Arterial spin labeling MR imaging for differentiation between high- and low-grade glioma—a meta-analysis

Arterial spin labeling MR imaging for differentiation between high- and low-grade glioma—a... Abstract Background Arterial spin labeling is an MR imaging technique that measures cerebral blood flow (CBF) non-invasively. The aim of the study is to assess the diagnostic performance of arterial spin labeling (ASL) MR imaging for differentiation between high-grade glioma and low-grade glioma. Methods Cochrane Library, Embase, Medline, and Web of Science Core Collection were searched. Study selection ended November 2017. This study was prospectively registered in PROSPERO (CRD42017080885). Two authors screened all titles and abstracts for possible inclusion. Data were extracted independently by 2 authors. Bivariate random effects meta-analysis was used to describe summary receiver operating characteristics. Trial sequential analysis (TSA) was performed. Results In total, 15 studies with 505 patients were included. The diagnostic performance of ASL CBF for glioma grading was 0.90 with summary sensitivity 0.89 (0.79–0.90) and specificity 0.80 (0.72–0.89). The diagnostic performance was similar between pulsed ASL (AUC 0.90) with a sensitivity 0.85 (0.71–0.91) and specificity 0.83 (0.69–0.92) and pseudocontinuous ASL (AUC 0.88) with a sensitivity 0.86 (0.79–0.91) and specificity 0.80 (0.65–0.87). In astrocytomas, the diagnostic performance was 0.89 with sensitivity 0.86 (0.79 to 0.91) and specificity 0.79 (0.63 to 0.89). Sensitivity analysis confirmed the robustness of the findings. TSA revealed that the meta-analysis was adequately powered. Conclusion Arterial spin labeling MR imaging had an excellent diagnostic accuracy for differentiation between high-grade and low-grade glioma. Given its low cost, non-invasiveness, and efficacy, ASL MR imaging should be considered for implementation in the routine workup of patients with glioma. arterial spin labeling, brain tumors, CNS, glioma, imaging Importance of the study Arterial spin labeling is an MR imaging technique for measurement of CBF without administration of contrast agent. The summary diagnostic performance of ASL to discriminate between glioma grades is unknown, the individual trials assessing this having been limited and hampered by a small number of participants. Accurate diagnosis is important for prognostication, treatment planning, and assessment of treatment response for glioma. Biopsy is highly invasive and carries the risk of undersampling. MRI plays an important role in this regard. This meta-analysis provides evidence that ASL MR imaging has an excellent diagnostic accuracy for differentiation between high-grade and low-grade glioma, in addition to its low cost and non-invasiveness. Its high diagnostic performance was independent of the specific ASL technique used. ASL can be considered for implementation in the routine workup of patients with glioma when gadolinium administration is unwanted or unnecessary. Gliomas are the most common primary brain tumors and comprise 80% of all malignant tumors in the brain. The average incidence of glioma was 6.61 (6.57% to 6.6%) per 100000 in 2007–2011 in the United States.1 Gliomas originate from the cerebral glial cells and are classified according to their cell type based on histological appearance and biomolecular status, most commonly into astrocytoma or oligodendroglioma, and according to their grade, low-grade (World Health Organization [WHO] grades I–II) or high-grade (III–IV).2 The 10-year survival for low-grade gliomas in adults is approximately 43%3; oligodendrogliomas have been reported associated with a relatively more favorable outcome compared with astrocytomas.4 Grade III glioma has a significantly worse prognosis compared with glioma grades I and II. The most common intracranial neoplasm, WHO grade IV glioma, also named glioblastoma, has a median survival of 15 months.5 Accurate assessment of tumor grade and type is crucial for optimal treatment planning. Tumor diagnosis, classification, and grading is established by biopsy or excision, which is associated with considerable risk, including death, especially in central areas of the brain.6 In addition, diagnosis from biopsy sampling or subtotal resection is potentially limited by the sample not being representative, leading to understaging of tumor grade7; in turn, this may mislead treatment planning toward a more conservative strategy. Furthermore, patients who are inoperable or unwilling to undergo surgery, or in whom tumors are located in eloquent areas, might not be suitable for surgical intervention even though they qualify for adjuvant therapy, including chemotherapy and or radiotherapy. Thus, there is a strong need for alternative methods for grading brain tumors preoperatively with imaging as a reasonable alternative even if there is no ideal imaging technique as of today. MRI, which is done routinely in the presurgical workup of patients with brain cancer, is a reasonable candidate for assessment of tumor grade in gliomas. Potential advantages of MR are its ability to non-invasively map the entire tumor and quantify cerebral blood flow (CBF), a marker of angiogenesis and thus of tumor grade. The diagnostic performance of dynamic susceptibility contrast (DSC) perfusion to discriminate between WHO grades II and III is limited8; in addition the technique requires administration of a gadolinium contrast agent