: 51 III: 40 35 (30) 63† Before NACRT: 26·7† Before resection: 25·8† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 5 5 5 Harimoto et al.41 Hepatocellular cancer, all consecutive patients I: 15·6 II: 51·1 III: 26·3 IV: 7·0 186 (145) – Sarcopenia: 20·5† No sarcopenia: 24·0† CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 7 6 6 Itoh et al.40 Hepatocellular cancer, all patients without simultaneous procedures n.a. 190 (146) – < 18·5: 7·9% ≥ 18·5 to < 25: 68·4% ≥ 25 to < 30: 21·6% ≥ 30: 2·1% CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 n.r. 6 6 Voron et al.39 Hepatocellular cancer, all consecutive patients n.a. 109 (92) 62† Overall: 24·6† Sarcopenia: 25·6† No sarcopenia: 26·9† CSAMM/m2 (L3) F 38·9 cm2 M 52·4 cm2 7 6 6 Peng et al.47 Pancreatic cancer, all consecutive patients IS: 0·2 I: 5·9 II: 16·9 III: 71·5 IV: 4·0 n.a.: 1·6 557 (296) 66† ≥ 30: 20·1% TPA (L3) F 362 mm2/m2 M 492 mm2/m2 6 n.r. 5 Jung et al.49 Colorectal cancer All stage III receiving adjuvant chemotherapy 229 (134) 61‡ Sarcopenia: 22·2† (< 30: 87·8%) No sarcopenia: 23·6† (< 30: 71·1%) TPA/m2 (L4) – n.r. 7 7 Lieffers et al.42 Colorectal cancer II: 31·6 III: 35·5 IV: 32·9 234 (135) 63† Overall: 28·5† Sarcopenia: 26·1† No sarcopenia: 30·0† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 6 n.r. n.r. Reisinger et al.43 Colorectal cancer, all consecutive patients I–II: 46·7 III–IV: 53·3 310 (155) 69† > 25: 58·7% CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 7 n.r. n.r. Sabel et al.48 Colorectal cancer, all consecutive patients I: 24 II: 33 II: 30 IV: 11 n.a.: 2 302 (157) 68† Overall: 28·7† PMD (L4) – 7 7 7 Peng et al.45 Colorectal liver metastases, all consecutive patients All stage IV 259 (155) 58‡ ≥ 30: 26·0% TPA/m2 (L3) 500 mm2 6 5 5 van Vledder et al.44 Colorectal liver metastases, all consecutive patients All stage IV 196 (120) 65‡ Sarcopenia: 23·7† No sarcopenia: 26·7† CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 n.r. 7 7 * Score from a maximum of 9 using the Newcastle–Ottawa quality assessment scale for cohort studies. † Mean; ‡ median. BMI, body mass index; DFS, disease-free survival; OS, overall survival; WHO, World Health Organization; NACRT, neoadjuvant chemotherapy; CSAMM, cross-sectional area of muscle mass; m2, squared body height; L3/4, at the level of the third/fourth lumbar vertebra; IS, in situ; TPA, total psoas area; PMD, psoas mean density; n.a., not available; n.r., not recorded. Open in new tab Table 1 Studies of the effects of sarcopenia in patients operated on for gastrointestinal and hepatopancreatobiliary malignancies Reference . Malignancy, patient selection . Disease stage (%) . n (men) . Age(years) . BMI(kg/m2) . Muscle(s) measured(level), cut-offs . Quality points by outcome* . . . . . . . . Short-term morbidity or mortality . DFS . OS . Awad et al.37 Oesophageal and gastric cancer, WHO performance status 0–2 Locally advanced 47 (34) 63† Before NACRT: 24·6† Before resection: 23·8† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 5 n.r. 4 Sheetz et al.46 Oesophageal cancer, all consecutive patients IS: 9·6 I: 24·3 II: 33·5 III: 27·4 IV: 5·2 230 (202) 62† Overall: 28·6† TPA, PMD (L4) – 7 6 7 Yip et al.38 Oesophageal cancer IS: 6 I: 3 II: 51 III: 40 35 (30) 63† Before NACRT: 26·7† Before resection: 25·8† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 5 5 5 Harimoto et al.41 Hepatocellular cancer, all consecutive patients I: 15·6 II: 51·1 III: 26·3 IV: 7·0 186 (145) – Sarcopenia: 20·5† No sarcopenia: 24·0† CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 7 6 6 Itoh et al.40 Hepatocellular cancer, all patients without simultaneous procedures n.a. 190 (146) – < 18·5: 7·9% ≥ 18·5 to < 25: 68·4% ≥ 25 to < 30: 21·6% ≥ 30: 2·1% CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 n.r. 6 6 Voron et al.39 Hepatocellular cancer, all consecutive patients n.a. 109 (92) 62† Overall: 24·6† Sarcopenia: 25·6† No sarcopenia: 26·9† CSAMM/m2 (L3) F 38·9 cm2 M 52·4 cm2 7 6 6 Peng et al.47 Pancreatic cancer, all consecutive patients IS: 0·2 I: 5·9 II: 16·9 III: 71·5 IV: 4·0 n.a.: 1·6 557 (296) 66† ≥ 30: 20·1% TPA (L3) F 362 mm2/m2 M 492 mm2/m2 6 n.r. 5 Jung et al.49 Colorectal cancer All stage III receiving adjuvant chemotherapy 229 (134) 61‡ Sarcopenia: 22·2† (< 30: 87·8%) No sarcopenia: 23·6† (< 30: 71·1%) TPA/m2 (L4) – n.r. 7 7 Lieffers et al.42 Colorectal cancer II: 31·6 III: 35·5 IV: 32·9 234 (135) 63† Overall: 28·5† Sarcopenia: 26·1† No sarcopenia: 30·0† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 6 n.r. n.r. Reisinger et al.43 Colorectal cancer, all consecutive patients I–II: 46·7 III–IV: 53·3 310 (155) 69† > 25: 58·7% CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 7 n.r. n.r. Sabel et al.48 Colorectal cancer, all consecutive patients I: 24 II: 33 II: 30 IV: 11 n.a.: 2 302 (157) 68† Overall: 28·7† PMD (L4) – 7 7 7 Peng et al.45 Colorectal liver metastases, all consecutive patients All stage IV 259 (155) 58‡ ≥ 30: 26·0% TPA/m2 (L3) 500 mm2 6 5 5 van Vledder et al.44 Colorectal liver metastases, all consecutive patients All stage IV 196 (120) 65‡ Sarcopenia: 23·7† No sarcopenia: 26·7† CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 n.r. 7 7 Reference . Malignancy, patient selection . Disease stage (%) . n (men) . Age(years) . BMI(kg/m2) . Muscle(s) measured(level), cut-offs . Quality points by outcome* . . . . . . . . Short-term morbidity or mortality . DFS . OS . Awad et al.37 Oesophageal and gastric cancer, WHO performance status 0–2 Locally advanced 47 (34) 63† Before NACRT: 24·6† Before resection: 23·8† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 5 n.r. 4 Sheetz et al.46 Oesophageal cancer, all consecutive patients IS: 9·6 I: 24·3 II: 33·5 III: 27·4 IV: 5·2 230 (202) 62† Overall: 28·6† TPA, PMD (L4) – 7 6 7 Yip et al.38 Oesophageal cancer IS: 6 I: 3 II: 51 III: 40 35 (30) 63† Before NACRT: 26·7† Before resection: 25·8† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 5 5 5 Harimoto et al.41 Hepatocellular cancer, all consecutive patients I: 15·6 II: 51·1 III: 26·3 IV: 7·0 186 (145) – Sarcopenia: 20·5† No sarcopenia: 24·0† CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 7 6 6 Itoh et al.40 Hepatocellular cancer, all patients without simultaneous procedures n.a. 190 (146) – < 18·5: 7·9% ≥ 18·5 to < 25: 68·4% ≥ 25 to < 30: 21·6% ≥ 30: 2·1% CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 n.r. 6 6 Voron et al.39 Hepatocellular cancer, all consecutive patients n.a. 109 (92) 62† Overall: 24·6† Sarcopenia: 25·6† No sarcopenia: 26·9† CSAMM/m2 (L3) F 38·9 cm2 M 52·4 cm2 7 6 6 Peng et al.47 Pancreatic cancer, all consecutive patients IS: 0·2 I: 5·9 II: 16·9 III: 71·5 IV: 4·0 n.a.: 1·6 557 (296) 66† ≥ 30: 20·1% TPA (L3) F 362 mm2/m2 M 492 mm2/m2 6 n.r. 5 Jung et al.49 Colorectal cancer All stage III receiving adjuvant chemotherapy 229 (134) 61‡ Sarcopenia: 22·2† (< 30: 87·8%) No sarcopenia: 23·6† (< 30: 71·1%) TPA/m2 (L4) – n.r. 7 7 Lieffers et al.42 Colorectal cancer II: 31·6 III: 35·5 IV: 32·9 234 (135) 63† Overall: 28·5† Sarcopenia: 26·1† No sarcopenia: 30·0† CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 6 n.r. n.r. Reisinger et al.43 Colorectal cancer, all consecutive patients I–II: 46·7 III–IV: 53·3 310 (155) 69† > 25: 58·7% CSAMM/m2 (L3) F 38·5 cm2 M 52·4 cm2 7 n.r. n.r. Sabel et al.48 Colorectal cancer, all consecutive patients I: 24 II: 33 II: 30 IV: 11 n.a.: 2 302 (157) 68† Overall: 28·7† PMD (L4) – 7 7 7 Peng et al.45 Colorectal liver metastases, all consecutive patients All stage IV 259 (155) 58‡ ≥ 30: 26·0% TPA/m2 (L3) 500 mm2 6 5 5 van Vledder et al.44 Colorectal liver metastases, all consecutive patients All stage IV 196 (120) 65‡ Sarcopenia: 23·7† No sarcopenia: 26·7† CSAMM/m2 (L3) F 41·1 cm2 M 43·75 cm2 n.r. 7 7 * Score from a maximum of 9 using the Newcastle–Ottawa quality assessment scale for cohort studies. † Mean; ‡ median. BMI, body mass index; DFS, disease-free survival; OS, overall survival; WHO, World Health Organization; NACRT, neoadjuvant chemotherapy; CSAMM, cross-sectional area of muscle mass; m2, squared body height; L3/4, at the level of the third/fourth lumbar vertebra; IS, in situ; TPA, total psoas area; PMD, psoas mean density; n.a., not available; n.r., not recorded. Open in new tab Prevalence of sarcopenia in different malignancies The prevalence of sarcopenia as assessed by CT-based skeletal muscle mass measurement in patients undergoing surgery for gastrointestinal and hepatopancreatobiliary malignancies was reported in ten studies37–45,47. None of the studies39,41,42,44,45,49 that compared characteristics in patients with and without sarcopenia reported on significant differences regarding cancer stage, differentiation grade or biomarkers. Despite comparable age and sex distribution between studies, there was a wide variation in the prevalence of sarcopenia, ranging from 17·0 per cent in a cohort of patients with hepatic colorectal metastases45 to 79 per cent in a cohort with oesophageal and gastric cancer37. In agreement, cohorts of patients with oesophageal and gastric cancer reported a widespread prevalence of sarcopenia before surgery, ranging from 43 to 79 per cent37,38. Less variation in the prevalence of sarcopenia was observed among patients undergoing surgical resection of hepatocellular carcinoma (40·3–54·1 per cent)39–41, colorectal cancer (38·9–47·7 per cent)42,43 and hepatic colorectal metastases (17·0–19·4 per cent)44,45. One study47 reported a prevalence of sarcopenia of 25·0 per cent in patients with pancreatic cancer. Two studies37,38 reported an increase in the prevalence of sarcopenia among patients with oesophageal and gastric cancer following neoadjuvant chemotherapy. The impact of neoadjuvant therapy on the prevalence of sarcopenia was not assessed in the colorectal cancer studies included in the present analysis. A possible impact of age or sex on the prevalence of sarcopenia could not be discerned. Detailed information regarding the prevalence of sarcopenia is shown in Table 2. Table 2 Studies reporting the prevalence of sarcopenia in gastrointestinal malignancies Reference . Malignancy . Prevalence (%) . Awad et al.37 Oesophageal and gastric cancer Before NACRT: 57 Before resection: 79 Yip et al.38 Oesophageal cancer Before NACRT: 26 Before resection: 43 Voron et al.39 Hepatocellular carcinoma 54·1 Itoh et al.40 Hepatocellular carcinoma 40·5 Harimoto et al.41 Hepatocellular carcinoma 40·3 Peng et al.47 Pancreatic cancer 25·0 Lieffers et al.42 Colorectal cancer 38·9 Reisinger et al.43 Colorectal cancer 47·7 van Vledder et al.44 Colorectal liver metastases 19·4 Peng et al.45 Colorectal liver metastases 17·0 Reference . Malignancy . Prevalence (%) . Awad et al.37 Oesophageal and gastric cancer Before NACRT: 57 Before resection: 79 Yip et al.38 Oesophageal cancer Before NACRT: 26 Before resection: 43 Voron et al.39 Hepatocellular carcinoma 54·1 Itoh et al.40 Hepatocellular carcinoma 40·5 Harimoto et al.41 Hepatocellular carcinoma 40·3 Peng et al.47 Pancreatic cancer 25·0 Lieffers et al.42 Colorectal cancer 38·9 Reisinger et al.43 Colorectal cancer 47·7 van Vledder et al.44 Colorectal liver metastases 19·4 Peng et al.45 Colorectal liver metastases 17·0 NACRT, neoadjuvant chemotherapy. Open in new tab Table 2 Studies reporting the prevalence of sarcopenia in gastrointestinal malignancies Reference . Malignancy . Prevalence (%) . Awad et al.37 Oesophageal and gastric cancer Before NACRT: 57 Before resection: 79 Yip et al.38 Oesophageal cancer Before NACRT: 26 Before resection: 43 Voron et al.39 Hepatocellular carcinoma 54·1 Itoh et al.40 Hepatocellular carcinoma 40·5 Harimoto et al.41 Hepatocellular carcinoma 40·3 Peng et al.47 Pancreatic cancer 25·0 Lieffers et al.42 Colorectal cancer 38·9 Reisinger et al.43 Colorectal cancer 47·7 van Vledder et al.44 Colorectal liver metastases 19·4 Peng et al.45 Colorectal liver metastases 17·0 Reference . Malignancy . Prevalence (%) . Awad et al.37 Oesophageal and gastric cancer Before NACRT: 57 Before resection: 79 Yip et al.38 Oesophageal cancer Before NACRT: 26 Before resection: 43 Voron et al.39 Hepatocellular carcinoma 54·1 Itoh et al.40 Hepatocellular carcinoma 40·5 Harimoto et al.41 Hepatocellular carcinoma 40·3 Peng et al.47 Pancreatic cancer 25·0 Lieffers et al.42 Colorectal cancer 38·9 Reisinger et al.43 Colorectal cancer 47·7 van Vledder et al.44 Colorectal liver metastases 19·4 Peng et al.45 Colorectal liver metastases 17·0 NACRT, neoadjuvant chemotherapy. Open in new tab Short-term postoperative morbidity and mortality Data regarding complication rate, length of ICU stay, length of hospital stay, postoperative morbidity and postoperative mortality were reported in ten37–39,41–43,45–48 of the 13 studies included in the analysis (Table 3). Table 3 Studies reporting the impact of sarcopenia on short-term outcome in patients operated on for gastrointestinal malignancies Reference . Malignancy . Complications . Length of stay (days) . . . All . Clavien–Dindo classification ≥ IIIa . Postoperative/in-hospital mortality . Anastomotic leakage . Infectious . ICU . Hospital . Awad et al.37* Oesophageal and gastric cancer P = 0·060 P = 0·51 Sheetz et al.46 Oesophageal cancer LPA: 1993 versus 1877 mm2, without versus with complications (P = 0·12) LPA: 1922 versus 1953 mm2, no leakage versus leakage (P = 0·40) Yip et al.38 Oesophageal cancer n.s. n.s. Harimoto et al.41 Hepatocellular carcinoma 32·0 versus 50·5% (P = 0·61) Voron et al.39 Hepatocellular carcinoma 39·0 versus 36·0% (P = 0·749) 20·3 versus 16·0% (P = 0·560) 6·8 versus 2·0% (P = 0·372) Peng et al.47 Pancreatic cancer OR 0·88 (0·60, 1·29) (P = 0·51) OR 0·72 (0·43, 1·21) (P = 0·21) HR 2·31 (0·78, 6·77) (P = 0·13) 0·4 versus 0·4 (P = 0·92) 12 versus 12 (P = 0·98) Lieffers et al.42 Colorectal cancer 23·1 versus 12·6% OR 4·6 (P = 0·007)† 15·9 versus 12·3(P = 0·038) Reisinger et al.43 Colorectal cancer OR 43·3 (2·74, 685·2) (P = 0·007)† OR 0·57 (0·28, 1·19) (P = 0·13) Sabel et al.48 Colorectal cancer OR 0·96 (0·94, 0·99) for every unit of increased psoas density (P = 0·004)† OR 0·95 (0·93, 0·98) for every unit of increased psoas density (P = 0·001)† Peng et al.45 Colorectal liver metastases OR 2·22 (P = 0·02) 22 versus 8% OR 3·12 (1·14, 8·49) (P = 0·02)† Prolonged stay (> 2 days): 15 versus 4% (P = 0·004) 6·6 versus 5·4 (P = 0·03) Reference . Malignancy . Complications . Length of stay (days) . . . All . Clavien–Dindo classification ≥ IIIa . Postoperative/in-hospital mortality . Anastomotic leakage . Infectious . ICU . Hospital . Awad et al.37* Oesophageal and gastric cancer P = 0·060 P = 0·51 Sheetz et al.46 Oesophageal cancer LPA: 1993 versus 1877 mm2, without versus with complications (P = 0·12) LPA: 1922 versus 1953 mm2, no leakage versus leakage (P = 0·40) Yip et al.38 Oesophageal cancer n.s. n.s. Harimoto et al.41 Hepatocellular carcinoma 32·0 versus 50·5% (P = 0·61) Voron et al.39 Hepatocellular carcinoma 39·0 versus 36·0% (P = 0·749) 20·3 versus 16·0% (P = 0·560) 6·8 versus 2·0% (P = 0·372) Peng et al.47 Pancreatic cancer OR 0·88 (0·60, 1·29) (P = 0·51) OR 0·72 (0·43, 1·21) (P = 0·21) HR 2·31 (0·78, 6·77) (P = 0·13) 0·4 versus 0·4 (P = 0·92) 12 versus 12 (P = 0·98) Lieffers et al.42 Colorectal cancer 23·1 versus 12·6% OR 4·6 (P = 0·007)† 15·9 versus 12·3(P = 0·038) Reisinger et al.43 Colorectal cancer OR 43·3 (2·74, 685·2) (P = 0·007)† OR 0·57 (0·28, 1·19) (P = 0·13) Sabel et al.48 Colorectal cancer OR 0·96 (0·94, 0·99) for every unit of increased psoas density (P = 0·004)† OR 0·95 (0·93, 0·98) for every unit of increased psoas density (P = 0·001)† Peng et al.45 Colorectal liver metastases OR 2·22 (P = 0·02) 22 versus 8% OR 3·12 (1·14, 8·49) (P = 0·02)† Prolonged stay (> 2 days): 15 versus 4% (P = 0·004) 6·6 versus 5·4 (P = 0·03) Data are shown for groups with versus without sarcopenia unless indicated otherwise. Odds ratios (ORs) and hazard ratios (HRs) are shown with 95 per cent c.i. * Using fat-free mass assessed by CT before resection. † Multivariable analysis. ICU, intensive care unit; LPA, lean psoas area; n.s., not significant. Open in new tab Table 3 Studies reporting the impact of sarcopenia on short-term outcome in patients operated on for gastrointestinal malignancies Reference . Malignancy . Complications . Length of stay (days) . . . All . Clavien–Dindo classification ≥ IIIa . Postoperative/in-hospital mortality . Anastomotic leakage . Infectious . ICU . Hospital . Awad et al.37* Oesophageal and gastric cancer P = 0·060 P = 0·51 Sheetz et al.46 Oesophageal cancer LPA: 1993 versus 1877 mm2, without versus with complications (P = 0·12) LPA: 1922 versus 1953 mm2, no leakage versus leakage (P = 0·40) Yip et al.38 Oesophageal cancer n.s. n.s. Harimoto et al.41 Hepatocellular carcinoma 32·0 versus 50·5% (P = 0·61) Voron et al.39 Hepatocellular carcinoma 39·0 versus 36·0% (P = 0·749) 20·3 versus 16·0% (P = 0·560) 6·8 versus 2·0% (P = 0·372) Peng et al.47 Pancreatic cancer OR 0·88 (0·60, 1·29) (P = 0·51) OR 0·72 (0·43, 1·21) (P = 0·21) HR 2·31 (0·78, 6·77) (P = 0·13) 0·4 versus 0·4 (P = 0·92) 12 versus 12 (P = 0·98) Lieffers et al.42 Colorectal cancer 23·1 versus 12·6% OR 4·6 (P = 0·007)† 15·9 versus 12·3(P = 0·038) Reisinger et al.43 Colorectal cancer OR 43·3 (2·74, 685·2) (P = 0·007)† OR 0·57 (0·28, 1·19) (P = 0·13) Sabel et al.48 Colorectal cancer OR 0·96 (0·94, 0·99) for every unit of increased psoas density (P = 0·004)† OR 0·95 (0·93, 0·98) for every unit of increased psoas density (P = 0·001)† Peng et al.45 Colorectal liver metastases OR 2·22 (P = 0·02) 22 versus 8% OR 3·12 (1·14, 8·49) (P = 0·02)† Prolonged stay (> 2 days): 15 versus 4% (P = 0·004) 6·6 versus 5·4 (P = 0·03) Reference . Malignancy . Complications . Length of stay (days) . . . All . Clavien–Dindo classification ≥ IIIa . Postoperative/in-hospital mortality . Anastomotic leakage . Infectious . ICU . Hospital . Awad et al.37* Oesophageal and gastric cancer P = 0·060 P = 0·51 Sheetz et al.46 Oesophageal cancer LPA: 1993 versus 1877 mm2, without versus with complications (P = 0·12) LPA: 1922 versus 1953 mm2, no leakage versus leakage (P = 0·40) Yip et al.38 Oesophageal cancer n.s. n.s. Harimoto et al.41 Hepatocellular carcinoma 32·0 versus 50·5% (P = 0·61) Voron et al.39 Hepatocellular carcinoma 39·0 versus 36·0% (P = 0·749) 20·3 versus 16·0% (P = 0·560) 6·8 versus 2·0% (P = 0·372) Peng et al.47 Pancreatic cancer OR 0·88 (0·60, 1·29) (P = 0·51) OR 0·72 (0·43, 1·21) (P = 0·21) HR 2·31 (0·78, 6·77) (P = 0·13) 0·4 versus 0·4 (P = 0·92) 12 versus 12 (P = 0·98) Lieffers et al.42 Colorectal cancer 23·1 versus 12·6% OR 4·6 (P = 0·007)† 15·9 versus 12·3(P = 0·038) Reisinger et al.43 Colorectal cancer OR 43·3 (2·74, 685·2) (P = 0·007)† OR 0·57 (0·28, 1·19) (P = 0·13) Sabel et al.48 Colorectal cancer OR 0·96 (0·94, 0·99) for every unit of increased psoas density (P = 0·004)† OR 0·95 (0·93, 0·98) for every unit of increased psoas density (P = 0·001)† Peng et al.45 Colorectal liver metastases OR 2·22 (P = 0·02) 22 versus 8% OR 3·12 (1·14, 8·49) (P = 0·02)† Prolonged stay (> 2 days): 15 versus 4% (P = 0·004) 6·6 versus 5·4 (P = 0·03) Data are shown for groups with versus without sarcopenia unless indicated otherwise. Odds ratios (ORs) and hazard ratios (HRs) are shown with 95 per cent c.i. * Using fat-free mass assessed by CT before resection. † Multivariable analysis. ICU, intensive care unit; LPA, lean psoas area; n.s., not significant. Open in new tab An increased postoperative morbidity rate was found in patients with sarcopenia in all studies where this was reported among patients undergoing surgical resection of colorectal cancer42,43,48 and hepatic colorectal metastases45. One study48 reported that an increase in psoas density protected against overall (odds ratio (OR) 0·96, 95 per cent c.i. 0·94 to 0·99; P = 0·004) and infectious (OR 0·95, 0·93 to 0·98; P = 0·001) complications in a cohort of 302 patients48. Another investigation42 observed an increase in infectious complications in patients with versus those without sarcopenia (23·1 versus 12·6 per cent; P = 0·036) in a cohort of 234 patients. Subgroup analysis revealed that the risk was especially pronounced in elderly patients (65 years or older) with sarcopenia (29·6 versus 8·8 per cent; P = 0·005). This difference remained significant in multivariable analysis (adjusted OR 4·6, 1·5 to 13·9; P = 0·007). The overall complication rate was not described. An increased risk of major postoperative complications (Clavien–Dindo grade IIIa or higher) in patients with sarcopenia compared with those without was reported among patients undergoing hepatic resection for colorectal metastases (22 versus 8 per cent respectively; OR 3·12; P = 0·02)45. However, the study did not specify the type of complications. Another investigation43 showed a strong association between sarcopenia and 30-day mortality combined with in-hospital mortality after elective colorectal cancer surgery (8·8 versus 0·6 per cent in patients with and without sarcopenia respectively; OR 43·3, 2·74 to 685·2, P = 0·007). No association between sarcopenia and postoperative morbidity and mortality was found in patients undergoing resection for oesophageal or hepatocellular cancer37,39,41,46. Specifically, in a cohort of 557 patients undergoing pancreatic cancer resection47, there was no difference in the rate of any postoperative complication (44·6 versus 51·8 per cent in men with and without sarcopenia respectively, P = 0·28; 41·5 versus 43·4 per cent respectively among women, P = 0·80), major postoperative complications (20·6 versus 24·8 per cent for men, P = 0·49; 12·1 versus 20·5 per cent for women, P = 0·15) or 30-day postoperative mortality (1·4 versus 0·5 per cent for men, P = 0·44; 0 versus 0·5 per cent for women, P = 1·00). However, the 90-day mortality rate differed between men with and without sarcopenia (9·5 versus 2·7 per cent respectively; P = 0·02). Two studies43,46 that reported on anastomotic leakage following surgical resection of colorectal and oesophageal cancer did not demonstrate an association with sarcopenia. Two studies adjusted for body mass index (BMI) in the multivariable analyses. One43 reported that sarcopenia was a risk factor for 30-day mortality, whereas BMI was not. Similarly, in another investigation45 sarcopenia, but not BMI, was a risk factor for postoperative complications. Length of intensive care unit and hospital stay Peng and colleagues reported a prolonged ICU admission (more than 2 days) for patients with sarcopenia undergoing resection with curative intent for hepatic colorectal metastases compared with those without sarcopenia (15 versus 4 per cent respectively; P = 0·004)45, but did not demonstrate a difference in the mean length of ICU stay in patients undergoing surgical resection of pancreatic cancer (mean(s.d.) 0·5(2·0) versus 0·5(1·7) days respectively for men, P = 1·00; 0·2(0·6) versus 0·2(0·6) days among women, P = 0·74)47. In two42,45 of five studies37,38,42,45,47 reporting length of hospital stay, patients with sarcopenia had a delayed discharge from hospital. Hospital stay was slightly prolonged in patients with sarcopenia undergoing resection with curative intent for hepatic colorectal metastases (6·6 versus 5·4 days; P = 0·03)45. The impact of sarcopenia on length of hospital stay may be greater in conjunction with other patient characteristics. For instance, hospital stay was significantly longer in patients with sarcopenia than in those without for all patients undergoing surgery for colorectal cancer (15·9 versus 12·3 days; P = 0·038)42. The corresponding rates for patients aged 65 years or older were 20·2 versus 13·1 days (P = 0·008). In addition, sarcopenia was an independent factor for the need for rehabilitation in patients aged 65 years and older (OR 3·1, 95 per cent c.i. 1·4 to 9·4; P < 0·040)42. The two studies42,45 that reported an increased length of hospital stay in patients with sarcopenia also observed an increased number of postoperative complications. Length of hospital stay did not differ significantly between patients with and those without sarcopenia in studies of pancreatic cancer47 and oesophageal and gastric cancer37,38. Disease-free survival Nine studies38–41,44–46,48,49 described the association between sarcopenia and disease-free survival. Data regarding disease-free survival rates and times in the individual studies are shown in Table 4 and Fig. 3. Table 4 Studies reporting the impact of sarcopenia on long-term outcomes in gastrointestinal malignancies Reference . Malignancy . Disease-free survival . Overall survival . Awad et al.37 Oesophageal and gastric cancer n.r. 12-month mortality equal in patients with low FFM and those with normal FFM after NACRT (P = 0·57) Sheetz et al.46 Oesophageal cancer NACRT: HR 0·83 (0·52, 1·33) for increasing LPA (P = 0·433)* NACRT: HR 0·77 (0·46, 1·28) for increasing LPA (P = 0·311)* No NACRT: HR 0·33 (0·14, 0·80) for increasing LPA (P = 0·014)* No NACRT: HR 0·31 (0·12, 0·82) for increasing LPA (P = 0·018)* Yip et al.38 Oesophageal cancer n.s. After chemotherapy: median 25·6 months versus median not reached (P = 0·063) Itoh et al.40 Hepatocellular carcinoma HR 1·30 (0·85, 2·00) (P = 0·215)* HR 1·96 (1·06, 2·83) (P = 0·031)* Harimoto et al.41 Hepatocellular carcinoma 5-year: 13·0 versus 33·2% (P = 0·013) 5-year: 71 versus 83·7% (P = 0·001) HR 0·97 (0·95, 1·00) for increasing muscle mass (P = 0·016)* HR 0·90 (0·84, 0·96) for increasing muscle mass (P = 0·002)* Voron et al.39 Hepatocellular carcinoma HR 3·03 (1·67, 5·49) (P < 0·001)* HR 3·19 (1·28, 7·96) (P = 0·013)* Peng et al.47 Pancreatic cancer n.r. M, 3-year: 20·3 versus 39·2% (P < 0·050) F, 3-year: 26·1 versus 40·8% (P < 0·050) 3-year: HR 1·63 (1·28, 2·07) (P < 0·001)* Jung et al.49 Colorectal cancer P = 0·946* HR 1·85 (1·10, 3·13) (P = 0·022)* Sabel et al.48 Colorectal cancer HR 0·97 (0·95, 1·00) (P = 0·03) for increasing PD HR 0·97 (0·95, 1·00) for increasing PD (P = 0·04) n.s.* n.s.* Peng et al.45 Colorectal liver metastases HR 1·07 (P = 0·78) HR 1·05 (P = 0·80) van Vledder et al.44 Colorectal liver metastases HR 1·96 (1·29, 2·97) (P = 0·002)* HR 2·69 (1·67, 4·32) (P < 0·001)* Reference . Malignancy . Disease-free survival . Overall survival . Awad et al.37 Oesophageal and gastric cancer n.r. 12-month mortality equal in patients with low FFM and those with normal FFM after NACRT (P = 0·57) Sheetz et al.46 Oesophageal cancer NACRT: HR 0·83 (0·52, 1·33) for increasing LPA (P = 0·433)* NACRT: HR 0·77 (0·46, 1·28) for increasing LPA (P = 0·311)* No NACRT: HR 0·33 (0·14, 0·80) for increasing LPA (P = 0·014)* No NACRT: HR 0·31 (0·12, 0·82) for increasing LPA (P = 0·018)* Yip et al.38 Oesophageal cancer n.s. After chemotherapy: median 25·6 months versus median not reached (P = 0·063) Itoh et al.40 Hepatocellular carcinoma HR 1·30 (0·85, 2·00) (P = 0·215)* HR 1·96 (1·06, 2·83) (P = 0·031)* Harimoto et al.41 Hepatocellular carcinoma 5-year: 13·0 versus 33·2% (P = 0·013) 5-year: 71 versus 83·7% (P = 0·001) HR 0·97 (0·95, 1·00) for increasing muscle mass (P = 0·016)* HR 0·90 (0·84, 0·96) for increasing muscle mass (P = 0·002)* Voron et al.39 Hepatocellular carcinoma HR 3·03 (1·67, 5·49) (P < 0·001)* HR 3·19 (1·28, 7·96) (P = 0·013)* Peng et al.47 Pancreatic cancer n.r. M, 3-year: 20·3 versus 39·2% (P < 0·050) F, 3-year: 26·1 versus 40·8% (P < 0·050) 3-year: HR 1·63 (1·28, 2·07) (P < 0·001)* Jung et al.49 Colorectal cancer P = 0·946* HR 1·85 (1·10, 3·13) (P = 0·022)* Sabel et al.48 Colorectal cancer HR 0·97 (0·95, 1·00) (P = 0·03) for increasing PD HR 0·97 (0·95, 1·00) for increasing PD (P = 0·04) n.s.* n.s.* Peng et al.45 Colorectal liver metastases HR 1·07 (P = 0·78) HR 1·05 (P = 0·80) van Vledder et al.44 Colorectal liver metastases HR 1·96 (1·29, 2·97) (P = 0·002)* HR 2·69 (1·67, 4·32) (P < 0·001)* Data are shown for groups with versus without sarcopenia unless indicated otherwise. Hazard ratios (HRs) are shown with 95 per cent c.i. * Multivariable analysis. n.r., Not reported; FFM, fat-free mass; NACRT, neoadjuvant chemoradiotherapy; LPA, lean psoas area; n.s., not significant; PD, psoas density. Open in new tab Table 4 Studies reporting the impact of sarcopenia on long-term outcomes in gastrointestinal malignancies Reference . Malignancy . Disease-free survival . Overall survival . Awad et al.37 Oesophageal and gastric cancer n.r. 12-month mortality equal in patients with low FFM and those with normal FFM after NACRT (P = 0·57) Sheetz et al.46 Oesophageal cancer NACRT: HR 0·83 (0·52, 1·33) for increasing LPA (P = 0·433)* NACRT: HR 0·77 (0·46, 1·28) for increasing LPA (P = 0·311)* No NACRT: HR 0·33 (0·14, 0·80) for increasing LPA (P = 0·014)* No NACRT: HR 0·31 (0·12, 0·82) for increasing LPA (P = 0·018)* Yip et al.38 Oesophageal cancer n.s. After chemotherapy: median 25·6 months versus median not reached (P = 0·063) Itoh et al.40 Hepatocellular carcinoma HR 1·30 (0·85, 2·00) (P = 0·215)* HR 1·96 (1·06, 2·83) (P = 0·031)* Harimoto et al.41 Hepatocellular carcinoma 5-year: 13·0 versus 33·2% (P = 0·013) 5-year: 71 versus 83·7% (P = 0·001) HR 0·97 (0·95, 1·00) for increasing muscle mass (P = 0·016)* HR 0·90 (0·84, 0·96) for increasing muscle mass (P = 0·002)* Voron et al.39 Hepatocellular carcinoma HR 3·03 (1·67, 5·49) (P < 0·001)* HR 3·19 (1·28, 7·96) (P = 0·013)* Peng et al.47 Pancreatic cancer n.r. M, 3-year: 20·3 versus 39·2% (P < 0·050) F, 3-year: 26·1 versus 40·8% (P < 0·050) 3-year: HR 1·63 (1·28, 2·07) (P < 0·001)* Jung et al.49 Colorectal cancer P = 0·946* HR 1·85 (1·10, 3·13) (P = 0·022)* Sabel et al.48 Colorectal cancer HR 0·97 (0·95, 1·00) (P = 0·03) for increasing PD HR 0·97 (0·95, 1·00) for increasing PD (P = 0·04) n.s.* n.s.* Peng et al.45 Colorectal liver metastases HR 1·07 (P = 0·78) HR 1·05 (P = 0·80) van Vledder et al.44 Colorectal liver metastases HR 1·96 (1·29, 2·97) (P = 0·002)* HR 2·69 (1·67, 4·32) (P < 0·001)* Reference . Malignancy . Disease-free survival . Overall survival . Awad et al.37 Oesophageal and gastric cancer n.r. 12-month mortality equal in patients with low FFM and those with normal FFM after NACRT (P = 0·57) Sheetz et al.46 Oesophageal cancer NACRT: HR 0·83 (0·52, 1·33) for increasing LPA (P = 0·433)* NACRT: HR 0·77 (0·46, 1·28) for increasing LPA (P = 0·311)* No NACRT: HR 0·33 (0·14, 0·80) for increasing LPA (P = 0·014)* No NACRT: HR 0·31 (0·12, 0·82) for increasing LPA (P = 0·018)* Yip et al.38 Oesophageal cancer n.s. After chemotherapy: median 25·6 months versus median not reached (P = 0·063) Itoh et al.40 Hepatocellular carcinoma HR 1·30 (0·85, 2·00) (P = 0·215)* HR 1·96 (1·06, 2·83) (P = 0·031)* Harimoto et al.41 Hepatocellular carcinoma 5-year: 13·0 versus 33·2% (P = 0·013) 5-year: 71 versus 83·7% (P = 0·001) HR 0·97 (0·95, 1·00) for increasing muscle mass (P = 0·016)* HR 0·90 (0·84, 0·96) for increasing muscle mass (P = 0·002)* Voron et al.39 Hepatocellular carcinoma HR 3·03 (1·67, 5·49) (P < 0·001)* HR 3·19 (1·28, 7·96) (P = 0·013)* Peng et al.47 Pancreatic cancer n.r. M, 3-year: 20·3 versus 39·2% (P < 0·050) F, 3-year: 26·1 versus 40·8% (P < 0·050) 3-year: HR 1·63 (1·28, 2·07) (P < 0·001)* Jung et al.49 Colorectal cancer P = 0·946* HR 1·85 (1·10, 3·13) (P = 0·022)* Sabel et al.48 Colorectal cancer HR 0·97 (0·95, 1·00) (P = 0·03) for increasing PD HR 0·97 (0·95, 1·00) for increasing PD (P = 0·04) n.s.* n.s.* Peng et al.45 Colorectal liver metastases HR 1·07 (P = 0·78) HR 1·05 (P = 0·80) van Vledder et al.44 Colorectal liver metastases HR 1·96 (1·29, 2·97) (P = 0·002)* HR 2·69 (1·67, 4·32) (P < 0·001)* Data are shown for groups with versus without sarcopenia unless indicated otherwise. Hazard ratios (HRs) are shown with 95 per cent c.i. * Multivariable analysis. n.r., Not reported; FFM, fat-free mass; NACRT, neoadjuvant chemoradiotherapy; LPA, lean psoas area; n.s., not significant; PD, psoas density. Open in new tab Fig. 3 Open in new tabDownload slide Forest plots showing studies that reported disease-free survival. Only studies reporting hazard ratios with 95 per cent c.i. are shown. NACRT, neoadjuvant chemoradiotherapy In patients with oesophageal cancer, sarcopenia was associated with impaired disease-free survival in those who underwent surgical resection without receiving neoadjuvant chemoradiotherapy independently of age, sex and tumour stage (hazard ratio (HR) 0·33, 95 per cent c.i. 0·14 to 0·80; P = 0·014)46. However, no association between sarcopenia and disease-free survival was observed in patients who underwent surgical resection following neoadjuvant chemoradiotherapy38,46. Patients with hepatocellular cancer who had sarcopenia had an increased risk of disease recurrence in two39,41 of three39–41 studies. One study39 reported a median disease-free survival of 10·1 months in patients with sarcopenia and 34·2 months in those without sarcopenia (P < 0·001), and an independent association between sarcopenia and disease-free survival (HR 3·03, 95 per cent c.i. 1·67 to 5·49; P < 0·001). Another study41 reported 5-year disease-free survival rates in patients with and without sarcopenia of 13·0 and 33·2 per cent respectively (P = 0·013). In multivariable analysis, a high skeletal muscle mass was independently associated with a lower risk of disease recurrence (HR 0·97, 0·95 to 1·00; P = 0·016). Yet another study40 reported reduced disease-free survival in patients undergoing hepatocellular cancer resection in univariable analysis (HR 1·62, 1·11 to 2·36; P = 0·012), but this association did not remain significant in the multivariable analysis (HR 1·30, 0·85 to 2·00; P = 0·215). In patients with primary colorectal cancer, sarcopenia impaired disease-free survival in one48 of the two studies48,49 reporting on disease recurrence. One study48 described a protective effect of high psoas muscle density (HR 0·97, 0·95 to 1·00; P = 0·03). However, there was no significant difference in disease-free survival between patients with normal and low skeletal muscle mass in another study49. No median survival times, or 1-, 3- or 5-year survival rates were reported. In patients with hepatic colorectal metastases, one study44 reported a median disease-free survival time of 8·7 months in patients with sarcopenia compared with 15·1 months in patients without sarcopenia (HR 1·96, 1·29 to 2·97; P = 0·002). However, another investigation45 found no association between sarcopenia and disease-free survival in patients with hepatic colorectal metastases; the 5-year recurrence-free survival rate was 23 and 27 per cent in patients with and without sarcopenia respectively (P = 0·78). Five studies39–41,44,49 made an adjustment for BMI in the analysis of the prognostic value of sarcopenia for disease-free survival. Whereas sarcopenia was associated with disease-free survival in four of nine studies, no association between BMI and disease-free survival was reported in patients with hepatocellular cancer, colorectal cancer or hepatic colorectal metastases. Overall survival Most authors reported a significant decrease in overall survival in patients with sarcopenia compared with those without sarcopenia. This effect was observed irrespective of cancer site or tumour origin39–41,44,46–49. Data regarding survival rates and median survival times in the individual studies are shown in Table 4 and Fig. 4. Fig. 4 Open in new tabDownload slide Forest plots showing studies that reported overall survival. Only studies reporting hazard ratios with 95 per cent c.i. are shown. NACRT, neoadjuvant chemoradiotherapy In patients with oesophageal cancer, a trend towards decreased survival among those with sarcopenia was reported in one study38 (median overall survival 25·6 months versus median not reached for patients without sarcopenia; P = 0·063). In another study46, overall survival was impaired in patients who had oesophageal cancer with sarcopenia and did not receive neoadjuvant chemotherapy (HR 0·31, 95 per cent c.i. 0·12 to 0·82; P = 0·018), whereas no significant association was found among patients who did receive neoadjuvant chemotherapy (HR 0·77, 0·46 to 1·28; P = 0·311). A study39 among patients with hepatocellular carcinoma reported a median survival time of 52·3 and 70·3 months in patients with and without sarcopenia respectively (P = 0·015), with a remarkably impaired 1-year survival rate (69·8 versus 95·5 per cent; P = 0·015). Another investigation41 described a less severe impact of sarcopenia on survival in patients with hepatocellular carcinoma, with a reduction in 5-year survival rate from 83·7 to 71 per cent (P = 0·001). In a study47 of patients who had surgery for pancreatic cancer, the 3-year survival rate was lower in patients with sarcopenia than in those without (20·3 versus 39·2 per cent in men, P = 0·003; 26·1 versus 40·8 per cent in women, P = 0·03). In the multivariable analysis, sarcopenia remained independently associated with an increased risk of death at 3 years (HR 1·63, 1·28 to 2·07; P < 0·001). Median overall survival times, or 1-, 3- or 5-year survival rates were not reported for patients with colorectal cancer in any of the included studies. In patients with hepatic colorectal metastases, one study44 reported a median survival time of 23·8 versus 59·8 months for patients with and without sarcopenia (HR 2·69, 1·67 to 4·32; P < 0·001). Two studies38,45 found a decreased overall survival in patients with sarcopenia in univariable but not in multivariable analyses. Five studies undertook multivariable analysis in which the predictive effect of sarcopenia on overall survival was adjusted for BMI. Sarcopenia was independently associated with overall survival in seven of ten studies, whereas no association between BMI and overall survival was reported in patients with hepatocellular cancer and hepatic colorectal metastases39–41,44. However, one study49 found that a BMI of 25 kg/m2 or higher was a risk factor for impaired overall survival independent of sarcopenia in patients with stage III colorectal cancer who received adjuvant chemotherapy. Discussion Several conclusions can be drawn from this systematic review of the impact of CT-assessed sarcopenia on short- and long-term outcomes in resectable gastrointestinal and hepatopancreatobiliary malignancies. Sarcopenia decreased overall survival, and increased recurrence rates following surgical resection in patients with hepatic colorectal metastases and hepatocellular cancer. Patients with sarcopenia undergoing surgery for colorectal cancer and hepatic colorectal metastases also had a prolonged length of stay and increased postoperative morbidity after surgery. Because of the heterogeneity of the included studies, the possible influence of age and sex on the prevalence of sarcopenia could not be assessed. A previous review50 described the relationship between CT-assessed core muscle size and mortality, postoperative morbidity and length of stay after major abdominal surgery. This systematic review included eight retrospective cohort studies, of which five investigated outcomes in oncological populations. As in the present investigation, sarcopenia was associated with increased morbidity, length of hospital stay and mortality. The relationship between sarcopenia and recurrence was not described. Preoperative risk stratification is of utmost importance in patient selection for surgery, as it may help physicians to identify patients with a high risk of worse outcome after surgery. A tool suitable for risk evaluation should be inexpensive, easily obtainable and reliable. Bioelectrical impedance analysis, dual-energy X-ray absorptiometry and skinfold measurement are often not performed routinely during the oncological evaluation, whereas the majority of patients undergo abdominal CT as part of preoperative investigations. Cross-sectional muscle area can be measured rapidly by single-slice analysis of abdominal CT images, and is linearly related to total body skeletal muscle mass51; this measurement has a low level of interobserver variability43,51. CT-based skeletal muscle mass measurement in patients with cancer may identify those at an early stage of frailty, which would otherwise have been undetected clinically52. It is still unknown whether treatment of sarcopenia may improve outcomes. Understanding of muscle wasting in cancer has greatly increased over the past decade53,54 and has led to new treatment options, such as myostatin inhibitors55,56. A phase II clinical trial on the efficacy of myostatin inhibitors in patients with advanced or metastatic pancreatic cancer receiving chemotherapy is ongoing, with overall survival as the primary endpoint. There are, however, several other ongoing clinical trials investigating stabilization or reversal of muscle loss in patients with cancer57. The present study has some limitations. The included studies were heterogeneous in design and were predominantly retrospective, observational studies, which precluded meta-analysis of the results. Consequently, no causative relationship between sarcopenia and outcome could be demonstrated. Furthermore, the present investigation is likely to have been influenced by submission or publication bias. As there is no standard definition of CT-based assessment of muscle mass, different methods were used, which also hampered evaluation of the results. Investigations that measured total cross-sectional area of muscle mass used distinct sex-specific cut-off values23,44. These cut-off values were obtained using the same method of stratification in two different patient populations, yielding two distinct sets of cut-off values. As such, these values may not be interchangeable and applicable to all populations. Another recent study58 developed a third set of cut-off values, which were both sex- and BMI-specific; these included muscle attenuation, based on Hounsfield units, as a marker for fat infiltration of muscle. These cut-off values remain to be validated. Sarcopenia impairs overall survival, and may increase postoperative morbidity. Muscle mass assessment by CT may assist in preoperative decision-making, particularly for patients who are thought to be unfit for surgery or who have a high risk of complications. Acknowledgements S.L. and J.L.A.v.V. contributed equally to this manuscript. Disclosure: The authors declare no conflict of interest. References 1 Kehlet H , Wilmore DW. Multimodal strategies to improve surgical outcome . Am J Surg 2002 ; 183 : 630 – 641 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Varadhan KK , Neal KR, Dejong CH, Fearon KC, Ljungqvist O, Lobo DN. The enhanced recovery after surgery (ERAS) pathway for patients undergoing major elective open colorectal surgery: a meta-analysis of randomized controlled trials . Clin Nutr 2010 ; 29 : 434 – 440 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Spanjersberg WR , Reurings J, Keus F, van Laarhoven CJ. Fast track surgery versus conventional recovery strategies for colorectal surgery . Cochrane Database Syst Rev 2011 ; ( 2 ) CD007635 . Google Scholar OpenURL Placeholder Text WorldCat 4 Kehlet H Fast-track colorectal surgery . Lancet 2008 ; 371 : 791 – 793 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Zhuang CL , Ye XZ, Zhang XD, Chen BC, Yu Z. Enhanced recovery after surgery programs versus traditional care for colorectal surgery: a meta-analysis of randomized controlled trials . Dis Colon Rectum 2013 ; 56 : 667 – 678 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Khuri SF , Henderson WG, DePalma RG, Mosca C, Healey NA, Kumbhani DJ; Participants in the VA National Surgical Quality Improvement Program. Determinants of long-term survival after major surgery and the adverse effect of postoperative complications . Ann Surg 2005 ; 242 : 326 – 341 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Kirchhoff P , Clavien PA, Hahnloser D. Complications in colorectal surgery: risk factors and preventive strategies . Patient Saf Surg 2010 ; 4 : 5 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Pedersen T Complications and death following anaesthesia. A prospective study with special reference to the influence of patient-, anaesthesia-, and surgery-related risk factors . Dan Med Bull 1994 ; 41 : 319 – 931 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 9 Tan KY , Kawamura YJ, Tokomitsu A, Tang T. Assessment for frailty is useful for predicting morbidity in elderly patients undergoing colorectal cancer resection whose comorbidities are already optimized . Am J Surg 2012 ; 204 : 139 – 143 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Saxton A , Velanovich V. Preoperative frailty and quality of life as predictors of postoperative complications . Ann Surg 2011 ; 253 : 1223 – 1229 . Google Scholar Crossref Search ADS PubMed WorldCat 11 DePinho RA . The age of cancer . Nature 2000 ; 408 : 248 – 254 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Papamichael D , Audisio RA, Glimelius B, de Gramont A, Glynne-Jones R, Haller D et al. Treatment of colorectal cancer in older patients: International Society of Geriatric Oncology (SIOG) consensus recommendations 2013 . Ann Oncol 2015 ; 26 : 463 – 476 . Google Scholar Crossref Search ADS PubMed WorldCat 13 Hamaker ME , Prins MC, Schiphorst AH, van Tuyl SA, Pronk A, van den Bos F. Long-term changes in physical capacity after colorectal cancer treatment . J Geriatr Oncol 2015 ; 6 : 153 – 164 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Fried LP , Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J et al. Frailty in older adults: evidence for a phenotype . J Gerontol A Biol Sci Med Sci 2001 ; 56 : M146 – M156 . Google Scholar Crossref Search ADS PubMed WorldCat 15 Milte R , Crotty M. Musculoskeletal health, frailty and functional decline . Best Pract Res Clin Rheumatol 2014 ; 28 : 395 – 410 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Cooper C , Dere W, Evans W, Kanis JA, Rizzoli R, Sayer AA et al. Frailty and sarcopenia: definitions and outcome parameters . Osteoporos Int 2012 ; 23 : 1839 – 1848 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Roubenoff R Sarcopenia: a major modifiable cause of frailty in the elderly . J Nutr Health Aging 2000 ; 4 : 140 – 142 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 18 Marzetti E , Leeuwenburgh C. Skeletal muscle apoptosis, sarcopenia and frailty at old age . Exp Gerontol 2006 ; 41 : 1234 – 1238 . Google Scholar Crossref Search ADS PubMed WorldCat 19 Cherin P , Voronska E, Fraoucene N, de Jaeger C. Prevalence of sarcopenia among healthy ambulatory subjects: the sarcopenia begins from 45 years . Aging Clin Exp Res 2014 ; 26 : 137 – 146 . Google Scholar Crossref Search ADS PubMed WorldCat 20 Cruz-Jentoft AJ , Baeyens JP, Bauer JM, Boirie Y, Cederholm T, Landi F et al. Sarcopenia: European consensus on definition and diagnosis: Report of the European Working Group on Sarcopenia in Older People . Age Ageing 2010 ; 39 : 412 – 423 . Google Scholar Crossref Search ADS PubMed WorldCat 21 Baumgartner RN , Koehler KM, Gallagher D, Romero L, Heymsfield SB, Ross RR et al. Epidemiology of sarcopenia among the elderly in New Mexico . Am J Epidemiol 1998 ; 147 : 755 – 163 . Google Scholar Crossref Search ADS PubMed WorldCat 22 Janssen I , Heymsfield SB, Ross R. Low relative skeletal muscle mass (sarcopenia) in older persons is associated with functional impairment and physical disability . J Am Geriatr Soc 2002 ; 50 : 889 – 896 . Google Scholar Crossref Search ADS PubMed WorldCat 23 Prado CM , Lieffers JR, McCargar LJ, Reiman T, Sawyer MB, Martin L et al. Prevalence and clinical implications of sarcopenic obesity in patients with solid tumours of the respiratory and gastrointestinal tracts: a population-based study . Lancet Oncol 2008 ; 9 : 629 – 635 . Google Scholar Crossref Search ADS PubMed WorldCat 24 Fearon K , Strasser F, Anker SD, Bosaeus I, Bruera E, Fainsinger RL et al. Definition and classification of cancer cachexia: an international consensus . Lancet Oncol 2011 ; 12 : 489 – 495 . Google Scholar Crossref Search ADS PubMed WorldCat 25 Evans WJ . Skeletal muscle loss: cachexia, sarcopenia, and inactivity . Am J Clin Nutr 2010 ; 91 : 1123S – 1127S . Google Scholar Crossref Search ADS PubMed WorldCat 26 Evans WJ , Morley JE, Argilés J, Bales C, Baracos V, Guttridge D et al. Cachexia: a new definition . Clin Nutr 2008 ; 27 : 793 – 799 . Google Scholar Crossref Search ADS PubMed WorldCat 27 Bachmann J , Heiligensetzer M, Krakowski-Roosen H, Büchler MW, Friess H, Martignoni ME. Cachexia worsens prognosis in patients with resectable pancreatic cancer . J Gastrointest Surg 2008 ; 12 : 1193 – 1201 . Google Scholar Crossref Search ADS PubMed WorldCat 28 Antoun S , Baracos VE, Birdsell L, Escudier B, Sawyer MB. Low body mass index and sarcopenia associated with dose-limiting toxicity of sorafenib in patients with renal cell carcinoma . Ann Oncol 2010 ; 21 : 1594 – 1598 . Google Scholar Crossref Search ADS PubMed WorldCat 29 Prado CM , Baracos VE, McCargar LJ, Reiman T, Mourtzakis M, Tonkin K et al. Sarcopenia as a determinant of chemotherapy toxicity and time to tumor progression in metastatic breast cancer patients receiving capecitabine treatment . Clin Cancer Res 2009 ; 15 : 2920 – 2926 . Google Scholar Crossref Search ADS PubMed WorldCat 30 Fearon KC . Cancer cachexia: developing multimodal therapy for a multidimensional problem . Eur J Cancer 2008 ; 44 : 1124 – 1132 . Google Scholar Crossref Search ADS PubMed WorldCat 31 Tisdale MJ . Mechanisms of cancer cachexia . Physiol Rev 2009 ; 89 : 381 – 410 . Google Scholar Crossref Search ADS PubMed WorldCat 32 Acharyya S , Butchbach ME, Sahenk Z, Wang H, Saji M, Carathers M et al. Dystrophin glycoprotein complex dysfunction: a regulatory link between muscular dystrophy and cancer cachexia . Cancer Cell 2005 ; 8 : 421 – 432 . Google Scholar Crossref Search ADS PubMed WorldCat 33 Loberg RD , Bradley DA, Tomlins SA, Chinnaiyan AM, Pienta KJ. The lethal phenotype of cancer: the molecular basis of death due to malignancy . CA Cancer J Clin 2007 ; 57 : 225 – 241 . Google Scholar Crossref Search ADS PubMed WorldCat 34 Prado CM , Birdsell LA, Baracos VE. The emerging role of computerized tomography in assessing cancer cachexia . Curr Opin Support Palliat Care 2009 ; 3 : 269 – 275 . Google Scholar Crossref Search ADS PubMed WorldCat 35 Moher D , Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement . Int J Surg 2010 ; 8 : 336 – 341 . Google Scholar Crossref Search ADS PubMed WorldCat 36 Wells GA , Shea B, O'Connell D, Peterson J, Welch V, Losos M et al. The Newcastle–Ottawa Scale (NOS) for Assessing the Quality of Nonrandomised Studies in Meta-Analyses . http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp [accessed 20 May 2015]. 37 Awad S , Tan BH, Cui H, Bhalla A, Fearon KC, Parsons CL et al. Marked changes in body composition following neoadjuvant chemotherapy for oesophagogastric cancer . Clin Nutr 2012 ; 31 : 74 – 77 . Google Scholar Crossref Search ADS PubMed WorldCat 38 Yip C , Goh V, Davies A, Gossage J, Mitchell-Hay R, Hynes O et al. Assessment of sarcopenia and changes in body composition after neoadjuvant chemotherapy and associations with clinical outcomes in oesophageal cancer . Eur Radiol 2014 ; 24 : 998 – 1005 . Google Scholar Crossref Search ADS PubMed WorldCat 39 Voron T , Tselikas L, Pietrasz D, Pigneur F, Laurent A, Compagnon P et al. Sarcopenia impacts on short- and long-term results of hepatectomy for hepatocellular carcinoma . Ann Surg 2015 ; 261 : 1173 – 1193 . Google Scholar Crossref Search ADS PubMed WorldCat 40 Itoh S , Shirabe K, Matsumoto Y, Yoshiya S, Muto J, Harimoto N et al. Effect of body composition on outcomes after hepatic resection for hepatocellular carcinoma . Ann Surg Oncol 2014 ; 21 : 3063 – 3068 . Google Scholar Crossref Search ADS PubMed WorldCat 41 Harimoto N , Shirabe K, Yamashita YI, Ikegami T, Yoshizumi T, Soejima Y et al. Sarcopenia as a predictor of prognosis in patients following hepatectomy for hepatocellular carcinoma . Br J Surg 2013 ; 100 : 1523 – 1530 . Google Scholar Crossref Search ADS PubMed WorldCat 42 Lieffers JR , Bathe OF, Fassbender K, Winget M, Baracos VE. Sarcopenia is associated with postoperative infection and delayed recovery from colorectal cancer resection surgery . Br J Cancer 2012 ; 107 : 931 – 936 . Google Scholar Crossref Search ADS PubMed WorldCat 43 Reisinger KW , van Vugt JLA, Tegels JJ, Snijders C, Hulsewé KW, Hoofwijk AG et al. Functional compromise reflected by sarcopenia, frailty, and nutritional depletion predicts adverse postoperative outcome after colorectal cancer surgery Ann Surg 2015 ; 261 : 345 – 352 . Google Scholar Crossref Search ADS PubMed WorldCat 44 Van Vledder MG , Levolger S, Ayez N, Verhoef C, Tran TC, IJzermans JN. Body composition and outcome in patients undergoing resection of colorectal liver metastases . Br J Surg 2012 ; 99 : 550 – 557 . Google Scholar Crossref Search ADS PubMed WorldCat 45 Peng PD , Van Vledder MG, Tsai S, de Jong MC, Makary M, Ng J et al. Sarcopenia negatively impacts short-term outcomes in patients undergoing hepatic resection for colorectal liver metastasis . HPB (Oxford) 2011 ; 13 : 439 – 446 . Google Scholar Crossref Search ADS PubMed WorldCat 46 Sheetz KH , Zhao L, Holcombe SA, Wang SC, Reddy RM, Lin J et al. Decreased core muscle size is associated with worse patient survival following esophagectomy for cancer . Dis Esophagus 2013 ; 26 : 716 – 722 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 47 Peng P , Hyder O, Firoozmand A, Kneuertz P, Schulick RD, Huang D et al. Impact of sarcopenia on outcomes following resection of pancreatic adenocarcinoma . J Gastrointest Surg 2012 ; 16 : 1478 – 1486 . Google Scholar Crossref Search ADS PubMed WorldCat 48 Sabel MS , Terjimanian M, Conlon AS, Griffith KA, Morris AM, Mulholland MW et al. Analytic morphometric assessment of patients undergoing colectomy for colon cancer . J Surg Oncol 2013 ; 108 : 169 – 175 . Google Scholar Crossref Search ADS PubMed WorldCat 49 Jung HW , Kim JW, Kim JY, Kim SW, Yang HK, Lee JW et al. Effect of muscle mass on toxicity and survival in patients with colon cancer undergoing adjuvant chemotherapy . Supportive Care Cancer 2015 ; 23 : 687 – 694 . Google Scholar Crossref Search ADS WorldCat 50 Hasselager R , Gogenur I. Core muscle size assessed by perioperative abdominal CT scan is related to mortality, postoperative complications, and hospitalization after major abdominal surgery: a systematic review . Langenbecks Arch Surg 2014 ; 399 : 287 – 295 . Google Scholar Crossref Search ADS PubMed WorldCat 51 Shen W , Punyanitya M, Wang Z, Gallagher D, St-Onge MP, Albu J et al. Total body skeletal muscle and adipose tissue volumes: estimation from a single abdominal cross-sectional image . J Appl Physiol (1985) 2004 ; 97 : 2333 – 2338 . Google Scholar Crossref Search ADS PubMed WorldCat 52 Lang PO , Michel JP, Zekry D. Frailty syndrome: a transitional state in a dynamic process . Gerontology 2009 ; 55 : 539 – 549 . Google Scholar Crossref Search ADS PubMed WorldCat 53 Fearon KC , Glass DJ, Guttridge DC. Cancer cachexia: mediators, signaling, and metabolic pathways . Cell Metab 2012 ; 16 : 153 – 166 . Google Scholar Crossref Search ADS PubMed WorldCat 54 Argiles JM , Busquets S, Stemmler B, López-Soriano FJ. Cancer cachexia: understanding the molecular basis . Nat Rev Cancer 2014 ; 14 : 754 – 762 . Google Scholar Crossref Search ADS PubMed WorldCat 55 Zhou X , Wang JL, Lu J, Song Y, Kwak KS, Jiao Q et al. Reversal of cancer cachexia and muscle wasting by ActRIIB antagonism leads to prolonged survival . Cell 2010 ; 142 : 531 – 543 . Google Scholar Crossref Search ADS PubMed WorldCat 56 Jameson GS , Hoff DD, Weiss GJ, Donald R, Smith DA, Becerra C et al. Safety of the antimyostatin monoclonal antibody LY2495655 in healthy subjects and patients with advanced cancer . J Clin Oncol 2012 ; 30 (Suppl): Abstract 2516. Google Scholar OpenURL Placeholder Text WorldCat 57 Dingemans AM , de Vos-Geelen J, Langen R, Schols AM. Phase II drugs that are currently in development for the treatment of cachexia . Expert Opin Investig Drugs 2014 ; 23 : 1655 – 1669 . Google Scholar Crossref Search ADS PubMed WorldCat 58 Martin L , Birdsell L, Macdonald N, Reiman T, Clandinin MT, McCargar LJ et al. Cancer cachexia in the age of obesity: skeletal muscle depletion is a powerful prognostic factor, independent of body mass index . J Clin Oncol 2013 ; 31 : 1539 – 1547 . Google Scholar Crossref Search ADS PubMed WorldCat © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd

Chandrasegaram, M D; Goldstein, D; Simes, J; Gebski, V; Kench, J G; Gill, A J; Samra, J S; Merrett, N D; Richardson, A J; Barbour, A P

doi: 10.1002/bjs.9892pmid: 26350029

Abstract Background R0 resection rates (complete tumour removal with negative resection margins) in pancreatic cancer are 70–80 per cent when a 0-mm margin is used, declining to 15–24 per cent with a 1-mm margin. This review evaluated the R0 resection rates according to different margin definitions and techniques. Methods Three databases (MEDLINE from 1946, PubMed from 1946 and Embase from 1949) were searched to mid-October 2014. The search terms included ‘pancreatectomy OR pancreaticoduodenectomy’ and ‘margin’. A meta-analysis was performed with studies in three groups: group 1, axial slicing technique (minimum 1-mm margin); group 2, other slicing techniques (minimum 1-mm margin); and group 3, studies with minimum 0-mm margin. Results The R0 rates were 29 (95 per cent c.i. 26 to 32) per cent in group 1 (8 studies; 882 patients) and 49 (47 to 52) per cent in group 2 (6 studies; 1568 patients). The combined R0 rate (groups 1 and 2) was 41 (40 to 43) per cent. The R0 rate in group 3 (7 studies; 1926 patients) with a 0-mm margin was 72 (70 to 74) per cent The survival hazard ratios (R1 resection/R0 resection) revealed a reduction in the risk of death of at least 22 per cent in group 1, 12 per cent in group 2 and 23 per cent in group 3 with an R0 compared with an R1 resection. Local recurrence occurred more frequently with an R1 resection in most studies. Conclusion Margin clearance definitions affect R0 resection rates in pancreatic cancer surgery. This review collates individual studies providing an estimate of achievable R0 rates, creating a benchmark for future trials. Introduction Pancreatic cancer has a poor prognosis, with only around 20 per cent of patients having potentially resectable disease after staging1. After full assessment including co-morbidities, only 7–12 per cent of patients undergo operative resection2,3. Worldwide rates of complete tumour resection with negative resection margins (R0 resection) on final pathology have ranged between 70 and 80 per cent for pancreatic cancer surgery in the past. Since 2005, with the advent of detailed three-dimensional pathological assessment in Europe, R0 resection rates have declined remarkably to 15–24 per cent4–7. Although many centres in Europe and Australia have adopted the minimum 1-mm margin to define an R0 resection, some centres still use the 0-mm minimum margin definition8,9. These differences make comparisons between studies and trials difficult. However, a recent consensus statement by the International Study Group of Pancreatic Surgery (ISGPS)10 on borderline resectable tumours has suggested a 1-mm margin for R0 resection with recommendations on minimum reporting on seven margins, which include the anterior, posterior, superior mesenteric vein (SMV) groove, superior mesenteric artery (SMA), bile duct (BD) and enteric margins. However, the paper does cite Jamieson and colleagues11, who found no prognostic significance associated with involvement of the anterior margin and posterior margins. Although the 1-mm margin is increasingly endorsed, a more rigorous margin clearance of 2 mm has been proposed as a superior prognostic factor for overall survival12. Recent studies have reported wide variation in R0 resection rates. The reasons for this lie in the lack of international consensus on the definition of microscopic margin involvement, the definition of what constitutes the circumferential resection margin in pancreaticoduodenectomy specimens, and the lack of a standard protocol for the examination of these specimens13. The aims of this meta-analysis were to evaluate the rates of R0 and R1 (resection margin involved by tumour cells) resection among patients undergoing pancreatic resection for pancreatic cancer according to the minimum 1- and 0-mm margin clearance definitions, and to assess whether differences in pathological examination affected the R0 rates and the sites of margin involvement. Correlations with recurrence and survival were also examined among studies reporting these outcomes. Methods This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines14. Three databases (MEDLINE from 1946, PubMed from 1946 and Embase from 1949) were searched to the second week of October 2014. The search terms included ‘pancreatectomy OR pancreaticoduodenectomy’ and ‘margin’. Study selection The studies included patients who underwent a pancreatic resection (pancreaticoduodenectomy, distal pancreatectomy and total pancreatectomy) for pancreatic adenocarcinoma. Studies were included only if their method of margin assessment was clearly described and margins were analysed with either a 0- or 1-mm margin. Studies that included combined results for other periampullary cancers such as ampullary, distal BD and duodenal cancers were excluded. Studies restricted to borderline or advanced pancreatic cancers, as defined by the authors, or that combined data from several different pathological assessment techniques during the course of the study, were excluded. Studies that were unclear in their pathological assessment of margins were also excluded, as were non-English-language studies. Data extraction Data extraction was done using a standard form. Information collected included: year of publication, origin, number of pancreatic resections, margin assessment, slicing techniques (axial versus other), R0/R1 rates by 0- versus 1-mm definitions, size of tumours if available, stage of tumours if available, survival data if available, use of neoadjuvant and adjuvant treatment, and vascular resection rates. Study groups by pathological margin assessment Studies were analysed according to their minimum margin assessment of 0 versus 1 mm. Studies that analysed a 1-mm margin using the axial slicing technique were grouped together, because this technique has been reported to be associated with the lowest R0 rates15, in order to form a more homogeneous group in terms of pathological technique. The data were analysed in three groups: group 1, axial slicing technique with minimum 1-mm margin for R0; group 2, other slicing techniques with minimum 1-mm margin for R0; and group 3, studies that used a minimum 0-mm margin for R0. In R1 resections, reported site of margin involvement was analysed. Margins reported in the studies were the medial margin, SMA margin, retroperitoneal margin, uncinate margin, SMV/portal vein (PV) margin, pancreatic transection margin, BD margin, proximal gastric or duodenal margin, distal duodenal/jejunal margin and anterior surface. Owing to the different terminology used for margin reporting, some margins were grouped together for analysis. Studies reporting involvement of the SMA, medial, uncinate and retroperitoneal margins had these margins grouped together as one category, the SMA/medial margin. Some studies described the medial margin as the vascular margin, and may have assessed the PV/SMV margin rather than the SMA margin. The PV/SMV margin, where reported, was analysed as a distinct margin category. Statistical analysis The R0 rates from each of the identified trials were pooled using the inverse-variance method to obtain the overall pooled proportion together with 95 per cent c.i. (fixed-effect model) using the statistical program StatsDirect16. The degree of heterogeneity present was quantified using the I2 statistic and tested using Cochran's Q test, with P < 0·050 indicating the presence of statistical heterogeneity. I2 values of 25, 50 and 75 per cent corresponded to low, moderate and high degrees of heterogeneity respectively17. Publication bias was quantified using Egger's regression model18. Differences in survival between R0 and R1 are described as hazard ratios (R1/ R0). Studies that reported recurrence rates are analysed with descriptive statistics to assess for trends in terms of recurrence with either an R0 or R1 resection. Results Nineteen studies4–6,12,19–33 met the inclusion criteria (Fig. 1). There were eight4–6,12,19–22 studies in group 1 (axial slicing technique; 882 patients) and six23–28 in group 2 (other slicing techniques; 1568) reporting the R0 resection rate using the minimum 1-mm margin (Table 1). There were seven studies19,23,29–33 in group 3 (1926 patients) that reported R0 resection using the minimum 0-mm margin (Table 2). The studies by Chang and colleagues23 and Delpero and co-workers19 included assessment of both 0- and 1-mm margins, and were each included in two groups. Fig. 1 Open in new tabDownload slide PRISMA flow diagram showing selection of articles for review Table 1 Studies reporting R0 resection with a minimum 1-mm margin according to pathological slicing technique Reference . Slicing technique . Margins examined . n . R0 (%) . Median overall survival (months) . Hazard ratio (R1/R0) . Group 1: axial slicing technique Campbell et al.6 (2009) Axial PT, MM, PM, PG, DD, BD 163 21 R0: 25·4 0·61 R1: 15·4 Delpero et al.19 (2014) Axial PT, SMA, PV/SMV, 150 32 n.r. n.a. (median 12 slices, studied in 0·5-mm increments) PM, PG, DD, BD Analysed with minimum 0-mm margin Esposito et al.5 (2008) Axial 3–5 mm slices PT, PV/SMV, PM, PG, DD, BD, AS 111 24 Median not reached 0·34 1-year overall survival R0: 86% R1: 64% Median overall survival from this and a larger cohort8 R0: 30·9 R1: 19·7 Gebauer et al.12 (2015) Axial PT, PV/SMV, PM, BD, AS 118 48 R0: 17·3 0·78 R1: 13·5 Jamieson et al.20 (2013) Transverse plane from D2 lumen into head of pancreas PT, MM, PM, PG, DD, BD, AS 217 28 R0: 26·6 0·62 R1: 16·5 John et al.21 (2013) Axial PT, PV/SMV, PM, PG, DD, BD, AS 70 26 R0: 22.4 0·73 R1: 16·3 Menon et al.22 (2009) Axial PT, PV/SMV, PM, PG, DD, BD, AS 27 19 R0: not reached at 55 months' follow-up n.a. R1: 14 Verbeke et al.4 (2006) Axial PT, PV/SMV, PM, PG, DD, BD, AS 26 15 R0: 37 0·30 R1: 11 Overall pooled R0 rate 882 29 (26, 32) Group 2: other slicing techniques Chang et al.23 (2009) Combination of longitudinal (perpendicular to pancreatic duct) and axial PT, SMA, PV/SMV, RPM (left-sided resections), PG, DD, BD 365 48 R0: 18·5 0·84 R1: 15·6 Gnerlich et al.24 (2012) n.a. PT, PV/SMV, PM, uncinate, BD 285 66 R0: 21·7R1: 16·4 0·76 Konstantinidis et al.25 (2013) Perpendicular sectioning PT, SMA, PM, BD 554 36 R0: 35 0·43 R1: 15 Pang et al.26(2014) Entire or near entire (minimum 3 or 4 blocks) periuncinate margin embedded PT, uncinate, PV/SMV, PM, PG, DD, BD, AS 116 42 R0: 29 R1: 23 0·79 Sugiura et al.27 (2013) Radial 5-mm sections PT, SMA, PM, BD 208 64 R0: 26 R1 (0 mm): 23 0·88 Westgaard et al.28 (2008) Serial perpendicular sectioning of RPM PT, RPM, PG, DD, BD 40 55 R0: 1.3 years 0·69 R1: 0·9 years Overall pooled R0 rate 1568 49 (47, 52) Reference . Slicing technique . Margins examined . n . R0 (%) . Median overall survival (months) . Hazard ratio (R1/R0) . Group 1: axial slicing technique Campbell et al.6 (2009) Axial PT, MM, PM, PG, DD, BD 163 21 R0: 25·4 0·61 R1: 15·4 Delpero et al.19 (2014) Axial PT, SMA, PV/SMV, 150 32 n.r. n.a. (median 12 slices, studied in 0·5-mm increments) PM, PG, DD, BD Analysed with minimum 0-mm margin Esposito et al.5 (2008) Axial 3–5 mm slices PT, PV/SMV, PM, PG, DD, BD, AS 111 24 Median not reached 0·34 1-year overall survival R0: 86% R1: 64% Median overall survival from this and a larger cohort8 R0: 30·9 R1: 19·7 Gebauer et al.12 (2015) Axial PT, PV/SMV, PM, BD, AS 118 48 R0: 17·3 0·78 R1: 13·5 Jamieson et al.20 (2013) Transverse plane from D2 lumen into head of pancreas PT, MM, PM, PG, DD, BD, AS 217 28 R0: 26·6 0·62 R1: 16·5 John et al.21 (2013) Axial PT, PV/SMV, PM, PG, DD, BD, AS 70 26 R0: 22.4 0·73 R1: 16·3 Menon et al.22 (2009) Axial PT, PV/SMV, PM, PG, DD, BD, AS 27 19 R0: not reached at 55 months' follow-up n.a. R1: 14 Verbeke et al.4 (2006) Axial PT, PV/SMV, PM, PG, DD, BD, AS 26 15 R0: 37 0·30 R1: 11 Overall pooled R0 rate 882 29 (26, 32) Group 2: other slicing techniques Chang et al.23 (2009) Combination of longitudinal (perpendicular to pancreatic duct) and axial PT, SMA, PV/SMV, RPM (left-sided resections), PG, DD, BD 365 48 R0: 18·5 0·84 R1: 15·6 Gnerlich et al.24 (2012) n.a. PT, PV/SMV, PM, uncinate, BD 285 66 R0: 21·7R1: 16·4 0·76 Konstantinidis et al.25 (2013) Perpendicular sectioning PT, SMA, PM, BD 554 36 R0: 35 0·43 R1: 15 Pang et al.26(2014) Entire or near entire (minimum 3 or 4 blocks) periuncinate margin embedded PT, uncinate, PV/SMV, PM, PG, DD, BD, AS 116 42 R0: 29 R1: 23 0·79 Sugiura et al.27 (2013) Radial 5-mm sections PT, SMA, PM, BD 208 64 R0: 26 R1 (0 mm): 23 0·88 Westgaard et al.28 (2008) Serial perpendicular sectioning of RPM PT, RPM, PG, DD, BD 40 55 R0: 1.3 years 0·69 R1: 0·9 years Overall pooled R0 rate 1568 49 (47, 52) Values in parentheses are 95 per cent c.i. PT, pancreatic transection margin; MM, medial margin; PM, posterior margin; PG, proximal gastric/duodenal margin; DD, distal duodenal/jejunal margin; BD, bile duct margin; SMA, superior mesenteric artery margin; PV, portal vein margin; SMV, superior mesenteric vein margin; n.r., not reported; n.a., not available; AS, anterior surface of pancreas; RPM, retroperitoneal margin. Fuller details can be found in Table S1 (supporting information). Open in new tab Table 1 Studies reporting R0 resection with a minimum 1-mm margin according to pathological slicing technique Reference . Slicing technique . Margins examined . n . R0 (%) . Median overall survival (months) . Hazard ratio (R1/R0) . Group 1: axial slicing technique Campbell et al.6 (2009) Axial PT, MM, PM, PG, DD, BD 163 21 R0: 25·4 0·61 R1: 15·4 Delpero et al.19 (2014) Axial PT, SMA, PV/SMV, 150 32 n.r. n.a. (median 12 slices, studied in 0·5-mm increments) PM, PG, DD, BD Analysed with minimum 0-mm margin Esposito et al.5 (2008) Axial 3–5 mm slices PT, PV/SMV, PM, PG, DD, BD, AS 111 24 Median not reached 0·34 1-year overall survival R0: 86% R1: 64% Median overall survival from this and a larger cohort8 R0: 30·9 R1: 19·7 Gebauer et al.12 (2015) Axial PT, PV/SMV, PM, BD, AS 118 48 R0: 17·3 0·78 R1: 13·5 Jamieson et al.20 (2013) Transverse plane from D2 lumen into head of pancreas PT, MM, PM, PG, DD, BD, AS 217 28 R0: 26·6 0·62 R1: 16·5 John et al.21 (2013) Axial PT, PV/SMV, PM, PG, DD, BD, AS 70 26 R0: 22.4 0·73 R1: 16·3 Menon et al.22 (2009) Axial PT, PV/SMV, PM, PG, DD, BD, AS 27 19 R0: not reached at 55 months' follow-up n.a. R1: 14 Verbeke et al.4 (2006) Axial PT, PV/SMV, PM, PG, DD, BD, AS 26 15 R0: 37 0·30 R1: 11 Overall pooled R0 rate 882 29 (26, 32) Group 2: other slicing techniques Chang et al.23 (2009) Combination of longitudinal (perpendicular to pancreatic duct) and axial PT, SMA, PV/SMV, RPM (left-sided resections), PG, DD, BD 365 48 R0: 18·5 0·84 R1: 15·6 Gnerlich et al.24 (2012) n.a. PT, PV/SMV, PM, uncinate, BD 285 66 R0: 21·7R1: 16·4 0·76 Konstantinidis et al.25 (2013) Perpendicular sectioning PT, SMA, PM, BD 554 36 R0: 35 0·43 R1: 15 Pang et al.26(2014) Entire or near entire (minimum 3 or 4 blocks) periuncinate margin embedded PT, uncinate, PV/SMV, PM, PG, DD, BD, AS 116 42 R0: 29 R1: 23 0·79 Sugiura et al.27 (2013) Radial 5-mm sections PT, SMA, PM, BD 208 64 R0: 26 R1 (0 mm): 23 0·88 Westgaard et al.28 (2008) Serial perpendicular sectioning of RPM PT, RPM, PG, DD, BD 40 55 R0: 1.3 years 0·69 R1: 0·9 years Overall pooled R0 rate 1568 49 (47, 52) Reference . Slicing technique . Margins examined . n . R0 (%) . Median overall survival (months) . Hazard ratio (R1/R0) . Group 1: axial slicing technique Campbell et al.6 (2009) Axial PT, MM, PM, PG, DD, BD 163 21 R0: 25·4 0·61 R1: 15·4 Delpero et al.19 (2014) Axial PT, SMA, PV/SMV, 150 32 n.r. n.a. (median 12 slices, studied in 0·5-mm increments) PM, PG, DD, BD Analysed with minimum 0-mm margin Esposito et al.5 (2008) Axial 3–5 mm slices PT, PV/SMV, PM, PG, DD, BD, AS 111 24 Median not reached 0·34 1-year overall survival R0: 86% R1: 64% Median overall survival from this and a larger cohort8 R0: 30·9 R1: 19·7 Gebauer et al.12 (2015) Axial PT, PV/SMV, PM, BD, AS 118 48 R0: 17·3 0·78 R1: 13·5 Jamieson et al.20 (2013) Transverse plane from D2 lumen into head of pancreas PT, MM, PM, PG, DD, BD, AS 217 28 R0: 26·6 0·62 R1: 16·5 John et al.21 (2013) Axial PT, PV/SMV, PM, PG, DD, BD, AS 70 26 R0: 22.4 0·73 R1: 16·3 Menon et al.22 (2009) Axial PT, PV/SMV, PM, PG, DD, BD, AS 27 19 R0: not reached at 55 months' follow-up n.a. R1: 14 Verbeke et al.4 (2006) Axial PT, PV/SMV, PM, PG, DD, BD, AS 26 15 R0: 37 0·30 R1: 11 Overall pooled R0 rate 882 29 (26, 32) Group 2: other slicing techniques Chang et al.23 (2009) Combination of longitudinal (perpendicular to pancreatic duct) and axial PT, SMA, PV/SMV, RPM (left-sided resections), PG, DD, BD 365 48 R0: 18·5 0·84 R1: 15·6 Gnerlich et al.24 (2012) n.a. PT, PV/SMV, PM, uncinate, BD 285 66 R0: 21·7R1: 16·4 0·76 Konstantinidis et al.25 (2013) Perpendicular sectioning PT, SMA, PM, BD 554 36 R0: 35 0·43 R1: 15 Pang et al.26(2014) Entire or near entire (minimum 3 or 4 blocks) periuncinate margin embedded PT, uncinate, PV/SMV, PM, PG, DD, BD, AS 116 42 R0: 29 R1: 23 0·79 Sugiura et al.27 (2013) Radial 5-mm sections PT, SMA, PM, BD 208 64 R0: 26 R1 (0 mm): 23 0·88 Westgaard et al.28 (2008) Serial perpendicular sectioning of RPM PT, RPM, PG, DD, BD 40 55 R0: 1.3 years 0·69 R1: 0·9 years Overall pooled R0 rate 1568 49 (47, 52) Values in parentheses are 95 per cent c.i. PT, pancreatic transection margin; MM, medial margin; PM, posterior margin; PG, proximal gastric/duodenal margin; DD, distal duodenal/jejunal margin; BD, bile duct margin; SMA, superior mesenteric artery margin; PV, portal vein margin; SMV, superior mesenteric vein margin; n.r., not reported; n.a., not available; AS, anterior surface of pancreas; RPM, retroperitoneal margin. Fuller details can be found in Table S1 (supporting information). Open in new tab Table 2 Studies reporting R0 resection with a minimum margin of 0 mm (group 3) Reference . Slicing technique . Margins examined . n . R0 (%) . Median overall survival (months) . Hazard ratio (R1/R0) . Chang et al.23 (2009) Combination of longitudinal (perpendicular to pancreatic duct) and axial PT, SMA, PV/SMV, RPM (left-sided resections), PG, DD, BD 365 64 R0: 19·6 0·67 R1: 13·2 Delpero et al.19 (2014) Axial PT, SMA, PV/SMV, PM, PG, DD, BD (PT, BD positivity not included) 150 70 R0: 32·9 (22·7, not reached) 0·54 R1: 17·7 (11·7, 36·4) Howard et al.29 (2006) n.a. PT, RPM, PG, DD, BD 226 (5 R2) 70 R0: 14 0·63 R1: 9 Above calculated from survival > 3 years R0: 17% R1: 6% Kimbrough et al.30 (2013) En face PT, RPM/uncinate 283 73 R0: 22·7 0·66 R1: 15·0 Mathur et al.31 (2014) Perpendicular, en face for BD, SMA PT, SMA, uncinate, PM, BD, PG, DD, AS 448 75 R0: 20 0·60 R1: 12 R1 → R0 (after intraop. frozen-section analysis and further resection in 40 patients): 14 Raut et al.32 (2007) En face section (2–3 mm) for BD/PT/SMA After 2000, SMA margin perpendicular sections PT, SMA, BD 360 83 R0: 27·8 0·77 R1: 21·5 Rau et al.33 (2012) Axial (median 4 (range 3–6) axial sections) PT, MM (includes SMV), PM, PG, DD, BD, AS 94 (1 Rx) 48 R0: 18·0 (3·8, 84·8) 0·77 R1: 13·8 (2·1, 48·2) Overall pooled R0 rate 1926 72 (70, 74) Reference . Slicing technique . Margins examined . n . R0 (%) . Median overall survival (months) . Hazard ratio (R1/R0) . Chang et al.23 (2009) Combination of longitudinal (perpendicular to pancreatic duct) and axial PT, SMA, PV/SMV, RPM (left-sided resections), PG, DD, BD 365 64 R0: 19·6 0·67 R1: 13·2 Delpero et al.19 (2014) Axial PT, SMA, PV/SMV, PM, PG, DD, BD (PT, BD positivity not included) 150 70 R0: 32·9 (22·7, not reached) 0·54 R1: 17·7 (11·7, 36·4) Howard et al.29 (2006) n.a. PT, RPM, PG, DD, BD 226 (5 R2) 70 R0: 14 0·63 R1: 9 Above calculated from survival > 3 years R0: 17% R1: 6% Kimbrough et al.30 (2013) En face PT, RPM/uncinate 283 73 R0: 22·7 0·66 R1: 15·0 Mathur et al.31 (2014) Perpendicular, en face for BD, SMA PT, SMA, uncinate, PM, BD, PG, DD, AS 448 75 R0: 20 0·60 R1: 12 R1 → R0 (after intraop. frozen-section analysis and further resection in 40 patients): 14 Raut et al.32 (2007) En face section (2–3 mm) for BD/PT/SMA After 2000, SMA margin perpendicular sections PT, SMA, BD 360 83 R0: 27·8 0·77 R1: 21·5 Rau et al.33 (2012) Axial (median 4 (range 3–6) axial sections) PT, MM (includes SMV), PM, PG, DD, BD, AS 94 (1 Rx) 48 R0: 18·0 (3·8, 84·8) 0·77 R1: 13·8 (2·1, 48·2) Overall pooled R0 rate 1926 72 (70, 74) Values in parentheses are 95 per cent c.i. PT, pancreatic transection margin; SMA, superior mesenteric artery margin; PV/SMV, portal vein/superior mesenteric vein margin; RPM, retroperitoneal margin; PG, proximal gastric/duodenal margin; DD, distal duodenal/jejunal margin; BD, bile duct margin; PM, posterior margin; n.a., not available; AS, anterior surface of pancreas; MM, medial margin. Fuller details can be found in Table S2 (supporting information). Open in new tab Table 2 Studies reporting R0 resection with a minimum margin of 0 mm (group 3) Reference . Slicing technique . Margins examined . n . R0 (%) . Median overall survival (months) . Hazard ratio (R1/R0) . Chang et al.23 (2009) Combination of longitudinal (perpendicular to pancreatic duct) and axial PT, SMA, PV/SMV, RPM (left-sided resections), PG, DD, BD 365 64 R0: 19·6 0·67 R1: 13·2 Delpero et al.19 (2014) Axial PT, SMA, PV/SMV, PM, PG, DD, BD (PT, BD positivity not included) 150 70 R0: 32·9 (22·7, not reached) 0·54 R1: 17·7 (11·7, 36·4) Howard et al.29 (2006) n.a. PT, RPM, PG, DD, BD 226 (5 R2) 70 R0: 14 0·63 R1: 9 Above calculated from survival > 3 years R0: 17% R1: 6% Kimbrough et al.30 (2013) En face PT, RPM/uncinate 283 73 R0: 22·7 0·66 R1: 15·0 Mathur et al.31 (2014) Perpendicular, en face for BD, SMA PT, SMA, uncinate, PM, BD, PG, DD, AS 448 75 R0: 20 0·60 R1: 12 R1 → R0 (after intraop. frozen-section analysis and further resection in 40 patients): 14 Raut et al.32 (2007) En face section (2–3 mm) for BD/PT/SMA After 2000, SMA margin perpendicular sections PT, SMA, BD 360 83 R0: 27·8 0·77 R1: 21·5 Rau et al.33 (2012) Axial (median 4 (range 3–6) axial sections) PT, MM (includes SMV), PM, PG, DD, BD, AS 94 (1 Rx) 48 R0: 18·0 (3·8, 84·8) 0·77 R1: 13·8 (2·1, 48·2) Overall pooled R0 rate 1926 72 (70, 74) Reference . Slicing technique . Margins examined . n . R0 (%) . Median overall survival (months) . Hazard ratio (R1/R0) . Chang et al.23 (2009) Combination of longitudinal (perpendicular to pancreatic duct) and axial PT, SMA, PV/SMV, RPM (left-sided resections), PG, DD, BD 365 64 R0: 19·6 0·67 R1: 13·2 Delpero et al.19 (2014) Axial PT, SMA, PV/SMV, PM, PG, DD, BD (PT, BD positivity not included) 150 70 R0: 32·9 (22·7, not reached) 0·54 R1: 17·7 (11·7, 36·4) Howard et al.29 (2006) n.a. PT, RPM, PG, DD, BD 226 (5 R2) 70 R0: 14 0·63 R1: 9 Above calculated from survival > 3 years R0: 17% R1: 6% Kimbrough et al.30 (2013) En face PT, RPM/uncinate 283 73 R0: 22·7 0·66 R1: 15·0 Mathur et al.31 (2014) Perpendicular, en face for BD, SMA PT, SMA, uncinate, PM, BD, PG, DD, AS 448 75 R0: 20 0·60 R1: 12 R1 → R0 (after intraop. frozen-section analysis and further resection in 40 patients): 14 Raut et al.32 (2007) En face section (2–3 mm) for BD/PT/SMA After 2000, SMA margin perpendicular sections PT, SMA, BD 360 83 R0: 27·8 0·77 R1: 21·5 Rau et al.33 (2012) Axial (median 4 (range 3–6) axial sections) PT, MM (includes SMV), PM, PG, DD, BD, AS 94 (1 Rx) 48 R0: 18·0 (3·8, 84·8) 0·77 R1: 13·8 (2·1, 48·2) Overall pooled R0 rate 1926 72 (70, 74) Values in parentheses are 95 per cent c.i. PT, pancreatic transection margin; SMA, superior mesenteric artery margin; PV/SMV, portal vein/superior mesenteric vein margin; RPM, retroperitoneal margin; PG, proximal gastric/duodenal margin; DD, distal duodenal/jejunal margin; BD, bile duct margin; PM, posterior margin; n.a., not available; AS, anterior surface of pancreas; MM, medial margin. Fuller details can be found in Table S2 (supporting information). Open in new tab Studies reporting a minimum 1-mm margin for R0 resection R0 rates with a 1-mm margin The majority of studies using the axial slicing technique assessed a minimum of six margins, with six4,5,19–22 of eight studies analysing seven margins (Table 1). The pooled R0 rate in group 1 using the axial slicing technique was 29 (95 per cent c.i. 26 to 32 per cent). There was high heterogeneity between studies (I2 = 76 per cent, P < 0·001) and little evidence of publication bias (P = 0·81) (Fig. 2). When the study by Gebauer and colleagues12, which is an outlier in this group, was excluded from analysis, the heterogeneity between studies disappeared (I2 = 14 per cent, P = 0·33) with only a small change in the results. The latter study assessed only five margins, unlike the majority. Given that neoadjuvant treatment could increase R0 rates, the R0 rate in group 1 was also analysed after excluding studies that employed neoadjuvant treatment19,20. The R0 rate dropped slightly from 29 per cent to 28 (24 to 32) per cent. Fig. 2 Open in new tabDownload slide Meta-analysis of R0 rates (1-mm margin) using the axial slicing technique. Rates are shown with 95 per cent c.i. A fixed-effect model was used for meta-analysis. Test for heterogeneity P < 0·001, I2 = 76 per cent; publication bias P = 0·81 Studies using other slicing techniques reported assessment of a minimum of four margins, with two23,26 of six studies analysing more than six margins (Table 1). The pooled R0 rate in group 2 with other slicing techniques was 49 (47 to 52) per cent. There was high heterogeneity between studies (I2 = 95 per cent, P < 0·001) and no publication bias (P = 0·49) (Fig. 3). Fig. 3 Open in new tabDownload slide Meta-analysis of R0 rates (1-mm margin) using the other slicing techniques. Rates are shown with 95 per cent c.i. A fixed-effect model was used for meta-analysis. Test for heterogeneity P < 0·001, I2 = 95 per cent; publication bias P = 0·49 In a study of 554 patients by Konstantinidis and colleagues25, 397 patients had an R0 resection and 157 had an R1 resection using a 0-mm margin. Some 339 (85 per cent) of 397 patients with an R0 resection were assessed using the 1-mm minimum margin clearance, of whom 170 (50 per cent) of 339 had an R0 resection. For the present analysis, it was assumed that half of the 397 patients had an R0 resection using the minimum 1-mm margin in order not to underestimate the R0 rate. This equated to 199 (36 per cent) of 554 R0 resections. Most commonly involved margins with R1 resections The most commonly involved margins in R1 resections with the axial slicing technique were the PV/SMV margin, which was involved in over 50 per cent of R1 resections in five4,5,19,21,22 of six studies reporting this, and the posterior margin, which was involved, in over 50 per cent of resections in five4,6,19,21,22 of eight studies (Table 3). In three studies6,19,20 that reported the SMA/medial margin, this was involved in 36–54 per cent of R1 resections with the axial slicing technique. All studies reported pancreatic neck margin involvement, which ranged from 4 to 30 per cent of R1 resections. Table 3 Margin involvement in R1 resections . Margins positive in R1 resections . . Pancreatic neck margin . Superior mesenteric artery/medial/uncinate/retroperitoneal margin . Portal vein/ superior mesenteric vein margin . Posterior margin . Proximal gastric/duodenal margin . Distal duodenal/ jejunal margin . Bile duct margin . Anterior surface of pancreas . Group 1 80 of 628 186 of 387 199 of 343 304 of 628 10 of 369 0 of 285 17 of 584 88 of 398 (13)4–6,12,19–22 (48)6,19,20 (58)4,5,12,19,21,22 (48)4–6,12,19–22 (3)5,6,20 (0)6,20 (3)5,6,12,19–21 (22)4,5,12,20–22 Group 2 50 of 256 143 of 256 67 of 164 80 of 238 n.r. n.r. 3 of 159 0 of 18 (20)24–27 (56)24–27 (41)24,25 (34)24,25,27 (2)26–28 (0)28 Group 3 96 of 285 188 of 285 31 of 177 11 of 45 3 of 180 0 of 132 10 of 237 n.r. (34)19,23,32,33 (66)19,23,32,33 (18)19,23 (24)19 (2)23,33 (0)23 (4)19,23,32 . Margins positive in R1 resections . . Pancreatic neck margin . Superior mesenteric artery/medial/uncinate/retroperitoneal margin . Portal vein/ superior mesenteric vein margin . Posterior margin . Proximal gastric/duodenal margin . Distal duodenal/ jejunal margin . Bile duct margin . Anterior surface of pancreas . Group 1 80 of 628 186 of 387 199 of 343 304 of 628 10 of 369 0 of 285 17 of 584 88 of 398 (13)4–6,12,19–22 (48)6,19,20 (58)4,5,12,19,21,22 (48)4–6,12,19–22 (3)5,6,20 (0)6,20 (3)5,6,12,19–21 (22)4,5,12,20–22 Group 2 50 of 256 143 of 256 67 of 164 80 of 238 n.r. n.r. 3 of 159 0 of 18 (20)24–27 (56)24–27 (41)24,25 (34)24,25,27 (2)26–28 (0)28 Group 3 96 of 285 188 of 285 31 of 177 11 of 45 3 of 180 0 of 132 10 of 237 n.r. (34)19,23,32,33 (66)19,23,32,33 (18)19,23 (24)19 (2)23,33 (0)23 (4)19,23,32 Values in parentheses are percentages. n.r., Not reported. Open in new tab Table 3 Margin involvement in R1 resections . Margins positive in R1 resections . . Pancreatic neck margin . Superior mesenteric artery/medial/uncinate/retroperitoneal margin . Portal vein/ superior mesenteric vein margin . Posterior margin . Proximal gastric/duodenal margin . Distal duodenal/ jejunal margin . Bile duct margin . Anterior surface of pancreas . Group 1 80 of 628 186 of 387 199 of 343 304 of 628 10 of 369 0 of 285 17 of 584 88 of 398 (13)4–6,12,19–22 (48)6,19,20 (58)4,5,12,19,21,22 (48)4–6,12,19–22 (3)5,6,20 (0)6,20 (3)5,6,12,19–21 (22)4,5,12,20–22 Group 2 50 of 256 143 of 256 67 of 164 80 of 238 n.r. n.r. 3 of 159 0 of 18 (20)24–27 (56)24–27 (41)24,25 (34)24,25,27 (2)26–28 (0)28 Group 3 96 of 285 188 of 285 31 of 177 11 of 45 3 of 180 0 of 132 10 of 237 n.r. (34)19,23,32,33 (66)19,23,32,33 (18)19,23 (24)19 (2)23,33 (0)23 (4)19,23,32 . Margins positive in R1 resections . . Pancreatic neck margin . Superior mesenteric artery/medial/uncinate/retroperitoneal margin . Portal vein/ superior mesenteric vein margin . Posterior margin . Proximal gastric/duodenal margin . Distal duodenal/ jejunal margin . Bile duct margin . Anterior surface of pancreas . Group 1 80 of 628 186 of 387 199 of 343 304 of 628 10 of 369 0 of 285 17 of 584 88 of 398 (13)4–6,12,19–22 (48)6,19,20 (58)4,5,12,19,21,22 (48)4–6,12,19–22 (3)5,6,20 (0)6,20 (3)5,6,12,19–21 (22)4,5,12,20–22 Group 2 50 of 256 143 of 256 67 of 164 80 of 238 n.r. n.r. 3 of 159 0 of 18 (20)24–27 (56)24–27 (41)24,25 (34)24,25,27 (2)26–28 (0)28 Group 3 96 of 285 188 of 285 31 of 177 11 of 45 3 of 180 0 of 132 10 of 237 n.r. (34)19,23,32,33 (66)19,23,32,33 (18)19,23 (24)19 (2)23,33 (0)23 (4)19,23,32 Values in parentheses are percentages. n.r., Not reported. Open in new tab The most commonly involved margin in R1 resections (tumour-positive resection margins) with the other slicing techniques was the SMA/medial margin, which was involved in 48–78 per cent of R1 resections in four studies24–27 reporting this margin. Three studies24,25,27 reported involvement of the posterior margin in over 30 per cent of R1 resections. Two studies24,25 reported involvement of the PV/SMV margin in 26 and 63 per cent of R1 resections respectively. Four studies24–27 reported pancreatic neck margin involvement, which ranged from 10 to 39 per cent. Survival hazard ratios The survival hazard ratios (R1/R0) with the axial slicing technique ranged from 0·30 to 0·78; patients with an R0 resection therefore had a minimum 22 per cent reduction in risk of death compared with patients with an R1 resection. The survival hazard ratios (R1/R0) with other slicing techniques in group 2 ranged from 0·43 to 0·88; patients with an R0 resection therefore had a minimum 12 per cent reduction in their risk of death compared with patients with an R1 resection. Recurrence rates Most of the studies did not provide data on recurrence. In the axial slicing group, Esposito and colleagues5 reported that local recurrence developed after 10 per cent of R1 resections. Jamieson et al.20 noted local recurrence after 38 per cent of R0 resections, and 44 per cent of R1 resections; distant recurrence rates were 28 and 40 per cent respectively. In the other slicing techniques group, Gnerlich and co-workers24 reported local recurrence after 27 per cent of R0 resections and 39 per cent of R1 resections. Sugiura and colleagues27 reported that local recurrence developed after 8 per cent of R0 resections and 34 per cent of R1 resections; respective distant recurrence rates were 88 and 81 per cent. Studies reporting a 0-mm margin for R0 resection R0 rates with a 0-mm margin Four19,23,31,33of seven studies reported assessment of a minimum of six margins (Table 2). The pooled R0 rate using a 0-mm margin (group 3) was 72 (95 per cent c.i. 70 to 74) per cent. There was high heterogeneity between studies (I2 = 91 per cent, P < 0·001) (a random-effects analysis yielded substantially similar results) and significant publication bias (P = 0·04) (Fig. 4). The study by Rau and colleagues33 had the lowest R0 rate, which may relate to their modified Verbeke technique for margin assessment, with a median of 4 sections (range 3–6 sections). Exclusion of the studies19,32 that included patients with neoadjuvant treatment resulted in the R0 rate falling from 72 per cent to 69 (67 to 72) per cent. Fig. 4 Open in new tabDownload slide Meta-analysis of R0 rates in studies that used a 0-mm margin. Rates are shown with 95 per cent c.i. A fixed-effect model was used for meta-analysis. Test for heterogeneity P < 0·001, I2 = 91 per cent; publication bias P = 0·04 Most commonly involved margins (R1 resections) The most commonly involved margin in R1 resections with a 0-mm margin was the SMA/medial margin, which was involved in 33–92 per cent of R1 resections in four19,23,32,33 studies reporting this (Table 3). Four studies19,23,32,33 reported involvement of the pancreatic neck margin in 18–49 per cent of R1 resections. Survival hazard ratios The survival hazard ratios (R1/R0) with a 0-mm margin ranged from 0·54 to 0·77; patients with an R0 resection therefore had a minimum 23 per cent reduction in risk of death compared with patients who had an R1 resection. Recurrence rates Most of the studies did not provide data on recurrence. Raut and colleagues32 reported local recurrence after 8 per cent of R0 resections and 7 per cent of R1 resections; distant recurrence rates were 42 and 45 per cent respectively. Rau et al.33 reported that local recurrence developed following 33 per cent of R0 resections and 58 per cent of R1 resections; corresponding distant recurrence rates were 58 and 51 per cent. Discussion This review has shown that the definition of margin clearance (0 versus 1 mm) and the method of pathological margin assessment is an important factor in R0 resection rates reported in different series. Studies with a minimum 1-mm margin employing the axial slicing technique (group 1) examined more margins and had the lowest R0 rate. These studies evaluated a minimum of six margins with R0 rates of 29 (95 per cent c.i. 26 to 32) per cent, whereas studies using other slicing techniques (group 2), which evaluated a minimum of four margins, had R0 rates of 49 (47 to 52) per cent. The combined R0 rate when a minimum 1-mm margin was used (groups 1 and 2) was 41 (40 to 43) per cent and may serve as a more relevant baseline in studies that employ a variation in histopathological assessment of margins. The R0 rates achieved with a 0-mm margin were much higher (group 3) at 72 (70 to 74) per cent. Although it is well known that the definition of margin clearance affects R0 rates, this review provides a comprehensive overview of studies giving some indication of achievable R0 rates in pancreatic cancer surgery depending on pathological assessment. A limitation of many of these studies, and hence the present combined analysis, is that it is not clear how much differences in patient selection and variation in surgical techniques may have contributed to the R0 rates. Neoadjuvant treatment did not seem to influence the R0 resection rate to a great extent. Based on studies reported by Chang and colleagues23 and Delpero et al.19, where R0 rates were reported for both the 0- and 1-mm margin clearance definitions in the same patient population, it is clear that margin definition is a major driver of R0 rates. Ultimately, technical factors and biology need to be factored in to assess the impact of margin outcomes more fully, but this was beyond the scope of this article given the limited data reported on this41. This review indicates that the most commonly involved margins are the SMA/medial, PV/SMV and posterior margins. However, these margins do not have equal prognostic significance. Multifocal margin involvement with an R1 resection was reported in 32–45 per cent of R1 procedures in three4–6 studies. Jamieson and colleagues11 analysed pancreatic margins by mobilization margins (anterior and posterior margins) and transection margins (pancreatic transection margin, medial margin and adjacent transection margins). Involvement of the mobilization margins alone was associated with a much longer median survival than involvement of the transection margin (median survival 18·9 versus 11·1 months; P < 0·001). Delpero et al.19 similarly found that a positive posterior margin had no impact on progression-free survival. It is clear from this that the anterior and posterior margins have a lesser impact on survival than the pancreatic transection margins, and that each margin may have different prognostic significance. It was not possible to analyse the data to extrapolate the prognostic role of each individual involved margin. This would be useful in future studies, but will depend on studies reporting a minimum set of margins. Increasing margin clearance has been shown to affect survival. The use of the 0-mm margin in adjuvant pancreatic cancer studies may explain why R1 status was not identified as a significant factor for survival in a meta-analysis of adjuvant randomized clinical trials1. The transition to using a minimum 1-mm margin has been driven by the observation that resection with an overtly involved margin at 0 mm has similar outcomes to resections in which tumour is found within 1 mm of the resection margin6,25. In the studies by Chang and colleagues23 and Jamieson et al.20, it was not until the resection margin was clear by more than 1·5 mm that long-term survival was achieved. Gebauer and co-workers12 reported that median overall survival in patients with a tumour margin clearance of less than 2 mm was lower than that in patients with a margin clearance of 2 mm or more (15·1 versus 22·2 months; P = 0·046). This may mean that dispersed cancer cells can remain despite a clear resection margin, requiring greater clearance to ensure no cancer is left behind. This may explain the improvement in survival with greater margin clearance15. Although increasing the minimum margin to 2 mm may well further define those with improved survival, it is clear that this will also reduce the achievable R0 rates. In a large series of 1071 consecutive patients, Hartwig and colleagues8 showed that the newly revised R0 rate using a 1-mm margin was an independent positive predictor for survival on multivariable analysis. This suggests that, in large series with standardized pathology, an R0 rate endpoint should be based on a minimum margin of 1 mm to assess the gains from surgery in improving survival in resectable pancreatic cancer. The survival hazard ratios (R1/R0) across both margin definitions in groups 1 and 3 were consistent with a minimum 22–23 per cent reduction in the risk of death from an R0 resection compared with an R1 resection. Although the survival hazard ratios show a similar proportional reduction in the risk of death, this does not take into account the baseline risk or therefore the absolute risk reduction, as a 22 per cent reduction in the risk of death at 10 months is not the same as a similar risk reduction at 20 months. Therefore, although the hazard ratios are similar across definitions, this does not mean that baseline survival across definitions or the absolute benefit from an R0 resection are similar. It was not possible to calculate a pooled estimate of median survival in those with an R0 or R1 resection because the data are heterogeneous in terms of follow-up time and use of neoadjuvant and adjuvant treatment, which significantly affects survival. Of note, studies with a high R0 rate do not necessarily have the longest survival, and a low R0 rate in a given study may not necessarily equate to poorer survival for the study group. This suggests that survival in pancreatic cancer is more affected by a complex interplay of numerous factors, including tumour characteristics and biology, as well as the use of adjuvant and neoadjuvant treatment, rather than the extent of surgery. Studies using neoadjuvant therapy in primarily resectable cancers tend to report better survival. This may in part be explained by the exclusion of patients with early systemic progression. Several studies have shown that pursuing negative margins after positive intraoperative frozen-section analysis portends a poorer survival than that in patients with negative margins on initial intraoperative frozen sections, and the pursuit of negative margins did not result in the intended survival benefit26,31,42–44. It has been proposed that R1 tumours may be inherently more biologically aggressive; this may relate to differences in tumour size or stage, but this finding has not been consistent across all the studies reviewed45. It would make sense that larger tumours are more likely to result in R1 resections. Kimbrough et al.30 found that R1 resections had a higher incidence of microvascular invasion, positive lymph node ratio and perineural invasion, without any differences in tumour size between R0 and R1 tumours. Similarly, Gebauer and colleagues12 reported that, although R1 tumours were more likely to have nodal and lymphovascular invasion, there was no statistical difference in the size of R0 and R1 tumours. However, in other studies25,31,32, R1 tumours were larger than R0 tumours. This was similarly found in the study by Campbell and co-workers6, where increasing tumour size significantly increased the likelihood of an R1 resection. The pattern of recurrence and failure following pancreatic resection offers insight into the poor survival with this disease. Most of the studies did not provide data on recurrence. Local recurrence developed more frequently after R1 than R0 resection in most studies. Although local recurrence has been shown to occur frequently from an autopsy study in patients who had curative resection of pancreatic cancer, this is rarely the cause of death46; most patients die from metastatic disease. The aim of radical surgery is to remove all site-specific macroscopic and microscopic tumour, but this has no effect on occult systemic disease. The aim of multimodal therapies is to eliminate this micrometastatic disease. As systemic therapies improve outcomes, durable local control becomes more important to the quality of patients' subsequent survival. The variable follow-up in all studies to date makes it difficult to evaluate the impact of achieving an R0 resection on recurrence. To make meaningful inferences from recurrence data, the assessment for recurrence and the follow-up time for this needs to be prospective, uniform and standardized. This is relevant for future studies because local control becomes more important as survival improves with systemic treatment; local recurrence causes substantial morbidity and compromises effective palliation and quality of life. The ability to predict the risk of local and distant recurrence by margin status will help guide the use of adjuvant local therapies such as radiation, and also guide the development of neoadjuvant treatments that increase resectability, and in so doing potentially affect both survival and local control. Although a high R0 rate is clearly desirable, it is evident that the more rigorous the pathological assessment, the less likely a high R0 rate is to be achieved. This review revealed that there were several different terms used for the same margins. The medial margin was also referred to as the uncinate or SMA margin in the reviewed studies. The retroperitoneal margin included the posterior margin in some studies in addition to the SMA margin. The PV/SMV margin was another margin reported by some studies; although this is known as the margin adjacent to the PV/SMV venous groove, this was not clearly defined in the studies. These different terminologies cause confusion and make comparisons between studies difficult, as the terms are synonymous in some instances but not in others44,47. Although examining more margins meticulously with extensive tissue sampling clearly increases the R1 rate, consensus on terminology, definition of microscopic margin involvement and the use of synoptic reporting for standard assessment15 is essential to allow valid and robust comparison between centres, and to avoid the current wide variations in reported R0 and R1 rates48,49. It is clear from Table 3 that studies do not report all the margins they analysed. Other unresolved issues include the number of sections examined, which can reduce the risk of underestimating margin involvement as a result of a sampling error. For example, if two standard tissue blocks are taken from a 1-cm area suspected of tumour involvement, only 1/1000th of the tissue of interest is examined15. Furthermore, the definition of a positive margin needs to be standardized; consideration needs to be given to the implications of tumour cells within blood vessels, lymphatics, perineural spaces and lymph nodes, and ‘isolated solitary ductal units’ that appear in the adipose tissue, on margin status6,15,50. Future pancreatic cancer trials should adopt uniform approaches to pathological assessment and interpretation of margins. These should include a standard approach to macroscopic dissection, and use of standard terminology for different anatomical margins, which should probably be a minimum of eight margins. Standard interpretation of involved margins should include a set cut-off for the definition of involved margins, noting both the 0- and 1-mm margin clearances (Table 4). This will require a collaborative effort from surgeons and pathologists in marking and staining the specimens adequately to identify these individual margins. Although the axial technique has gained popularity in Europe, it is practised less elsewhere, and it would be useful if this technique were adopted internationally to allow comparisons between trials. Although some margins clearly have a greater prognostic role than others, standard reporting of a minimum of eight margins – the pancreatic neck margin, the SMA margin (and doing away with other terms such as the medial or uncinate margin, as previously suggested51), PV/SMV margin, anterior surface, posterior margin, BD margin, proximal gastric/duodenal margin and distal duodenal/jejunal margin – will allow more robust data analysis to assess the prognostic significance of each individual margin and also whether these relate to recurrence patterns. The initiators of this study identified the most active hepatopancreatobiliary surgeons in Australia to ensure the broadest diversity of views so that a consensus was likely to be broadly acceptable. This study will hopefully lead closer towards standard margin reporting and assessment; once this is in place, it will be possible to assess properly whether margin status is an independent measure of recurrence and metastatic risk. Table 4 Proposal for standardized pancreatic margin reporting Minimum margin Clearly defined and reported Both 0- and 1-mm margin clearances noted Slicing techniques Axial slicing Number of sections 3–5-mm sections Minimum margin assessment 8 margins Individual margin reporting Pancreatic neck margin Superior mesenteric artery margin (not using terms such as medial or uncinate margin) Portal vein/superior mesenteric vein margin Anterior surface Posterior margin (mobilization margin) Bile duct margin Proximal gastric/duodenal margin Distal duodenal/jejunal margin Minimum margin Clearly defined and reported Both 0- and 1-mm margin clearances noted Slicing techniques Axial slicing Number of sections 3–5-mm sections Minimum margin assessment 8 margins Individual margin reporting Pancreatic neck margin Superior mesenteric artery margin (not using terms such as medial or uncinate margin) Portal vein/superior mesenteric vein margin Anterior surface Posterior margin (mobilization margin) Bile duct margin Proximal gastric/duodenal margin Distal duodenal/jejunal margin Open in new tab Table 4 Proposal for standardized pancreatic margin reporting Minimum margin Clearly defined and reported Both 0- and 1-mm margin clearances noted Slicing techniques Axial slicing Number of sections 3–5-mm sections Minimum margin assessment 8 margins Individual margin reporting Pancreatic neck margin Superior mesenteric artery margin (not using terms such as medial or uncinate margin) Portal vein/superior mesenteric vein margin Anterior surface Posterior margin (mobilization margin) Bile duct margin Proximal gastric/duodenal margin Distal duodenal/jejunal margin Minimum margin Clearly defined and reported Both 0- and 1-mm margin clearances noted Slicing techniques Axial slicing Number of sections 3–5-mm sections Minimum margin assessment 8 margins Individual margin reporting Pancreatic neck margin Superior mesenteric artery margin (not using terms such as medial or uncinate margin) Portal vein/superior mesenteric vein margin Anterior surface Posterior margin (mobilization margin) Bile duct margin Proximal gastric/duodenal margin Distal duodenal/jejunal margin Open in new tab Margin reporting was examined in neoadjuvant studies in resectable pancreatic cancer to compare the impact of different regimens on R0 rate (Table S3, supporting information). It was found that neoadjuvant studies in resectable pancreatic cancer are not necessarily clear or similar in the assessment of pathological margins. Because this review demonstrates that these definitions affect the ultimate reported R0 rate, comparisons of R0 rates across studies are difficult. Future trials must address inconsistent terminology and pathological definitions to enable useful international multicentre comparisons to be made. This review has highlighted that inconsistent terminology, lack of agreement on synoptic reporting guidelines, variation in pathological techniques and inconsistent pathological definitions are hampering international comparative analysis of outcomes and assessment of multimodal treatments for these difficult tumours. An international consensus definition for margin assessment and reporting needs to be agreed and, based on this analysis, it is recommended that a margin of 1 mm be adopted as the internationally accepted norm. Collaborators Collaborators in this study were: J. Fawcett (University of Queensland, Princess Alexandra Hospital, Brisbane, Queensland, Australia); P. S. Grimison (Department of Medical Oncology, Chris O'Brien Lifehouse and University of Sydney, New South Wales, Australia); C. Christophi (University of Melbourne, Department of Surgery, Austin Health, Melbourne, Victoria, Australia); R. Padbury (Flinders Medical Centre, Bedford Park, Adelaide, South Australia, Australia); K. S. Haghighi (Department of Upper Gastrointestinal and Transplant Surgery, Prince of Wales Hospital, Sydney, New South Wales, Australia); J. W. Chen (Flinders Medical Centre, Bedford Park, Adelaide, South Australia, Australia); M. Nikfarjam (University of Melbourne, Department of Surgery, Austin Health, Melbourne, Victoria, Australia); N. O'Rourke (Hepatopancreatobiliary Unit, Department of Surgery, Royal Brisbane and Women's Hospital, Brisbane, Queensland, Australia); and N. Spry (Faculty of Medicine, University of Western Australia, Perth, Western Australia, Australia) with the support of the NHMRC Clinical Trials Centre and the Australasian Gastro-Intestinal Trials Group. Acknowledgements The authors acknowledge grant support from Cancer Australia and Cancer Institute New South Wales for M.D.C. and the statistical analyses respectively. Disclosure: The authors declare no conflict of interest. References 1 Butturini G , Stocken DD, Wente MN, Jeekel H, Klinkenbijl JH, Bakkevold KE et al. Influence of resection margins and treatment on survival in patients with pancreatic cancer: metaanalysis of randomized controlled trials . Arch Surg 2008 ; 143 : 75 – 83 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Speer AG , Thursfield VJ, Torn-Broers Y, Jefford M. Pancreatic cancer: surgical management and outcomes after 6 years of follow-up . Med J Aust 2012 ; 196 : 511 – 515 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Sharp L , Carsin AE, Cronin-Fenton DP, O'Driscoll D, Comber H. Is there under-treatment of pancreatic cancer? Evidence from a population-based study in Ireland . Eur J Cancer 2009 ; 45 : 1450 – 1459 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Verbeke CS , Leitch D, Menon KV, McMahon MJ, Guillou PJ, Anthoney A. Redefining the R1 resection in pancreatic cancer . Br J Surg 2006 ; 93 : 1232 – 1237 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Esposito I , Kleef J, Bergmann F, Reiser C, Herpel E, Friess H et al. Most pancreatic resections are R1 resections . Ann Surg Oncol 2008 ; 15 : 1651 – 1660 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Campbell F , Smith RA, Whelan P, Sutton R, Raraty M, Neoptolemos JP et al. Classification of R1 resections for pancreatic cancer: the prognostic relevance of tumour involvement within 1 mm of a resection margin . Histopathology 2009 ; 55 : 277 – 283 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Büchler MW , Werner J, Weitz J. R0 in pancreatic cancer surgery: surgery, pathology, biology, or definition matters? Ann Surg 2010 ; 251 : 1011 – 1012 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Hartwig W , Hackert T, Hinz U, Gluth A, Bergmann F, Strobel O et al. Pancreatic cancer surgery in the new millennium: better prediction of outcome . Ann Surg 2011 ; 254 : 311 – 319 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Royal College of Pathologists of Australasia (RCPA) . Cancer of the Exocrine Pancreas, Ampulla of Vater and Distal Common Bile Duct. Structured Reporting Protocol, 1st Edition 2014 . https://www.rcpa.edu.au/getattachment/3b6a41df-939d-492e-bc6b-35fc264bd89b/Protocol-pancreatic-cancer.aspx [accessed 2 July 2015 ]. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 10 Bockhorn M , Uzunoglu FG, Adham M, Imrie C, Milicevic M, Sandberg AA et al. Borderline resectable pancreatic cancer: a consensus statement by the International Study Group of Pancreatic Surgery (ISGPS) . Surgery 2014 ; 155 : 977 – 988 . Google Scholar Crossref Search ADS PubMed WorldCat 11 Jamieson NB , Foulis AK, Oien KA, Going JJ, Glen P, Dickson EJ et al. Positive mobilization margins alone do not influence survival following pancreatico-duodenectomy for pancreatic ductal adenocarcinoma . Ann Surg 2010 ; 251 : 1003 – 1010 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Gebauer F , Tachezy M, Vashist YK, Marx AH, Yekebas E, Izbicki JR et al. Resection margin clearance in pancreatic cancer after implementation of the Leeds Pathology Protocol (LEEPP): clinically relevant or just academic? World J Surg 2015 ; 39 : 493 – 499 . Google Scholar Crossref Search ADS PubMed WorldCat 13 Verbeke CS . Resection margins and R1 rates in pancreatic cancer – are we there yet? Histopathology 2008 ; 52 : 787 – 796 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Moher D , Liberati A, Tetzlaff J, Altmann DG. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement . J Clin Epidemiol 2009 ; 62 : 1006 – 1012 . Google Scholar Crossref Search ADS PubMed WorldCat 15 Verbeke CS . Resection margins in pancreatic cancer . Surg Clin N Am 2013 ; 93 : 647 – 662 . Google Scholar Crossref Search ADS PubMed WorldCat 16 StatsDirect . StatsDirect version 3 . http://www.statsdirect.com [accessed 24 June 2015 ]. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 17 Higgins JP , Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses . BMJ 2003 ; 327 : 557 – 560 . Google Scholar Crossref Search ADS PubMed WorldCat 18 Egger M , Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test . BMJ 1997 ; 315 : 629 – 634 . Google Scholar Crossref Search ADS PubMed WorldCat 19 Delpero JR , Bachellier P, Regenet N, Le Treut YP, Paye F, Carrere N et al. Pancreaticoduodenectomy for pancreatic ductal adenocarcinoma: a French multicentre prospective evaluation of resection margins in 150 evaluable specimens . HPB (Oxford) 2014 ; 16 : 20 – 33 . Google Scholar Crossref Search ADS PubMed WorldCat 20 Jamieson NB , Chan NIJ, Foulis AK, Dickson EJ, McKay CJ, Carter CR. The prognostic influence of resection margin clearance following pancreaticoduodenectomy for pancreatic ductal adenocarcinoma . J Gastrointest Surg 2013 ; 17 : 511 – 521 . Google Scholar Crossref Search ADS PubMed WorldCat 21 John BJ , Naik P, Ironside A, Davidson BR, Fusai G, Gillmore R et al. Redefining the R1 resection for pancreatic ductal adenocarcinoma: tumour lymph nodal burden and lymph node ratio are the only prognostic factors associated with survival . HPB (Oxford) 2013 ; 15 : 674 – 680 . Google Scholar Crossref Search ADS PubMed WorldCat 22 Menon KV , Gomez D, Smith AM, Anthoney A, Verbeke CS. Impact of margin status on survival following pancreatoduodenectomy for cancer: the Leeds Pathology Protocol (LEEPP) . HPB (Oxford) 2009 ; 11 : 18 – 24 . Google Scholar Crossref Search ADS PubMed WorldCat 23 Chang DK , Johns AL, Merrett ND, Gill AJ, Colvin EK, Scarlett CJ et al. Margin clearance and outcome in resected pancreatic cancer . J Clin Oncol 2009 ; 27 : 2855 – 2863 . Google Scholar Crossref Search ADS PubMed WorldCat 24 Gnerlich JL , Luka SR, Deshpande AD, Dubray BJ, Weir JS, Carpenter DH et al. Microscopic margins and patterns of treatment failure in resected pancreatic adenocarcinoma . Arch Surg 2012 ; 147 : 753 – 760 . Google Scholar Crossref Search ADS PubMed WorldCat 25 Konstantinidis IT , Warshaw AL, Allen JN, Blaszkowsky LS, Castillo CF, Deshpande V et al. Pancreatic ductal adenocarcinoma: is there a survival difference for R1 resections versus locally advanced unresectable tumors? What is a ‘true’ R0 resection? Ann Surg 2013 ; 257 : 731 – 736 . Google Scholar Crossref Search ADS PubMed WorldCat 26 Pang TC , Wilson O, Argueta MA, Hugh TJ, Chou A, Samra JS et al. Frozen section of the pancreatic neck margin in pancreatoduodenectomy for pancreatic adenocarcinoma is of limited utility . Pathology 2014 ; 46 : 188 – 192 . Google Scholar Crossref Search ADS PubMed WorldCat 27 Sugiura T , Uesaka K, Mihara K, Sasaki K, Kanemoto H, Mizuno T et al. Margin status, recurrence pattern, and prognosis after resection of pancreatic cancer . Surgery 2013 ; 154 : 1078 – 1086 . Google Scholar Crossref Search ADS PubMed WorldCat 28 Westgaard A , Tafjord S, Farstad IN, Cvancarova M, Eide TJ, Mathisen O et al. Resectable adenocarcinomas in the pancreatic head: the retroperitoneal resection margin is an independent prognostic factor . BMC Cancer 2008 ; 8 : 5 . Google Scholar Crossref Search ADS PubMed WorldCat 29 Howard TJ , Krug JE, Yu J, Zyromski NJ, Schmidt CM, Jacobson LE et al. A margin-negative R0 resection accomplished with minimal postoperative complications is the surgeon's contribution to long-term survival in pancreatic cancer . J Gastrointest Surg 2006 ; 10 : 1338 – 1345 . Google Scholar Crossref Search ADS PubMed WorldCat 30 Kimbrough CW , St Hill CR, Martin RCG, McMasters KM, Scoggins CR. Tumor-positive resection margins reflect an aggressive tumor biology in pancreatic cancer . J Surg Oncol 2013 ; 107 : 602 – 607 . Google Scholar Crossref Search ADS PubMed WorldCat 31 Mathur A , Ross SB, Luberice K, Kurian T, Vice M, Toomey P et al. Margin status impacts survival after pancreaticoduodenectomy but negative margins should not be pursued . Am Surg 2014 ; 80 : 353 – 360 . Google Scholar Crossref Search ADS PubMed WorldCat 32 Raut CP , Tseng JF, Sun CC, Wang H, Wolff RA, Crane CH et al. Impact of resection status on pattern of failure and survival after pancreaticoduodenectomy for pancreatic adenocarcinoma . Ann Surg 2007 ; 246 : 52 – 60 . Google Scholar Crossref Search ADS PubMed WorldCat 33 Rau BM , Moritz K, Schuschan S, Alsfasser G, Prall F, Klar E. R1 resection in pancreatic cancer has significant impact on long-term outcome in standardized pathology modified for routine use . Surgery 2012 ; 152 : S103 – S111 . Google Scholar Crossref Search ADS PubMed WorldCat 34 Campbell F , Bennett M, Foulis AJ. Minimum Dataset for Histopathological Reporting of Pancreatic, Ampulla of Vater and Bile Duct Carcinoma . Royal College of Pathologists : London , 2002 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 35 Gill AJ , Johns AL, Eckstein R, Samra JS, Kaufman A, Chang DK et al. Synoptic reporting improves histopathological assessment of pancreatic resection speciments . Pathology 2008 ; 6 : 1 – 7 . Google Scholar OpenURL Placeholder Text WorldCat 36 Staley CA , Cleary KR, Abbruzzese JL, Lee JE, Ames FC, Fenoglio CJ et al. The need for standardised pathologic staging of pancreaticoduodenectomy specimens . Pancreas 1996 ; 12 : 373 – 380 . Google Scholar Crossref Search ADS PubMed WorldCat 37 Japan Pancreas Society (ed.) General Rules for the Study of Pancreatic Cancer (2nd edn). Kanehara : Tokyo , 2003 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 38 Greene FI , Page DL, Fleming ID (eds). Exocrine pancreas. In AJCC Cancer Staging Manuel (6th edn). Springer : New York , 2002 ; 157 – 164 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 39 Washington K , Berlin J, Branton P, Burgart LJ, Carter DK, Fitzgibbons P et al. ; College of American Pathologists. Protocol for the Examination of Specimens from Patients with Carcinoma of the Exocrine Pancreas ; 2013 . http://www.cap.org/ShowProperty?nodePath=/UCMCon/Contribution%20Folders/WebContent/pdf/pancreasexo-13protocol-3201.pdf [accessed 2 July 2015]. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 40 Sobin L International Union Against Cancer: TNM Classification of Malignant Tumours . Wiley-Liss : New York , 1997 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 41 Noto M , Miwa K, Kitagawa H, Kayahara M, Takamura H, Shimizu K et al. Pancreas head carcinoma: frequency of invasion to soft tissue adherent to the superior mesenteric artery . Am J Surg Pathol 2005 ; 29 : 1056 – 1061 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 42 Kooby DA , Lad NL, Squires MH, Maithel SK, Sarmiento JM, Staley CA et al. Value of intraoperative neck margin analysis during Whipple for pancreatic adenocarcinoma. A multicenter analysis of 1399 patients . Ann Surg 2014 ; 260 : 494 – 501 . Google Scholar Crossref Search ADS PubMed WorldCat 43 Hernandez J , Mullinax J, Clark W, Toomey P, Villadolid D, Morton C et al. Survival after pancreaticoduodenectomy is not improved by extending resections to achieve negative margins . Ann Surg 2009 ; 250 : 76 – 80 . Google Scholar Crossref Search ADS PubMed WorldCat 44 Khalifa MA , Maksymov V, Rowsell C. Retroperitoneal margin of the pancreaticoduodenectomy specimen: anatomic mapping for the surgical pathologist . Virchows Arch 2009 ; 454 : 125 – 131 . Google Scholar Crossref Search ADS PubMed WorldCat 45 Neoptolemos JP , Stocken DD, Dunn JA, Almond J, Beger HG, Pederzoli P et al. Influence of resection margins on survival for patients with pancreatic cancer treated by adjuvant chemoradiation and/or chemotherapy in the ESPAC-1 randomized controlled trial . Ann Surg 2001 ; 234 : 758 – 768 . Google Scholar Crossref Search ADS PubMed WorldCat 46 Hishinuma S , Ogata Y, Tomikawa M, Ozawa I, Hirabayashi K, Igarashi S. Patterns of recurrence after curative resection of pancreatic cancer, based on autopsy findings . J Gastrointest Surg 2006 ; 10 : 511 – 518 . Google Scholar Crossref Search ADS PubMed WorldCat 47 Feakins R , Campbell F, Verbeke CS. Survey of UK histopathologists' approach to the reporting of resection specimens for carcinomas of the pancreatic head . J Clin Pathol 2013 ; 66 : 715 – 717 . Google Scholar Crossref Search ADS PubMed WorldCat 48 Verbeke CS , Gladhaug IP. Resection margin involvement and tumour origin in pancreatic head cancer . Br J Surg 2012 ; 99 : 1036 – 1049 . Google Scholar Crossref Search ADS PubMed WorldCat 49 Schlitter AM , Esposito I. Definition of microscopic tumour clearance (R0) in pancreatic cancer resections . Cancers (Basel) 2010 ; 2 : 2001 – 2010 . Google Scholar Crossref Search ADS PubMed WorldCat 50 Bandyopadhyay S , Basturk O, Coban I, Thirabanjasak D, Liang H, Altinel D et al. Isolated solitary ducts (naked ducts) in adipose tissue: a specific but underappreciated finding of pancreatic adenocarcinoma and one of the potential reasons of understaging and high recurrence rate . Am J Surg Pathol 2009 ; 33 : 425 – 429 . Google Scholar Crossref Search ADS PubMed WorldCat 51 Evans DB , Farnell MB, Lillemoe KD, Vollmer C, Strasberg SM, Schulick RD. Surgical treatment of resectable and borderline resectable pancreas cancer: expert consensus statement . Ann Surg Oncol 2009 ; 16 : 1736 – 1744 . Google Scholar Crossref Search ADS PubMed WorldCat 52 Evans DB , Rich TA, Byrd DR, Cleary KR, Connelly JH, Levin B et al. Preoperative chemoradiation and pancreaticoduodenectomy for adenocarcinoma of the pancreas . Arch Surg 1992 ; 127 : 1335 – 1339 . Google Scholar Crossref Search ADS PubMed WorldCat 53 Pisters PW , Abbruzzese JL, Janjan NA, Cleary KR, Charnsangavej C, Goswitz MS et al. Rapid-fractionation preoperative chemoradiation, pancreaticoduodenectomy, and intraoperative radiation therapy for resectable pancreatic adenocarcinoma . J Clin Oncol 1998 ; 12 : 3843 – 3850 . Google Scholar OpenURL Placeholder Text WorldCat 54 Hoffman JP , Lipsitz S, Pisansky T, Weese JL, Solin L, Benson AB. Phase II trial of preoperative radiation therapy and chemotherapy for patients with localized, resectable adenocarcinoma of the pancreas: an Eastern Cooperative Oncology Group Study . J Clin Oncol 1998 ; 16 : 317 – 323 . Google Scholar Crossref Search ADS PubMed WorldCat 55 White RR , Hurwitz HI, Morse MA, Lee C, Anscher MS, Paulson EK et al. Neoadjuvant chemoradiation for localized adenocarcinoma of the pancreas . Ann Surg Oncol 2001 ; 8 : 758 – 765 . Google Scholar Crossref Search ADS PubMed WorldCat 56 Pisters PW , Wolff RA, Janjan NA, Cleary KR, Charnsangavej C, Crane CN et al. Preoperative paclitaxel and concurrent rapid-fractionation radiation for resectable pancreatic adenocarcinoma: toxicities, histologic response rates and event-free outcome . J Clin Oncol 2002 ; 20 : 2537 – 2544 . Google Scholar Crossref Search ADS PubMed WorldCat 57 Calvo FA , Matute R, Garcia-Sabrido JL, Gomez-Espi M, Martinez NE, Lozano MA et al. Neoadjuvant chemoradiation with tegafur in cancer of the pancreas: initial analysis of clinical tolerance and outcome . Am J Clin Oncol 2004 ; 27 : 343 – 349 . Google Scholar Crossref Search ADS PubMed WorldCat 58 Talamonti MS , Small W, Mulcahy MF, Wayne JD, Attaluri V, Colletti LM et al. A multi-institutional phase II trial of preoperative full-dose gemcitabine and concurrent radiation for patients with potentially resectable pancreatic carcinoma . Ann Surg Oncol 2005 ; 13 : 150 – 158 . Google Scholar Crossref Search ADS WorldCat 59 Le Scodan R , Mornex F, Partensky C, Mercier C, Valette PJ, Ychou M et al. Histopathological response to preoperative chemoradiation for resectable pancreatic adenocarcinoma: the French Phase II FFCD 9704-SFRO Trial . Am J Clin Oncol 2008 ; 31 : 545 – 552 . Google Scholar Crossref Search ADS PubMed WorldCat 60 Evans DB , Varadhachary GR, Crane CH, Sun CC, Lee JE, Pisters PW et al. Pre-operative gemcitabine-based chemoradiation for patients with resectable adenocarcinoma of the pancreatic head . J Clin Oncol 2008 ; 26 : 3496 – 3502 . Google Scholar Crossref Search ADS PubMed WorldCat 61 Varadhachary GR , Wolff RA, Crane CH, Sun CC, Lee JE, Pisters PW et al. Preoperative gemcitabine and cisplatin followed by gemcitabine-based chemoradiation for resectable adenocarcinoma of the pancreatic head . J Clin Oncol 2008 ; 26 : 3487 – 3495 . Google Scholar Crossref Search ADS PubMed WorldCat 62 Turrini O , Viret F, Moureau-Zabotto L, Guiramand J, Moutardier V, Lelong B et al. Neoadjuvant 5 fluorouracil–cisplatin chemoradiation effect on survival in patients with resectable pancreatic head adenocarcinoma: a ten-year single institution experience . Oncology 2009 ; 76 : 413 – 419 . Google Scholar Crossref Search ADS PubMed WorldCat 63 Turrini O , Ychou M, Moureau-Zabotto L, Rouanet P, Giovanni M, Moutardier V et al. Neoadjuvant docetaxel-based chemoradiation for resectable adenocarcinoma of the pancreas: new neoadjuvant regimen was safe and provided an interesting pathologic response . Eur J Surg Oncol 2010 ; 36 : 987 – 992 . Google Scholar Crossref Search ADS PubMed WorldCat 64 Takahashi H , Ohigashi H, Gotoh K, Marubashi S, Yamada T, Murata M et al. Preoperative gemcitabine-based chemoradiation therapy for resectable and borderline resectable pancreatic cancer . Ann Surg 2013 ; 258 : 1040 – 1050 . Google Scholar Crossref Search ADS PubMed WorldCat 65 Palmer DH , Stocken DD, Hewitt H, Markham CE, Hassan B, Johnson PJ et al. A randomized phase 2 trial of neoadjuvant chemotherapy in resectable pancreatic cancer: gemcitabine alone versus gemcitabine combined with cisplatin . Ann Surg Oncol 2007 ; 14 : 2088 – 2096 . Google Scholar Crossref Search ADS PubMed WorldCat 66 Heinrich S , Pestalozzi BC, Schafer M, Weber A, Bauerfeind P, Knuth A et al. Prospective phase II trial of neoadjuvant chemotherapy with gemcitabine and cisplatin for resectable adenocarcinoma of the pancreatic head . J Clin Oncol 2008 ; 26 : 2526 – 2531 . Google Scholar Crossref Search ADS PubMed WorldCat 67 Greene FL . TNM staging for malignancies of the digestive tract: 2003 changes and beyond . Semin Surg Oncol 2003 ; 21 : 23 – 29 . Google Scholar Crossref Search ADS PubMed WorldCat 68 Sobin LH , Wittekind C (eds). TNM Classification of Malignant Tumours (6th edn). Wiley-Liss : New York , 2002 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 69 O'Reilly EM , Perelshteyn A, Jarnagin WR, Schattner M, Gerdes H, Capanu M et al. A single-arm, nonrandomised phase II trial of neoadjuvant gemcitabine and oxaliplatin in patients with resectable pancreas adenocarcinoma . Ann Surg 2014 ; 260 : 142 – 148 . Google Scholar Crossref Search ADS PubMed WorldCat 70 Barbour N , O'Rourke N, Samra JS, Haghighi KS, Kench J, Mitchell J et al. A multicenter, phase II trial of preoperative gemcitabine and nab-paclitaxel for resectable pancreas cancer: the AGITG GAP study . J Clin Oncol 2015 ; 33 ( Suppl 3 ); abstract 387. Google Scholar OpenURL Placeholder Text WorldCat Snapshot quiz 15/11 Open in new tabDownload slide Author notes Co-authors can be found under the heading Collaborators. © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd

Day, A R; Smith, R V P; Scott, M J P; Fawcett, W J; Rockall, T A

doi: 10.1002/bjs.9936pmid: 26395762

Abstract Background One of the key elements of managed recovery is thought to be suppression of the neuroendocrine response using regional analgesics. This may be superfluous in laparoscopic colorectal surgery with small wounds. This trial assessed the effects of spinal analgesia versus intravenous patient-controlled analgesia (PCA) on neuroendocrine responses in that setting. Methods A randomized clinical trial was conducted with participation of patients undergoing laparoscopic colorectal surgery within a managed recovery programme. Consenting patients were allocated randomly to spinal analgesia or morphine PCA as primary postoperative analgesia. The primary outcome was interleukin (IL) 6 levels; secondary outcomes were levels of cortisol, glucose, insulin and other cytokines, pain scores, morphine use and length of hospital stay. Stress response analysis was conducted before operation, and 3, 6, 12, 24 and 48 h after surgery. Results Of 143 eligible patients, 133 were randomized and 120 completed the study. Baseline patient characteristics were similar in the two groups. There were no significant differences in median levels of insulin, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12, interferon γ, tumour necrosis factor α or vascular endothelial growth factor between the spinal analgesia and PCA groups at any time point. Three hours after surgery (but at no other time point) median (i.q.r.) levels of cortisol (468 (329–678) versus 701 (429–820) nmol/l; P = 0·004) and glucose (6·1 (5·4–7·5) versus 7·0 (6·0–7·7) mmol/l; P = 0·012) were lower in the spinal analgesia group than in the PCA group. Median (i.q.r.) levels of total intravenous morphine were lower in the spinal analgesia group (10·0 (3·3–15·8) versus 45·5 (34·0–60·5) mg; P < 0·001). Conclusion Spinal analgesia reduced early neuroendocrine responses and overall parenteral morphine use. Registration number: NCT01128088 (http://www.clinicaltrials.gov). Introduction The stress response to surgery has been characterized extensively, as have techniques to modify it. Broadly, there are two major components to this response. The first is a systemic neuroendocrine response with the concomitant metabolic sequelae. This is characterized by sympathetic nervous system and pituitary activation, resulting in many predictable metabolic consequences characterized by catabolism, insulin resistance and hyperglycaemia1,2. The second component comprises inflammatory and immunological changes3,4. Modification of the neuroendocrine (and metabolic) responses from afferent neural blockade by thoracic epidural anaesthesia (TEA) has been studied extensively4,5. Although TEA is useful in open colorectal operations, it may be unnecessary in laparoscopic surgery6. When given in addition to general anaesthesia, spinal analgesia also provides afferent neural blockade, but of much more limited duration as it is a ‘one shot’ approach. It may be associated with improved analgesia and shorter hospital stay than TEA6–8. The short duration of the block avoids prolonged immobilization and hypotension. However, it might be anticipated that the limited duration of neural blockade may result in relatively brief modification of the neuroendocrine stress response. To date, the duration and magnitude of the stress response to laparoscopic surgery using spinal analgesia has not been characterized. This study investigated the stress response to laparoscopic colorectal surgery in patients randomized to receive spinal analgesia or intravenous patient-controlled analgesia (PCA) as the principal pain management modality. Methods This single-site study randomized patients scheduled for laparoscopic colorectal resection for benign or malignant disease. Patients were ineligible or excluded if there was a planned or unforeseen stoma, conversion to an open procedure (defined by an incision larger than that required for specimen extraction), a contraindication to spinal analgesia, a contraindication to use of an oesophageal Doppler probe, diabetes requiring insulin therapy, exogenous steroid use in the preceding 3 months, or if oral bowel preparation was to be used. The trial was registered with www.clinicaltrials.gov (NCT01128088). Ethical approval for the trial was obtained from the Kent and Brighton West Research Ethics Committees, and the hospital research and development department. The trial was conducted in compliance with the Declaration of Helsinki and the principles of the International Conference on Harmonization of Good Clinical Practice. The chief investigator was responsible for enrolment and assignment of individuals to the treatment groups as determined by randomization. Randomization codes were prepared by a University of Surrey statistician and sealed in sequentially numbered opaque envelopes; these were kept in an off-site building. Once patients had consented to the trial they were randomized equally in a factorial design into one of four groups depending on the type of primary postoperative analgesia (spinal analgesia or PCA) and intravenous fluid (Hartmann's solution or 6 per cent Volulyte® (Fresnius Kabi, Manor Park, UK) to be used. Data on type of intravenous fluid are not reported here. Spinal analgesia was administered before induction by inserting a spinal needle through the L2–3 or L3–4 interspace with the patient sitting, and injecting 2·5 ml of 0·5 per cent hyperbaric bupivacaine with 0·25 mg diamorphine. PCA involved administering 10 mg morphine towards the end of the procedure and then connecting a pump to deliver morphine on demand at a dose of 1 mg, with a 5-min lockout and a maximum dose of 20 mg every 4 h. Neither the participants nor the clinicians involved in the trial were blinded to the treatment administered. The primary outcome of this study was plasma interleukin (IL) 6 levels. The null hypothesis was that there was no statistical difference between the four groups in plasma IL-6 levels at 6 h after the start of surgery. Secondary outcomes were: levels of cortisol, glucose, insulin and other cytokines, morphine use, physiological variables and length of hospital stay. The aim of the trial was twofold: to ascertain the effect that choice of analgesia may have on the stress response, and the effect the type of fluid used may have on secondary outcomes when administered with guidance from the oesophageal Doppler monitor; the second aim is not addressed in this article. Preoperative protocol The majority of patients were admitted on the morning of surgery. In keeping with the principles of enhanced recovery, all patients were allowed to eat and drink the night before surgery and to have clear fluids up to 2 h before induction of anaesthesia. An oral carbohydrate drink (Preload™; Vitaflo, Liverpool, UK) was administered as 100 g in 800 ml water at 21.00 hours the night before surgery, and a further 50 g in 400 ml water at 06.00 hours on the day of surgery. Patients undergoing right-sided colonic resection received no bowel preparation and those having a left-sided resection received one phosphate enema at 07.00 hours on the day of surgery. Baseline measurements were taken before induction of anaesthesia. Bodyweight, height, abdominal circumference at the level of the anterior superior iliac spines, and variables to calculate the Portsmouth modification of the Physiological and Operative Severity Score for the EnUmeration of Mortality and morbidity (P-POSSUM) score were recorded. Between 07.00 and 08.00 hours venous blood was drawn by venepuncture into glucose, serum and heparin containers for the stress response analysis. For patients not admitted until later on the day of surgery, blood was drawn when practically possible. Glucose levels were analysed immediately; the serum sample was centrifuged, separated and stored at −20°C in the hospital laboratory for later analysis of insulin and cortisol levels. The heparin tube was transferred to the university oncology laboratory and placed on a rocker for analysis later that day. Perioperative protocol Anaesthesia was induced using propofol (2–3 mg/kg), alfentanil (10 µg/kg) and rocuronium (0·6 mg/kg). No premedications were administered. The patient was intubated and ventilated mechanically; anaesthesia was maintained with sevoflurane at a mean alveolar concentration of 1·0–1·4, oxygen-enriched air and a remifentanil infusion (40 µg/ml) titrated to the depth of anaesthesia. A nasogastric tube was not inserted, unless the stomach required decompression; if used, it was removed at the end of the procedure. All patients had a 16-Fr peripheral line, 20-Fr arterial line and a four-lumen right internal jugular central venous line inserted. A Bair-HuggerTM (3M, Bracknell, UK) and urinary catheter were used in all patients. Antibiotic prophylaxis comprising 1·5 g cefuroxime and 500 mg metronidazole was administered intravenously before skin incision, with two subsequent doses of 750 mg cefuroxime and 500 mg metronidazole. Patients received intravenous fluid during the perioperative phase as determined by the randomization code. Fluid was administered in order to optimize stroke volume as guided by an oesophageal Doppler monitor (Deltex, Chichester, UK) following the modified Wakeling protocol as described previously6. The oesophageal probe was placed immediately after induction, and readings were recorded throughout the procedure; the probe was removed before extubation. At the end of surgery any residual neuromuscular blockade was reversed with 2·5 mg neostigmine and 0·5 mg glycopyrrolate. All patients received 1 g paracetamol and 4 mg ondansetron; those who had PCA also received local wound infiltration with 20 ml levobupivacaine (5 mg/ml). Postoperative protocol In addition to the randomly assigned primary postoperative analgesia, patients received oral paracetamol 1 g four times daily, oral diclofenac 50 mg three times daily and oral omeprazole 20 mg once daily. Rescue analgesia was prescribed as tramadol 50 mg and morphine 5–10 mg, both orally as required. Antiemetics available were intravenous or oral cyclizine 50 mg three times daily and ondansetron 4 mg four times daily as required. All patients received lactulose 10 ml twice daily, 200 ml Fortisip (NutriDrinks, Perivale, UK) three times daily, and subcutaneous enoxaparin 40 mg once daily. Hartmann's solution was infused after surgery, 1 litre over 8 h followed by another litre over 12 h; intravenous fluid was then discontinued. Patients were allowed to eat and drink freely as tolerated, and encouraged to mobilize. The urinary catheter was removed the day after surgery unless there was a clinical indication not to (for example when surgery was for a colovesical fistula). Further blood was drawn from the central venous catheter for stress response analysis at 3, 6, 12 and 24 h after the start of surgery. At 48 h blood was drawn by venepuncture as the central venous catheter was removed the day after surgery. The blood was prepared as described earlier. At 6 and 24 h after surgery, central venous blood was analysed for oxygen saturation. If a value of less 60 per cent was obtained, a bolus of 500 ml Hartmann's solution or 6 per cent Volulyte® (according to randomization) was administered and the venous blood saturations were rechecked, as required by the ethics committee for patient safety6. At 24-h intervals from the start of surgery, until discharge, bodyweight, abdominal circumference and pain score (determined using a visual analogue scale from 0 to 10) were noted. The times to first bowel sounds, passage of flatus, passage of stool and discharge from hospital were recorded, calculated from the end of surgery. Patients were considered fit for discharge once they were comfortable with simple oral analgesia, had passed urine following removal of the urinary catheter, could tolerate a normal diet and had fully mobilized. The total amount of analgesia, intravenous fluid and antiemetic required during the hospital stay was recorded. All patients were reviewed in the outpatient clinic between 2 and 6 weeks after discharge. Laboratory cytokine analysis of plasma The heparin blood tubes were centrifuged at 660g for 10 min. The plasma was then drawn off and centrifuged for a further 10 min, before being aliquoted and stored at −80°C. Cytokine analysis was carried out using a Bio-Plex® Human Cytokine assay reagent kit (Bio-Rad Laboratories, Hemel Hempstead, UK). Use of a 96-well plate allowed duplicate analysis of 39 samples, with eight standards. All samples for an individual patient were analysed on the same plate. The samples were thawed and prepared in accordance with the manufacturer's instructions. The plates were analysed on a calibrated Bio-Plex® plate reader system. Statistical analysis To determine whether there was a difference of 200 pg/ml in IL-6 concentration between the four study groups over a 6-h interval (from 0 to 6 h), the number of patients required in each group was calculated as 30, with a power of 80 per cent and a significance of 5 per cent, using a two-sided test. This was based on the level of variation (s.d. 269·4) determined previously in a pilot study undertaken in this unit. However, to allow for patient drop-out owing to conversion to open surgery (rate 5·6 per cent, obtained from a prospectively maintained database for the colorectal unit), 32 patients were required in each treatment group. These group sizes would also detect a difference of 100 pg/ml for insulin and 300 pg/ml for cortisol between 0 and 6 h. Continuous data were assessed for normality using the Kolmogorov–Smirnov test. The majority of data were distributed asymmetrically and these data are presented as median (i.q.r.). Asymmetrical data from more than two groups were compared by means of independent-samples Kruskal–Wallis test. If a difference was detected, Mann–Whitney U test was used in pairwise fashion to identify the groups that were statistically different. Asymmetrical data from two groups were analysed with an independent-samples Mann–Whitney U test. A related Wilcoxon signed-rank test was used to analyse medians within a group. Categorical variables were compared using Pearson's χ2 test. Where significance was identified between the two analgesia groups, analysis of co-variance (ANCOVA) was performed to identify any influence of fluid type; data for these analyses are presented as mean(s.d.). No change to the study design was made following commencement of the trial. Statistical analysis was performed using SPSS® version 19 (IBM, Armonk, New York, USA). Results Of the 143 patients approached to participate in this trial, 133 were randomized and 120 patients completed the trial (Fig. 1). There were no significant differences in patient characteristics between the two analgesia groups (Table 1). Fig. 1 Open in new tabDownload slide CONSORT diagram for the trial. PCA, patient-controlled analgesia Table 1 Demographic, clinical and operative data . PCA (n = 60) . Spinal analgesia (n = 60) . Age (years)* 66 (48–77) 65 (51–73) Sex ratio (M : F) 27 : 33 32 : 28 Weight (kg)* 75 (62–85) 74 (64–86) Body mass index (kg/m2)* 26 (24–28) 26 (23–29) P-POSSUM score* 26 (23–30) 24 (23–29) ASA fitness grade I 21 (35) 25 (42) II 39 (65) 34 (57) III 0 (0) 1 (2) Disease Cancer 30 (50) 35 (58) Non-cancer 30 (50) 25 (42) Procedure Right hemicolectomy 25 (42) 19 (32) Left hemicolectomy 5 (8) 4 (7) Sigmoid colectomy 4 (7) 5 (8) Anterior resection (not full TME) 21 (35) 29 (48) Anterior resection (full TME) 3 (5) 2 (3) Other 2 (3) 1 (2) Side of resection Right 25 (42) 20 (33) Left 35 (58) 40 (67) Duration of surgery (min)* 95 (71–134) 102 (70–122) Incision length (cm)* 6 (5–7) 6 (5–7) . PCA (n = 60) . Spinal analgesia (n = 60) . Age (years)* 66 (48–77) 65 (51–73) Sex ratio (M : F) 27 : 33 32 : 28 Weight (kg)* 75 (62–85) 74 (64–86) Body mass index (kg/m2)* 26 (24–28) 26 (23–29) P-POSSUM score* 26 (23–30) 24 (23–29) ASA fitness grade I 21 (35) 25 (42) II 39 (65) 34 (57) III 0 (0) 1 (2) Disease Cancer 30 (50) 35 (58) Non-cancer 30 (50) 25 (42) Procedure Right hemicolectomy 25 (42) 19 (32) Left hemicolectomy 5 (8) 4 (7) Sigmoid colectomy 4 (7) 5 (8) Anterior resection (not full TME) 21 (35) 29 (48) Anterior resection (full TME) 3 (5) 2 (3) Other 2 (3) 1 (2) Side of resection Right 25 (42) 20 (33) Left 35 (58) 40 (67) Duration of surgery (min)* 95 (71–134) 102 (70–122) Incision length (cm)* 6 (5–7) 6 (5–7) Values in parentheses are percentages unless indicated otherwise; * values are median (i.q.r.). PCA, patient-controlled analgesia; ASA, American Society of Anesthesiologists; P-POSSUM, Portsmouth modification of the Physiological and Operative Severity Score for the EnUmeration of Mortality and morbidity; TME, total mesorectal excision. Open in new tab Table 1 Demographic, clinical and operative data . PCA (n = 60) . Spinal analgesia (n = 60) . Age (years)* 66 (48–77) 65 (51–73) Sex ratio (M : F) 27 : 33 32 : 28 Weight (kg)* 75 (62–85) 74 (64–86) Body mass index (kg/m2)* 26 (24–28) 26 (23–29) P-POSSUM score* 26 (23–30) 24 (23–29) ASA fitness grade I 21 (35) 25 (42) II 39 (65) 34 (57) III 0 (0) 1 (2) Disease Cancer 30 (50) 35 (58) Non-cancer 30 (50) 25 (42) Procedure Right hemicolectomy 25 (42) 19 (32) Left hemicolectomy 5 (8) 4 (7) Sigmoid colectomy 4 (7) 5 (8) Anterior resection (not full TME) 21 (35) 29 (48) Anterior resection (full TME) 3 (5) 2 (3) Other 2 (3) 1 (2) Side of resection Right 25 (42) 20 (33) Left 35 (58) 40 (67) Duration of surgery (min)* 95 (71–134) 102 (70–122) Incision length (cm)* 6 (5–7) 6 (5–7) . PCA (n = 60) . Spinal analgesia (n = 60) . Age (years)* 66 (48–77) 65 (51–73) Sex ratio (M : F) 27 : 33 32 : 28 Weight (kg)* 75 (62–85) 74 (64–86) Body mass index (kg/m2)* 26 (24–28) 26 (23–29) P-POSSUM score* 26 (23–30) 24 (23–29) ASA fitness grade I 21 (35) 25 (42) II 39 (65) 34 (57) III 0 (0) 1 (2) Disease Cancer 30 (50) 35 (58) Non-cancer 30 (50) 25 (42) Procedure Right hemicolectomy 25 (42) 19 (32) Left hemicolectomy 5 (8) 4 (7) Sigmoid colectomy 4 (7) 5 (8) Anterior resection (not full TME) 21 (35) 29 (48) Anterior resection (full TME) 3 (5) 2 (3) Other 2 (3) 1 (2) Side of resection Right 25 (42) 20 (33) Left 35 (58) 40 (67) Duration of surgery (min)* 95 (71–134) 102 (70–122) Incision length (cm)* 6 (5–7) 6 (5–7) Values in parentheses are percentages unless indicated otherwise; * values are median (i.q.r.). PCA, patient-controlled analgesia; ASA, American Society of Anesthesiologists; P-POSSUM, Portsmouth modification of the Physiological and Operative Severity Score for the EnUmeration of Mortality and morbidity; TME, total mesorectal excision. Open in new tab Cytokines There were no statistically significant differences in median IL-6 levels between the two analgesia groups at any time point (Fig. 2). Postoperative IL-6 levels were significantly higher than preoperative levels at all time points in both groups. In the spinal analgesia group, levels peaked at 6 h before returning towards preoperative values. In the PCA group, IL-6 levels reached a plateau between 6 and 24 h, then decreased towards preoperative levels. Fig. 2 Open in new tabDownload slide Median levels of interleukin (IL) 6 before, and 3, 6, 12, 24 and 48 h after surgery in spinal analgesia and patient-controlled analgesia (PCA) groups There were no significant differences at 6 h between the spinal analgesia and PCA groups in median levels of IL-2 (1·8 (0·9–2·7) versus 1·8 (0·3–13·3) pg/ml), IL-4 (0·3 (0·1–0·7) versus 0·3 (0·2–0·6) pg/ml), IL-8 (29·5 (17·0–102·4) versus 31·7 (17·6–81·0) pg/ml), IL-10 (14·7 (6·9–26·9) versus 16·2 (9·3–29·8) pg/ml), IL-12 (14·2 (6·1–25·3) versus 16·8 (8·4–35·5) pg/ml), interferon γ (16·5 (7·1–27·1) versus 15·4 (7·6–29·0) pg/ml), tumour necrosis factor α (3·1 (1·5–7·5) versus 3·3 (1·3–5·8) pg/ml) or vascular endothelial growth factor (94 (58–169) versus 113 (72–206) pg/ml). Cortisol, insulin and glucose levels The spinal analgesia group had significantly lower median cortisol levels at 3 h compared with the PCA group (468 (329–678) versus 701 (429–820) nmol/l; P = 0·004) (Fig. 3). This effect was not influenced by the fluid type used, as assessed by ANCOVA (mean(s.d.) 491·4(256·7) versus 629·8(278·2) nmol/l for spinal analgesia versus PCA; P = 0·005). No significant difference was detected between the groups at the other time points measured. Fig. 3 Open in new tabDownload slide Median levels of cortisol before, and 3, 6, 12, 24 and 48 h after surgery in spinal analgesia and patient-controlled analgesia (PCA) groups Median insulin levels decreased in the postoperative period in the two groups, returning to slightly above preoperative levels by 48 h (Fig. 4). There were no significant differences between the two analgesia groups. Fig. 4 Open in new tabDownload slide Median levels of insulin before, and 3, 6, 12, 24 and 48 h after surgery in spinal analgesia and patient-controlled analgesia (PCA) groups Median glucose levels increased during the postoperative period in the two analgesia groups. At 3 h the median glucose level was significantly higher in the PCA group than in the spinal analgesia group (7·0 (6·0–7·7) versus 6·1 (5·4–7·5) mmol/l; P = 0·012) (Fig. 5). This effect was not influenced by the fluid type used (mean(s.d.) 7·1(1·4) versus 6·5(1·3) mmol/l for spinal analgesia versus PCA; P = 0·013). There were no other significant differences. Fig. 5 Open in new tabDownload slide Median levels of glucose before, and 3, 6, 12, 24 and 48 h after surgery in spinal analgesia and patient-controlled analgesia (PCA) groups Pain scores and analgesia Median pain scores measured at rest were significantly lower in patients who received spinal analgesia compared with those using PCA at 3 h (4 (2–5) versus 8 (5–8); P < 0·001), 6 h (3 (1–5) versus 5 (3–6·5); P = 0·001), 12 h (2 (1–3) versus 3·5 (2–5); P = 0·008), 24 h (2 (0·6–4) versus 3 (2–5); P = 0·027) and 48 h (1 (0–2) versus 1·5 (0–3); P = 0·045). This effect was not influenced by the fluid type used, as shown by ANCOVA at 3 h (mean(s.d.) 3·8(2·7) versus 6·6(2·4) for spinal analgesia versus PCA; P < 0·001), 6 h (3·1(2·4) versus 4·8(2·4); P = 0·001), 12 h (2·1(1·7) versus 3·7(2·5); P = 0·005), 24 h (2·6(2·3) versus 3·4(2·0); P = 0·050) and 48 h (1·2(1·5) versus 1·9(1·9); P = 0·035). There was no significant difference between spinal analgesia and PCA groups in the median amount of paracetamol (8 (7–12) versus 9·5 (7–14) g respectively; P = 0·473), diclofenac (250 (200–350) versus 250 (150–400) mg; P = 0·893) or tramadol (250 (75–450) versus 300 (200–450) mg; P = 0·805) used. Patients receiving PCA used significantly greater volumes of morphine compared with those who had spinal analgesia (45·5 (34·0–60·5) versus 10·0 (3·3–15·8) mg; P < 0·001). This effect was not influenced by the fluid type used (mean(s.d.) 60·9(77·6) versus 12·5(15·1) mg for spinal analgesia versus PCA; P < 0·001). Other secondary outcomes Patients who received spinal analgesia required a significantly greater amount of phenylephrine to maintain the mean arterial BP within the assigned study parameters compared with those who received PCA (median 900 (300–1600) versus 300 (200–500) mg respectively; P = 0·001). There was a trend towards a reduced length of hospital stay in patients who received spinal analgesia, but this did not reach statistical significance (median time to discharge 54 (45–75) h versus 70 (50–94) h in PCA group; P = 0·059). There were no significant differences between spinal analgesia and PCA groups in median time to passage of first flatus (28 (17–44) versus 30 (21–41) h respectively; P = 0·349), passage of first stool (50 (37–69) versus 51 (42–86) h; P = 0·345) and ability to tolerate normal diet (21 (16–39) versus 26 (19–49) h; P = 0·081). Bodyweight increased in all patients after operation, but there was no significant difference in maximum weight gain before discharge between the spinal analgesia and PCA group (2·0 (1·1–3·2) versus 1·8 (0·9–3·4) kg respectively; P = 0·640). Complications Thirty-two patients experienced a complication during recovery. There was one death in the study (total mortality rate 0·8 per cent). The 30-day readmission rate was 8·3 per cent (10 of 120). Patients who received PCA had significantly more complications than those who received spinal analgesia (22 of 60 versus 10 of 60; P = 0·013) (Table 2). The difference was explained largely by a higher incidence of postoperative ileus in the PCA group (11 of 60 versus 2 of 60; P = 0·008). Table 2 Complications categorized according to the Dindo–Demartines–Clavien classification Grade . PCA (n = 60) . Spinal analgesia (n = 60) . 1 13 6 2 4 1 3 4 2 4 1 0 5 0 1 Grade . PCA (n = 60) . Spinal analgesia (n = 60) . 1 13 6 2 4 1 3 4 2 4 1 0 5 0 1 PCA, patient-controlled analgesia. Open in new tab Table 2 Complications categorized according to the Dindo–Demartines–Clavien classification Grade . PCA (n = 60) . Spinal analgesia (n = 60) . 1 13 6 2 4 1 3 4 2 4 1 0 5 0 1 Grade . PCA (n = 60) . Spinal analgesia (n = 60) . 1 13 6 2 4 1 3 4 2 4 1 0 5 0 1 PCA, patient-controlled analgesia. Open in new tab Discussion Patients receiving spinal analgesia exhibited a transient blockade of the spinal afferents, as reflected by suppression of the hypophyseal–pituitary–adrenal axis. At 3 h there was a statistically significant reduction in cortisol levels and glucose levels in the spinal analgesia group. The difference ceased to exist by 6 h, which is a reflection of the expected duration of the effect of spinal analgesia. There was no difference in insulin levels between the groups, with levels being reduced in both groups during the first 12 h after surgery. In common with the inflammatory response, the metabolic/endocrine markers had returned to preoperative levels by 48 h. Other studies9–12 have characterized the effects of longer-term neuroaxial blockade with an epidural infusion for major surgery, but no previous study has investigated such a block following a spinal anaesthetic for laparoscopic surgery. The authors were unsure whether any significant effects of spinal anaesthesia on the stress response would be evident given that the surgical stimulus (up to the cervical dermatomes) would have been considerably higher than the height of the spinal anaesthesia (up to the mid-thoracic dermatomes). In addition, it has been confirmed that the offset of neuroaxial blockade leads to any stress response modification being lost. These findings complement those of Fawcett and colleagues13, namely that both the onset and offset of the neuroaxial block is mirrored rapidly in the onset and offset of the neuroendocrine modification. Patients receiving spinal analgesia experienced a reduced rate of postoperative complications. This was explained largely by the number of patients experiencing postoperative ileus in the morphine group. There was a non-significant trend towards a reduced length of stay in the spinal analgesia group. The postoperative inflammatory response was not attenuated in the spinal analgesia group compared with that in the PCA group in this study. This is not unexpected as the magnitude of the response is governed largely by tissue damage and was unchanged by the use of regional anaesthesia. The global inflammatory response was reduced in magnitude and time in comparison with data available from patients undergoing open colorectal resection4,14. The use of spinal analgesia with concomitant simple analgesia has been shown previously to be sufficient to control postoperative pain adequately in the majority of patients6,8. The significant reduction in pain scores in the spinal analgesia group lasted well beyond the expected duration of action of the bupivacaine–diamorphine mixture, albeit with a decreasing level of significance. Although use of opioids has been linked to immunosuppression during the postoperative phase15, there was no evidence of this, as reflected by the cytokine measurements in this study. This study may be criticized for not including an epidural group; however, based on previous work6,8 the authors feel this would have been to the detriment of the patient. The dose of diamorphine used here was lower than in other published work8 (0·25 mg versus 1–1·5 mg); the dose in this study was selected as higher doses have been associated with postoperative nausea and vomiting, with no additional analgesic benefit16. In addition, the i.q.r. values are wide in the study groups, which reflects the wide variation from patient to patient when measuring the study variables; there could be the potential to ameliorate these differences with greater numbers in each group. This study was powered to 80 per cent based on identifying a significant difference in IL-6 levels between the four study groups, but the data have been presented in combined analgesia groups for clarity. Acknowledgements The study received research grant funding from the European Association of Endoscopic Surgeons, and equipment was provided by Deltex Medical, Chichester, UK. Disclosure: The authors declare no conflict of interest. References 1 Ljungqvist O , Nygren J, Thorell A. Modulation of post-operative insulin resistance by pre-operative carbohydrate loading . Proc Nutr Soc 2002 ; 61 : 329 – 336 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Wang ZG , Wang Q, Wang WJ, Qin HL. Randomized clinical trial to compare the effects of preoperative oral carbohydrate versus placebo on insulin resistance after colorectal surgery . Br J Surg 2010 ; 97 : 317 – 327 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Ishikawa M , Nishioka M, Hanaki N, Miyauchi T, Kashiwagi Y, Ioki H et al. Perioperative immune responses in cancer patients undergoing digestive surgeries . World J Surg Oncol 2009 ; 7 : 7 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Veenhof AA , Vlug MS, van der Pas MH, Sietses C, van der Peet DL, de Lange-de Klerk ES et al. Surgical stress response and postoperative immune function after laparoscopy or open surgery with fast track or standard perioperative care: a randomized trial . Ann Surg 2012 ; 255 : 216 – 221 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Kehlet H Fast-track surgery – an update on physiological care principles to enhance recovery . Langenbecks Arch Surg 2011 ; 396 : 585 – 590 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Levy BF , Scott MJ, Fawcett W, Fry C, Rockall TA. Randomized clinical trial of epidural, spinal or patient-controlled analgesia for patients undergoing laparoscopic colorectal surgery . Br J Surg 2011 ; 98 : 1068 – 1078 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Levy BF , Scott MJ, Fawcett WJ, Rockall TA. 23-hour-stay laparoscopic colectomy . Dis Colon Rectum 2009 ; 52 : 1239 – 1243 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Virlos I , Clements D, Beynon J, Ratnalikar V, Khot U. Short-term outcomes with intrathecal versus epidural analgesia in laparoscopic colorectal surgery . Br J Surg 2010 ; 97 : 1401 – 1406 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Møller W , Rem J, Brandt R, Kehlet H. Effect of posttraumatic epidural analgesia on the cortisol and hyperglycaemic response to surgery . Acta Anaesthesiol Scand 1982 ; 26 : 56 – 58 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Kouraklis G , Glinavou A, Raftopoulos L, Alevisou V, Lagos G, Karatzas G. Epidural analgesia attenuates the systemic stress response to upper abdominal surgery: a randomized trial . Int Surg 2000 ; 85 : 353 – 357 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 11 Holte K , Kehlet H. Epidural anaesthesia and analgesia – effects on surgical stress responses and implications for postoperative nutrition . Clin Nutr 2002 ; 21 : 199 – 206 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Lattermann R , Schricker T, Wachter U, Goertz A, Georgieff M. Intraoperative epidural blockade prevents the increase in protein breakdown after abdominal surgery . Acta Anaesthesiol Scand 2001 ; 45 : 1140 – 1146 . Google Scholar Crossref Search ADS PubMed WorldCat 13 Fawcett WJ , Edwards RE, Quinn AC, MacDonald IA, Hall GM. Thoracic epidural analgesia started after cardiopulmonary bypass. Adrenergic, cardiovascular and respiratory sequelae . Anaesthesia 1997 ; 52 : 294 – 299 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Sammour T , Kahokehr A, Chan S, Booth RJ, Hill AG. The humoral response after laparoscopic versus open colorectal surgery: a meta-analysis . J Surg Res 2010 ; 164 : 28 – 37 . Google Scholar Crossref Search ADS PubMed WorldCat 15 Afsharimani B , Cabot P, Parat MO. Morphine and tumor growth and metastasis . Cancer Metastasis Rev 2011 ; 30 : 225 – 238 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Saravanan S , Robinson AP, Qayoum Dar A, Columb MO, Lyons GR. Minimum dose of intrathecal diamorphine required to prevent intraoperative supplementation of spinal anaesthesia for caesarean section . Br J Anaesth 2003 ; 91 : 368 – 372 . Google Scholar Crossref Search ADS PubMed WorldCat Author notes Presented to the Annual Meeting of the Association of Surgeons of Great Britain and Ireland, Glasgow, UK, May 2013, and the Annual Meeting of the Association of Coloproctology of Great Britain and Ireland, Liverpool, UK, July 2013; published in abstract form as Br J Surg 2013; 100(Suppl 7): 2 and Colorectal Dis 2013; 15(Suppl 1): 1 © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd

Lederle, F A; Noorbaloochi, S; Nugent, S; Taylor, B C; Grill, J P; Kohler, T R; Cole, L

doi: 10.1002/bjs.9895pmid: 26331269

Abstract Background No effective treatment is currently available to prevent progression of small and medium-sized abdominal aortic aneurysms (AAAs). Identification of drugs with sufficient promise to justify large expensive randomized trials remains challenging. One potentially useful strategy is to look for associations between commonly used drugs and AAA enlargement in appropriately adjusted observational studies. Methods Potential AAA measurements were identified from abdominal imaging reports in the electronic data files of three medical centres from 1995 to 2010. AAA measurements were extracted manually and patients with an aneurysm of 3 cm or larger, who had at least two measurements over an interval of at least 6 months, were identified. Other data were obtained from the electronic data files (demographics, co-morbidities, smoking status, drug use) to conduct a propensity analysis of the associations of drugs and other factors with AAA enlargement. Results From 52 962 abdominal imaging studies, 5362 patients with an AAA of 3 cm or more were identified, of whom 2428 had at least two measurements over at least 6 months. Mean AAA follow-up was 3·4 years and the mean AAA enlargement rate was 2·0 mm per year. Propensity analysis demonstrated no significant association of AAA enlargement with statins, beta-blockers, angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers. Diabetes was associated with a reduction in AAA enlargement of 1·2 mm per year (P = 0·008), and chronic obstructive pulmonary disease was associated with increased enlargement (0·5 mm per year; P = 0·050). Moderate AAA measurement variation and substantial terminal digit preference were also observed, but the digit preference became less pronounced after 2000. Conclusion This study confirms the negative association of diabetes with AAA progression. There was no evidence that commonly used cardiovascular drugs affect AAA enlargement. Introduction Large randomized trials have demonstrated that one-time ultrasound screening decreases mortality from abdominal aortic aneurysm (AAA) and all causes, in selected high-risk populations1. Based on these studies and resulting recommendations from the US Preventive Services Task Force2 and other groups, ultrasound screening for AAA has become widespread. Screening, together with incidental imaging findings, results in the discovery of many AAAs, most of them smaller than 5·5 cm, the diameter threshold for elective repair established by randomized trials3. This presents an opportunity to intervene with treatments to inhibit aneurysm progression, but unfortunately no such treatments are available; these individuals are currently managed with imaging surveillance alone4. Few treatments to slow AAA enlargement have been tested adequately. Identification of drugs with sufficient promise to justify large expensive randomized trials is challenging, partly because positive animal studies have not been predictive of success in humans. An alternative strategy is to look for associations between commonly used drugs and aneurysm enlargement rates in appropriately adjusted observational studies. A variety of drugs have been proposed as potentially effective at reducing AAA enlargement, including statins, beta-blockers, angiotensin-converting enzyme (ACE) inhibitors and angiotensin II receptor blockers (ARBs)5. This approach has been limited by a scarcity of data sets that contain both sufficient numbers of AAA measurements and the necessary information on drugs and confounders. In the present study, AAA measurements were extracted from a large, comprehensive electronic medical record system, along with information on drug use and other factors. These data were used to assess AAA measurement variability and enlargement patterns, and the association of patient factors and several classes of drugs with AAA enlargement. AAA enlargement was selected as the outcome of interest because it represents disease progression, is strongly associated with rupture, is the main determinant of AAA repair and is the only one potentially modifiable by drugs. It is the usual primary outcome of randomized trials of medical therapy for AAA. Methods Electronic data files were searched from the US Department of Veterans Affairs (VA) medical centres in Minneapolis, Seattle and West Los Angeles. The study was approved by the institutional review boards of all three centres. Abdominal CT and ultrasound reports from 1995 to 2009 were identified using Current Procedural Terminology (CPT®) codes (74150, 74160, 74170–74175, 76700–76705, 76770–76775). Because some studies (such as ‘CT of the abdomen and pelvis’) had no CPT® code, studies without CPT® codes were also included if the procedure name contained ‘abdom’ plus a term indicating CT or ultrasonography (CT, US, U/S, ultrasound, echo). Using a trial-and-error approach with sample sets, a strategy to identify AAA measurements was developed from the text of the radiology reports (clinical history, body of report and impression), whereby reports were included if they contained: (a) ‘AAA’ or ‘aneu’, or (b) ‘aort’ within 60 characters of ‘mm’, ‘cm’, ‘millimeters’ or ‘centimeters’, and not ‘velocit’ (used in flow studies of occlusive disease). For individual patients with tests that met either of these criteria, all other abdominal CT or ultrasound reports from any of the three study centres were included, as well as abdominal CT or ultrasound reports from other VA medical centres throughout the nation if these outside reports themselves met criterion (a) or (b) above. Data extraction Research technicians reviewed each report and recorded maximum infrarenal aortic diameter in any cross-sectional plane, as reported by the radiologist, and any indication of aortic repair or rupture, using an application developed by the investigators that presented each report with key terms highlighted, and provided spaces to enter the desired information (Fig. 1). Each report was reviewed by two technicians who were unaware of each other's findings, and who could also submit a question. Discrepancies between the two technicians' reports and all questions were resolved by one of the authors. Fig. 1 Open in new tabDownload slide Data extraction application developed by the authors AAA was defined as an infrarenal aortic diameter of 3·0 cm or larger. For each individual, imaging studies from before the first report of AAA and those done after AAA repair were excluded. AAA repair was identified from the radiology reports and from the electronic records using CPT® (0078T–0081T, 34800–34805, 34825–34832, 35081–35103, 75952 or 75953) and ICD-9 (38.14, 38.44, 39.71) procedure codes. One of the authors resolved discrepancies between administrative evidence of repair or rupture and subsequent extracted reports, and instances of a large (1 cm) decrease in AAA diameter without reported repair. Other data on study patients, such as demographics, visits and hospital admissions, distance from residence to the medical centre, diagnoses and medications were obtained from VA databases. The diagnosis of diabetes was determined by ICD-9 diagnosis code or use of drugs to treat diabetes. Blood pressure measurements and dates were extracted for use as a time-varying co-variable. Smoking status (never, ever, current or unknown) at the time of each measurement was determined from a variety of data sources: attendance at a smoking cessation clinic, ICD-9 diagnosis codes 305.1X or V15.82, prescriptions for smoking cessation aids, data collected on patients having procedures for the National VA Surgical Quality Improvement Program, and the VA tobacco use clinical reminder field, which was searched for common text strings. All prescription data were collected on the drugs of interest. Drug use was defined as receipt of a VA prescription for a drug in that class during the study. Analysis For the AAA enlargement analyses, all individuals with at least two measurements of an unrepaired AAA at least 6 months apart were included. The first infrarenal aortic diameter of 3·0 cm or larger after 1994 and all subsequent aortic diameter measurements before abdominal aortic surgery or AAA rupture were included until the end of the study in 2010. Multiple measurements made within 30 days for a given patient were replaced with their mean, entered at the date of the last measurement. A correction term derived from the study data was used to make CT measurements comparable with ultrasound measurements. The Akaike and Bayesian information criteria (AIC and BIC respectively) were used to compare goodness-of-fit of several patterns of AAA enlargement, including linear, exponential, logistic and Gompertz models6. Individual drugs within a class were converted to an equivalent dose of a class archetype. For each interval, the proportion of time with medication and the total number of milligrams taken were calculated, for use as co-variables in the propensity models (Appendix S1, supporting information). For each factor examined, a propensity score was constructed for the likelihood of having the factor given the other co-variables. The other co-variables included demographics, diagnoses, smoking status, drug use and dose, and healthcare utilization (Appendix S1, supporting information). A generalized linear mixed model with logistic link was used to construct the longitudinal propensity model. Nearest-neighbour matching based on propensity score was used for each factor to create matched data sets for the final mixed models. The matched models were also adjusted for baseline AAA diameter and any co-variables that were not balanced (Appendix S2, supporting information). All the matchings were done using MatchIt7 and the mixed modelling was performed using Zelig (http://GKing.harvard.edu/zelig)8. Missing values were imputed using a Bayesian bootstrap method (Appendix S2, supporting information). To plan the appropriate sample size for the study, an absolute reduction of 1·2 mm per year was used as the minimum important difference in AAA expansion rate that would justify drug therapy, based on a survey of physicians9. Using data from the Aneurysm Detection and Management (ADAM) trial10, the standard deviation of change in AAA measurement from first to last measurement was estimated to be 5·8 mm. To have 90 per cent power to detect a mean change in AAA diameter of 1·2 mm per year from each drug with α = 0·05, discounted for presumably unbalanced propensity strata, at least 2000 individuals were needed. Data were therefore collected from three VA medical centres. Results From 180 404 abdominal imaging reports on 83 226 patients between 1995 and 2009 at the three VA medical centres, the search identified 19 597 patients with at least one study meeting the search criteria. All 52 962 abdominal radiology studies on these patients were obtained for manual extraction. The two reviewers disagreed or had questions in 6·0 per cent of instances, and these were resolved by a third reviewer. This extraction process identified 23 432 studies with an AAA of 3 cm or more. To assess the sensitivity of the search, 4638 of the abdominal imaging studies that did not meet the search criteria were also reviewed (3·6 per cent). This identified 19 studies with an AAA diameter of at least 3 cm, implying that the search missed 522 AAA measurements, and indicating that the sensitivity of the search to identify measurements of AAA diameter 3 cm or greater was 97·8 per cent. As finding one AAA measurement resulted in inclusion of all other imaging on that patient, it is unlikely that a patient with more than one AAA measurement was missed. Some 5362 individuals with an unrepaired AAA were identified. This group was used to examine the agreement of reported measurements. There were 81 instances of an individual having two ultrasound measurements of aortic diameter within 30 days (Fig. 2a). The mean(s.d.) difference between the two measurements was 3·7(7·0) mm. Of 524 instances of two CT measurements within 30 days, the mean difference was 3·0(4·5) mm (Fig. 2b). There were 887 instances of an individual having ultrasonography and CT within 30 days, with a mean difference between the two measurements of 4·1(6·1) mm (Fig. 2c). The mean group difference was 0·9 mm larger for CT than for ultrasound imaging. Limiting these comparisons to measurements after 1 January 2000 did not result in substantial differences (Figs S1–S3, supporting information). The CT measurements show marked rounding to the half-centimetre (P < 0·001) (Fig. 3), but this rounding became significantly less pronounced after 2000 (P < 0·001) (Figs S4 and S5, supporting information). Fig. 2 Open in new tabDownload slide Differences between two abdominal aortic aneurysm (AAA) measurements done within 30 days, by type of test. a Two ultrasound measurements in 81 individuals; mean(s.d.) difference 3·7(7·0) mm. b Two CT measurements in 524 individuals; mean(s.d.) difference 3·0(4·5) mm. c CT versus ultrasound measurements in 887 individuals; mean difference (CT – ultrasonography) 0·9 mm; mean(s.d.) absolute difference 4·1(6·1) mm Fig. 3 Open in new tabDownload slide Frequency distribution of abdominal aortic aneurysm (AAA) diameter among a 1021 CT measurements and b 382 ultrasound measurements. Extreme values are not shown for ease of display There were 2428 patients with an AAA who had at least two measurements over at least 6 months; this group was used to conduct the analyses of AAA enlargement. These individuals had a mean follow-up of 3·4 years (maximum 14·4 years) with median follow-up of 2·7 (i.q.r. 1·4–4·7) years, and a total of 11 879 measurements (mean 4·9, median 4·0 per patient). Of these 2428 individuals, 484 had all of their measurements by ultrasound imaging, 539 had all of their measurements by CT, and the remainder had measurements by both examinations. Their mean age was 71·2 years, and 99·0 per cent were men (Table 1). Some 26·3 per cent were identified as current smokers at study entry, 3·7 per cent had never smoked, and smoking status was unknown for 23·5 per cent. Statins, beta-blockers and ACE inhibitors were each received by about 40 per cent of the individuals at some time during the study, whereas ARBs were received by only 4·7 per cent. AAA rupture was identified in 57 patients in the follow-up group, of whom 30 had an aortic diameter of 5·5 cm or larger a week or more before rupture. Table 1 Characteristics of 2428 individuals with an abdominal aortic aneurysm of 3 cm or more followed for at least 6 months . No. of individuals* (n = 2428) . Age (years)† 71·2(7·9) Sex ratio (M : F) 2403 : 25 Smoking status Current smoker 639 (26·3) Never smoked 91 (3·8) Unknown 570 (23·5) No. of AAA measurements† 4·9(3·6) AAA diameter at first measurement (cm)† 4·0(0·8) AAA diameter at last measurement (cm)† 4·6(1·2) Co-morbidities Coronary artery disease 894 (36·8) Cerebrovascular disease 290 (11·9) Peripheral vascular disease 352 (14·5) Hypertension 1368 (56·3) Diabetes 263 (10·8) Chronic obstructive pulmonary disease 671 (27·6) Depression 323 (13·3) Renal failure 157 (6·5) Any drug use Statins 1013 (41·7) Beta-blockers 992 (40·9) Angiotensin-converting enzyme inhibitors 994 (40·9) Angiotensin II receptor blockers 115 (4·7) . No. of individuals* (n = 2428) . Age (years)† 71·2(7·9) Sex ratio (M : F) 2403 : 25 Smoking status Current smoker 639 (26·3) Never smoked 91 (3·8) Unknown 570 (23·5) No. of AAA measurements† 4·9(3·6) AAA diameter at first measurement (cm)† 4·0(0·8) AAA diameter at last measurement (cm)† 4·6(1·2) Co-morbidities Coronary artery disease 894 (36·8) Cerebrovascular disease 290 (11·9) Peripheral vascular disease 352 (14·5) Hypertension 1368 (56·3) Diabetes 263 (10·8) Chronic obstructive pulmonary disease 671 (27·6) Depression 323 (13·3) Renal failure 157 (6·5) Any drug use Statins 1013 (41·7) Beta-blockers 992 (40·9) Angiotensin-converting enzyme inhibitors 994 (40·9) Angiotensin II receptor blockers 115 (4·7) * With percentages in parentheses unless indicated otherwise; † values are mean(s.d.). Data are for characteristics at baseline, except drug use which is at any time during the study. AAA, abdominal aortic aneurysm. Open in new tab Table 1 Characteristics of 2428 individuals with an abdominal aortic aneurysm of 3 cm or more followed for at least 6 months . No. of individuals* (n = 2428) . Age (years)† 71·2(7·9) Sex ratio (M : F) 2403 : 25 Smoking status Current smoker 639 (26·3) Never smoked 91 (3·8) Unknown 570 (23·5) No. of AAA measurements† 4·9(3·6) AAA diameter at first measurement (cm)† 4·0(0·8) AAA diameter at last measurement (cm)† 4·6(1·2) Co-morbidities Coronary artery disease 894 (36·8) Cerebrovascular disease 290 (11·9) Peripheral vascular disease 352 (14·5) Hypertension 1368 (56·3) Diabetes 263 (10·8) Chronic obstructive pulmonary disease 671 (27·6) Depression 323 (13·3) Renal failure 157 (6·5) Any drug use Statins 1013 (41·7) Beta-blockers 992 (40·9) Angiotensin-converting enzyme inhibitors 994 (40·9) Angiotensin II receptor blockers 115 (4·7) . No. of individuals* (n = 2428) . Age (years)† 71·2(7·9) Sex ratio (M : F) 2403 : 25 Smoking status Current smoker 639 (26·3) Never smoked 91 (3·8) Unknown 570 (23·5) No. of AAA measurements† 4·9(3·6) AAA diameter at first measurement (cm)† 4·0(0·8) AAA diameter at last measurement (cm)† 4·6(1·2) Co-morbidities Coronary artery disease 894 (36·8) Cerebrovascular disease 290 (11·9) Peripheral vascular disease 352 (14·5) Hypertension 1368 (56·3) Diabetes 263 (10·8) Chronic obstructive pulmonary disease 671 (27·6) Depression 323 (13·3) Renal failure 157 (6·5) Any drug use Statins 1013 (41·7) Beta-blockers 992 (40·9) Angiotensin-converting enzyme inhibitors 994 (40·9) Angiotensin II receptor blockers 115 (4·7) * With percentages in parentheses unless indicated otherwise; † values are mean(s.d.). Data are for characteristics at baseline, except drug use which is at any time during the study. AAA, abdominal aortic aneurysm. Open in new tab The AIC and BIC for the various models of AAA enlargement were: linear 6767·4 and 6802·2, exponential 6343·9 and 6378·6, logistic 6293·0 and 6333·5, and Gompertz 7944·3 and 7984·9, with lower numbers signifying better fit. These data most closely fitted the logistic model, but the linear model produced sufficiently similar results. Because of its simplicity, it was used for subsequent analyses. Using the linear model, the mean AAA enlargement rate for the follow-up group was 2·0 (95 per cent c.i. 1·9 to 2·3) mm per year. Table 2 shows the results of the propensity analysis. Any use of each of the four classes of drug (statins, beta-blockers, ACE inhibitors, ARBs) was associated with a small change in AAA enlargement of 0·5 mm per year or less that was not statistically significant. Current smoking also had no statistically significant association with AAA enlargement, although chronic obstructive pulmonary disease was associated with more rapid enlargement of 0·5 mm per year (P = 0·050). Diabetes was associated with a decrease in AAA enlargement of 1·2 mm per year (P = 0·008). Because of changes in practice over time, the analysis was repeated, limited to measurements since 1 January 2000, with no important differences. The association of current smoking with AAA enlargement remained non-significant in models that placed all those with unknown smoking status into the current smoking group, or alternatively into the not currently smoking group. Table 2 Propensity analysis of factors potentially affecting abdominal aortic aneurysm enlargement rate . No. of matched pairs . Enlargement rate without factor (mm/year) . Change in rate with factor (mm/year) . P . Age > 72 years* 459 2·1 0·0 (−0·7, 0·7) 0·934 After 2000 310 1·9 −1·1 (−2·0, −0·2) 0·016 Current smoking 639 2·1 – 0·2 (−0·6, 0·2) 0·358 After 2000 461 2·0 −0·2 (−0·7, 0·2) 0·338 Coronary artery disease 894 2·1 −0·3 (−0·8, 0·2) 0·198 After 2000 561 2·1 −0·5 (−1·2, 0·2) 0·154 Chronic obstructive pulmonary disease 671 1·7 0·5 (0·0, 1·0) 0·050 After 2000 430 1·7 0·7 (−0·1, 1·6) 0·109 Diabetes 263 2·4 −1·2 (−2·0, −0·3) 0·008 After 2000 185 2·1 −1·1 (−2·0, −0·2) 0·020 Statins 1013 2·1 0·1 (−0·2, 0·5) 0·510 After 2000 538 1·8 0·4 (0·3, 1·1) 0·290 Beta-blockers 828 2·1 −0·5 (−1·2, 0·3) 0·242 After 2000 605 2·0 0·0 (−0·6, 0·6) 0·997 Angiotensin-converting enzyme inhibitors 994 2·0 0·1 (−0·3, 0·4) 0·656 After 2000 669 2·0 0·1 (−0·4, 0·7) 0·613 Angiotensin II receptor blockers 115 1·8 −0·2 (−1·3, 0·9) 0·608 After 2000 107 1·8 0·1 (−1·3, 1·1) 0·823 . No. of matched pairs . Enlargement rate without factor (mm/year) . Change in rate with factor (mm/year) . P . Age > 72 years* 459 2·1 0·0 (−0·7, 0·7) 0·934 After 2000 310 1·9 −1·1 (−2·0, −0·2) 0·016 Current smoking 639 2·1 – 0·2 (−0·6, 0·2) 0·358 After 2000 461 2·0 −0·2 (−0·7, 0·2) 0·338 Coronary artery disease 894 2·1 −0·3 (−0·8, 0·2) 0·198 After 2000 561 2·1 −0·5 (−1·2, 0·2) 0·154 Chronic obstructive pulmonary disease 671 1·7 0·5 (0·0, 1·0) 0·050 After 2000 430 1·7 0·7 (−0·1, 1·6) 0·109 Diabetes 263 2·4 −1·2 (−2·0, −0·3) 0·008 After 2000 185 2·1 −1·1 (−2·0, −0·2) 0·020 Statins 1013 2·1 0·1 (−0·2, 0·5) 0·510 After 2000 538 1·8 0·4 (0·3, 1·1) 0·290 Beta-blockers 828 2·1 −0·5 (−1·2, 0·3) 0·242 After 2000 605 2·0 0·0 (−0·6, 0·6) 0·997 Angiotensin-converting enzyme inhibitors 994 2·0 0·1 (−0·3, 0·4) 0·656 After 2000 669 2·0 0·1 (−0·4, 0·7) 0·613 Angiotensin II receptor blockers 115 1·8 −0·2 (−1·3, 0·9) 0·608 After 2000 107 1·8 0·1 (−1·3, 1·1) 0·823 Values in parentheses are 95 per cent c.i. * Age at first measurement; other factors could be present at any time during follow-up. After 2000 refers to measurements after 1 January 2000. Open in new tab Table 2 Propensity analysis of factors potentially affecting abdominal aortic aneurysm enlargement rate . No. of matched pairs . Enlargement rate without factor (mm/year) . Change in rate with factor (mm/year) . P . Age > 72 years* 459 2·1 0·0 (−0·7, 0·7) 0·934 After 2000 310 1·9 −1·1 (−2·0, −0·2) 0·016 Current smoking 639 2·1 – 0·2 (−0·6, 0·2) 0·358 After 2000 461 2·0 −0·2 (−0·7, 0·2) 0·338 Coronary artery disease 894 2·1 −0·3 (−0·8, 0·2) 0·198 After 2000 561 2·1 −0·5 (−1·2, 0·2) 0·154 Chronic obstructive pulmonary disease 671 1·7 0·5 (0·0, 1·0) 0·050 After 2000 430 1·7 0·7 (−0·1, 1·6) 0·109 Diabetes 263 2·4 −1·2 (−2·0, −0·3) 0·008 After 2000 185 2·1 −1·1 (−2·0, −0·2) 0·020 Statins 1013 2·1 0·1 (−0·2, 0·5) 0·510 After 2000 538 1·8 0·4 (0·3, 1·1) 0·290 Beta-blockers 828 2·1 −0·5 (−1·2, 0·3) 0·242 After 2000 605 2·0 0·0 (−0·6, 0·6) 0·997 Angiotensin-converting enzyme inhibitors 994 2·0 0·1 (−0·3, 0·4) 0·656 After 2000 669 2·0 0·1 (−0·4, 0·7) 0·613 Angiotensin II receptor blockers 115 1·8 −0·2 (−1·3, 0·9) 0·608 After 2000 107 1·8 0·1 (−1·3, 1·1) 0·823 . No. of matched pairs . Enlargement rate without factor (mm/year) . Change in rate with factor (mm/year) . P . Age > 72 years* 459 2·1 0·0 (−0·7, 0·7) 0·934 After 2000 310 1·9 −1·1 (−2·0, −0·2) 0·016 Current smoking 639 2·1 – 0·2 (−0·6, 0·2) 0·358 After 2000 461 2·0 −0·2 (−0·7, 0·2) 0·338 Coronary artery disease 894 2·1 −0·3 (−0·8, 0·2) 0·198 After 2000 561 2·1 −0·5 (−1·2, 0·2) 0·154 Chronic obstructive pulmonary disease 671 1·7 0·5 (0·0, 1·0) 0·050 After 2000 430 1·7 0·7 (−0·1, 1·6) 0·109 Diabetes 263 2·4 −1·2 (−2·0, −0·3) 0·008 After 2000 185 2·1 −1·1 (−2·0, −0·2) 0·020 Statins 1013 2·1 0·1 (−0·2, 0·5) 0·510 After 2000 538 1·8 0·4 (0·3, 1·1) 0·290 Beta-blockers 828 2·1 −0·5 (−1·2, 0·3) 0·242 After 2000 605 2·0 0·0 (−0·6, 0·6) 0·997 Angiotensin-converting enzyme inhibitors 994 2·0 0·1 (−0·3, 0·4) 0·656 After 2000 669 2·0 0·1 (−0·4, 0·7) 0·613 Angiotensin II receptor blockers 115 1·8 −0·2 (−1·3, 0·9) 0·608 After 2000 107 1·8 0·1 (−1·3, 1·1) 0·823 Values in parentheses are 95 per cent c.i. * Age at first measurement; other factors could be present at any time during follow-up. After 2000 refers to measurements after 1 January 2000. Open in new tab Possible associations of any use of statins, beta-blockers and ACE inhibitors up to the time of each measurement with AAA enlargement from the first measurement to the measurement in question, and from the immediately preceding measurement to the measurement in question, were also examined, but none was found to be statistically significant. Discussion In this study of 2428 individuals with an AAA followed up for a mean of 3·4 years, there was no reduction in AAA enlargement rate associated with use of statins, beta-blockers, ACE inhibitors or ARBs. The 95 per cent confidence intervals exclude the prespecified minimum important difference of 1·2 mm per year in AAA enlargement rate for all drugs, except the less frequently used ARBs. These findings are similar to those reported by an adjusted meta-analysis11 of previous studies of AAA enlargement and to a recent analysis of ADAM trial data12, both of which reported no significant drug associations. Of the drugs studied, randomized trials are available only for the beta-blocker propranolol, and the three trials13–15 that have been reported all had high rates of drug discontinuation and found non-significant reductions in AAA enlargement. The AAA enlargement rate of 2·0 mm per year observed in this study was similar to the rate of 2·2 mm per year in the meta-analysis11; the latter value falls within the 95 per cent c.i. of the present study. Unlike the meta-analysis and the ADAM study, the present study did not find a significant association between current smoking and AAA enlargement. The only factors associated with AAA enlargement in this study were chronic obstructive pulmonary disease and diabetes. The reduction in AAA enlargement rate by diabetes was 1·2 mm per year, compared with 1·1 mm per year in the ADAM data12 and 0·5 mm per year in the meta-analysis11. The negative association between diabetes and AAA was first described in the ADAM screening study and subsequently confirmed by a variety of studies of AAA diagnosed through screening or clinical events16. Although the negative association between diabetes and AAA now seems convincing, the mechanism remains unexplained. It is nevertheless strong evidence that AAA is not a manifestation of atherosclerosis16, which may help explain the lack of associations with AAA progression by drugs that are effective against atherosclerosis. In this study, CT measurements were a mean of 0·9 mm larger than ultrasound measurements done within 30 days. This is similar to the difference observed in the ADAM study17, in which local CT measurements were an average of 1·2 mm larger than ultrasound measurements done within 30 days17. The slightly larger readings by CT may reflect the ability of CT to identify the maximum cross-sectional diameter in any plane, whereas ultrasound imaging usually measures only the anterior–posterior diameter. The measurement variability observed in this study for CT compared with ultrasound is similar to the ADAM study comparison of ultrasound with central CT reading17. The variability between unblinded measurements of two separate CT images in this study was similar to that seen for the local versus blinded central readings of the same CT image in the ADAM study. The rounding to the half-centimetre seen in the CT measurements, termed terminal digit preference, was described previously in the ADAM study17, and in the present study was seen to improve from 2000, possibly influenced by the ADAM study report. This study has several potential limitations. First, it relies on observational rather than randomized data, which could affect the observed drug associations. The findings of no drug associations could represent residual confounding by drug indication (such as occlusive vascular disease and hypertension), although every effort was made to account for this in the analysis. Had there been evidence of reduced AAA enlargement associated with one or more of the study drugs, development of a randomized trial was planned. Second, use of VA pharmacy data may have failed to reflect the medications taken by these individuals, owing to either non-adherence to prescribed medications or use of prescriptions from sources other than VA. However, available data do not support these concerns. Wannemacher and colleagues18 calculated adherence (defined as mean percentage of time with drug) for the year 1998 in a VA region to be 93 per cent for beta-blockers and 95 per cent for ACE inhibitors. Piette and Heisler19 found that cost-related medication non-adherence was lower among VA patients than among other groups. The use of VA pharmacy refill data as a measure of adherence has been validated by comparison with clinical drug effects20. A 1997 systematic review21 noted that 98–100 per cent of VA pharmacy users reported no non-VA pharmacy use. Third, drug indications may have changed during the study, such as increased use of the study drugs for coronary disease and heart failure in the 1990s. However, limiting the analyses to data collected after 2000 did not result in substantial changes to the findings. Fourth, routine clinically reported measurements of AAA diameter were relied on, rather than measurements done for research, which resulted in suboptimal accuracy as evidenced by the substantial terminal digit preference noted above. The large numbers of patients and long duration of follow-up should minimize the impact of minor measurement inaccuracy. Finally, administrative data were used for identification of co-variables, which entails some inaccuracy, particularly in the ability to identify baseline smoking status in older records. This may partially explain why no association between current smoking and AAA enlargement was detected, although models that placed all those with unknown smoking status into the current smoking group or into the not currently smoking group obtained the same result. Another possible reason for no association between current smoking and AAA enlargement was that there were few never smokers, so comparison was largely with former smokers, which may result in less of a difference. Consistent with previous studies, diabetes was associated with slower AAA enlargement, whereas use of statins, beta-blockers, ACE inhibitors or ARBs was not. Based on these findings, the expense of a large randomized trial of these drugs to reduce AAA enlargement is difficult to justify. Acknowledgements Research data were extracted by: W. Cumming, P. Larson, E. Guter, C. Lederle, L. Lederle, M. Prior, E. Duane, J. Bennett, E. Davis, S. Murphy, S. Monono, J. Barnes, C. Robert and T. Leighton. Funding was provided by the US Department of Veterans Affairs, Office of Research and Development (Clinical Science award EPID-004-06F). Disclosure: The authors declare no conflict of interest. References 1 Takagi H , Niwa M, Mizuno Y, Goto SN, Umemoto T; All-Literature Investigation of Cardiovascular Evidence Group . The Last Judgment upon abdominal aortic aneurysm screening . Int J Cardiol 2013 ; 167 : 2331 – 2332 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Preventive US Services Task Force. Screening for abdominal aortic aneurysm: recommendation statement . Ann Intern Med 2005 ; 142 : 198 – 202 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 3 Filardo G , Powell JT, Martinez MA, Ballard DJ. Surgery for small asymptomatic abdominal aortic aneurysms . Cochrane Database Syst Rev 2012 ; ( 3 ) CD001835 . 4 Lederle FA . Abdominal aortic aneurysm: still no pill . Ann Intern Med 2013 ; 159 : 852 – 853 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Golledge J , Norman PE. Current status of medical management for abdominal aortic aneurysm . Atherosclerosis 2011 ; 217 : 57 – 63 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Burnham KP , Anderson DR. Model Selection and Multimodel Inference: a Practical Information–Theoretic Approach (2nd edn). Springer : New York , 2002 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 7 Ho DE , Imai K, King G, Stuart EA. MatchIt: nonparametric preprocessing for parametric causal inference . J Stat Software 2011 ; 42 : 1 – 28 . Google Scholar Crossref Search ADS WorldCat 8 Imai K , King G, Lau O. Toward a common framework for statistical analysis and development . Journal of Computational and Graphical Statistics 2008 ; 17 : 892 – 913 . Google Scholar Crossref Search ADS WorldCat 9 van Walraven C , Mahon JL, Moher D, Bohm C, Laupacis A. Surveying physicians to determine the minimal important difference: implications for sample-size calculation . J Clin Epidemiol 1999 ; 52 : 717 – 723 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Lederle FA , Wilson SE, Johnson GR, Reinke DB, Littooy FN, Acher CW et al. ; Aneurysm Detection and Management (ADAM) Veterans Affairs Cooperative Study Group . Immediate repair compared with surveillance of small abdominal aortic aneurysms . N Engl J Med 2002 ; 346 : 1437 – 1444 . Google Scholar Crossref Search ADS PubMed WorldCat 11 Sweeting MJ , Thompson SG, Brown LC, Powell JT; RESCAN collaborators . Meta-analysis of individual patient data to examine factors affecting growth and rupture of small abdominal aortic aneurysms . Br J Surg 2012 ; 99 : 655 – 665 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Bhak RH , Wininger M, Johnson GR, Lederle FA, Messina LM, Ballard DJ et al. ; Aneurysm Detection and Management (ADAM) Study Group . Factors associated with small abdominal aortic aneurysm expansion rate . JAMA Surg 2015 ; 150 : 44 – 50 . Google Scholar Crossref Search ADS PubMed WorldCat 13 Lindholt JS , Henneberg EW, Juul S, Fasting H. Impaired results of a randomised double blinded clinical trial of propranolol versus placebo on the expansion rate of small abdominal aortic aneurysms . Int Angiol 1999 ; 18 : 52 – 57 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 14 Propranolol Aneurysm Trial Investigators . Propranolol for small abdominal aortic aneurysms: results of a randomized trial . J Vasc Surg 2002 ; 35 : 72 – 79 . Crossref Search ADS PubMed WorldCat 15 Wilmink ABM , Hubbard CSFF, Day NE, Quick CRG. Effect of propranolol on the expansion of abdominal aortic aneurysms: a randomized study . Br J Surg 2000 ; 87 : 499 . Google Scholar Crossref Search ADS WorldCat 16 Lederle FA . The strange relationship between diabetes and abdominal aortic aneurysm . Eur J Vasc Endovasc Surg 2012 ; 43 : 254 – 256 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Lederle FA , Wilson SE, Johnson GR, Reinke DB, Littooy FN, Acher C et al. Variability in measurement of abdominal aortic aneurysms. Abdominal Aortic Aneurysm Detection and Management Veterans Administration Cooperative Study Group . J Vasc Surg 1995 ; 21 : 945 – 952 . Google Scholar Crossref Search ADS PubMed WorldCat 18 Wannemacher AJ , Schepers GP, Townsend KA. Antihypertensive medication compliance in a Veterans Affairs Healthcare System . Ann Pharmacother 2002 ; 36 : 986 – 991 . Google Scholar Crossref Search ADS PubMed WorldCat 19 Piette JD , Heisler M. Problems due to medication costs among VA and non-VA patients with chronic illnesses . Am J Manag Care 2004 ; 10 : 861 – 868 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 20 Steiner JF , Koepsell TD, Fihn SD, Inui TS. A general method of compliance assessment using centralized pharmacy records. Description and validation . Med Care 1988 ; 26 : 814 – 823 . Google Scholar Crossref Search ADS PubMed WorldCat 21 Steiner JF , Prochazka AV. The assessment of refill compliance using pharmacy records: methods, validity, and applications . J Clin Epidemiol 1997 ; 50 : 105 – 116 . Google Scholar Crossref Search ADS PubMed WorldCat Published 2015. This article is a U.S. Government work and is in the public domain in the USA. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) Published 2015. This article is a U.S. Government work and is in the public domain in the USA.

Findlay, J M; Bradley, K M; Maile, E J; Braden, B; Maw, J; Phillips-Hughes, J; Gillies, R S; Maynard, N D; Middleton, M R

doi: 10.1002/bjs.9905pmid: 26458070

Abstract Background Following CT, guidelines for staging oesophageal and gastro-oesophageal junction (GOJ) cancer recommend endoscopic ultrasonography (EUS), PET–CT and laparoscopy for T3–T4 GOJ tumours. These recommendations are based on generic utilities, but it is unclear whether the test risk outweighs the potential benefit for some patients. This study sought to quantify investigation risks, benefits and utilities, in order to develop pragmatic, personalized staging recommendations. Methods All patients with a histological diagnosis of oesophageal or GOJ cancer staged between May 2006 and July 2013 comprised a development set; those staged from July 2013 to July 2014 formed the prospective validation set. Probability thresholds of altering management were calculated and predictive factors identified. Algorithms and models (decision tree analysis, logistic regression, artificial neural networks) were validated internally and independently. Results Some 953 patients were staged following CT, by [18F]fluorodeoxyglucose PET–CT (918), EUS (798) and laparoscopy (458). Of these patients, 829 comprised the development set (800 PET–CT, 698 EUS, 397 laparoscopy) and 124 the validation set (118 PET–CT, 100 EUS, 61 laparoscopy). EUS utility in the 71·8 per cent of patients with T2–T4a disease on CT was minimal (0·4 per cent), its risk exceeding benefit. EUS was moderately accurate for pT1 N0 disease. A number of factors predicted metastases on PET–CT and laparoscopy, although none could inform an algorithm. PET–CT altered management in 23·0 per cent, and laparoscopy in 7·1 per cent, including those with T2 and distal oesophageal tumours. Conclusion Although EUS provided additional information on T and N category, its risk outweighed potential benefit in patients with T2–T4a disease on CT. Laparoscopy seemed justified for distal oesophageal tumours of T2 or greater. Introduction There is consensus internationally regarding initial staging algorithms for oesophageal cancer. For patients in whom potentially curative treatment may be both feasible and appropriate, the Association of Upper Gastrointestinal Surgeons of Great Britain and Ireland (in conjunction with the British Society of Gastroenterology and British Association of Surgical Oncology), the USA Society of Thoracic Surgeons and National Comprehensive Cancer Network, and the European Society of Medical Oncology all recommend CT and endoscopic ultrasonography (EUS) for all patients, followed by PET–CT for T1b–T4 disease1–4. In the USA and Europe, laparoscopy is reserved mainly for T3–T4 disease involving the gastro-oesophageal junction (GOJ), although in the UK it is also recommended for distal oesophageal tumours. These recommendations reflect the complementary roles of each modality: EUS is most accurate for assignment of T and N category5, PET–CT for distant metastases6,7 and laparoscopy for peritoneal disease8. Use of all modalities, however, consumes resources, delays treatment, and may confer small but clinically significant risks, while not necessarily altering management decisions. The primary utility of staging modalities is to divide patients into optimal management groups, identifying patients with early (T1a N0 intramucosal cancer for endoscopic resection (ER); T1b N0 for resection of submucosal cancer without neoadjuvant therapy), locally advanced (T1b N1 to T4a; typically resection with neoadjuvant therapy), and unresectable T4b and metastatic disease. Use of all modalities provides optimal staging precision and useful prognostic information (for example, distinguishing between T2 N1 and T3 N3 disease), but typically without altering management. It is not known whether patient subgroups exist for which these primary test utilities differ. Similarly, decision analytic metrics are not available to guide staging (whether this utility justifies test risk). The primary aim of the present study was to calculate the net benefits and risks of EUS, PET–CT and laparoscopy, their primary utilities (probability of altering management) and probability thresholds (Pt; at which test benefit equals risk), using decision theory in a development data set. A secondary aim was to determine whether clinical, radiological and histopathological factors could be identified that were related to these endpoints, in order to generate predictive models to identify patient subgroups for selective staging. The final aim was then to refine existing staging algorithms on the basis of optimal pragmatism, maximal efficiency and minimal patient risk, evaluated using a validation data set. Methods Consecutive patients with oesophageal/GOJ cancer staged beyond CT in a UK centre were identified from four parallel databases. These patients represented those either diagnosed at the centre, or referred to the central oesophagogastric multidisciplinary team from three regional referring National Health Service (NHS) hospitals. Patients staged between May 2006 and July 2013 comprised a development set, and those staged from July 2013 to July 2014 an independent validation set. All investigations were reported and reviewed by a specialist multidisciplinary team, in accordance with the contemporary American Joint Committee on Cancer TNM staging manuals (6th9 or 7th10 edition)8. Patients without unequivocal metastases on CT were routinely staged sequentially using [18F]fluorodeoxyglucose (FDG) PET–CT, EUS and laparoscopy, with oesophagogastroduodenoscopy (OGD) for GOJ tumours and distal oesophageal tumours extending below the diaphragm. Investigations were performed out of sequence in a small number of patients to maximize expediency. Neoadjuvant chemotherapy was considered for disease beyond T1 N0; ER was used from 2008 for possible T1a tumours. The study was registered with the Oxford University Hospitals NHS Trust clinical governance department as a service provision audit (number 2516), and as such was exempt from ethical approval. CT was performed locally (367) or regionally (586) at three referring hospital trusts using a standard protocol8 (11 scans were performed elsewhere). [18F]FDG PET–CT was carried out using a General Electric Discovery STE 16-slice instrument (General Electric Healthcare, Milwaukee, Wisconsin, USA) (scanner 1; 60 min after 400 MBq [18F]FDG) before November 2009, and thereafter a Discovery 690 64-slice system (scanner 2; 90 min after 4 MBq/kg [18F]FDG), without intravenous contrast using standard iterative reconstruction, and reported independently by two dedicated PET–CT radiologists. EUS and ER were performed locally using 5–10-MHz radial or linear echoscopes (Hitachi, Wellingborough, UK), by one of three consultants and reported using the contemporary TNM edition and British Society of Gastroenterology guidelines11. Miniprobe EUS was performed using 20-Hz high-frequency catheter probes (Olympus UM-2R; Olympus, Southend-on-Sea, UK). ER was achieved using multiband ligation mucosectomy (DuetteTM; Cook Medical, Limerick, Ireland). Nodal stage was determined by ultrasonography, with fine-needle aspiration (FNA) if the result was considered likely to alter management. Potential bias was assessed for referring centre and PET scanner, but not EUS operator, as possible T1 N0 tumours were referred selectively for EUS with or without ER. At laparoscopy, visual examination of the peritoneal cavity was carried out without opening the lesser sac. Lavage cytology was not undertaken routinely. Data collected were: patient age at diagnosis, sex, endoscopic findings, radiological findings (CT, EUS, PET–CT TNM staging; PET–CT maximum standardized uptake value (SUVmax), length, presence of avid nodes; additional PET–CT findings; subsequent investigations), pathological (pretreatment cell type; grade10; pathological stage, grade) and multidisciplinary team management decisions. Two TNM editions were used within this period; N status and overall stage were classified using the sixth edition as insufficient data were available for conversion to the seventh edition. Management was uniform for both (for example, patients with M1a nodes based on the 6th edition were considered for radical treatment). The primary utility of EUS was considered to be that altering management by identifying or refuting T1 N0 or T4b disease, or unsuspected metastases (distant organ, or nodal outside a standard resection field). For PET–CT, this was by identifying and confirming metastases, or unsuspected lesions altering management following appropriate specific investigation. The latter included treatment of a synchronous cancer or detection of high-risk lesions such as colonic polyps requiring surveillance (for example 3–4 adenomatous polyps, or those larger than 1 cm12). For laparoscopy, this was the identification of unresectable disease, either metastatic or T4b, with histopathological confirmation. Statistical analysis Analysis was performed using R version 3.0.213. P < 0·050 was corrected for multiple comparisons by the Bonferroni method14. Multivariable logistic regression included all variables following exclusion of perfect separators. Continuous variables were assessed (kernel density plots) and transformed (age2; logSUVmax). Decision analytic measures and cost analysis Pt values were calculated to inform decision-making and models15 (Appendix S1, supporting information16–31). Decision curve analysis (DCA) was performed to calculate the effect of varying Pt (0·496–2·719 per cent) reflecting EUS perforation risk (0·02–1·00 per cent)32. Model and algorithm development, validation and performance Modelling was undertaken to corroborate proposed algorithms for selective EUS, PET–CT and laparoscopy, and to determine whether unidentified interactions could refine them. Three modelling techniques were used: logistic regression models (LRMs; backwards stepwise binary logistic), decision tree analysis (DTA; recursive partitioning) and artificial neural networks (ANNs; feed-forward back-propagation multilayer perceptron; Appendix S1, supporting information). Internal validation was carried out using bootstrapping/random forests (Appendix S1, supporting information); temporal validation was undertaken using the temporal validation data set. Performance was quantified using calibration (observed : expected ratio), accuracy (Brier and κ scores), discrimination (area under the receiver operating characteristic (ROC) curve), and utility (sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, and Peirce's net benefit33). Sensitivities and specificities were compared using McNemar's t test (DTComPair version 1.0.331). Results A total of 953 consecutive patients were staged following CT by PET–CT (918), EUS (798) and laparoscopy (458). Of these, 829 comprised the development data set (800, 698 and 397 respectively) and 124 the validation set (118, 100 and 61) (Table 1). Table 1 Patient and tumour characteristics . Development set (n = 829) . Validation set (n = 124) . Age (years)*§ 66 (13, 29–88) 68 (13, 40–80) Sex ratio (F : M) 204 : 625 35 : 89 Cell type Adenocarcinoma 639 (77·1) 104 (83·9) Squamous cell carcinoma 165 (19·9) 18 (14·5) Adenosquamous carcinoma 7 (0·8) 2 (1·6) Neuroendocrine carcinoma 3 (0·4) 0 (0) Small cell carcinoma 3 (0·4) 0 (0) Undifferentiated 12 (1·4) 0 (0) Grade Well 71 (8·6) 5 (4·0) Moderate 347 (41·9) 58 (46·8) Poor 389 (46·9) 60 (48·4) Undifferentiated 22 (2·7) 1 (0·8) Tumour site Proximal 23 (2·8) 1 (0·8) Mid 67 (8·1) 12 (9·7) Distal 237 (28·6) 42 (33·9) Siewert 1 197 (23·8) 38 (30·6) Siewert 2 195 (23·5) 16 (12·9) Siewert 3 105 (12·7) 14 (11·3) Multifocal 5 (0·6) 1 (0·8) Pretreatment stage (TNM 6th edition) I 52 (6·3) 11 (8·9) IIA 157 (18·9) 21 (16·9) IIB 57 (6·9) 15 (12·1) III 373 (45·0) 55 (44·4) IV 190 (22·9) 22 (17·7) Impassable at OGD No 699 (84·3) 107 (86·3) Yes 130 (15·7) 17 (13·7) Management decisions Palliative (performance status) 143 (17·2) 12 (9·7) Palliative (incurable) 185 (22·3) 22 (17·7) Endoscopic therapy 26 (3·2) 11 (8·9) Surgery 58 (7·0) 5 (4·0) Neoadjuvant therapy + surgery 378 (45·6) 67 (54·0) DCRT 37 (4·5) 7 (5·6) Died during staging 2 (0·2) 0 (0) Endoscopic ultrasonography Not requested 127 (15·3)† 24 (19·4)‡ Performed (passable) 543 (65·5) 75 (60·5) Performed (impassable + miniprobe) 49 (5·9) 1 (0·8) Performed (impassable, no miniprobe) 106 (12·8) 24 (19·4) Abandoned 4 (0·5) 0 (0) . Development set (n = 829) . Validation set (n = 124) . Age (years)*§ 66 (13, 29–88) 68 (13, 40–80) Sex ratio (F : M) 204 : 625 35 : 89 Cell type Adenocarcinoma 639 (77·1) 104 (83·9) Squamous cell carcinoma 165 (19·9) 18 (14·5) Adenosquamous carcinoma 7 (0·8) 2 (1·6) Neuroendocrine carcinoma 3 (0·4) 0 (0) Small cell carcinoma 3 (0·4) 0 (0) Undifferentiated 12 (1·4) 0 (0) Grade Well 71 (8·6) 5 (4·0) Moderate 347 (41·9) 58 (46·8) Poor 389 (46·9) 60 (48·4) Undifferentiated 22 (2·7) 1 (0·8) Tumour site Proximal 23 (2·8) 1 (0·8) Mid 67 (8·1) 12 (9·7) Distal 237 (28·6) 42 (33·9) Siewert 1 197 (23·8) 38 (30·6) Siewert 2 195 (23·5) 16 (12·9) Siewert 3 105 (12·7) 14 (11·3) Multifocal 5 (0·6) 1 (0·8) Pretreatment stage (TNM 6th edition) I 52 (6·3) 11 (8·9) IIA 157 (18·9) 21 (16·9) IIB 57 (6·9) 15 (12·1) III 373 (45·0) 55 (44·4) IV 190 (22·9) 22 (17·7) Impassable at OGD No 699 (84·3) 107 (86·3) Yes 130 (15·7) 17 (13·7) Management decisions Palliative (performance status) 143 (17·2) 12 (9·7) Palliative (incurable) 185 (22·3) 22 (17·7) Endoscopic therapy 26 (3·2) 11 (8·9) Surgery 58 (7·0) 5 (4·0) Neoadjuvant therapy + surgery 378 (45·6) 67 (54·0) DCRT 37 (4·5) 7 (5·6) Died during staging 2 (0·2) 0 (0) Endoscopic ultrasonography Not requested 127 (15·3)† 24 (19·4)‡ Performed (passable) 543 (65·5) 75 (60·5) Performed (impassable + miniprobe) 49 (5·9) 1 (0·8) Performed (impassable, no miniprobe) 106 (12·8) 24 (19·4) Abandoned 4 (0·5) 0 (0) Values in parentheses are percentages unless indicated otherwise; * values are median (i.q.r., range). Endoscopic ultrasonography not requested owing to: † metastases on PET–CT (79) or laparoscopy (5), patient choice (22), prohibitive anatomy (10), decision to proceed to radical chemoradiotherapy irrespective of findings (7), death during staging (2) and unclear reasons (2); ‡ metastases on PET–CT (10) or laparoscopy (1), patient choice/performance status (9), prohibitive anatomy (1) and decision to proceed to definitive chemoradiotherapy (DCRT) (3). OGD, oesophagogastroduodenoscopy. § P < 0·001 (Shapiro–Wilk test). Open in new tab Table 1 Patient and tumour characteristics . Development set (n = 829) . Validation set (n = 124) . Age (years)*§ 66 (13, 29–88) 68 (13, 40–80) Sex ratio (F : M) 204 : 625 35 : 89 Cell type Adenocarcinoma 639 (77·1) 104 (83·9) Squamous cell carcinoma 165 (19·9) 18 (14·5) Adenosquamous carcinoma 7 (0·8) 2 (1·6) Neuroendocrine carcinoma 3 (0·4) 0 (0) Small cell carcinoma 3 (0·4) 0 (0) Undifferentiated 12 (1·4) 0 (0) Grade Well 71 (8·6) 5 (4·0) Moderate 347 (41·9) 58 (46·8) Poor 389 (46·9) 60 (48·4) Undifferentiated 22 (2·7) 1 (0·8) Tumour site Proximal 23 (2·8) 1 (0·8) Mid 67 (8·1) 12 (9·7) Distal 237 (28·6) 42 (33·9) Siewert 1 197 (23·8) 38 (30·6) Siewert 2 195 (23·5) 16 (12·9) Siewert 3 105 (12·7) 14 (11·3) Multifocal 5 (0·6) 1 (0·8) Pretreatment stage (TNM 6th edition) I 52 (6·3) 11 (8·9) IIA 157 (18·9) 21 (16·9) IIB 57 (6·9) 15 (12·1) III 373 (45·0) 55 (44·4) IV 190 (22·9) 22 (17·7) Impassable at OGD No 699 (84·3) 107 (86·3) Yes 130 (15·7) 17 (13·7) Management decisions Palliative (performance status) 143 (17·2) 12 (9·7) Palliative (incurable) 185 (22·3) 22 (17·7) Endoscopic therapy 26 (3·2) 11 (8·9) Surgery 58 (7·0) 5 (4·0) Neoadjuvant therapy + surgery 378 (45·6) 67 (54·0) DCRT 37 (4·5) 7 (5·6) Died during staging 2 (0·2) 0 (0) Endoscopic ultrasonography Not requested 127 (15·3)† 24 (19·4)‡ Performed (passable) 543 (65·5) 75 (60·5) Performed (impassable + miniprobe) 49 (5·9) 1 (0·8) Performed (impassable, no miniprobe) 106 (12·8) 24 (19·4) Abandoned 4 (0·5) 0 (0) . Development set (n = 829) . Validation set (n = 124) . Age (years)*§ 66 (13, 29–88) 68 (13, 40–80) Sex ratio (F : M) 204 : 625 35 : 89 Cell type Adenocarcinoma 639 (77·1) 104 (83·9) Squamous cell carcinoma 165 (19·9) 18 (14·5) Adenosquamous carcinoma 7 (0·8) 2 (1·6) Neuroendocrine carcinoma 3 (0·4) 0 (0) Small cell carcinoma 3 (0·4) 0 (0) Undifferentiated 12 (1·4) 0 (0) Grade Well 71 (8·6) 5 (4·0) Moderate 347 (41·9) 58 (46·8) Poor 389 (46·9) 60 (48·4) Undifferentiated 22 (2·7) 1 (0·8) Tumour site Proximal 23 (2·8) 1 (0·8) Mid 67 (8·1) 12 (9·7) Distal 237 (28·6) 42 (33·9) Siewert 1 197 (23·8) 38 (30·6) Siewert 2 195 (23·5) 16 (12·9) Siewert 3 105 (12·7) 14 (11·3) Multifocal 5 (0·6) 1 (0·8) Pretreatment stage (TNM 6th edition) I 52 (6·3) 11 (8·9) IIA 157 (18·9) 21 (16·9) IIB 57 (6·9) 15 (12·1) III 373 (45·0) 55 (44·4) IV 190 (22·9) 22 (17·7) Impassable at OGD No 699 (84·3) 107 (86·3) Yes 130 (15·7) 17 (13·7) Management decisions Palliative (performance status) 143 (17·2) 12 (9·7) Palliative (incurable) 185 (22·3) 22 (17·7) Endoscopic therapy 26 (3·2) 11 (8·9) Surgery 58 (7·0) 5 (4·0) Neoadjuvant therapy + surgery 378 (45·6) 67 (54·0) DCRT 37 (4·5) 7 (5·6) Died during staging 2 (0·2) 0 (0) Endoscopic ultrasonography Not requested 127 (15·3)† 24 (19·4)‡ Performed (passable) 543 (65·5) 75 (60·5) Performed (impassable + miniprobe) 49 (5·9) 1 (0·8) Performed (impassable, no miniprobe) 106 (12·8) 24 (19·4) Abandoned 4 (0·5) 0 (0) Values in parentheses are percentages unless indicated otherwise; * values are median (i.q.r., range). Endoscopic ultrasonography not requested owing to: † metastases on PET–CT (79) or laparoscopy (5), patient choice (22), prohibitive anatomy (10), decision to proceed to radical chemoradiotherapy irrespective of findings (7), death during staging (2) and unclear reasons (2); ‡ metastases on PET–CT (10) or laparoscopy (1), patient choice/performance status (9), prohibitive anatomy (1) and decision to proceed to definitive chemoradiotherapy (DCRT) (3). OGD, oesophagogastroduodenoscopy. § P < 0·001 (Shapiro–Wilk test). Open in new tab PET–CT Some 800 PET–CT examinations were performed in the development data set. Previous CT had suggested metastatic disease in 100 of these patients. This was confirmed in 57 (57·0 per cent) and refuted in 43 (43·0 per cent). In 28 (49 per cent) of the 57 patients with confirmed metastases, additional unsuspected metastases were demonstrated. Metastases were identified in 104 (14·9 per cent) of the 700 patients without a previous suggestion of metastatic disease, In total, therefore, metastatic disease was demonstrated in 161 patients (144 unequivocal, 17 requiring confirmatory biopsy). Five further lesions were biopsied, which demonstrated benign disease (4) and chronic lymphocytic leukaemia (1). Some 119 non-metastatic lesions were identified in 103 patients. Seventy-eight patients were investigated, confirming significant pathology in 17 (2·1 per cent): nine synchronous cancers involving the thyroid (2), colon (2), lung (2), prostate (2) and bladder (1), and eight patients with high-risk colonic polyps. In six patients (0·8 per cent), PET–CT also staged known synchronous cancers. Consequently, PET–CT altered management in 23·0 per cent: confirming metastases (7·1 per cent), identifying unsuspected metastases (13·0 per cent) and additional pathology (2·1 per cent), and staging synchronous cancers (0·8 per cent). Higher SUVmax was associated with squamous cell carcinoma (SCC), advanced stage, female sex and younger age; longer length with advanced stage; and FDG-avid nodes with SCC, advanced stage and age. The later PET–CT scanner was associated with a marginally higher SUVmax and avid nodes, but all reported associations were independent of this (Tables S1–S3, supporting information). Predicting unsuspected metastases Analysis was restricted to the 700 patients with CT M0 examinations (Table S4, supporting information), to identify variables associated with the demonstration of metastases on PET–CT. Although advancing EUS T and CT N category independently predicted metastases, no factors could be used to identify patients with a probability below the Pt (0·083 per cent), that is patients in whom the risk of demonstrating metastases was sufficiently low not to justify the risk of PET–CT. Although there was zero incidence in EUS T1 disease, the 95 per cent c.i. was broad (0–6·12 per cent), suggesting that, contrary to common clinical practice, PET–CT may have utility in tumours staged by EUS as T1. Endoscopic ultrasonography T1 N0 In the development study, Tx/T1 N0 disease was reported in 51 (7·3 per cent) of 698 patients. In four tumours impassable to the standard echoscope, possible T1 N0 disease was demonstrated in the visualized tumour. However, in none was T1 N0 status confirmed by miniprobe. T1 N0 disease was associated with earlier CT T category, CT N0 status and a passable tumour at OGD (Table 2). Table 2 Factors associated with T1 N0 disease on endoscopic ultrasonography: overall and multivariable logistic regression . EUS T1 N0 (n = 51) . EUS ≥ T1 N1 (n = 647) . P (corrected α = 0·0042)§ . Multivariable odds ratio† . P . Age (years)*‡ 69 (59–72, 32–84) 66 (64–75, 29–88) 0·019¶ 1·00 (1·00, 1·00) 0·649 Sex 1·000# F 12 (24) 158 (24·4) Reference Reference M 39 (76) 489 (75·6) 0·70 (0·18, 2·80) 0·617 Cell type 0·209 Adenocarcinoma 46 (90) 494 (76·4) Reference Reference Squamous cell carcinoma 5 (10) 133 (20·6) 0·21 (0·02, 2·20) 0·192 Adenosquamous carcinoma 0 (0) 7 (1·1) n.a. n.a. Neuroendocrine carcinoma 0 (0) 3 (0·5) n.a. n.a. Small cell carcinoma 0 (0) 2 (0·3) n.a. n.a. Undifferentiated 0 (0) 8 (1·2) n.a. n.a. Grade 0·525 Well 6 (12) 59 (9·1) Reference Reference Moderate 25 (49) 269 (41·6) 2·54 (0·32, 20·10) 0·367 Poor 19 (37) 302 (46·7) 3·03 (0·37, 24·76) 0·302 Undifferentiated 1 (2) 17 (2·6) 0·00 (0, ∞) 0·996 Tumour site 0·050 Proximal 0 (0) 14 (2·2) n.a. n.a. Mid 4 (8) 56 (8·7) Reference Reference Distal 21 (41) 173 (26·7) 0·71 (0·03, 14·71) 0·822 Siewert 1 17 (33) 160 (24·7) 1·67 (0·07, 39·77) 0·750 Siewert 2 8 (16) 160 (24·7) 0·42 (0·02, 9·88) 0·591 Siewert 3 1 (2) 80 (12·4) 2·16 (0·50, 9·25) 0·534 Multifocal 0 (0) 4 (0·6) n.a. n.a. Impassable (OGD) 0·002 No 51 (100) 566 (87·5) n.a. n.a. Yes 0 (0) 81 (12·5) CT T category < 0·001 Possible T1 49 (96) 79 (12·2) Reference Reference T2–T4a 2 (4) 499 (77·1) 0·03 (0·00, 0·16) < 0·001 Possible T4b 0 (0) 69 (10·7) n.a. n.a. CT N category < 0·001 N0 49 (96) 274 (42·3) Reference Reference N1 2 (4) 373 (57·7) 0·75 (0·08, 7·39) 0·803 CT centre 0·954 OUH 26 (51) 226 (34·9) 2·61 (0·62, 10·95) 0·188 Centre 1 8 (16) 157 (24·3) Reference Reference Centre 2 13 (25) 150 (23·2) 2·16 (0·50, 9·25) 0·301 Centre 3 1 (2) 92 (14·2) 0·44 (0·03, 6·24) 0·542 Other 3 (6) 22 (3·4) 29·4 (0·58, 1503·75) 0·100 SUVmax* 4·15 10·50 < 0·001¶ 0·06 (0·00, 3·37) 0·168 (2·50–5·13, (7·08–15·9, 2·50–11·80) 2·50–52·60) Not performed 13 (25) 10 (1·5) PET length (cm)* 1·00 5·40 < 0·001¶ 0·53 (0·29, 0·95) 0·034 (0–1·78, (3·70–7·00, 0–6·90) 0–20·00) Not performed 13 (25) 10 (1·5) PET avid nodes < 0·001 n.a. n.a. No 38 (75) 376 (58·1) Yes 0 (0) 259 (40·0) Not performed 13 (25) 12 (1·9) PET scanner n.a. 1 14 (27) 304 (47·0) Reference Reference 2 24 (47) 331 (51·2) 2·11 (0·67, 6·66) 0·203 Not performed 13 (25) 12 (1·9) . EUS T1 N0 (n = 51) . EUS ≥ T1 N1 (n = 647) . P (corrected α = 0·0042)§ . Multivariable odds ratio† . P . Age (years)*‡ 69 (59–72, 32–84) 66 (64–75, 29–88) 0·019¶ 1·00 (1·00, 1·00) 0·649 Sex 1·000# F 12 (24) 158 (24·4) Reference Reference M 39 (76) 489 (75·6) 0·70 (0·18, 2·80) 0·617 Cell type 0·209 Adenocarcinoma 46 (90) 494 (76·4) Reference Reference Squamous cell carcinoma 5 (10) 133 (20·6) 0·21 (0·02, 2·20) 0·192 Adenosquamous carcinoma 0 (0) 7 (1·1) n.a. n.a. Neuroendocrine carcinoma 0 (0) 3 (0·5) n.a. n.a. Small cell carcinoma 0 (0) 2 (0·3) n.a. n.a. Undifferentiated 0 (0) 8 (1·2) n.a. n.a. Grade 0·525 Well 6 (12) 59 (9·1) Reference Reference Moderate 25 (49) 269 (41·6) 2·54 (0·32, 20·10) 0·367 Poor 19 (37) 302 (46·7) 3·03 (0·37, 24·76) 0·302 Undifferentiated 1 (2) 17 (2·6) 0·00 (0, ∞) 0·996 Tumour site 0·050 Proximal 0 (0) 14 (2·2) n.a. n.a. Mid 4 (8) 56 (8·7) Reference Reference Distal 21 (41) 173 (26·7) 0·71 (0·03, 14·71) 0·822 Siewert 1 17 (33) 160 (24·7) 1·67 (0·07, 39·77) 0·750 Siewert 2 8 (16) 160 (24·7) 0·42 (0·02, 9·88) 0·591 Siewert 3 1 (2) 80 (12·4) 2·16 (0·50, 9·25) 0·534 Multifocal 0 (0) 4 (0·6) n.a. n.a. Impassable (OGD) 0·002 No 51 (100) 566 (87·5) n.a. n.a. Yes 0 (0) 81 (12·5) CT T category < 0·001 Possible T1 49 (96) 79 (12·2) Reference Reference T2–T4a 2 (4) 499 (77·1) 0·03 (0·00, 0·16) < 0·001 Possible T4b 0 (0) 69 (10·7) n.a. n.a. CT N category < 0·001 N0 49 (96) 274 (42·3) Reference Reference N1 2 (4) 373 (57·7) 0·75 (0·08, 7·39) 0·803 CT centre 0·954 OUH 26 (51) 226 (34·9) 2·61 (0·62, 10·95) 0·188 Centre 1 8 (16) 157 (24·3) Reference Reference Centre 2 13 (25) 150 (23·2) 2·16 (0·50, 9·25) 0·301 Centre 3 1 (2) 92 (14·2) 0·44 (0·03, 6·24) 0·542 Other 3 (6) 22 (3·4) 29·4 (0·58, 1503·75) 0·100 SUVmax* 4·15 10·50 < 0·001¶ 0·06 (0·00, 3·37) 0·168 (2·50–5·13, (7·08–15·9, 2·50–11·80) 2·50–52·60) Not performed 13 (25) 10 (1·5) PET length (cm)* 1·00 5·40 < 0·001¶ 0·53 (0·29, 0·95) 0·034 (0–1·78, (3·70–7·00, 0–6·90) 0–20·00) Not performed 13 (25) 10 (1·5) PET avid nodes < 0·001 n.a. n.a. No 38 (75) 376 (58·1) Yes 0 (0) 259 (40·0) Not performed 13 (25) 12 (1·9) PET scanner n.a. 1 14 (27) 304 (47·0) Reference Reference 2 24 (47) 331 (51·2) 2·11 (0·67, 6·66) 0·203 Not performed 13 (25) 12 (1·9) Values in parentheses are percentages unless indicated otherwise; values are * median (i.q.r., range) and † 95 per cent c.i. The analysis was done following exclusion of perfect separators. EUS, endoscopic ultrasonography; n.a., not applicable; OGD, oesophagogastroduodenoscopy; OUH, Oxford University Hospitals; SUVmax, maximum standardized uptake value. ‡ P < 0·001 (Shapiro–Wilk test). § Pearson's χ2 test, except ¶ Mann–Whitney U test and # Fisher's exact test. Open in new tab Table 2 Factors associated with T1 N0 disease on endoscopic ultrasonography: overall and multivariable logistic regression . EUS T1 N0 (n = 51) . EUS ≥ T1 N1 (n = 647) . P (corrected α = 0·0042)§ . Multivariable odds ratio† . P . Age (years)*‡ 69 (59–72, 32–84) 66 (64–75, 29–88) 0·019¶ 1·00 (1·00, 1·00) 0·649 Sex 1·000# F 12 (24) 158 (24·4) Reference Reference M 39 (76) 489 (75·6) 0·70 (0·18, 2·80) 0·617 Cell type 0·209 Adenocarcinoma 46 (90) 494 (76·4) Reference Reference Squamous cell carcinoma 5 (10) 133 (20·6) 0·21 (0·02, 2·20) 0·192 Adenosquamous carcinoma 0 (0) 7 (1·1) n.a. n.a. Neuroendocrine carcinoma 0 (0) 3 (0·5) n.a. n.a. Small cell carcinoma 0 (0) 2 (0·3) n.a. n.a. Undifferentiated 0 (0) 8 (1·2) n.a. n.a. Grade 0·525 Well 6 (12) 59 (9·1) Reference Reference Moderate 25 (49) 269 (41·6) 2·54 (0·32, 20·10) 0·367 Poor 19 (37) 302 (46·7) 3·03 (0·37, 24·76) 0·302 Undifferentiated 1 (2) 17 (2·6) 0·00 (0, ∞) 0·996 Tumour site 0·050 Proximal 0 (0) 14 (2·2) n.a. n.a. Mid 4 (8) 56 (8·7) Reference Reference Distal 21 (41) 173 (26·7) 0·71 (0·03, 14·71) 0·822 Siewert 1 17 (33) 160 (24·7) 1·67 (0·07, 39·77) 0·750 Siewert 2 8 (16) 160 (24·7) 0·42 (0·02, 9·88) 0·591 Siewert 3 1 (2) 80 (12·4) 2·16 (0·50, 9·25) 0·534 Multifocal 0 (0) 4 (0·6) n.a. n.a. Impassable (OGD) 0·002 No 51 (100) 566 (87·5) n.a. n.a. Yes 0 (0) 81 (12·5) CT T category < 0·001 Possible T1 49 (96) 79 (12·2) Reference Reference T2–T4a 2 (4) 499 (77·1) 0·03 (0·00, 0·16) < 0·001 Possible T4b 0 (0) 69 (10·7) n.a. n.a. CT N category < 0·001 N0 49 (96) 274 (42·3) Reference Reference N1 2 (4) 373 (57·7) 0·75 (0·08, 7·39) 0·803 CT centre 0·954 OUH 26 (51) 226 (34·9) 2·61 (0·62, 10·95) 0·188 Centre 1 8 (16) 157 (24·3) Reference Reference Centre 2 13 (25) 150 (23·2) 2·16 (0·50, 9·25) 0·301 Centre 3 1 (2) 92 (14·2) 0·44 (0·03, 6·24) 0·542 Other 3 (6) 22 (3·4) 29·4 (0·58, 1503·75) 0·100 SUVmax* 4·15 10·50 < 0·001¶ 0·06 (0·00, 3·37) 0·168 (2·50–5·13, (7·08–15·9, 2·50–11·80) 2·50–52·60) Not performed 13 (25) 10 (1·5) PET length (cm)* 1·00 5·40 < 0·001¶ 0·53 (0·29, 0·95) 0·034 (0–1·78, (3·70–7·00, 0–6·90) 0–20·00) Not performed 13 (25) 10 (1·5) PET avid nodes < 0·001 n.a. n.a. No 38 (75) 376 (58·1) Yes 0 (0) 259 (40·0) Not performed 13 (25) 12 (1·9) PET scanner n.a. 1 14 (27) 304 (47·0) Reference Reference 2 24 (47) 331 (51·2) 2·11 (0·67, 6·66) 0·203 Not performed 13 (25) 12 (1·9) . EUS T1 N0 (n = 51) . EUS ≥ T1 N1 (n = 647) . P (corrected α = 0·0042)§ . Multivariable odds ratio† . P . Age (years)*‡ 69 (59–72, 32–84) 66 (64–75, 29–88) 0·019¶ 1·00 (1·00, 1·00) 0·649 Sex 1·000# F 12 (24) 158 (24·4) Reference Reference M 39 (76) 489 (75·6) 0·70 (0·18, 2·80) 0·617 Cell type 0·209 Adenocarcinoma 46 (90) 494 (76·4) Reference Reference Squamous cell carcinoma 5 (10) 133 (20·6) 0·21 (0·02, 2·20) 0·192 Adenosquamous carcinoma 0 (0) 7 (1·1) n.a. n.a. Neuroendocrine carcinoma 0 (0) 3 (0·5) n.a. n.a. Small cell carcinoma 0 (0) 2 (0·3) n.a. n.a. Undifferentiated 0 (0) 8 (1·2) n.a. n.a. Grade 0·525 Well 6 (12) 59 (9·1) Reference Reference Moderate 25 (49) 269 (41·6) 2·54 (0·32, 20·10) 0·367 Poor 19 (37) 302 (46·7) 3·03 (0·37, 24·76) 0·302 Undifferentiated 1 (2) 17 (2·6) 0·00 (0, ∞) 0·996 Tumour site 0·050 Proximal 0 (0) 14 (2·2) n.a. n.a. Mid 4 (8) 56 (8·7) Reference Reference Distal 21 (41) 173 (26·7) 0·71 (0·03, 14·71) 0·822 Siewert 1 17 (33) 160 (24·7) 1·67 (0·07, 39·77) 0·750 Siewert 2 8 (16) 160 (24·7) 0·42 (0·02, 9·88) 0·591 Siewert 3 1 (2) 80 (12·4) 2·16 (0·50, 9·25) 0·534 Multifocal 0 (0) 4 (0·6) n.a. n.a. Impassable (OGD) 0·002 No 51 (100) 566 (87·5) n.a. n.a. Yes 0 (0) 81 (12·5) CT T category < 0·001 Possible T1 49 (96) 79 (12·2) Reference Reference T2–T4a 2 (4) 499 (77·1) 0·03 (0·00, 0·16) < 0·001 Possible T4b 0 (0) 69 (10·7) n.a. n.a. CT N category < 0·001 N0 49 (96) 274 (42·3) Reference Reference N1 2 (4) 373 (57·7) 0·75 (0·08, 7·39) 0·803 CT centre 0·954 OUH 26 (51) 226 (34·9) 2·61 (0·62, 10·95) 0·188 Centre 1 8 (16) 157 (24·3) Reference Reference Centre 2 13 (25) 150 (23·2) 2·16 (0·50, 9·25) 0·301 Centre 3 1 (2) 92 (14·2) 0·44 (0·03, 6·24) 0·542 Other 3 (6) 22 (3·4) 29·4 (0·58, 1503·75) 0·100 SUVmax* 4·15 10·50 < 0·001¶ 0·06 (0·00, 3·37) 0·168 (2·50–5·13, (7·08–15·9, 2·50–11·80) 2·50–52·60) Not performed 13 (25) 10 (1·5) PET length (cm)* 1·00 5·40 < 0·001¶ 0·53 (0·29, 0·95) 0·034 (0–1·78, (3·70–7·00, 0–6·90) 0–20·00) Not performed 13 (25) 10 (1·5) PET avid nodes < 0·001 n.a. n.a. No 38 (75) 376 (58·1) Yes 0 (0) 259 (40·0) Not performed 13 (25) 12 (1·9) PET scanner n.a. 1 14 (27) 304 (47·0) Reference Reference 2 24 (47) 331 (51·2) 2·11 (0·67, 6·66) 0·203 Not performed 13 (25) 12 (1·9) Values in parentheses are percentages unless indicated otherwise; values are * median (i.q.r., range) and † 95 per cent c.i. The analysis was done following exclusion of perfect separators. EUS, endoscopic ultrasonography; n.a., not applicable; OGD, oesophagogastroduodenoscopy; OUH, Oxford University Hospitals; SUVmax, maximum standardized uptake value. ‡ P < 0·001 (Shapiro–Wilk test). § Pearson's χ2 test, except ¶ Mann–Whitney U test and # Fisher's exact test. Open in new tab Of 128 patients with possible T1 disease on CT, 49 (38·3 (95 per cent c.i. 30·0 to 46·9) per cent) were identified by EUS, and two of 501 with CT T2–T4a disease (0·4 (0 to 1·5) per cent). There was zero incidence in the 69 patients with possible T4b disease on CT (95 per cent c.i. 0 to 4·4 per cent) and 81 with impassable tumours at OGD (95 per cent c.i. 0 to 3·7 per cent). Multivariable regression was therefore not possible for impassable tumours. Earlier CT T category independently predicted EUS T1 N0 disease (Table 2). Some 675 of 698 patients subsequently underwent PET–CT (including 38 patients staged by EUS as T1 N0). There was zero incidence of EUS T1 N0 disease among the 259 patients with FDG-avid nodes (95 per cent c.i. 0 to 10·9 per cent). In addition to this, multivariable regression demonstrated shorter avid length and earlier CT T category to predict T1 N0 disease. T4b Among 69 patients with T4b disease suggested on CT, this was confirmed in 26 (38 per cent) patients (4 miniprobe EUS) and refuted in 31 (3 miniprobe EUS). A further 12 patients with impassable tumours were managed as T3. No additional patients were identified. On multivariable analysis (unadjusted for CT T category as a perfect separator) cell type (SCC and adenosquamous carcinoma), site (non-GOJ tumour), impassability and longer PET length independently predicted T4b disease (Table 3). Table 3 Factors associated with T4b disease on endoscopic ultrasonography: overall and binary logistic regression . EUS T4b (n = 26) . EUS < T4b (n = 672) . P (corrected α = 0·0042)§ . Multivariable odds ratio† . P . Age (years)*‡ 67·5 66 0·475¶ 1·00 (1·00, 1·00) 0·790 (58·3–69·8, (59·8–72·3, 48·0–80·0) 44·0–83·0) Sex 0·059# F 2 (8) 168 (25·0) Reference Reference M 24 (92) 504 (75·0) 0·49 (1·72, 1·40) 0·182 Cell type < 0·001 Adenocarcinoma 7 (27) 533 (79·3) Reference Reference Squamous cell carcinoma 17 (65) 121 (18·0) 4·98 (1·27, 19·49) 0·021 Adenosquamous carcinoma 1 (4) 6 (0·9) 54·1 (2·68, 1091·91) 0·009 Neuroendocrine carcinoma 0 (0) 3 (0·4) n.a. n.a. Small cell carcinoma 1 (4) 1 (0·2) n.a. n.a. Undifferentiated 0 (0) 8 (1·2) n.a. n.a. Grade < 0.001 Well 4 (15) 61 (9·1) Reference Reference Moderate 12 (46) 282 (42·0) 0·25 (0·05, 1·18) 0·081 Poor 8 (31) 313 (46·6) 0·20 (0·04, 1·10) 0·065 Undifferentiated 2 (8) 16 (2·4) 0·00 (0, ∞) 0·991 Tumour site < 0·001 Proximal 2 (8) 12 (1·8) Reference Reference Mid 12 (46) 48 (7·1) 1·24 (0·20, 7·89) 0·865 Distal 7 (27) 187 (27·8) 0·31 (0·05, 2·07) 0·228 Siewert 1 3 (12) 174 (25·9) 0·05 (0·00, 0·68) 0·025 Siewert 2 0 (0) 168 (25·0) n.a. n.a. Siewert 3 1 (4) 80 (11·9) 0·13 (0·01, 2·30) 0·164 Multifocal 1 (4) 3 (0·5) 0·52 (0·02, 14·68) 0·699 Impassable (OGD) 0·002 No 17 (65) 600 (89·3) Reference Reference Yes 9 (35) 72 (10·7) 4·15 (1·31, 13·12) 0·015 CT T category < 0·001 Possible T1 0 (0) 128 (19·0) n.a. n.a. T2–T4a 0 (0) 501 (74·6) n.a. n.a. Possible T4b 26 (100) 43 (6·4) n.a. n.a. CT N category 1·000# N0 12 (46) 311 (46·3) Reference Reference N1 14 (54) 361 (53·7) 0·90 (0·30, 2·70) 0·846 CT centre 0·664 OUH 9 (35) 243 (36·2) 0·44 (0·12, 1·57) 0·205 Centre 1 9 (35) 156 (23·2) Reference Reference Centre 2 4 (15) 159 (23·7) 0·22 (0·04, 1·06) 0·059 Centre 3 3 (12) 90 (13·4) 0·44 (0·08, 2·51) 0·358 Other 1 (4) 24 (3·6) 1·42 (0·12, 17·60) 0·785 SUVmax* 13·80 10·30 0·062¶ 0·36 (0·12, 1·11) 0·074 (9·70–16·80, (7·00–15·90, 2·90–32·50) 2·50–52·60) PET length* 7·00 5·00 < 0·001¶ 1·63 (1·29, 2·06) < 0·001 (5·93–8·28, (3·50–7·00, 2·00–14·40) 0–20·00) PET avid nodes No 12 (46) 402 (59·8) 0·337# Reference Reference Yes 12 (46) 247 (36·8) 1·03 (0·32, 3·34) 0·965 Not performed 2 (8) 23 (3·4) PET scanner n.a. 1 17 (65) 301 (46·5) Reference Reference 2 7 (27) 348 (53·8) 0·36 (0·12, 1·10) 0·074 Not performed 2 (8) 23 (3·6) . EUS T4b (n = 26) . EUS < T4b (n = 672) . P (corrected α = 0·0042)§ . Multivariable odds ratio† . P . Age (years)*‡ 67·5 66 0·475¶ 1·00 (1·00, 1·00) 0·790 (58·3–69·8, (59·8–72·3, 48·0–80·0) 44·0–83·0) Sex 0·059# F 2 (8) 168 (25·0) Reference Reference M 24 (92) 504 (75·0) 0·49 (1·72, 1·40) 0·182 Cell type < 0·001 Adenocarcinoma 7 (27) 533 (79·3) Reference Reference Squamous cell carcinoma 17 (65) 121 (18·0) 4·98 (1·27, 19·49) 0·021 Adenosquamous carcinoma 1 (4) 6 (0·9) 54·1 (2·68, 1091·91) 0·009 Neuroendocrine carcinoma 0 (0) 3 (0·4) n.a. n.a. Small cell carcinoma 1 (4) 1 (0·2) n.a. n.a. Undifferentiated 0 (0) 8 (1·2) n.a. n.a. Grade < 0.001 Well 4 (15) 61 (9·1) Reference Reference Moderate 12 (46) 282 (42·0) 0·25 (0·05, 1·18) 0·081 Poor 8 (31) 313 (46·6) 0·20 (0·04, 1·10) 0·065 Undifferentiated 2 (8) 16 (2·4) 0·00 (0, ∞) 0·991 Tumour site < 0·001 Proximal 2 (8) 12 (1·8) Reference Reference Mid 12 (46) 48 (7·1) 1·24 (0·20, 7·89) 0·865 Distal 7 (27) 187 (27·8) 0·31 (0·05, 2·07) 0·228 Siewert 1 3 (12) 174 (25·9) 0·05 (0·00, 0·68) 0·025 Siewert 2 0 (0) 168 (25·0) n.a. n.a. Siewert 3 1 (4) 80 (11·9) 0·13 (0·01, 2·30) 0·164 Multifocal 1 (4) 3 (0·5) 0·52 (0·02, 14·68) 0·699 Impassable (OGD) 0·002 No 17 (65) 600 (89·3) Reference Reference Yes 9 (35) 72 (10·7) 4·15 (1·31, 13·12) 0·015 CT T category < 0·001 Possible T1 0 (0) 128 (19·0) n.a. n.a. T2–T4a 0 (0) 501 (74·6) n.a. n.a. Possible T4b 26 (100) 43 (6·4) n.a. n.a. CT N category 1·000# N0 12 (46) 311 (46·3) Reference Reference N1 14 (54) 361 (53·7) 0·90 (0·30, 2·70) 0·846 CT centre 0·664 OUH 9 (35) 243 (36·2) 0·44 (0·12, 1·57) 0·205 Centre 1 9 (35) 156 (23·2) Reference Reference Centre 2 4 (15) 159 (23·7) 0·22 (0·04, 1·06) 0·059 Centre 3 3 (12) 90 (13·4) 0·44 (0·08, 2·51) 0·358 Other 1 (4) 24 (3·6) 1·42 (0·12, 17·60) 0·785 SUVmax* 13·80 10·30 0·062¶ 0·36 (0·12, 1·11) 0·074 (9·70–16·80, (7·00–15·90, 2·90–32·50) 2·50–52·60) PET length* 7·00 5·00 < 0·001¶ 1·63 (1·29, 2·06) < 0·001 (5·93–8·28, (3·50–7·00, 2·00–14·40) 0–20·00) PET avid nodes No 12 (46) 402 (59·8) 0·337# Reference Reference Yes 12 (46) 247 (36·8) 1·03 (0·32, 3·34) 0·965 Not performed 2 (8) 23 (3·4) PET scanner n.a. 1 17 (65) 301 (46·5) Reference Reference 2 7 (27) 348 (53·8) 0·36 (0·12, 1·10) 0·074 Not performed 2 (8) 23 (3·6) Values in parentheses are percentages unless indicated otherwise; values are * median (i.q.r., range) and † 95 per cent c.i. The analysis was done following exclusion of perfect separators. EUS, endoscopic ultrasonography; n.a., not applicable; OGD, oesophagogastroduodenoscopy; OUH, Oxford University Hospitals; SUVmax, maximum standardized uptake value. ‡ P < 0.001 (Shapiro–Wilk test). § Pearson's χ2 test, except ¶ Mann–Whitney U test and # Fisher's exact test. Open in new tab Table 3 Factors associated with T4b disease on endoscopic ultrasonography: overall and binary logistic regression . EUS T4b (n = 26) . EUS < T4b (n = 672) . P (corrected α = 0·0042)§ . Multivariable odds ratio† . P . Age (years)*‡ 67·5 66 0·475¶ 1·00 (1·00, 1·00) 0·790 (58·3–69·8, (59·8–72·3, 48·0–80·0) 44·0–83·0) Sex 0·059# F 2 (8) 168 (25·0) Reference Reference M 24 (92) 504 (75·0) 0·49 (1·72, 1·40) 0·182 Cell type < 0·001 Adenocarcinoma 7 (27) 533 (79·3) Reference Reference Squamous cell carcinoma 17 (65) 121 (18·0) 4·98 (1·27, 19·49) 0·021 Adenosquamous carcinoma 1 (4) 6 (0·9) 54·1 (2·68, 1091·91) 0·009 Neuroendocrine carcinoma 0 (0) 3 (0·4) n.a. n.a. Small cell carcinoma 1 (4) 1 (0·2) n.a. n.a. Undifferentiated 0 (0) 8 (1·2) n.a. n.a. Grade < 0.001 Well 4 (15) 61 (9·1) Reference Reference Moderate 12 (46) 282 (42·0) 0·25 (0·05, 1·18) 0·081 Poor 8 (31) 313 (46·6) 0·20 (0·04, 1·10) 0·065 Undifferentiated 2 (8) 16 (2·4) 0·00 (0, ∞) 0·991 Tumour site < 0·001 Proximal 2 (8) 12 (1·8) Reference Reference Mid 12 (46) 48 (7·1) 1·24 (0·20, 7·89) 0·865 Distal 7 (27) 187 (27·8) 0·31 (0·05, 2·07) 0·228 Siewert 1 3 (12) 174 (25·9) 0·05 (0·00, 0·68) 0·025 Siewert 2 0 (0) 168 (25·0) n.a. n.a. Siewert 3 1 (4) 80 (11·9) 0·13 (0·01, 2·30) 0·164 Multifocal 1 (4) 3 (0·5) 0·52 (0·02, 14·68) 0·699 Impassable (OGD) 0·002 No 17 (65) 600 (89·3) Reference Reference Yes 9 (35) 72 (10·7) 4·15 (1·31, 13·12) 0·015 CT T category < 0·001 Possible T1 0 (0) 128 (19·0) n.a. n.a. T2–T4a 0 (0) 501 (74·6) n.a. n.a. Possible T4b 26 (100) 43 (6·4) n.a. n.a. CT N category 1·000# N0 12 (46) 311 (46·3) Reference Reference N1 14 (54) 361 (53·7) 0·90 (0·30, 2·70) 0·846 CT centre 0·664 OUH 9 (35) 243 (36·2) 0·44 (0·12, 1·57) 0·205 Centre 1 9 (35) 156 (23·2) Reference Reference Centre 2 4 (15) 159 (23·7) 0·22 (0·04, 1·06) 0·059 Centre 3 3 (12) 90 (13·4) 0·44 (0·08, 2·51) 0·358 Other 1 (4) 24 (3·6) 1·42 (0·12, 17·60) 0·785 SUVmax* 13·80 10·30 0·062¶ 0·36 (0·12, 1·11) 0·074 (9·70–16·80, (7·00–15·90, 2·90–32·50) 2·50–52·60) PET length* 7·00 5·00 < 0·001¶ 1·63 (1·29, 2·06) < 0·001 (5·93–8·28, (3·50–7·00, 2·00–14·40) 0–20·00) PET avid nodes No 12 (46) 402 (59·8) 0·337# Reference Reference Yes 12 (46) 247 (36·8) 1·03 (0·32, 3·34) 0·965 Not performed 2 (8) 23 (3·4) PET scanner n.a. 1 17 (65) 301 (46·5) Reference Reference 2 7 (27) 348 (53·8) 0·36 (0·12, 1·10) 0·074 Not performed 2 (8) 23 (3·6) . EUS T4b (n = 26) . EUS < T4b (n = 672) . P (corrected α = 0·0042)§ . Multivariable odds ratio† . P . Age (years)*‡ 67·5 66 0·475¶ 1·00 (1·00, 1·00) 0·790 (58·3–69·8, (59·8–72·3, 48·0–80·0) 44·0–83·0) Sex 0·059# F 2 (8) 168 (25·0) Reference Reference M 24 (92) 504 (75·0) 0·49 (1·72, 1·40) 0·182 Cell type < 0·001 Adenocarcinoma 7 (27) 533 (79·3) Reference Reference Squamous cell carcinoma 17 (65) 121 (18·0) 4·98 (1·27, 19·49) 0·021 Adenosquamous carcinoma 1 (4) 6 (0·9) 54·1 (2·68, 1091·91) 0·009 Neuroendocrine carcinoma 0 (0) 3 (0·4) n.a. n.a. Small cell carcinoma 1 (4) 1 (0·2) n.a. n.a. Undifferentiated 0 (0) 8 (1·2) n.a. n.a. Grade < 0.001 Well 4 (15) 61 (9·1) Reference Reference Moderate 12 (46) 282 (42·0) 0·25 (0·05, 1·18) 0·081 Poor 8 (31) 313 (46·6) 0·20 (0·04, 1·10) 0·065 Undifferentiated 2 (8) 16 (2·4) 0·00 (0, ∞) 0·991 Tumour site < 0·001 Proximal 2 (8) 12 (1·8) Reference Reference Mid 12 (46) 48 (7·1) 1·24 (0·20, 7·89) 0·865 Distal 7 (27) 187 (27·8) 0·31 (0·05, 2·07) 0·228 Siewert 1 3 (12) 174 (25·9) 0·05 (0·00, 0·68) 0·025 Siewert 2 0 (0) 168 (25·0) n.a. n.a. Siewert 3 1 (4) 80 (11·9) 0·13 (0·01, 2·30) 0·164 Multifocal 1 (4) 3 (0·5) 0·52 (0·02, 14·68) 0·699 Impassable (OGD) 0·002 No 17 (65) 600 (89·3) Reference Reference Yes 9 (35) 72 (10·7) 4·15 (1·31, 13·12) 0·015 CT T category < 0·001 Possible T1 0 (0) 128 (19·0) n.a. n.a. T2–T4a 0 (0) 501 (74·6) n.a. n.a. Possible T4b 26 (100) 43 (6·4) n.a. n.a. CT N category 1·000# N0 12 (46) 311 (46·3) Reference Reference N1 14 (54) 361 (53·7) 0·90 (0·30, 2·70) 0·846 CT centre 0·664 OUH 9 (35) 243 (36·2) 0·44 (0·12, 1·57) 0·205 Centre 1 9 (35) 156 (23·2) Reference Reference Centre 2 4 (15) 159 (23·7) 0·22 (0·04, 1·06) 0·059 Centre 3 3 (12) 90 (13·4) 0·44 (0·08, 2·51) 0·358 Other 1 (4) 24 (3·6) 1·42 (0·12, 17·60) 0·785 SUVmax* 13·80 10·30 0·062¶ 0·36 (0·12, 1·11) 0·074 (9·70–16·80, (7·00–15·90, 2·90–32·50) 2·50–52·60) PET length* 7·00 5·00 < 0·001¶ 1·63 (1·29, 2·06) < 0·001 (5·93–8·28, (3·50–7·00, 2·00–14·40) 0–20·00) PET avid nodes No 12 (46) 402 (59·8) 0·337# Reference Reference Yes 12 (46) 247 (36·8) 1·03 (0·32, 3·34) 0·965 Not performed 2 (8) 23 (3·4) PET scanner n.a. 1 17 (65) 301 (46·5) Reference Reference 2 7 (27) 348 (53·8) 0·36 (0·12, 1·10) 0·074 Not performed 2 (8) 23 (3·6) Values in parentheses are percentages unless indicated otherwise; values are * median (i.q.r., range) and † 95 per cent c.i. The analysis was done following exclusion of perfect separators. EUS, endoscopic ultrasonography; n.a., not applicable; OGD, oesophagogastroduodenoscopy; OUH, Oxford University Hospitals; SUVmax, maximum standardized uptake value. ‡ P < 0.001 (Shapiro–Wilk test). § Pearson's χ2 test, except ¶ Mann–Whitney U test and # Fisher's exact test. Open in new tab Metastases Possible M1 nodal metastases were identified by EUS in five patients; they were excluded by FNA cytology in three and confirmed in two (both evident on PET–CT). Utility EUS altered management decisions in 77 patients (11·0 per cent). Although PET–CT was typically performed before EUS, in a limited number it was performed on the same or subsequent days. Metastases were demonstrated in 96 patients (14·2 per cent) by further staging with PET–CT with or without laparoscopy. In patients downstaged by EUS from possible T4b disease, metastases on PET–CT were subsequently demonstrated in ten of 43 patients. Stratification by CT T category demonstrated distinct utilities: in patients with Tx/possible T1 (early) disease (128, 18·3 per cent), EUS confirmed T1 N0 in 49 (38·3 per cent). In patients with possible T4b disease (69, 9·9 per cent) EUS was confirmatory in 26 (38 per cent). However, in 501 patients (71·8 per cent) without possible T1 or T4b disease on CT, EUS altered management in just two (0·4 per cent). In the 81 patients with impassable tumours, EUS altered management in three (4 per cent), confirming T4b with miniprobe EUS. Comparison with pathological staging Among 367 patients who underwent resection, 81 (22·1 per cent) received no neoadjuvant therapy, of whom 19 underwent ER, 7 ER plus surgery and 55 surgery. Of the 51 EUS T1 N0 tumours, 46 were resected (17, 7 and 22 respectively). Of the 17 patients who had ER alone, pT1 disease was confirmed in 11. Among the 29 treated with surgery with or without ER, pT1 N0 disease was confirmed in 24. Five tumours were upstaged by T category, two also to N1 disease. Five other tumours staged by EUS as greater than T1 N0 were downstaged to pT1 N0, from EUS T3 (3) and EUS T1 N1 (2). Excluding the 17 patients who, after EUS, underwent ER without surgical resection (in whom pN status could not be assessed), EUS was 83 per cent sensitive and 84 per cent specific for pT1 N0 (PPV 83 per cent; NPV 84 per cent). Factors associated with pT1 N0 disease are presented in Table S5 (supporting information). Endoscopic ultrasonography and decision theory There was one instance of perforation (0·1 per cent). The Pt for EUS T1 N0 disease was 0·95 per cent (the probability of identifying T1 N0 disease at which the benefit of EUS equals its risk). As the probability in patients with impassable tumours (0 per cent) or T2–T4a disease on CT (0·4 per cent) was lower, the risk of EUS to these patients outweighed its potential benefit of altering management. The Pt for EUS T4b disease was 2·02 per cent (based on T4 disease overall). Staging laparoscopy Some 397 patients underwent laparoscopy, and metastases were demonstrated in 28 (7·1 per cent). In 341 patients undergoing laparoscopy without feeding jejunostomy there was one major complication (pneumonia, 0·3 per cent) and four minor complications (1·2 per cent, urinary retention). Metastases were demonstrated in two (4 per cent) of 54 distal oesophageal tumours not involving the GOJ endoscopically. An impassable tumour, undifferentiated grade, possible T4b disease on CT, and lower SUVmax predicted unsuspected peritoneal metastases (Table S6, supporting information). No factor could identify patients below the Pt (0·38 per cent). Refinement of existing algorithm As a result of the findings that the incidence of T1 N0 disease on EUS among patients staged as T2–T4a by CT was minimal, and insufficient to justify the EUS test risk, it is proposed that EUS should be reserved only for patients with possible T1 or T4b disease on CT. This would have been 96·1 per cent sensitive and 87·8 per cent specific for T1 N0, and 100 per cent sensitive and 93·6 per cent specific for T4b disease on EUS, with very high NPV (99·7 and 100 per cent respectively) (Table 4). Table 4 Apparent, internally validated and temporally validated performance metrics of existing and novel algorithms, plus decision tree analyses, logistic regression models and artificial neural networks in predicting T1 N0, T4b and pT1 N0 disease by endoscopic ultrasonography Algorithm/model . Sensitivity . Specificity . PPV . NPV . Brier . κ . O : E . AUC . Net benefit . Total cost (€) . EUS T1 N0 (n = 51 development; n = 11 temporal validation) Default 1·000 0·000 0·738 0·000 n.a. 0·000 0·731 0·500 0·050 534 803 Novel (apparent) 0·961 0·878 0·383 0·997 0·046 0·495 0·398 0·919 0·071 118 671 Novel (internal) 0·941 0·881 0·384 0·995 0·137 0·493 0·349 0·928 0·067 128 971 Novel (independent) 1·000 0·855 0·478 1·000 0·048 0·581 0·478 0·928 0·107 n.a. Models before PET–CT DTA 1 (apparent) 0·961 0·878 0·383 0·997 0·046 0·495 0·398 0·919 0·071 118 671 DTA 1 (internal) 0·941 0·881 0·384 0·995 0·137 0·493 0·349 0·928 0·070 128 971 DTA 1 (independent) 1·000 0·855 0·478 1·000 0·048 0·581 0·478 0·928 0·107 n.a. LRM1 (apparent) 0·980 0·535 0·157 0·997 0·055 0·152 0·160 0·758 0·068 324 243 LRM 1 (internal) 0·980 0·535 0·157 0·997 0·055 0·152 0·160 0·758 0·068 324 243 LRM 1 (independent) 1·000 0·723 0·324 1·000 0·052 0·379 0·324 0·861 0·107 n.a. ANN 1 (apparent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN 1 (internal) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN 1 (independent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Models after PET–CT DTA 2 (apparent) 1·000 0·840 0·331 1·000 0·031 0·436 0·331 0·920 0·072 159 368 DTA 2 (internal) 0·944 0·866 0·225 0·996 0·038 0·401 0·298 0·907 0·051 179 967 LRM 2 (apparent) 0·947 0·734 0·310 0·991 0·049 0·359 0·301 0·953 0·100* (0·069) 179 967 LRM 2 (internal) 0·947 0·734 0·310 0·991 0·052 0·036 0·301 0·946 0·100* (0·069) 179 967 ANN 2 (apparent) 0·974 0·835 0·261 0·998 0·022 0·354 0·261 0·904 0·069 169 667 ANN 2 (internal) 0·974 0·835 0·261 0·998 0·018 0·354 0·261 0·904 0·068 169 667 EUS T4b (n = 26 development; n = 0 temporal validation) Default 1·000 0·000 0·038 0·000 n.a. n.a. n.a. 0·500 0·009 534 803 Novel (apparent) 1·000 0·936 0·377 1·000 0·038 0·522 0·377 0·968 0·035 52 867 Novel (internal) 0·962 0·926 0·333 0·998 0·029 0·466 0·333 0·944 0·034 91 995 DTA (apparent) 1·000 0·936 0·377 1·000 0·038 0·522 0·377 0·968 0·035 52 867 DTA (internal) 0·962 0·926 0·333 0·998 0·029 0·466 0·333 0·944 0·034 91 995 ANN (apparent) 1·000 0·700 0·114 1·000 0·019 0·147 0·114 0·849 0·028 52 867 ANN (internal) 1·000 0·702 0·115 1·000 0·085 0·147 0·115 0·851 0·028 52 867 pT1 N0 Default 1·000 0·000 0·564 0·000 n.a. n.a. n.a. 0·500 0·466 n.a. DTA (apparent) 1·000 0·348 0·580 1·000 0·200 0·333 0·580 0·790 0·470 n.a. DTA (internal 0·667 0·708 0·741 0·630 0·421 0·370 0·889 0·685 0·308 n.a. LRM (apparent) 1·000 0·111 0·590 1·000 0·102 0·139 0·583 0·556 0·440 n.a. LRM (internal) 1·000 0·111 0·590 1·000 0·102 0·139 0·583 0·556 0·440 n.a. ANN (apparent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN (internal) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Algorithm/model . Sensitivity . Specificity . PPV . NPV . Brier . κ . O : E . AUC . Net benefit . Total cost (€) . EUS T1 N0 (n = 51 development; n = 11 temporal validation) Default 1·000 0·000 0·738 0·000 n.a. 0·000 0·731 0·500 0·050 534 803 Novel (apparent) 0·961 0·878 0·383 0·997 0·046 0·495 0·398 0·919 0·071 118 671 Novel (internal) 0·941 0·881 0·384 0·995 0·137 0·493 0·349 0·928 0·067 128 971 Novel (independent) 1·000 0·855 0·478 1·000 0·048 0·581 0·478 0·928 0·107 n.a. Models before PET–CT DTA 1 (apparent) 0·961 0·878 0·383 0·997 0·046 0·495 0·398 0·919 0·071 118 671 DTA 1 (internal) 0·941 0·881 0·384 0·995 0·137 0·493 0·349 0·928 0·070 128 971 DTA 1 (independent) 1·000 0·855 0·478 1·000 0·048 0·581 0·478 0·928 0·107 n.a. LRM1 (apparent) 0·980 0·535 0·157 0·997 0·055 0·152 0·160 0·758 0·068 324 243 LRM 1 (internal) 0·980 0·535 0·157 0·997 0·055 0·152 0·160 0·758 0·068 324 243 LRM 1 (independent) 1·000 0·723 0·324 1·000 0·052 0·379 0·324 0·861 0·107 n.a. ANN 1 (apparent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN 1 (internal) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN 1 (independent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Models after PET–CT DTA 2 (apparent) 1·000 0·840 0·331 1·000 0·031 0·436 0·331 0·920 0·072 159 368 DTA 2 (internal) 0·944 0·866 0·225 0·996 0·038 0·401 0·298 0·907 0·051 179 967 LRM 2 (apparent) 0·947 0·734 0·310 0·991 0·049 0·359 0·301 0·953 0·100* (0·069) 179 967 LRM 2 (internal) 0·947 0·734 0·310 0·991 0·052 0·036 0·301 0·946 0·100* (0·069) 179 967 ANN 2 (apparent) 0·974 0·835 0·261 0·998 0·022 0·354 0·261 0·904 0·069 169 667 ANN 2 (internal) 0·974 0·835 0·261 0·998 0·018 0·354 0·261 0·904 0·068 169 667 EUS T4b (n = 26 development; n = 0 temporal validation) Default 1·000 0·000 0·038 0·000 n.a. n.a. n.a. 0·500 0·009 534 803 Novel (apparent) 1·000 0·936 0·377 1·000 0·038 0·522 0·377 0·968 0·035 52 867 Novel (internal) 0·962 0·926 0·333 0·998 0·029 0·466 0·333 0·944 0·034 91 995 DTA (apparent) 1·000 0·936 0·377 1·000 0·038 0·522 0·377 0·968 0·035 52 867 DTA (internal) 0·962 0·926 0·333 0·998 0·029 0·466 0·333 0·944 0·034 91 995 ANN (apparent) 1·000 0·700 0·114 1·000 0·019 0·147 0·114 0·849 0·028 52 867 ANN (internal) 1·000 0·702 0·115 1·000 0·085 0·147 0·115 0·851 0·028 52 867 pT1 N0 Default 1·000 0·000 0·564 0·000 n.a. n.a. n.a. 0·500 0·466 n.a. DTA (apparent) 1·000 0·348 0·580 1·000 0·200 0·333 0·580 0·790 0·470 n.a. DTA (internal 0·667 0·708 0·741 0·630 0·421 0·370 0·889 0·685 0·308 n.a. LRM (apparent) 1·000 0·111 0·590 1·000 0·102 0·139 0·583 0·556 0·440 n.a. LRM (internal) 1·000 0·111 0·590 1·000 0·102 0·139 0·583 0·556 0·440 n.a. ANN (apparent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN (internal) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. * LRM net benefit derived on smaller subset; adjusted benefit for whole cohort in parentheses. PPV, positive predictive value; NPV, negative predictive value; 0 : E, observed : expected; AUC, area under curve; EUS, endoscopic ultrasonography; n.a., not applicable; internal, internal validation; independent, independent validation; DTA, decision tree analysis; LRM, logistic regression model; ANN, artificial neural network. Open in new tab Table 4 Apparent, internally validated and temporally validated performance metrics of existing and novel algorithms, plus decision tree analyses, logistic regression models and artificial neural networks in predicting T1 N0, T4b and pT1 N0 disease by endoscopic ultrasonography Algorithm/model . Sensitivity . Specificity . PPV . NPV . Brier . κ . O : E . AUC . Net benefit . Total cost (€) . EUS T1 N0 (n = 51 development; n = 11 temporal validation) Default 1·000 0·000 0·738 0·000 n.a. 0·000 0·731 0·500 0·050 534 803 Novel (apparent) 0·961 0·878 0·383 0·997 0·046 0·495 0·398 0·919 0·071 118 671 Novel (internal) 0·941 0·881 0·384 0·995 0·137 0·493 0·349 0·928 0·067 128 971 Novel (independent) 1·000 0·855 0·478 1·000 0·048 0·581 0·478 0·928 0·107 n.a. Models before PET–CT DTA 1 (apparent) 0·961 0·878 0·383 0·997 0·046 0·495 0·398 0·919 0·071 118 671 DTA 1 (internal) 0·941 0·881 0·384 0·995 0·137 0·493 0·349 0·928 0·070 128 971 DTA 1 (independent) 1·000 0·855 0·478 1·000 0·048 0·581 0·478 0·928 0·107 n.a. LRM1 (apparent) 0·980 0·535 0·157 0·997 0·055 0·152 0·160 0·758 0·068 324 243 LRM 1 (internal) 0·980 0·535 0·157 0·997 0·055 0·152 0·160 0·758 0·068 324 243 LRM 1 (independent) 1·000 0·723 0·324 1·000 0·052 0·379 0·324 0·861 0·107 n.a. ANN 1 (apparent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN 1 (internal) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN 1 (independent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Models after PET–CT DTA 2 (apparent) 1·000 0·840 0·331 1·000 0·031 0·436 0·331 0·920 0·072 159 368 DTA 2 (internal) 0·944 0·866 0·225 0·996 0·038 0·401 0·298 0·907 0·051 179 967 LRM 2 (apparent) 0·947 0·734 0·310 0·991 0·049 0·359 0·301 0·953 0·100* (0·069) 179 967 LRM 2 (internal) 0·947 0·734 0·310 0·991 0·052 0·036 0·301 0·946 0·100* (0·069) 179 967 ANN 2 (apparent) 0·974 0·835 0·261 0·998 0·022 0·354 0·261 0·904 0·069 169 667 ANN 2 (internal) 0·974 0·835 0·261 0·998 0·018 0·354 0·261 0·904 0·068 169 667 EUS T4b (n = 26 development; n = 0 temporal validation) Default 1·000 0·000 0·038 0·000 n.a. n.a. n.a. 0·500 0·009 534 803 Novel (apparent) 1·000 0·936 0·377 1·000 0·038 0·522 0·377 0·968 0·035 52 867 Novel (internal) 0·962 0·926 0·333 0·998 0·029 0·466 0·333 0·944 0·034 91 995 DTA (apparent) 1·000 0·936 0·377 1·000 0·038 0·522 0·377 0·968 0·035 52 867 DTA (internal) 0·962 0·926 0·333 0·998 0·029 0·466 0·333 0·944 0·034 91 995 ANN (apparent) 1·000 0·700 0·114 1·000 0·019 0·147 0·114 0·849 0·028 52 867 ANN (internal) 1·000 0·702 0·115 1·000 0·085 0·147 0·115 0·851 0·028 52 867 pT1 N0 Default 1·000 0·000 0·564 0·000 n.a. n.a. n.a. 0·500 0·466 n.a. DTA (apparent) 1·000 0·348 0·580 1·000 0·200 0·333 0·580 0·790 0·470 n.a. DTA (internal 0·667 0·708 0·741 0·630 0·421 0·370 0·889 0·685 0·308 n.a. LRM (apparent) 1·000 0·111 0·590 1·000 0·102 0·139 0·583 0·556 0·440 n.a. LRM (internal) 1·000 0·111 0·590 1·000 0·102 0·139 0·583 0·556 0·440 n.a. ANN (apparent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN (internal) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Algorithm/model . Sensitivity . Specificity . PPV . NPV . Brier . κ . O : E . AUC . Net benefit . Total cost (€) . EUS T1 N0 (n = 51 development; n = 11 temporal validation) Default 1·000 0·000 0·738 0·000 n.a. 0·000 0·731 0·500 0·050 534 803 Novel (apparent) 0·961 0·878 0·383 0·997 0·046 0·495 0·398 0·919 0·071 118 671 Novel (internal) 0·941 0·881 0·384 0·995 0·137 0·493 0·349 0·928 0·067 128 971 Novel (independent) 1·000 0·855 0·478 1·000 0·048 0·581 0·478 0·928 0·107 n.a. Models before PET–CT DTA 1 (apparent) 0·961 0·878 0·383 0·997 0·046 0·495 0·398 0·919 0·071 118 671 DTA 1 (internal) 0·941 0·881 0·384 0·995 0·137 0·493 0·349 0·928 0·070 128 971 DTA 1 (independent) 1·000 0·855 0·478 1·000 0·048 0·581 0·478 0·928 0·107 n.a. LRM1 (apparent) 0·980 0·535 0·157 0·997 0·055 0·152 0·160 0·758 0·068 324 243 LRM 1 (internal) 0·980 0·535 0·157 0·997 0·055 0·152 0·160 0·758 0·068 324 243 LRM 1 (independent) 1·000 0·723 0·324 1·000 0·052 0·379 0·324 0·861 0·107 n.a. ANN 1 (apparent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN 1 (internal) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN 1 (independent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Models after PET–CT DTA 2 (apparent) 1·000 0·840 0·331 1·000 0·031 0·436 0·331 0·920 0·072 159 368 DTA 2 (internal) 0·944 0·866 0·225 0·996 0·038 0·401 0·298 0·907 0·051 179 967 LRM 2 (apparent) 0·947 0·734 0·310 0·991 0·049 0·359 0·301 0·953 0·100* (0·069) 179 967 LRM 2 (internal) 0·947 0·734 0·310 0·991 0·052 0·036 0·301 0·946 0·100* (0·069) 179 967 ANN 2 (apparent) 0·974 0·835 0·261 0·998 0·022 0·354 0·261 0·904 0·069 169 667 ANN 2 (internal) 0·974 0·835 0·261 0·998 0·018 0·354 0·261 0·904 0·068 169 667 EUS T4b (n = 26 development; n = 0 temporal validation) Default 1·000 0·000 0·038 0·000 n.a. n.a. n.a. 0·500 0·009 534 803 Novel (apparent) 1·000 0·936 0·377 1·000 0·038 0·522 0·377 0·968 0·035 52 867 Novel (internal) 0·962 0·926 0·333 0·998 0·029 0·466 0·333 0·944 0·034 91 995 DTA (apparent) 1·000 0·936 0·377 1·000 0·038 0·522 0·377 0·968 0·035 52 867 DTA (internal) 0·962 0·926 0·333 0·998 0·029 0·466 0·333 0·944 0·034 91 995 ANN (apparent) 1·000 0·700 0·114 1·000 0·019 0·147 0·114 0·849 0·028 52 867 ANN (internal) 1·000 0·702 0·115 1·000 0·085 0·147 0·115 0·851 0·028 52 867 pT1 N0 Default 1·000 0·000 0·564 0·000 n.a. n.a. n.a. 0·500 0·466 n.a. DTA (apparent) 1·000 0·348 0·580 1·000 0·200 0·333 0·580 0·790 0·470 n.a. DTA (internal 0·667 0·708 0·741 0·630 0·421 0·370 0·889 0·685 0·308 n.a. LRM (apparent) 1·000 0·111 0·590 1·000 0·102 0·139 0·583 0·556 0·440 n.a. LRM (internal) 1·000 0·111 0·590 1·000 0·102 0·139 0·583 0·556 0·440 n.a. ANN (apparent) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ANN (internal) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. * LRM net benefit derived on smaller subset; adjusted benefit for whole cohort in parentheses. PPV, positive predictive value; NPV, negative predictive value; 0 : E, observed : expected; AUC, area under curve; EUS, endoscopic ultrasonography; n.a., not applicable; internal, internal validation; independent, independent validation; DTA, decision tree analysis; LRM, logistic regression model; ANN, artificial neural network. Open in new tab Compared with existing practice there was no significant change in sensitivity, but specificity improved (P < 0·001) and net benefits increased. This would be accompanied by a 71·8 per cent reduction in examinations, procedural risk and expenditure. DCA demonstrated minimal change in net algorithm benefits with extreme variations in perforation risk (Fig. S1, supporting information). The zero incidence of T1 N0 disease in impassable tumours also suggests that EUS could be omitted in these patients. However, this would have applied to only two of 128 patients with possible T1 disease on CT. Similarly, the zero incidence in patients with FDG-avid nodes applied to only 17 patients with FDG-avid nodes staged by CT as possible T1 disease, for whom confirmatory assessment by EUS with or without FNA may be warranted (Fig. 1) Fig. 1 Open in new tabDownload slide CT-guided algorithm (and decision tree analysis 1) for performing endoscopic ultrasonography (EUS) for T1 N0 disease before PET–CT. Numbers relate to development set This CT-guided algorithm was validated internally (using bootstrapping) and independently. Internal validation demonstrated minimal overfitting and sampling bias; the adjusted T1 N0 sensitivity was 94·1 per cent and the specificity 88·1 per cent; for T4b, respective values were 96·2 and 92·6 per cent (Table 4). Independent validation demonstrated 100·0 per cent sensitivity and 85·5 per cent specificity for T1 N0 (11 EUS T1 N0). Validation of new endoscopic ultrasonography algorithm Some 91 patients in the validation set underwent PET–CT and EUS. No patient was staged by EUS as having T1 N0 disease among the 60 with avid nodes. Twelve had possible T4b disease on CT; seven underwent EUS refuting T4b and EUS was omitted in five. EUS was performed in eight of 17 patients with impassable tumours; two were downstaged from possible T4b disease. Modelling The optimal model for identifying T1 N0 disease by EUS before PET–CT was a decision tree (DTA 1); this reserved EUS for those with possible T1 disease on CT, and was identical to the pragmatic CT-guided algorithm (Fig. 1 and Table 4; Table S7, supporting information). After PET–CT, the optimal model was a modified decision tree (DTA 2); this reserved EUS for patients with possible T1 disease on CT without FDG-avid nodes, or CT T2–T4a disease with SUVmax below 6·38 and length less than 3·4 cm on PET–CT (Fig. 2). This was 100 per cent sensitive and 84·0 per cent specific with minimal overfitting on internal validation (Table 4). This could not be validated independently, as no patient in the validation set with CT T2–T4a disease had EUS T1 N0 disease. DCA demonstrated variation in EUS perforation rate to have minimal effects on model net benefits (Fig. S1, supporting information) Fig. 2 Open in new tabDownload slide Decision tree analysis 2 for performing endoscopic ultrasonography (EUS) for T1 N0 disease after PET–CT. SUVmax, maximum standardized uptake value. Numbers relate to development set Again, the optimal model for identifying T4b disease by EUS was a decision tree identical to the proposed algorithm; this reserved EUS for patients with possible T4b disease on CT (100 per cent sensitivity) (Table 4). ANNs could be generated, but no model had clinical utility. Similarly, no model had clinical utility in predicting pT1 N0 disease, and unsuspected metastases on PET–CT and at laparoscopy. Suggested staging algorithm Based on these findings, the following staging algorithm is proposed when considering patients for resection (Fig. 3). Following CT, EUS (with or without FNA or staging ER) should be reserved for patients with either: Tx/possible T1 disease on CT, passable at OGD; or possible T4b disease without metastases on PET–CT. For all other patients EUS can be omitted, thereby reducing risk, delay and expenditure. Fig. 3 Open in new tabDownload slide Pragmatic algorithm for staging oesophageal cancer. OGD, oesophagogastroduodenectomy; EUS, endoscopic ultrasonography; FNA, fine-needle aspiration; ER, endoscopic resection; GOJ, gastro-oesophageal junction Discussion This study sought to quantify the staging utilities of EUS, PET–CT and laparoscopy for oesophageal/GOJ cancer, within the context of risk, benefit and probability thresholds. A number of factors predicted EUS T1 N0 and T4b disease, identifying patients above and below these thresholds. Although EUS provided additional precision and information regarding T and N categories over CT and PET–CT, this had no added utility in the majority of patients. For these patients, EUS risk exceeded potential benefit (as defined by its primary utility). This pragmatic algorithm was validated internally and independently. The main findings regarding utilities of EUS, PET–CT and laparoscopy, and CT T category grouping and impassable tumours, are likely to be robust, owing to the size of the cohorts. Although the lower resolution of CT confers lower sensitivity and specificity for T and N category than EUS34,35, it typically demonstrates early Tx/possible T1, possible T4b or locally advanced T2–T4a disease, and this appears sufficient to guide use of EUS. As the relative merits and results of investigations vary with local resources, imaging and software platforms, however, external evaluation of these algorithms is required to assess generalizability. These recommendations are based on pragmatic definitions of the primary utilities of EUS: the identification of patients for endoscopic resection and neoadjuvant therapy. The benefit of the latter in patients with T2 N0 disease has not been established unequivocally (as reflected by variations in study protocols). In selected patients, EUS may also have secondary utilities, such as providing additional prognostic or anatomical information (for example, distinguishing between T2 N0 and T3 N3 disease). Owing to the limitations of predicting outcome on the basis of EUS T and N categories before potentially downstaging neoadjuvant therapy, such information is likely to alter management infrequently. However, in such instances EUS may have a role to play in patients with T2–T4a disease on CT. Although variables were associated with metastases on PET–CT and at laparoscopy, comparable groups could not be identified. No benefit was seen for PET–CT in the few tumours staged as T1 N0, but owing to the significant EUS false-positive rate (8·9 per cent) the possibility of metastases in this group exists, reinforcing the need for staging ER4. Indeed, PET variables could predict false-positive T1 N0 disease. Notably, there was benefit of laparoscopy for T2 and distal oesophageal tumours, which conflicts with American and European guidelines1–3. The study also confirmed that adenocarcinoma and SCC differ in their FDG avidity8, and identified novel associations with FDG-avid nodes, SUVmax and length. The role of miniprobe EUS in impassable tumours remains unclear. There was benefit in patients with possible T4b disease on CT, but none for those without, although the 95 per cent c.i. for the latter (0 to 3·70 per cent) exceeded the Pt. Although the present study is retrospective, bias was minimized by collecting data from four parallel databases to ensure accuracy and that no data were missing. Having CT performed in multiple centres was not a significant confounder and was adjusted for. The advantages and disadvantages of DTA, LRM and ANN were mitigated by comparing all three. On the basis of pragmatic primary staging utilities, the risk of EUS typically exceeded its benefit in patients with either T2–T4a disease on CT, impassable tumours at OGD or FDG-avid nodes on PET–CT. Laparoscopy was beneficial in patients with distal oesophageal tumours and T2 disease. There may be further roles for DTA and LRMs in guiding further selection of patients with oesophagogastric cancers for specific treatment pathways. Acknowledgements The authors thank R. Marshall, B. Sgromo and F. Gleeson for critical review of the manuscript; R. Styles for her work in coordinating the Oxford OesophagoGastric multidisciplinary team and maintenance of its database; and D. Lunn for statistical advice. Disclosure: The authors declare no conflict of interest. References 1 Varghese TK Jr, Hofstetter WL, Rizk NP, Low DE, Darling GE, Watson TJ et al. The Society of Thoracic Surgeons guidelines on the diagnosis and staging of patients with esophageal cancer . Ann Thorac Surg 2013 ; 96 : 346 – 356 . Google Scholar Crossref Search ADS PubMed WorldCat 2 National Comprehensive Cancer Network (NCCN) . NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines): Esophageal and Esophagogastric Junction Cancers . Version 1; 2014 . https://www.nccn.org/store/login/login.aspx?ReturnURL=http://www.nccn.org/professionals/physician_gls/pdf/esophageal.pdf [accessed 12 July 2017]. 3 Stahl M , Budach W, Meyer HJ, Cervantes A; ESMO Guidelines Working Group. Esophageal cancer: Clinical Practice Guidelines for diagnosis, treatment and follow-up . Ann Oncol 2010 ; 21 ( Suppl 5 ): v46 – v49 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Allum WH , Blazeby JM, Griffin SM, Cunningham D, Jankowski JA, Wong R et al. Guidelines for the management of oesophageal and gastric cancer . Gut 2011 ; 60 : 1449 – 1472 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Puli SR , Reddy JB, Bechtold ML, Antillon D, Ibdah JA, Antillon MR. Staging accuracy of esophageal cancer by endoscopic ultrasound: a meta-analysis and systematic review . World J Gastroenterol 2008 ; 14 : 1479 – 1490 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Gillies RS , Middleton MR, Han C, Marshall RE, Maynard ND, Bradley KM et al. Role of positron emission tomography–computed tomography in predicting survival after neoadjuvant chemotherapy and surgery for oesophageal adenocarcinoma . Br J Surg 2012 ; 99 : 239 – 245 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Richardson JR , Khan OA. In patients with radiologically-staged resectable oesophago-gastric junctional tumours, is diagnostic laparoscopy useful as an additional staging procedure? Int J Surg 2012 ; 10 : 198 – 202 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Gillies RS , Middleton MR, Maynard ND, Bradley KM, Gleeson FV. Additional benefit of 18F-fluorodeoxyglucose integrated positron emission tomography/computed tomography in the staging of oesophageal cancer . Eur Radiol 2011 ; 21 : 274 – 280 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Greene FL . The American Joint Committee on Cancer: updating the strategies in cancer staging . Bull Am Coll Surg 2002 ; 87 : 13 – 15 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 10 Rice TW , Blackstone EH, Rusch VW. 7th edition of the AJCC Cancer Staging Manual: esophagus and esophagogastric junction . Ann Surg Oncol 2010 ; 17 : 1721 – 1724 . Google Scholar Crossref Search ADS PubMed WorldCat 11 UK EUS Users Group , British Society of Gastroenterology. Staging of Oesophago-gastric Carcinoma by Endoscopic Ultrasonography: Guidance and Minimum Standards . British Society of Gastroenterology : London , 2004 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 12 Cairns SR , Scholefield JH, Steele RJ, Dunlop MG, Thomas HJ, Evans GD et al. Guidelines for colorectal cancer screening and surveillance in moderate and high risk groups (update from 2002) . Gut 2010 ; 59 : 666 – 689 . Google Scholar Crossref Search ADS PubMed WorldCat 13 R Foundation for Statistical Computing . R: a Language and Environment for Statistical Computing . R Foundation for Statistical Computing : Vienna , 2013 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 14 Bland JM , Altman DG. Multiple significance tests: the Bonferroni method . BMJ 1995 ; 310 : 170 . Google Scholar Crossref Search ADS PubMed WorldCat 15 Pauker SG , Kassirer JP. The threshold approach to clinical decision making . N Engl J Med 1980 ; 302 : 1109 – 1117 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Pech O , May A, Manner H, Behrans A, Pohl J, Weferling M et al. Long-term efficacy and safety of endoscopic resection for patients with mucosal adenocarcinoma of the esophagus . Gastroenterology 2014 ; 146 : 652 – 660 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Medical Research Council Oesophageal Cancer Working Group . Surgical resection with or without preoperative chemotherapy in oesophageal cancer: a randomised controlled trial . Lancet 2002 ; 359 : 1727 – 1733 . Crossref Search ADS PubMed WorldCat 18 Committee to Assess Health Risks from Exposure to Low Levels of Ionizing Radiation; Board on Radiation Effects Research; Division on Earth and Life Studies; National Research Council . Health Risks from Exposure to Low Levels of Ionizing Radiation: BEIR VII, Phase 2 . National Academies Press : Washington, DC , 2006 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 19 Rouvelas I , Zeng W, Lindblad M, Viklund P, Ye W, Lagergren J. Survival after surgery for oesophageal cancer: a population-based study . Lancet Oncol 2005 ; 6 : 864 – 870 . Google Scholar Crossref Search ADS PubMed WorldCat 20 Jamieson GG , Mathew G, Ludemann R, Wayman J, Myers JC, Devitt PG. Postoperative mortality following oesophagectomy and problems in reporting its rate . Br J Surg 2004 ; 91 : 2983 – 2990 . Google Scholar OpenURL Placeholder Text WorldCat 21 Findlay JM , Tilson RC, Harikrishnan A, Sqromo B, Marshall RE, Maynard ND et al. Attempted validation of the NUn score and inflammatory markers as predictors of esophageal anastomotic leak and major complications . Dis Esophagus 2014 ; [Epub ahead of print]. 22 Nelson LM , Bloch DA, Longstreth WT, Shi H. Recursive partitioning for the identification of disease risk subgroups: a case–control study of subarachnoid hemorrhage . J Clin Epidemiol 1998 ; 51 : 199 – 209 . Google Scholar Crossref Search ADS PubMed WorldCat 23 Tu JV . Advantages and disadvantages of using artificial neural networks versus logistic regression for predicting medical outcomes . J Clin Epidemiol 1996 ; 49 : 1225 – 1231 . Google Scholar Crossref Search ADS PubMed WorldCat 24 Steyerberg EW , Harrell FE Jr, Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis . J Clin Epidemiol 2001 ; 54 : 774 – 781 . Google Scholar Crossref Search ADS PubMed WorldCat 25 Harrell FE RMS: Regression Modeling Strategies. R package version 4.1-0 . R Foundation for Statistical Computing : Vienna , 2013 . http://CRAN.R-project.org/package=rms [accessed 12 July 2015]. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 26 Therneau T , Atkinson B, Ripley B rpart: Recursive Partitioning. R package version 4.1-4 . R Foundation for Statistical Computing : Vienna , 2013 . http://CRAN.R-project.org/package=rpart [accessed 12 July 2015]. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 27 Liaw A , Wiener M. Classification and regression by randomForest . R News 2002 ; 2 : 18 – 22 . Google Scholar OpenURL Placeholder Text WorldCat 28 Venables WN , Ripley BD. Modern Applied Statistics with S (4th edn). Springer : New York , 2002 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 29 Kuhn M , Wing J, Weston S, Williams A, Keefer C, Engelhardt A et al. Caret: Classification and Regression Training. R package version 6.0-35 . R Foundation for Statistical Computing : Vienna , 2014 . http://CRAN.R-project.org/package=caret [accessed 12 July 2015]. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 30 Gevrey M , Dimopoulos I, Lek S. Review and comparison of methods to study the contribution of variables in artificial neural network models . Ecological Modelling 2003 ; 160 : 249 – 264 . Google Scholar Crossref Search ADS WorldCat 31 Stock C , Hielscher T. DTComPair: Comparison of Binary Diagnostic Tests in a Paired Study Design. R package version 1.0.3 . R Foundation for Statistical Computing : Vienna , 2014 . http://CRAN.R-project.org/package=DTComPair [accessed 12 July 2015]. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 32 Vickers AJ , Elkin EB. Decision curve analysis: a novel method for evaluating prediction models . Med Decis Making 2006 ; 26 : 565 – 574 . Google Scholar Crossref Search ADS PubMed WorldCat 33 Peirce CS . The numerical measure of the success of predictions . Science 1884 ; 4 : 453 – 454 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 34 Lowe VJ , Booya F, Fletcher JG, Nathan M, Jensen E, Mullan B et al. Comparison of positron emission tomography, computed tomography, and endoscopic ultrasound in the initial staging of patients with esophageal cancer . Mol Imaging Biol 2005 ; 7 : 422 – 430 . Google Scholar Crossref Search ADS PubMed WorldCat 35 van Vliet EP , Heijenbrok-Kal MH, Hunink MG, Kuipers EJ, Siersema PD. Staging investigations for oesophageal cancer: a meta-analysis . Br J Cancer 2008 ; 98 : 547 – 557 . Google Scholar Crossref Search ADS PubMed WorldCat Author notes Presented in part to meetings of the Association of Upper Gastrointestinal Surgeons of Great Britain and Ireland, Newcastle upon Tyne, UK, September 2013, the Society of Academic and Research Surgery, Cambridge, UK, January 2014, and the Association of Surgeons of Great Britain and Ireland, Harrogate, UK, May 2014; published in abstract form as Br J Surg 2013; 100(Suppl 8): 44, Br J Surg 2014; 101(Suppl 4): 18 and Br J Surg 2015; 102(Suppl 1): 34 © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd

Lee, J H; Nam, B-H; Ryu, K W; Ryu, S Y; Park, Y K; Kim, S; Kim, Y W

doi: 10.1002/bjs.9902pmid: 26398912

Abstract Background The aim of this study was to compare the results of laparoscopy-assisted total gastrectomy with those of open total gastrectomy for early gastric cancer. Methods Patients with gastric cancer who underwent total gastrectomy with curative intent in three Korean tertiary hospitals between January 2003 and December 2010 were included in this multicentre, retrospective, propensity score-matched cohort study. Cox proportional hazards regression models were used to evaluate the association between operation method and survival. Results A total of 753 patients with early gastric cancer were included in the study. There were no significant differences in the matched cohort for overall survival (hazard ratio (HR) for laparoscopy-assisted versus open total gastrectomy 0·96, 95 per cent c.i. 0·57 to 1·65) or recurrence-free survival (HR 2·20, 0·51 to 9·52). The patterns of recurrence were no different between the two groups. The severity of complications, according to the Clavien–Dindo classification, was similar in both groups. The most common complications were anastomosis-related in the laparoscopy-assisted group (8·0 per cent versus 4·2 per cent in the open group; P = 0·015) and wound-related in the open group (1·6 versus 5·6 per cent respectively; P = 0·003). Postoperative death was more common in the laparoscopy-assisted group (1·6 versus 0·2 per cent; P = 0·045). Conclusion Laparoscopy-assisted total gastrectomy for early gastric cancer is feasible in terms of long-term results, including survival and recurrence. However, a higher postoperative mortality rate and an increased risk of anastomotic leakage after laparoscopic-assisted total gastrectomy are of concern. Introduction Laparoscopy-assisted distal gastrectomy is considered safe and the patients' quality of life has been reported to be superior to that of patients who undergo open distal gastrectomy1–3. With the development of better laparoscopic instruments and accumulated experience of laparoscopic surgery, laparoscopy-assisted distal gastrectomy has become a widely used treatment worldwide for patients with early gastric cancer3,4. Recent data have shown that long-term outcomes of laparoscopy-assisted distal gastrectomy are comparable to those of open distal gastrectomy5. In contrast to laparoscopy-assisted distal gastrectomy, laparoscopy-assisted total gastrectomy is not popular because of its technical complexity6. The main obstacle is the difficult reconstruction. Some investigators7,8 have reported that laparoscopy-assisted total gastrectomy is feasible in terms of safety and survival, but the sample sizes in these studies were so small that the results were not conclusive. Recent meta-analyses9,10 of laparoscopy-assisted total gastrectomy have shown similar results, but there have been no high-quality controlled clinical trials and the existing studies have limitations of potential bias and heterogeneity1. It would be difficult to recruit patients for a high-quality controlled clinical trial comparing laparoscopy-assisted with open total gastrectomy. Moreover, there is a high possibility of selection bias because of the technical complexity of laparoscopy-assisted total gastrectomy. The propensity score matching method might be an alternative approach to obtain matched results11. This study used the propensity score matching method to compare survival of patients with gastric cancer who underwent laparoscopy-assisted total gastrectomy with that of patients who had open total gastrectomy. Methods A retrospective review of prospectively maintained gastric cancer databases was carried out. The study cohort included patients with early gastric cancer who were treated by open or laparoscopic total gastrectomy between January 2003 and December 2010 at one of three institutions in Korea: Chunnam Whasoon Hospital, National Cancer Centre and Samsung Medical Centre. Patients who had a histologically confirmed gastric adenocarcinoma, a mucosal or submucosal tumour, and newly diagnosed cancer without previous treatments were included in the analysis. All information was obtained with appropriate institutional review board waivers and data were collected without revealing any personal information. Patient characteristics and clinical data Characteristics of patients were obtained from a review of medical records. Demographic data included age, sex and body mass index. Clinicopathological characteristics included tumour location, tumour size, differentiation, gross type, depth of invasion, lymphatic invasion, lymph node metastasis, stage at diagnosis and operation method. Stage at diagnosis was determined according to the sixth edition of the International Union Against Cancer (UICC)/American Joint Committee on Cancer (AJCC) classification system12. In patients with multiple synchronous gastric cancers, the lesion with the deepest infiltration of the gastric wall was considered to be the main lesion and any others were regarded as accessory lesions. The clinicopathological characteristics of the main lesion were used in the analysis. Operative procedures and adjuvant treatment A total of 12 gastric cancer surgeons were involved in the study; all staff surgeons had previously conducted more than 200 open gastric cancer operations. Gastrectomy was performed with a tumour-free margin of 2 cm. The extent of lymph node dissection was determined using the recommendations of the Japanese Research Society for Gastric Carcinoma13. After laparotomy or laparoscopy, surgeons examined the intra-abdominal cavities, and inspected the peritoneum, diaphragm, liver capsule and pelvic cavity. All patients enrolled in the present study underwent gastrectomy with D1 + β or more lymph node dissection. Roux-en-Y oesophagojejunostomy was performed using a 25-mm circular stapler after open total gastrectomy. Extracorporeal Roux-en-Y oesophagojejunostomy using a 25-mm circular stapler or intracorporeal Roux-en-Y oesophagojejunostomy using an OrVilTM device (Covidien, Mansfield, Massachusetts, USA) was carried out after laparoscopy-assisted total gastrectomy. An extracorporeal Roux-en-Y oesophagojejunostomy was made using a circular stapler in the same manner as for open total gastrectomy, after creating a 5–6-cm minilaparotomy on the epigastrium5. A jejunojejunostomy was prepared manually through the minilaparotomy wound. Intracorporeal Roux-en-Y oesophagojejunostomy using the OrVilTM stapler was done as described previously14. The OrVil™ anvil was passed transorally by the anaesthetist through the larynx to the stapled oesophageal stump. A small hole was created in the corresponding position in the stapled oesophageal stump, and the tube was pulled through the hole into the abdominal cavity until the white plastic ring was fully revealed. The connecting thread was cut and the orogastric tube was disconnected from the anvil; then the spike was connected to the oesophageal anvil to create an oesophagojejunal anastomosis. A 3-cm longitudinal minilaparotomy incision was then made at the midline of the epigastric region. After a wound protector had been placed, the stomach was delivered through the incision for pathological examination. The jejunum was cut off at 15 cm from the ligament of Treitz and the stump of proximal jejunum was sutured. A 25-mm circular stapler was inserted into the distal limb of the jejunum and introduced into the abdominal cavity after a second pneumoperitoneum had been created. The anvil and circular stapler were connected and an end-to-side oesophageal jejunostomy anastomosis was fashioned under direct laparoscopic view. After restoring the continuity of the oesophagus and jejunum, the Peterson defect was closed. Because all patients included in this study were diagnosed with early gastric cancer, no patient was received adjuvant chemotherapy. Follow-up schedule Follow-up was conducted according to accepted clinical practice at each institution. In general, follow-up consisted of abdominopelvic CT every 6 months for 5 years after surgery and oesophagogastroduodenoscopy annually for 5 years. Cancer recurrence was diagnosed when there was positive radiological evidence. Patients were followed until death or until the cut-off date of 31 December 2013. Those lost to follow-up and operative deaths were treated as censored. Outcome data The primary endpoints of the study were death and tumour recurrence. Deaths from any cause and disease-related deaths (defined as death from recurrence) were analysed. Peritoneal recurrences were defined as carcinomatosis or ovarian metastasis. All recurrences were documented pathologically and/or by radiological imaging. Morbidity was defined as complications that required an extended hospital stay or readmission, and was graded according to the Clavien–Dindo classification15. Overall survival was calculated from the time of surgery to death from any cause. Recurrence-free survival (RFS) was calculated from the time of surgery to tumour recurrence or death with evidence of recurrence. For RFS, patients who died without known tumour recurrence were censored at the last documented evaluation. Statistical analysis Continuous variables were compared using the t test or Mann–Whitney U test, and categorical variables were analysed using the χ2 test. To make this study as close as possible to a randomized clinical trial setting, propensity score matching was employed11. To generate the propensity score, a multiple logistic regression model was used. The dependent variable was the treatment received, and variables included in the multivariable model were: age, sex, tumour differentiation, depth of tumour invasion, tumour size and presence of lymphatic invasion. Using the SAS® Greedy 5 → 1 digit match macro (SAS Institute, Cary, North Carolina, USA), propensity score-matched pairs were created without replacement (1 : 2 match for laparoscopic : open). A mixed-effect model was used, where matched-pair effects were considered as random. Statistical significance and the effect of treatment on outcomes were estimated using appropriate statistical methods for matched data. In the propensity score-matched cohort, continuous variables were compared using the mixed linear model and categorical variables by conditional logistic regression. The risks of death, recurrence and metachronous gastric cancer were compared using Cox proportional hazards regression models with robust standard errors that accounted for the clustering of matched pairs. The proportional hazards assumption was confirmed by examination of log (−log [survival]) curves; no relevant violations were found. Survival curves were generated by the Kaplan–Meier method and analysed using the log rank test. Statistical significance was set at P < 0·050. Results From January 2003 to December 2010, 1513 patients with early gastric cancer underwent total gastrectomy at one of three institutions. Of these, patients who had another cancer (63) or needed combined resection other than stomach (157) were excluded; the final study population comprised 1293 patients (Fig. 1). After propensity score matching, a total of 753 patients were included in this study. Clinical and pathological characteristics of the tumours are shown in Table 1. Median age was 59 (range 23–84) years and 479 patients (63·6 per cent) were men. All tumours were located in the upper or middle third of the stomach. Sixty-two tumours (8·2 per cent) had metastasized to lymph nodes. There was no significant difference in baseline characteristics other than tumour size between the open and laparoscopy-assisted total gastrectomy groups. Fig. 1 Open in new tabDownload slide Study flow chart Table 1 Baseline characteristics of propensity score-matched patients who underwent open or laparoscopy-assisted total gastrectomy . Open group (n = 502) . Laparoscopy group (n = 251) . P* . Demographic characteristics Age (years) Mean(s.d.) 57·6(11·6) 58·4(12·7) Median (range) 57 (23–84) 58 (23–84) 0·392† Sex ratio (M : F) 319 : 183 160 : 91 0·957 Body mass index (kg/m2) Mean(s.d.) 23·1(11·6) 23·1(3·0) Median (range) 23·6 (15–37) 23·6 (15–37) 0·768† ASA fitness grade 0·480 I 206 (41·0) 92 (36·7) II 278 (55·4) 148 (59·0) III 18 (3·6) 11 (4·4) Tumour characteristics Location 0·081 Upper third 371 (73·9) 200 (79·7) Middle third 131 (26·1) 51 (20·3) Size (cm) Mean(s.d.) 3·9(2·6) 3·3(2·4) Median (range) 2·4 (0·1–9·0) 2·4 (0·1–16) 0·004† Histology 0·796 Differentiated 229 (45·6) 112 (44·6) Undifferentiated 273 (54·4) 139 (55·4) Morphology 0·243 Ulcer 304 (60·6) 163 (64·9) No ulcer 198 (39·4) 88 (35·1) No lymphovascular invasion 429 (85·5) 219 (87·3) 0·679 Depth of tumour invasion 0·834 Mucosa 208 (41·4) 106 (42·2) Submucosa 294 (58·6) 145 (57·8) Lymph node metastasis 42 (8·4) 20 (8·0) 0·851 . Open group (n = 502) . Laparoscopy group (n = 251) . P* . Demographic characteristics Age (years) Mean(s.d.) 57·6(11·6) 58·4(12·7) Median (range) 57 (23–84) 58 (23–84) 0·392† Sex ratio (M : F) 319 : 183 160 : 91 0·957 Body mass index (kg/m2) Mean(s.d.) 23·1(11·6) 23·1(3·0) Median (range) 23·6 (15–37) 23·6 (15–37) 0·768† ASA fitness grade 0·480 I 206 (41·0) 92 (36·7) II 278 (55·4) 148 (59·0) III 18 (3·6) 11 (4·4) Tumour characteristics Location 0·081 Upper third 371 (73·9) 200 (79·7) Middle third 131 (26·1) 51 (20·3) Size (cm) Mean(s.d.) 3·9(2·6) 3·3(2·4) Median (range) 2·4 (0·1–9·0) 2·4 (0·1–16) 0·004† Histology 0·796 Differentiated 229 (45·6) 112 (44·6) Undifferentiated 273 (54·4) 139 (55·4) Morphology 0·243 Ulcer 304 (60·6) 163 (64·9) No ulcer 198 (39·4) 88 (35·1) No lymphovascular invasion 429 (85·5) 219 (87·3) 0·679 Depth of tumour invasion 0·834 Mucosa 208 (41·4) 106 (42·2) Submucosa 294 (58·6) 145 (57·8) Lymph node metastasis 42 (8·4) 20 (8·0) 0·851 Values in parentheses are percentages unless indicated otherwise. ASA, American Society of Anesthesiologists. * χ2 test, except † Mann–Whitney U test. Open in new tab Table 1 Baseline characteristics of propensity score-matched patients who underwent open or laparoscopy-assisted total gastrectomy . Open group (n = 502) . Laparoscopy group (n = 251) . P* . Demographic characteristics Age (years) Mean(s.d.) 57·6(11·6) 58·4(12·7) Median (range) 57 (23–84) 58 (23–84) 0·392† Sex ratio (M : F) 319 : 183 160 : 91 0·957 Body mass index (kg/m2) Mean(s.d.) 23·1(11·6) 23·1(3·0) Median (range) 23·6 (15–37) 23·6 (15–37) 0·768† ASA fitness grade 0·480 I 206 (41·0) 92 (36·7) II 278 (55·4) 148 (59·0) III 18 (3·6) 11 (4·4) Tumour characteristics Location 0·081 Upper third 371 (73·9) 200 (79·7) Middle third 131 (26·1) 51 (20·3) Size (cm) Mean(s.d.) 3·9(2·6) 3·3(2·4) Median (range) 2·4 (0·1–9·0) 2·4 (0·1–16) 0·004† Histology 0·796 Differentiated 229 (45·6) 112 (44·6) Undifferentiated 273 (54·4) 139 (55·4) Morphology 0·243 Ulcer 304 (60·6) 163 (64·9) No ulcer 198 (39·4) 88 (35·1) No lymphovascular invasion 429 (85·5) 219 (87·3) 0·679 Depth of tumour invasion 0·834 Mucosa 208 (41·4) 106 (42·2) Submucosa 294 (58·6) 145 (57·8) Lymph node metastasis 42 (8·4) 20 (8·0) 0·851 . Open group (n = 502) . Laparoscopy group (n = 251) . P* . Demographic characteristics Age (years) Mean(s.d.) 57·6(11·6) 58·4(12·7) Median (range) 57 (23–84) 58 (23–84) 0·392† Sex ratio (M : F) 319 : 183 160 : 91 0·957 Body mass index (kg/m2) Mean(s.d.) 23·1(11·6) 23·1(3·0) Median (range) 23·6 (15–37) 23·6 (15–37) 0·768† ASA fitness grade 0·480 I 206 (41·0) 92 (36·7) II 278 (55·4) 148 (59·0) III 18 (3·6) 11 (4·4) Tumour characteristics Location 0·081 Upper third 371 (73·9) 200 (79·7) Middle third 131 (26·1) 51 (20·3) Size (cm) Mean(s.d.) 3·9(2·6) 3·3(2·4) Median (range) 2·4 (0·1–9·0) 2·4 (0·1–16) 0·004† Histology 0·796 Differentiated 229 (45·6) 112 (44·6) Undifferentiated 273 (54·4) 139 (55·4) Morphology 0·243 Ulcer 304 (60·6) 163 (64·9) No ulcer 198 (39·4) 88 (35·1) No lymphovascular invasion 429 (85·5) 219 (87·3) 0·679 Depth of tumour invasion 0·834 Mucosa 208 (41·4) 106 (42·2) Submucosa 294 (58·6) 145 (57·8) Lymph node metastasis 42 (8·4) 20 (8·0) 0·851 Values in parentheses are percentages unless indicated otherwise. ASA, American Society of Anesthesiologists. * χ2 test, except † Mann–Whitney U test. Open in new tab Effect of operation method on short-term outcomes The mean number of dissected lymph nodes was lower and the mean duration of operation was longer in the laparoscopy-assisted group (Table 2). Table 2 Short-term surgical outcomes after open and laparoscopic surgery among the propensity score-matched patients . Open group (n = 502) . Laparoscopy group (n = 251) . P† . No. of dissected lymph nodes Mean(s.d.) 45·7(17·0) 41·8(16·8) 0·003‡ Median (range) 35 (4–103) 35 (3–100) Duration of operation (min) Mean(s.d.) 185·8(63·0) 237·6(81·4) < 0·001‡ Median (range) 184 (84–490) 225 (70–555) Complications Clavien–Dindo classification 0·118 None 414 (82·5) 197 (78·5) I 12 (2·4) 12 (4·8) II 29 (5·8) 15 (6·0) IIIA 37 (7·4) 15 (6·0) IIIB 8 (1·6) 7 (2·8) IV 1 (0·2) 1 (0·4) V 1 (0·2) 4 (1·6) Site of complication 0·007 None 414 (82·5) 197 (78·5) Anastomosis 21 (4·2) 20 (8·0) Wound 28 (5·6) 4 (1·6) Intra-abdominal 11 (2·2) 9 (3·6) Other 18 (3·6) 15 (6·0) Length of hospital stay (days)* 9·4(0·5) 11·2(0·6) 0·001‡ No. receiving ICU care 33 (7·2) 13 (13·1) 0·054 Length of ICU stay (days)* 0·2(1·0) 0·6(3·3) 0·083‡ . Open group (n = 502) . Laparoscopy group (n = 251) . P† . No. of dissected lymph nodes Mean(s.d.) 45·7(17·0) 41·8(16·8) 0·003‡ Median (range) 35 (4–103) 35 (3–100) Duration of operation (min) Mean(s.d.) 185·8(63·0) 237·6(81·4) < 0·001‡ Median (range) 184 (84–490) 225 (70–555) Complications Clavien–Dindo classification 0·118 None 414 (82·5) 197 (78·5) I 12 (2·4) 12 (4·8) II 29 (5·8) 15 (6·0) IIIA 37 (7·4) 15 (6·0) IIIB 8 (1·6) 7 (2·8) IV 1 (0·2) 1 (0·4) V 1 (0·2) 4 (1·6) Site of complication 0·007 None 414 (82·5) 197 (78·5) Anastomosis 21 (4·2) 20 (8·0) Wound 28 (5·6) 4 (1·6) Intra-abdominal 11 (2·2) 9 (3·6) Other 18 (3·6) 15 (6·0) Length of hospital stay (days)* 9·4(0·5) 11·2(0·6) 0·001‡ No. receiving ICU care 33 (7·2) 13 (13·1) 0·054 Length of ICU stay (days)* 0·2(1·0) 0·6(3·3) 0·083‡ Values in parentheses are percentages unless indicated otherwise; * values are mean(s.d.). ICU, intensive care unit. † χ2 test, except ‡ t test. Open in new tab Table 2 Short-term surgical outcomes after open and laparoscopic surgery among the propensity score-matched patients . Open group (n = 502) . Laparoscopy group (n = 251) . P† . No. of dissected lymph nodes Mean(s.d.) 45·7(17·0) 41·8(16·8) 0·003‡ Median (range) 35 (4–103) 35 (3–100) Duration of operation (min) Mean(s.d.) 185·8(63·0) 237·6(81·4) < 0·001‡ Median (range) 184 (84–490) 225 (70–555) Complications Clavien–Dindo classification 0·118 None 414 (82·5) 197 (78·5) I 12 (2·4) 12 (4·8) II 29 (5·8) 15 (6·0) IIIA 37 (7·4) 15 (6·0) IIIB 8 (1·6) 7 (2·8) IV 1 (0·2) 1 (0·4) V 1 (0·2) 4 (1·6) Site of complication 0·007 None 414 (82·5) 197 (78·5) Anastomosis 21 (4·2) 20 (8·0) Wound 28 (5·6) 4 (1·6) Intra-abdominal 11 (2·2) 9 (3·6) Other 18 (3·6) 15 (6·0) Length of hospital stay (days)* 9·4(0·5) 11·2(0·6) 0·001‡ No. receiving ICU care 33 (7·2) 13 (13·1) 0·054 Length of ICU stay (days)* 0·2(1·0) 0·6(3·3) 0·083‡ . Open group (n = 502) . Laparoscopy group (n = 251) . P† . No. of dissected lymph nodes Mean(s.d.) 45·7(17·0) 41·8(16·8) 0·003‡ Median (range) 35 (4–103) 35 (3–100) Duration of operation (min) Mean(s.d.) 185·8(63·0) 237·6(81·4) < 0·001‡ Median (range) 184 (84–490) 225 (70–555) Complications Clavien–Dindo classification 0·118 None 414 (82·5) 197 (78·5) I 12 (2·4) 12 (4·8) II 29 (5·8) 15 (6·0) IIIA 37 (7·4) 15 (6·0) IIIB 8 (1·6) 7 (2·8) IV 1 (0·2) 1 (0·4) V 1 (0·2) 4 (1·6) Site of complication 0·007 None 414 (82·5) 197 (78·5) Anastomosis 21 (4·2) 20 (8·0) Wound 28 (5·6) 4 (1·6) Intra-abdominal 11 (2·2) 9 (3·6) Other 18 (3·6) 15 (6·0) Length of hospital stay (days)* 9·4(0·5) 11·2(0·6) 0·001‡ No. receiving ICU care 33 (7·2) 13 (13·1) 0·054 Length of ICU stay (days)* 0·2(1·0) 0·6(3·3) 0·083‡ Values in parentheses are percentages unless indicated otherwise; * values are mean(s.d.). ICU, intensive care unit. † χ2 test, except ‡ t test. Open in new tab The overall morbidity rate was 18·9 per cent. The most common type of morbidity was wound-related complications, including infection (23 patients, 3·1 per cent), dehiscence (6 patients, 0·8 per cent) and hernia (3 patients, 0·4 per cent), followed by anastomosis-related complications, including stricture (22 patients, 2·9 per cent), leakage (11 patients, 1·5 per cent) and bleeding of the oesophagojejunostomy (8 patients, 1·1 per cent). Patterns of complications were different in the two groups. The most common complication in the open group was wound infection, whereas stricture of the oesophagojejunostomy was the most common complication after laparoscopy-assisted total gastrectomy (Table 2). Comparison of the severity of complications, according to the Clavien–Dindo classification, revealed no difference between the two groups. There were five operation-related deaths overall (1 in the open group, 4 in the laparoscopy group). Operation-related deaths were significantly more common after laparoscopy-assisted total gastrectomy (1·6 versus 0·2 per cent; P = 0·045). Two patients died from internal herniation of the small bowel 22 and 27 months after operation. Other causes of death were aspiration pneumonia, luminal bleeding and anastomosis leakage. Effect of operation method on long-term outcomes Thirty patients (4·0 per cent) were not available at the time of last follow-up. Median (i.q.r.) follow-up was 55 (37–69) and 58 (44–67) months for the open and laparoscopy-assisted groups respectively. During follow-up, 33 patients experienced recurrences (Table S1, supporting information). There was no difference between the groups in the rate or patterns of recurrence. Liver metastases were the most common form of recurrence in both groups, followed by distant lymph node metastases in the open group. The risk of death did not differ significantly between the two groups (hazard ratio (HR) for laparoscopy-assisted versus open total gastrectomy 0·96, 95 per cent c.i. 0·57 to 1·65; P = 0·894). The 5-year overall survival rate was 99·7 per cent after open and 99·0 per cent after laparoscopy-assisted procedures (Fig. S1, supporting information). There was no significant difference in the risk of recurrence during follow-up (HR 2·20, 0·51 to 9·52; P = 0·291). The 5-year RFS rate was 96·5 per cent in the open total gastrectomy group and 92·6 per cent in the laparoscopy-assisted group (Fig. S2, supporting information). Discussion This study compared short- and long-term results after open and laparoscopy-assisted total gastrectomy for early gastric cancer. A recent large-scale study5 performed after laparoscopic gastrectomy in Korea showed similar long-term outcomes to open gastrectomy. However, the exact number of patients who underwent laparoscopy-assisted total gastrectomy was not presented and most patients underwent laparoscopy-assisted distal gastrectomy, so interpretation of long-term outcomes after laparoscopy-assisted versus open total gastrectomy was difficult. Favourable long-term outcomes have been reported in the limited number of studies comparing laparoscopy-assisted with open total gastrectomy for early gastric cancer. Five-year overall survival rates following laparoscopy-assisted total gastrectomy were reported to be 98·9 and 91·5 per cent in previous studies6,16. Similarly, the 5-year overall survival rate of laparoscopy-assisted total gastrectomy was 99·0 per cent in the present study. However, previous reports7,8,16 of laparoscopy-assisted total gastrectomy have limited data, with considerable selection bias for long-term outcome. Although survival after laparoscopy-assisted total gastrectomy was comparable to that following open total gastrectomy and the median number of dissected lymph nodes was similar, the operating time was longer for laparoscopy-assisted total gastrectomy. This might have been due to the more difficult reconstruction and did not appear to increase the risk of complications other than those related to anastomosis. However, anastomotic complications are mostly serious and a frequent cause of delayed discharge or death. Comparison of complications according to the Clavien–Dindo classification showed that the severity of postoperative complications was similar after laparoscopy-assisted and open total gastrectomy. However, the pattern of complications was different; anastomosis-related complications, including leakage, stenosis and bleeding, were more common in the laparoscopy-assisted group. These complications require a longer hospital stay, explaining why the hospital stay was longer in the laparoscopy-assisted group in this study. These results differ from those of recent meta-analyses9,10 that showed a reduced risk of postoperative complications after laparoscopy-assisted total gastrectomy compared with open total gastrectomy, similar to the risk after laparoscopy-assisted distal gastrectomy. Postoperative deaths were more common after laparoscopy-assisted than open total gastrectomy. Four of five postoperative deaths were associated with anastomosis-related complications, including bleeding, leakage, and internal hernia through the afferent limb. Use of effective methods to restore continuity of the oesophagus and jejunum may decrease the postoperative mortality rate after laparoscopy-assisted total gastrectomy. There is currently no standard method for restoration of continuity of the oesophagus and jejunum. Recently reported17,18 reconstruction methods after laparoscopic surgery might help in reducing such anastomosis-related complications. For example, Okabe and colleagues17 reported that anastomosis using linear staplers might reduce anastomosis-related complications. It is evident that reconstruction after laparoscopy-assisted total gastrectomy is more difficult than that after open total gastrectomy. The major limitation of this study is that it is retrospective and as such the treatment strategy was not based on random assignment. As a result, selection bias might have occurred in choosing the treatment modality even though a propensity-matched cohort was studied. Another limitation is that the level of experience in laparoscopy-assisted total gastrectomy was different from that in open total gastrectomy for most surgeons. Data were analysed from the initial experiences of laparoscopy-assisted total gastrectomy and this might have influenced the results. Finally, other important outcomes such as quality of life were not compared in this study. Because survival rates in both groups were similar, other outcomes might more meaningfully inform the decision regarding the method of operation. Nonetheless, the present data indicate that laparoscopy-assisted total gastrectomy is not inferior to open total gastrectomy in terms of long-term outcomes. However, considering the higher anastomosis-related complication rate, further development of anastomotic methods and technological innovations in laparoscopy-assisted total gastrectomy are needed. Disclosure The authors declare no conflict of interest. References 1 Kim YW , Baik YH, Yun YH, Nam BH, Kim DH, Choi IJ et al. Improved quality of life outcomes after laparoscopy-assisted distal gastrectomy for early gastric cancer: results of a prospective randomized clinical trial . Ann Surg 2008 ; 248 : 721 – 727 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Takiguchi S , Fujiwara Y, Yamasaki M, Miyata H, Nakajima K, Sekimoto M et al. Laparoscopy-assisted distal gastrectomy versus open distal gastrectomy. A prospective randomized single-blind study . World J Surg 2013 ; 37 : 2379 – 2386 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Kim HH , Hyung WJ, Cho GS, Kim MC, Han SU, Kim W et al. Morbidity and mortality of laparoscopic gastrectomy versus open gastrectomy for gastric cancer: an interim report – a phase III multicenter, prospective, randomized Trial (KLASS Trial) . Ann Surg 2010 ; 251 : 417 – 420 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Zeng YK , Yang ZL, Peng JS, Lin HS, Cai L. Laparoscopy-assisted versus open distal gastrectomy for early gastric cancer: evidence from randomized and nonrandomized clinical trials . Ann Surg 2012 ; 256 : 39 – 52 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Kim HH , Han SU, Kim MC, Hyung WJ, Kim W, Lee HJ et al. Long-term results of laparoscopic gastrectomy for gastric cancer: a large-scale case–control and case-matched Korean multicenter study . J Clin Oncol 2014 ; 32 : 627 – 633 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Jeong GA , Cho GS, Kim HH, Lee HJ, Ryu SW, Song KY et al. Laparoscopy-assisted total gastrectomy for gastric cancer: a multicenter retrospective analysis . Surgery 2009 ; 146 : 469 – 474 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Lee MS , Lee JH, Park do J, Lee HJ, Kim HH, Yang HK. Comparison of short- and long-term outcomes of laparoscopic-assisted total gastrectomy and open total gastrectomy in gastric cancer patients . Surg Endosc 2013 ; 27 : 2598 – 2605 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Wada N , Kurokawa Y, Takiguchi S, Takahashi T, Yamasaki M, Miyata H et al. Feasibility of laparoscopy-assisted total gastrectomy in patients with clinical stage I gastric cancer . Gastric Cancer 2014 ; 17 : 137 – 140 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Chen K , Xu XW, Zhang RC, Pan Y, Wu D, Mou YP. Systematic review and meta-analysis of laparoscopy-assisted and open total gastrectomy for gastric cancer . World J Gastroenterol 2013 ; 19 : 5365 – 5376 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Shen H , Shan C, Liu S, Qiu M. Laparoscopy-assisted versus open total gastrectomy for gastric cancer: a meta-analysis . J Laparoendosc Adv Surg Tech A 2013 ; 23 : 832 – 840 . Google Scholar Crossref Search ADS PubMed WorldCat 11 D'Agostino RB Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group . Stat Med 1998 ; 17 : 2265 – 2281 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Sobin LH , Wittekind C (eds). International Union Against Cancer (UICC) TNM Classification of Malignant Tumours (6th edn). Wiley-Liss : New York , 2002 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 13 Japanese Gastric Cancer Association . Japanese gastric cancer treatment guidelines 2010 (ver. 3) . Gastric Cancer 2011 ; 14 : 113 – 123 . Crossref Search ADS PubMed WorldCat 14 Jeong O , Park YK. Intracorporeal circular stapling esophagojejunostomy using the transorally inserted anvil (OrVil) after laparoscopic total gastrectomy . Surg Endosc 2009 ; 23 : 2624 – 2630 . Google Scholar Crossref Search ADS PubMed WorldCat 15 Clavien PA , Barkun J, de Oliveira ML, Vauthey JN, Dindo D, Schulick RD et al. The Clavien–Dindo classification of surgical complications: five-year experience . Ann Surg 2009 ; 250 : 187 – 196 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Eom BW , Kim YW, Lee SE, Ryu KW, Lee JH, Yoon HM et al. Survival and surgical outcomes after laparoscopy-assisted total gastrectomy for gastric cancer: case–control study . Surg Endosc 2012 ; 26 : 3273 – 3281 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Okabe H , Obama K, Tsunoda S, Tanaka E, Sakai Y. Advantage of completely laparoscopic gastrectomy with linear stapled reconstruction: a long-term follow-up study . Ann Surg 2014 ; 259 : 109 – 116 . Google Scholar Crossref Search ADS PubMed WorldCat 18 Kim HS , Kim MG, Kim BS, Lee IS, Lee S, Yook JH et al. Comparison of totally laparoscopic total gastrectomy and laparoscopic-assisted total gastrectomy methods for the surgical treatment of early gastric cancer near the gastroesophageal junction . J Laparoendosc Adv Surg Tech A 2013 ; 23 : 204 – 210 . Google Scholar Crossref Search ADS PubMed WorldCat © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd

Hong, G-S; Schwandt, T; Stein, K; Schneiker, B; Kummer, M P; Heneka, M T; Kitamura, K; Kalff, J C; Wehner, S

doi: 10.1002/bjs.9907pmid: 26313905

Abstract Background The pathophysiology of adhesion formation after abdominal and pelvic surgery is still largely unknown. The aim of the study was to investigate the role of macrophage polarization and the effect of peroxisome proliferator-activated receptor (PPAR) γ stimulation on adhesion formation in an animal model. Methods Peritoneal adhesion formation was induced by the creation of ischaemic buttons within the peritoneal wall and the formation of a colonic anastomosis in wild-type, interleukin (IL) 10-deficient (IL-10−/−), IL-4-deficient (IL-4−/−) and CD11b-Cre/PPARγfl/fl mice. Adhesions were assessed at regular intervals, and cell preparations were isolated from ischaemic buttons and normal peritoneum. These samples were analysed for macrophage differentiation and its markers, and expression of cytokines by quantitative PCR, fluorescence microscopy, arginase activity and pathological examination. Some animals underwent pioglitazone (PPAR-γ agonist) or vehicle treatment to inhibit adhesion formation. Anastomotic healing was evaluated by bursting pressure measurement and collagen gene expression. Results Macrophage M2 marker expression and arginase activity were raised in buttons without adhesions compared with buttons with adhesions. IL-4−/− and IL-10−/− mice were not affected, whereas CD11b-Cre/PPARγfl/fl mice showed decreased arginase activity and increased adhesion formation. Perioperative pioglitazone treatment increased arginase activity and decreased adhesion formation in wild-type but not CD11b-Cre/PPARγfl/fl mice. Pioglitazone had no effect on anastomotic healing. Conclusion Endogenous macrophage-specific PPAR-γ signalling affected arginase activity and macrophage polarization, and counter-regulated peritoneal adhesion manifestation. Pharmacological PPAR-γ agonism induced a shift towards macrophage M2 polarization and ameliorated adhesion formation in a macrophage-dependent manner. Surgical relevance Postoperative adhesion formation is frequently seen after abdominal surgery and occurs in response to peritoneal trauma. The pathogenesis is still unknown but includes an imbalance in fibrinolysis, collagen production and inflammatory mechanisms. Little is known about the role of macrophages during adhesion formation. In an experimental model, macrophage M2 marker expression was associated with reduced peritoneal adhesion formation and involved PPAR-γ-mediated arginase activity. Macrophage-specific PPAR-γ deficiency resulted in reduced arginase activity and aggravated adhesion formation. Pioglitazone, a PPAR-γ agonist, induced M2 polarization and reduced postoperative adhesion formation without compromising anastomotic healing in mice. Pioglitazone ameliorated postoperative adhesion formation without compromising intestinal wound healing. Therefore, perioperative PPAR-γ agonism might be a promising strategy for prevention of adhesion formation after abdominal surgery. Introduction Intra-abdominal adhesions may cause bowel obstruction, chronic abdominal pain, infertility and dyspareunia. More than one-third of patients are readmitted to hospital because of adhesions within 10 years after open abdominal or pelvic surgery1. Adhesions may also be responsible for the higher complication rates observed in reoperations2 that may lead to poor quality of life. The pathophysiology of intra-abdominal adhesion formation is, however, still poorly understood3. It is now widely accepted that injury to the peritoneal surface and the subsequent wound healing processes contribute to adhesion formation. Peritoneal adhesions do not originate exclusively from dysregulated collagen production and fibrinolysis, but also involve inflammatory responses3. Recent studies4–7 have demonstrated the presence of leucocyte populations, including mast cells, T lymphocytes and macrophages, within adhesions. Macrophages probably play a key role, as systematic depletion of macrophages aggravated adhesion formation in a mouse model8. Macrophage function depends mainly on differentiation status. Cytokine and mediator expression profiles can be used to evaluate whether macrophage differentiation is towards a proinflammatory M1 phenotype, important during host defence, or an M2 phenotype, primarily involved in downregulation of inflammation, initiating wound repair and phagocytosis of apoptotic neutrophils and cell debris9. Levels of proinflammatory cytokines, such as tumour necrosis factor (TNF) α, interleukin (IL) 6 and interferon γ, have been shown to be raised in the M1 phenotype and in adhesive tissue10, but are also expressed to some extent within the M2 phenotype. The cytokine and marker panel indicating changes to the M2 phenotype includes mannose receptor (MR) 1 (CD206), chitinase 3-like protein 3 (YM1) and arginase (Arg) 19. Drivers of the M2 phenotype include Th2-derived IL-4 or IL-109 but also peroxisome proliferator-activated receptor (PPAR) γ11, a ligand-activated nuclear receptor with potent anti-inflammatory properties that can be activated by endogenous ligands, such as lipid mediators or synthetic ligands from the class of thiazolidinediones including pioglitazone. Although macrophage differentiation affects wound healing in several organs, such as skin, liver and lung12, the role of macrophage differentiation is still controversial. Some reports have shown a profibrotic impact by Arg-113 and MR-114 expressing macrophages in fibrotic tissue9, whereas others have described antifibrotic functions in animal models of cardiac fibrosis, Crohn's disease, renal fibrosis and arteriosclerosis15. Additionally, deletion of PPAR-γ in macrophages has been shown to suppress M2-like differentiation and to aggravate liver fibrosis in mice16. Here, the role of macrophage differentiation during intra-abdominal adhesion formation was investigated. Methods Experiments were performed with 6–8-week-old male wild-type (WT) C57BL/6 J (Janvier, Saint Berthevin, France), and IL-10−/− and IL-4−/− (The Jackson Laboratory, Bar Harbor, Maine, USA) mice. CD11b lineage (predominantly affecting macrophage/monocyte) depletion of PPAR-γ was analysed in CD11b-Cre/PPARγfl/fl mice. PPARγfl/fl mice, without any cell-specific Cre expression and normal PPAR-γ activity, were used as controls. All mice had a mean bodyweight of 20–25 g at the time of use. The study was approved by the committee for animal experiments of North-Rhine Westphalia and performed in accordance with federal law regarding the protection of animals. Animals were maintained in specific pathogen-free housing on a 12-h light–dark cycle, and had free access to commercially available rodent chow and tap water. They were grouped up to five per cage. Animals were observed for signs of pain (weight loss, quality of fur). No adverse advent was observed during the experiments. The manuscript was written according to the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines. All chemicals were purchased from Sigma (Taufkirchen, Germany) unless stated otherwise. Pioglitazone treatment Animals were gavaged daily with 1·25 mg pioglitazone (Import Gerke Pharma, Grevenbroich, Germany) dissolved in 0·5 per cent methylcellulose for 6 days, beginning 3 days before surgery. Control groups received 0·5 per cent methylcellulose as vehicle. Pioglitazone-treated animals and control groups were treated simultaneously, but housed in separate cages. Surgery Ischaemic button experiments Surgery was performed under aseptic conditions. Anaesthesia was induced using isoflurane (Abbott, Wiesbaden, Germany). For analgesia, animals received carprofen 5 mg/kg bodyweight subcutaneously. Peritoneal adhesion formation was induced by the construction of four buttons on the peritoneal wall (Fig. S1, supporting information)17. Via a median laparotomy, the peritoneum was lifted with a clamp, and a ligature was applied by first stitching through the base of the button and then ligating the peritoneum. Two buttons were placed on both sides of the peritoneum using a Vicryl® 6/0 suture (Ethicon, Somerville, New Jersey, USA). The abdomen was closed with a double-layered suture of the peritoneum (Vicryl® 5/0) and skin (silk 5/0; Braun, Sempach, Switzerland). Surgery was performed in all groups on day 0. Mice were killed on postoperative day (POD) 1, 3 or 7. Tissue samples were divided into buttons with and without adhesions. Before harvesting, adhesive tissue was stripped off buttons with adhesions. Unmanipulated peritoneal wall served as control. Eight mice were used per group for ischaemic button surgery, and five in the control group. Colonic anastomosis A standardized anastomosis model in mice was used, as described previously18. In brief, the abdomen was opened via a 3-cm midline incision. The ascending colon (1 cm from the caecum) was transected, strictly avoiding damage to the vessels. An end-to-end anastomosis was constructed with 12 interrupted full-thickness sutures (Vicryl® 8/0) before closure of the belly with a double-layered suture of the peritoneum (Vicryl® 5/0) and skin (silk 5/0). Eight animals were included per group for these experiments. Functional analysis Adhesion formation Adhesion formation after ischaemic button surgery was quantified by counting adhesions on POD 7. In addition, an adhesion score was used19: score 0, no adhesions; score 1, thin, pellucid adhesions; score 2, tensile adhesion; score 3, inseparable and vascularized adhesion; score 4, entire abdomen linked by adhesions. Adhesion formation in the colonic anastomosis experiments was quantified as the percentage of adhesions at the site of anastomosis; 0 per cent indicated absence of adhesions and 100 per cent indicated complete coverage of the circumference of the anastomosis. Adhesions around the anastomosis were also scored according to the method of Zuhlke and colleagues20, whereby grade 0 represents no adhesions and grade 4 indicates manifest adhesions, dissectible only with sharp instruments. Anastomotic bursting pressure Anastomotic bursting pressure was measured directly after death by sampling a 3-cm colonic segment including the anastomotic site18. The anastomosis specimen was ligated at the distal end and connected via a catheter at the proximal end to a pressure transducer. Krebs–Henseleit buffer was infused at a constant rate via an infusion pump, and intraluminal pressure was recorded by means of a pressure transducer (Biopac Systems, Goleta, California, USA). A sudden loss of pressure indicated that the anastomosis had burst, and anastomotic bursting pressure was defined as the maximum intraluminal pressure before leakage. Immunofluorescence Cells from the control group, and buttons with and without adhesions were isolated after enzymatic digestion in a solution containing collagenase II (Worthington, Lakewood, New Jersey, USA), Dispase® II (La Roche, Mannheim, Germany), DNase (La Roche), bovine serum albumin and trypsin inhibitor. Cells were centrifuged on to glass slides (cytospin method) and stained with rat antimouse F4/80 (1 : 200) antibody (BM8; Life Technologies, Darmstadt, Germany) and Arg-1 (1 : 200) antibody (N20; Santa Cruz Biotechnology, Dallas, Texas, USA), followed by secondary donkey antirat Alexa 488 (Life Technologies) and donkey antigoat Cy3 (Dianova, Hamburg, Germany) antibodies. The nucleus was stained using 4′,6-diamidino-2-phenylindole (DAPI) (Life Technologies). F4/80+ and Arg-1+ cells were counted in five randomly chosen areas in each specimen at a magnification of ×200. Quantitative PCR Gene expression of M1 and M2 markers was analysed by PCR. Reagents were from Life Technologies unless specified otherwise. Total RNA was extracted with Trizol® reagent using a tissue homogenizer (Precellys® 24; Peqlab, Erlangen, Germany) followed by DNase I treatment. cDNA was synthesized using a High Capacity cDNA rt kit. Expression of mRNA was quantified in triplicate by reverse transcription–PCR with specific probes/primers (Table S1, supporting information). The PCR was performed in Power SYBR® Green or Universal PCR Master Mix by amplification of 10 ng cDNA for 40 cycles (95°C for 15 s, 60°C for 1 min) on an AbiPrism® 7900HT (Life Technologies). Data quantification was performed by the ΔΔCT method and value normalized with respect to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) levels. IL-4 data were displayed as the ratio of IL-4 to GAPDH. Arginase assay Arginase activity was determined as described previously21. In brief, proteins were isolated with 0·1 per cent Triton X-100, activated with 10 mmol/l manganese chloride and 50 mmol/l Tris–hydrochloric acid, and 0·5 mol/l l-arginine was added as substrate. The reaction was stopped by addition of 10 per cent sulphuric acid and 30 per cent phosphoric acid in water. Absorbance of the assay product urea was measured photometrically at 450 nm. Statistical analysis Continuous data are presented as mean(s.d.). Statistical analysis was performed using one- or two-way ANOVA with Bonferroni's post hoc test, or Students t test, with Prism® 5.02 software (GraphPad, La Jolla, California, USA). P < 0·050 was considered statistically significant. Results Infiltrated M2 macrophages are associated with diminished adhesion formation Adhesion formation started 24 h after ischaemic button placement, and the number and strength of adhesions increased up to POD 7. Immunofluorescence analyses showed equal numbers of infiltrated F4/80+ macrophages within cytospin specimens of buttons with and without adhesions, whereas no macrophages were present in the unmanipulated peritoneal wall (Fig. 1a). Furthermore, transcription of the inflammatory mediators monocyte chemoattractant protein 1 and IL-6 was analysed, demonstrating the presence of inflammation up to POD 7 (Fig. S2, supporting information). Fig. 1 Open in new tabDownload slide Macrophage infiltration and markers of M2 differentiation in ischaemic buttons. Ischaemic button (IB) tissue was collected from wild-type mice on days 1, 3 and 7 after surgery, and separated into buttons with (IB+A) and without (IB–A) adhesions. Unmanipulated peritoneal wall served as control tissue (CTL). a Representative immunofluorescence staining of F4/80+ macrophages and nuclei (stained with 4′,6-diamidino-2-phenylindole, DAPI), and quantification of F4/80+ and F4/80− cell populations relative to total cell counts in IB+A and IB–A groups on day 3 after surgery (original magnification ×400; scale bar 100 µm). Images are representative of five independent experiments. b Gene expression of mannose receptor (MR) 1 mRNA, relative to levels in CTL, and arginase activity. c Representative immunofluorescence staining for macrophage surface marker F4/80, M2 macrophage marker arginase (Arg-1) and nuclei (DAPI) in single-cell cytospin preparations of ischaemic button specimens (original magnification ×400; scale bar 100 µm). Images are representative of five independent experiments. d Quantification of F4/80+Arg-1+ and F4/80+Arg-1− cells in IB+A and IB–A groups on day 3. Values are mean(s.d.) (n = 5–8 for all groups). *P < 0·050, †P < 0·010, ‡P < 0·001 versus CTL unless indicated otherwise (one-way ANOVA with Bonferroni's post hoc test) Ischaemic button specimens were examined for M2 differentiation markers including MR-1 and arginase. Raised gene expression of MR-1 was present in buttons without adhesions compared with controls (P = 0·002) and buttons with adhesions (P = 0·039) on POD 3 (Fig. 1b). Arginase activity also increased in the early postoperative phase in buttons without adhesions compared with those with adhesions (0·15(0·05) versus 0·07(0·05) µg urea per mg per h; P = 0·012) on POD 3 (Fig. 1b). On POD 7 increased arginase activity was observed in buttons with adhesions compared with controls (P < 0·001). Immunofluorescence analyses identified F4/80+ macrophages as the exclusive cellular target of arginase expression (Fig. 1c). Corresponding to the arginase activity levels, larger numbers of arginase-expressing macrophages were found in buttons without versus with adhesions in WT mice (52·5(9·7) versus 28·0(10·7) per cent; P = 0·013) (Fig. 1d). PPAR-γ modulates macrophage differentiation and affects adhesion formation Gene expression analysis of three possible inducers of M2 differentiation (IL-4, IL-10 and PPAR-γ) showed total absence of IL-4 transcription in ischaemic buttons, and no differences in adhesion formation in IL-4−/− compared with WT mice (Fig. S3, supporting information). However, IL-10 and PPAR-γ mRNAs were significantly upregulated compared with levels in control tissue (Fig. 2a,b). Levels of IL-10 mRNA did not differ between buttons with and without adhesions, but PPAR-γ mRNA levels were increased in those without adhesions on POD 3 (P = 0·012) (Fig. 2b). The role of IL-10 and PPAR-γ in adhesion formation was analysed by use of IL-10−/− and CD11b-Cre/PPARγfl/fl mice. The latter lack PPAR-γ exclusively in CD11b+ cells such as macrophages and monocytes. Numbers of adhesions and adhesion score were not altered in IL-10−/− compared with WT mice (Fig. 2c), but were raised in CD11b-Cre/PPARγfl/fl mice compared with PPARγfl/fl mice on POD 7 (36·1(18·2) versus 16·7(12·9) per cent; P = 0·042) (Fig. 2d; Fig. S4, supporting information). Fig. 2 Open in new tabDownload slide Contribution of interleukin (IL) 10 and macrophage-specific peroxisome proliferator-activated receptor (PPAR) γ activity to postoperative adhesion formation. Ischaemic button (IB) tissue was collected from wild-type (WT), IL-10 deficient (IL-10−/−) and CD11b-Cre/PPARγfl/fl mice on days 1, 3 and 7 after surgery, and separated into buttons with (IB+A) and without (IB–A) adhesions. Unmanipulated peritoneal wall served as control tissue (CTL). a,b Gene expression of a IL-10 and b PPAR-γ mRNA levels relative to CTL. c,d Adhesion formation in c IL-10−/−versus WT and b CD11b-Cre/PPARγfl/flversus PPARγfl/fl mice on day 7. Values are mean(s.d.) (n = 5–8 for all groups). *P < 0·050 versus CTL unless indicated otherwise (a,b one-way ANOVA with Bonferroni's post hoc test, d Student's t test) It was questioned whether PPAR-γ affects macrophage differentiation during adhesion formation. Expression of IL-6 and TNF-α mRNA was not altered in buttons without adhesions (Fig. 3a,b). However, Arg-1 mRNA expression and arginase activity were reduced in CD11b-Cre/PPARγfl/fl mice compared with PPARγfl/fl control mice (P = 0·005 and P = 0·013 respectively) (Fig. 3c,d). This suggests that Arg-1 is expressed in a PPAR-γ-dependent manner and might have a functional impact on adhesion formation. Fig. 3 Open in new tabDownload slide Impact of macrophage-specific peroxisome proliferator-activated receptor (PPAR) γ deficiency on mRNA expression of M1 (interleukin (IL) 6 and tumour necrosis factor (TNF) α) and M2 (arginase (Arg) 1) markers, and arginase activity. Ischaemic button (IB) tissue was collected from CD11b-Cre/PPARγfl/fl mice and corresponding PPARγfl/fl control mice on day 3 after surgery, and separated into buttons with (IB+A) and without (IB–A) adhesions. Unmanipulated peritoneal wall served as control tissue (CTL). a–c Gene expression of a IL-6, b TNF-α and c Arg-1 mRNA levels relative to CTL. d Arginase activity measured indirectly by l-arginine to urea conversion. Values are mean(s.d.) (n = 5–8 for all groups). *P < 0·050, †P < 0·010 (two-way ANOVA with Bonferroni's post hoc test) Pioglitazone drives arginine 1 expression in macrophages and prevents adhesion formation Gene expression analysis in pioglitazone-treated mice revealed a significant reduction in IL-6 mRNA and TNF-α mRNA in buttons with and without adhesions on POD 3 (all P < 0·001) (Fig. 4a,b), whereas MR-1 mRNA was upregulated compared with values in the vehicle-treated group (P < 0·001) (Fig. 4c). Arginase activity was also upregulated in buttons without adhesions after pioglitazone treatment compared with the vehicle group (P = 0·020) (Fig. 4d). By POD 7, pioglitazone treatment resulted in a reduced number of adhesions in WT mice (21·9(20·9) per cent versus 55·0(19·7) per cent in vehicle-treated group; P = 0·003) (Fig. 4e). Pioglitazone also decreased the severity of adhesions, as indicated by a lower adhesion score than in the vehicle-treated group (0·37(0·29) versus 1·05(0·42); P = 0·015) (Fig. 4f). Pioglitazone did not reduce numbers of adhesions (54·2(18·2) versus 37·5(14·1) per cent for pioglitazone versus vehicle-treated groups), nor did it affect adhesion scoring (1·08(0·28) versus 0·74(0·35)) in CD11b-Cre/PPARγfl/fl mice (Fig. 4e,f). Fig. 4 Open in new tabDownload slide Effect of pioglitazone on mRNA expression of M1 (interleukin (IL) 6 and tumour necrosis factor (TNF) α) and M2 (mannose receptor (MR) 1 and arginase activity) markers, and adhesion formation in wild-type (WT) mice and CD11b-Cre/PPARγfl/fl mice. Mice received a daily gavage of pioglitazone for 7 consecutive days, starting 3 days before surgery. Controls were treated with vehicle. Ischaemic button (IB) tissue was collected on days 3 and 7 after surgery, and separated into buttons with (IB+A) and without (IB–A) adhesions. Unmanipulated peritoneal wall served as control tissue (CTL). a–c Gene expression of a IL-6, b TNF-α and c MR-1 mRNA on day 3 relative to CTL. d Arginase activity was measured indirectly by l-arginine to urea conversion on day 3. e Adhesion formation was determined on day 7 and f verified by an adhesion score. Values are mean(s.d.) (n = 5–7 for all groups). *P < 0·050, †P < 0·010, ‡P < 0·001 (a–d two-way ANOVA with Bonferroni's post hoc test, e,f Student's t test) Pioglitazone did not compromise anastomotic healing but reduced anastomotic adhesion formation There was no difference in anastomotic bursting pressure between pioglitazone- and vehicle-treated mice on POD 7 (197·5(106·4) versus 190·7(94·2) mmHg) (Fig. 5a). Additionally, anastomotic gene expression of collagen type I and III transcripts was not altered by pioglitazone (Fig. 5b,c). The adhesion-covered anastomotic circumference and strength of adhesions was significantly lower after pioglitazone treatment than in the vehicle-treated group (P < 0·001 and P = 0·016 respectively) (Fig. 5d,e). Fig. 5 Open in new tabDownload slide Effect of pioglitazone on healing and adhesion formation at the colonic anastomosis. Wild-type mice received a daily gavage of pioglitazone or vehicle for 7 consecutive days and underwent colonic anastomosis surgery on day 4. Results for the anastomotic tissue were compared with those for an unmanipulated colonic specimen (CTL) from the same animal. a Anastomotic bursting pressure measured 7 days after surgery. b,c Gene expression of b collagen I and c collagen III mRNA measured within the anastomosis relative to CTL. Collagen expression was compared between anastomosis and CTL after pioglitazone or vehicle treatment. d Adhesion formation at the circumference of the anastomosis 7 days after surgery and e scored according to ease of dissection. Values are mean(s.d.) (n = 8 for all groups). *P < 0·050, †P < 0·010, ‡P < 0·001 versus CTL unless indicated otherwise (a,d,e Student's t test, b,c two-way ANOVA with Bonferroni's post hoc test) Discussion In this study, perioperative pharmacological and endogenous PPAR-γ agonism reduced adhesion formation in mice. This correlated with alteration of macrophage differentiation shown by increased arginase activity and MR-1 expression. Furthermore, PPAR-γ agonism did not compromise anastomotic healing. Macrophages represent a dynamic and functionally heterogeneous cell population; their function strongly depends on the differentiation status and their plasticity accounts for significant variation during healing processes9,15,22. Previous studies suggested an involvement of macrophages in adhesion formation8, but their exact function and differentiation status is still unknown. In the present study, adhesion formation correlated inversely with M2 marker expression, as buttons without adhesions contained higher transcript levels of Arg-1 and MR-1. Furthermore, increased arginase activity and higher levels of Arg-1-expressing macrophages were observed in buttons without than in those with adhesions. Arg-1 is a key molecule in macrophage polarization23. The finding of diminished arginase activity on POD 3 and prolonged increased activity on POD 7 in buttons with adhesions, compared with those without adhesions, suggests putative resolution or at least a modulatory effect by altered macrophage differentiation during adhesion formation. This is supported by a recent study24 showing a reduction in hepatic fibrosis and collagen deposition linked to Arg-1-expressing M2 macrophage induction. Suppression of fibrosis by a shift to M2 differentiation is a promising therapeutic target for fibrotic diseases15. This could be due to the competitive effect of arginase activity and inducible nitric oxide synthase. As release of nitric oxide is known to increase collagen production in wound fibroblasts25, increased arginase activity could suppress nitric oxide release and therefore reduce collagen production during adhesion formation. Pesce and colleagues26 also hypothesized that arginase activity within macrophages competes with fibroblasts for l-arginine, a substrate for collagen production, resulting in reduced collagen production. However, contradictory results have also been published, describing a profibrotic role of M2 macrophages, by increasing collagen production and deposition during fibrosis9. Given the inverse correlation between M2 marker expression and adhesion formation in the present model, an analysis was undertake to determine which molecules drive macrophage differentiation towards an M2 phenotype during peritoneal healing. Neither IL-4- nor IL-10-mediated pathways, which are both well known to induce different M2-like macrophage subtypes9, were involved in adhesion formation, although IL-10 was at least expressed and may have contributed to overall inflammation during the healing process. This is in agreement with the findings of Daley and co-workers27, who demonstrated IL-4/IL-13-independent polarization of monocytes into M2 macrophages during murine wound healing after surgery27. In contrast, in the present study, expression of PPAR-γ, another molecule that induces M2 differentiation28, was greater in buttons without adhesions than in buttons with adhesions. PPAR-γ modulates immune and metabolic functions in macrophages by inhibiting proinflammatory responses and upregulating anti-inflammatory cytokine expression11. Reduced arginase activity and expression was observed in CD11b-Cre/PPARγfl/fl mice, which lack PPAR-γ specifically in monocytes and macrophages, indicating reduced M2 macrophage function. Furthermore, adhesion formation was increased in CD11b-Cre/PPARγfl/fl mice, providing an axis between M2 differentiation, arginase activity and adhesion formation. Importantly, Arg-1 expression and arginase activity have been found to depend on PPAR-γ activation29. Pharmacological PPAR-γ agonism with pioglitazone prevented adhesion formation, and increased arginase activity and M2 marker gene expression. This was also shown in the vicinity of the colonic anastomosis. These data indicate that exogenous PPAR-γ agonism is powerful enough to overcome the detrimental impact of an additional surgical trauma. The protective effect of PPAR-γ agonism probably originates from the macrophage compartment, as pioglitazone failed to prevent adhesion formation in CD11b-Cre/PPARγfl/fl mice. The role of the PPAR-γ pathway in adhesion formation is summarized inFig. 6. Fig. 6 Open in new tabDownload slide Role of the peroxisome proliferator-activated receptor (PPAR) γ pathway and its intervention in M2 macrophage polarization during abdominal adhesion formation. Macrophage activation and infiltration to the site of a peritoneal lesion occurs during abdominal surgery. Excessive expression of interleukin (IL) 6, tumour necrosis factor (TNF) α and inducible ntiric oxide synthese (iNOS) indicates M1 macrophage polarization (left pathway) and increases adhesion formation, presumably via local induction of dysbalanced collagen formation/degradation. Counter-regulatory mechanisms include PPAR-γ-dependent M2 macrophage polarization (right pathway), indicated by increased mannose receptor (MR) 1 expression and arginase activity. Dysbalanced macrophage polarization with an M1 greater than M2 phenotype may originate from disturbed or inappropriate PPAR-γ activity and coincides with augmented adhesion formation. However, the mechanism leading to the insufficient local PPAR-γ activation remains to be determined. Nevertheless, pioglitazone treatment reduces postoperative adhesion formation, indicating that PPAR-γ agonism corrects the disturbances in macrophage polarization by supporting M2 polarization, as shown by increased MR-1 expression and arginase activity Of note, PPARγfl/fl mice showed reduced adhesion formation compared with the C57BL6/J WT counterparts that were used as controls in all other experiments. This difference can be explained by their genetic background. C57BL6/J WT, but not the genetically modified mice with a C57BL/6NCrJ background, carry a deletion in the Nnt gene driving the metabolism leading to glucose intolerance30. As glucose intolerance is known to impair wound healing31, this may be responsible for differences in postoperative adhesion formation between the two mouse strains. This study suggests that PPAR-γ-mediated adhesion prevention depends on macrophage arginase function. This was confirmed by a recent study32 showing a requirement for local macrophage-dependent arginase activity in cutaneous wound healing. Controversially, other authors33 have demonstrated improved excisional wound healing by topical arginase inhibition. Further analyses of the role of arginase in peritoneal wound healing are required. Perioperative pioglitazone treatment could be a strategy for the prevention of abdominal adhesion formation. The present data show that PPAR-γ agonism does not compromise anastomotic healing in mice, which generally involves arginase activity34. However, the anastomotic leakage model used in this study cannot be extrapolated to the clinical setting. Nevertheless, use of systemic pioglitazone to prevent adhesion formation may be complementary to the currently available antiadhesive agents such as biodegradable barriers. These barriers are limited to local application, and some are suspected to impair anastomotic healing35. Acknowledgements This study was funded by BONFOR (O-112.0053) and the Else Kröner-Forschungskolleg Bonn (Q-605.0812). Disclosure: The authors declare no conflict of interest. References 1 Ellis H , Moran BJ, Thompson JN, Parker MC, Wilson MS, Menzies D et al. Adhesion-related hospital readmissions after abdominal and pelvic surgery: a retrospective cohort study . Lancet 1999 ; 353 : 1476 – 1480 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Ten Broek RP , Issa Y, van Santbrink EJ, Bouvy ND, Kruitwagen RF, Jeekel J et al. Burden of adhesions in abdominal and pelvic surgery: systematic review and meta-analysis . BMJ 2013 ; 347 : f5588 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Hellebrekers BW , Kooistra T. Pathogenesis of postoperative adhesion formation . Br J Surg 2011 ; 98 : 1503 – 1516 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Binnebosel M , Rosch R, Junge K, Lynen-Jansen P, Schumpelick V, Klinge U. Macrophage and T-lymphocyte infiltrates in human peritoneal adhesions indicate a chronic inflammatory disease . World J Surg 2008 ; 32 : 296 – 304 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Dizerega GS , Campeau JD. Peritoneal repair and post-surgical adhesion formation . Hum Reprod Update 2001 ; 7 : 547 – 555 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Xu X , Rivkind A, Pappo O, Pikarsky A, Levi-Schaffer F. Role of mast cells and myofibroblasts in human peritoneal adhesion formation . Ann Surg 2002 ; 236 : 593 – 601 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Ohashi K , Yoshimoto T, Kosaka H, Hirano T, Iimuro Y, Nakanishi K et al. Interferon gamma and plasminogen activator inhibitor 1 regulate adhesion formation after partial hepatectomy . Br J Surg 2014 ; 101 : 398 – 407 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Burnett SH , Beus BJ, Avdiushko R, Qualls J, Kaplan AM, Cohen DA. Development of peritoneal adhesions in macrophage depleted mice . J Surg Res 2006 ; 131 : 296 – 301 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Gordon S , Martinez FO. Alternative activation of macrophages: mechanism and functions . Immunity 2010 ; 32 : 593 – 604 . Google Scholar Crossref Search ADS PubMed WorldCat 10 ten Broek RP , Stommel MW, Strik C, van Laarhoven CJ, Keus F, van Goor H. Benefits and harms of adhesion barriers for abdominal surgery: a systematic review and meta-analysis . Lancet 2014 ; 383 : 48 – 59 . Google Scholar Crossref Search ADS PubMed WorldCat 11 Nagy L , Szanto A, Szatmari I, Szeles L. Nuclear hormone receptors enable macrophages and dendritic cells to sense their lipid environment and shape their immune response . Physiol Rev 2012 ; 92 : 739 – 789 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Sindrilaru A , Scharffetter-Kochanek K. Disclosure of the culprits: macrophages – versatile regulators of wound healing . Adv Wound Care (New Rochelle) 2013 ; 2 : 357 – 368 . Google Scholar Crossref Search ADS PubMed WorldCat 13 El Kasmi KC , Qualls JE, Pesce JT, Smith AM, Thompson RW, Henao-Tamayo M et al. Toll-like receptor-induced arginase 1 in macrophages thwarts effective immunity against intracellular pathogens . Nat Immunol 2008 ; 9 : 1399 – 1406 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Stein M , Keshav S, Harris N, Gordon S. Interleukin 4 potently enhances murine macrophage mannose receptor activity: a marker of alternative immunologic macrophage activation . J Exp Med 1992 ; 176 : 287 – 292 . Google Scholar Crossref Search ADS PubMed WorldCat 15 Wynn TA , Ramalingam TR. Mechanisms of fibrosis: therapeutic translation for fibrotic disease . Nat Med 2012 ; 18 : 1028 – 1040 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Morán-Salvador E , Titos E, Rius B, González-Périz A, García-Alonso V, López-Vicario C et al. Cell-specific PPARγ deficiency establishes anti-inflammatory and anti-fibrogenic properties for this nuclear receptor in non-parenchymal liver cells . J Hepatol 2013 ; 59 : 1045 – 1053 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Reed KL , Fruin AB, Bishop-Bartolomei KK, Gower AC, Nicolaou M, Stucchi AF et al. Neurokinin-1 receptor and substance P messenger RNA levels increase during intraabdominal adhesion formation . J Surg Res 2002 ; 108 : 165 – 172 . Google Scholar Crossref Search ADS PubMed WorldCat 18 Pantelis D , Beissel A, Kahl P, Vilz TO, Stoffels B, Wehner S et al. Colonic anastomotic healing in the context of altered macrophage function and endotoxemia . Int J Colorectal Dis 2011 ; 26 : 737 – 746 . Google Scholar Crossref Search ADS PubMed WorldCat 19 Sikkink CJ , de Man B, Bleichrodt RP, van Goor H. Auto-cross-linked hyaluronic acid gel does not reduce intra-abdominal adhesions or abscess formation in a rat model of peritonitis . J Surg Res 2006 ; 136 : 255 – 259 . Google Scholar Crossref Search ADS PubMed WorldCat 20 Zühlke HV , Lorenz EM, Straub EM, Savvas V. [ Pathophysiology and classification of adhesions .] Langenbecks Arch Chir Suppl II Verh Dtsch Ges Chir 1990 ; 1009 – 1016 . Google Scholar OpenURL Placeholder Text WorldCat 21 Corraliza IM , Campo ML, Soler G, Modolell M. Determination of arginase activity in macrophages: a micromethod . J Immunol Methods 1994 ; 174 : 231 – 235 . Google Scholar Crossref Search ADS PubMed WorldCat 22 Schürmann C , Seitz O, Sader R, Pfeilschifter J, Goren I, Frank S. Role of wound macrophages in skin flap loss or survival in an experimental diabetes model . Br J Surg 2010 ; 97 : 1437 – 1451 . Google Scholar Crossref Search ADS PubMed WorldCat 23 Rath M , Müller I, Kropf P, Closs EI, Munder M. Metabolism via arginase or nitric oxide synthase: two competing arginine pathways in macrophages . Front Immunol 2014 ; 5 : 532 . Google Scholar Crossref Search ADS PubMed WorldCat 24 Manuelpillai U , Lourensz D, Vaghjiani V, Tchongue J, Lacey D, Tee JY et al. Human amniotic epithelial cell transplantation induces markers of alternative macrophage activation and reduces established hepatic fibrosis . PLoS One 2012 ; 7 : e38631 . Google Scholar Crossref Search ADS PubMed WorldCat 25 Schaffer MR , Efron PA, Thornton FJ, Klingel K, Gross SS, Barbul A. Nitric oxide, an autocrine regulator of wound fibroblast synthetic function . J Immunol 1997 ; 158 : 2375 – 2381 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 26 Pesce JT , Ramalingam TR, Mentink-Kane MM, Wilson MS, El Kasmi KC, Smith AM et al. Arginase-1-expressing macrophages suppress Th2 cytokine-driven inflammation and fibrosis . PLoS Pathog 2009 ; 5 : e1000371 . Google Scholar Crossref Search ADS PubMed WorldCat 27 Daley JM , Brancato SK, Thomay AA, Reichner JS, Albina JE. The phenotype of murine wound macrophages . J Leukoc Biol 2010 ; 87 : 59 – 67 . Google Scholar Crossref Search ADS PubMed WorldCat 28 Odegaard JI , Ricardo-Gonzalez RR, Goforth MH, Morel CR, Subramanian V, Mukundan L et al. Macrophage-specific PPARgamma controls alternative activation and improves insulin resistance . Nature 2007 ; 447 : 1116 – 1120 . Google Scholar Crossref Search ADS PubMed WorldCat 29 Gallardo-Soler A , Gómez-Nieto C, Campo ML, Marathe C, Tontonoz P, Castrillo A et al. Arginase I induction by modified lipoproteins in macrophages: a peroxisome proliferator-activated receptor-gamma/delta-mediated effect that links lipid metabolism and immunity . Mol Endocrinol 2008 ; 22 : 1394 – 1402 . Google Scholar Crossref Search ADS PubMed WorldCat 30 Freeman HC , Hugill A, Dear NT, Ashcroft FM, Cox RD. Deletion of nicotinamide nucleotide transhydrogenase: a new quantitive trait locus accounting for glucose intolerance in C57BL/6 J mice . Diabetes 2006 ; 55 : 2153 – 2156 . Google Scholar Crossref Search ADS PubMed WorldCat 31 Greer N , Foman NA, MacDonald R, Dorrian J, Fitzgerald P, Rutks I et al. Advanced wound care therapies for nonhealing diabetic, venous, and arterial ulcers: a systematic review . Ann Intern Med 2013 ; 159 : 532 – 542 . Google Scholar Crossref Search ADS PubMed WorldCat 32 Campbell L , Saville CR, Murray PJ, Cruickshank SM, Hardman MJ. Local arginase 1 activity is required for cutaneous wound healing . J Invest Dermatol 2013 ; 133 : 2461 – 2470 . Google Scholar Crossref Search ADS PubMed WorldCat 33 Kavalukas SL , Uzgare AR, Bivalacqua TJ, Barbul A. Arginase inhibition promotes wound healing in mice . Surgery 2012 ; 151 : 287 – 295 . Google Scholar Crossref Search ADS PubMed WorldCat 34 Witte MB , Vogt N, Stuelten C, Gotoh T, Mori M, Becker HD. Arginase acts as an alternative pathway of l-arginine metabolism in experimental colon anastomosis . J Gastrointest Surg 2003 ; 7 : 378 – 385 . Google Scholar Crossref Search ADS PubMed WorldCat 35 Beck DE , Cohen Z, Fleshman JW, Kaufman HS, van Goor H, Wolff BG. A prospective, randomized, multicenter, controlled study of the safety of Seprafilm adhesion barrier in abdominopelvic surgery of the intestine . Dis Colon Rectum 2003 ; 46 : 1310 – 1319 . Google Scholar Crossref Search ADS PubMed WorldCat © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © 2015 BJS Society Ltd Published by John Wiley & Sons Ltd