with considerable costs; and more importantly, recent evidence of gadolinium deposition in the brain parenchyma has been reported.9 Arterial spin labeling (ASL) is a completely non-invasive MR imaging method mainly for measurement of CBF.10,11 The ASL technique is based on magnetic labeling of inflowing protons, which subsequently are measured in the brain parenchyma; roughly, CBF maps are acquired by subtraction of background non-labeled tissue. Based on the labeling method, the ASL technique is categorized into continuous ASL (CASL), pulsed ASL (PASL), or pseudocontinuous ASL (pCASL). At present mainly multiphase PASL and 3D pCASL are recommended12,13; however, there is no consensus regarding the optimal ASL technique for brain tumor evaluation. A large number of publications have suggested a promising role for ASL for discrimination between high-grade glioma (HGG) and low-grade glioma (LGG),14–16 with a strong correlation between DSC-derived CBF and ASL-derived CBF.17 However, concerns have been raised regarding the interpatient variability in the measurements,18 as well as the use of divergent parameters and techniques, WHO classification system, and postprocessing schemes. This lack of consensus and guidelines has probably contributed to the reluctance to implement ASL in clinical routine for the management of glioma patients. One meta-analysis concluded that there is a significant difference in blood flow between glioma grades,19 but the diagnostic performance of ASL to discriminate between tumor grades was not assessed in the study. Therefore the aim of this meta-analysis is to fully assess the diagnostic performance of ASL for differentiation of HGG from LGG, including potential sources of heterogeneity of its diagnostic performance. Materials and Methods This meta-analysis was conducted with adherence to the the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines20 and the Cochrane handbook for systematic reviews of diagnostic test accuracy.21 The study protocol was prospectively registered in PROSPERO (CRD42017080885). Search Strategy and Selection Criteria Literature search A literature search was performed in the following databases: Medline (OVID), Embase.com (Elsevier), Web of Science Core Collection (Clarivate), and Cochrane Library (Wiley). The MeSH (Medical Subject Headings) terms identified for searching Medline (OVID) were adapted in accordance to the corresponding vocabulary in Embase. Each search concept was also complemented with relevant free-text terms like: brain tumor, astrocytoma, arterial spin labeling, and ASL. The free-text terms were, if appropriate, truncated and/or combined with proximity operators. Conference abstracts were excluded in the search strategy. To include all eligible studies in the search, no language restriction was applied. Databases were searched from inception. All searches were performed by a librarian at the library of Karolinska Institutet in Stockholm, Sweden, in November 2017; the search strategies are available in Supplementary Tables S1–S4. Study eligibility criteria We included retrospective and prospective studies that evaluated patients with primary brain tumors, who received diagnoses according to the WHO classification, who were preoperatively examined using ASL, and for whom subsequent neuropathological diagnosis was available. Inclusion criteria were availability of (i) absolute CBF values and normalized CBF values stratified for tumor, and (ii) individual patient data or sensitivity and specificity for discrimination between high- and low-grade glioma. All types of ASL techniques were eligible for inclusion: PASL, pCASL, and CASL were eligible for inclusion. All tumor types as listed in the WHO classification were eligible, since the latest update of the WHO 2016 classification mainly added a strengthened biomolecular classification of tumor types with none or only slight changes in tumor grade classification.2 Furthermore, only untreated gliomas were eligible for inclusion. The following exclusion criteria were applied in the meta-analysis: studies with a strict pediatric cohort; studies grouping metastases or meningioma with intracranial tumors, studies reporting on recurrent or treated gliomas only, meningiomas, studies in other languages than English, duplicate or overlapping cohorts, abstracts or editorials. Oligoastrocytomas were excluded if possible from the analysis, since this diagnosis has been removed from the WHO classification 2016 and these tumors are now classified as either oligodendrogliomas or astrocytomas based on biomolecular mutational status. In the case of a duplicate cohort, the largest was included and the overlapping study was excluded from the meta-analysis. Study selection After the electronic search was conducted, 2 investigators independently screened and reviewed all titles and abstracts for possible inclusion in the qualitative assessment. Any disagreement between the investigators was solved through discussion with a third investigator, who cast the deciding vote. In addition, the third investigator hand-searched reference lists to identify additional studies. Quality Assessment Quality assessment was performed according to the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool,22 which was adjusted and developed to be used for this particular meta-analysis. Each study was graded according to prespecified criteria in 2 domains. Domain 1 included 4 internal bias items concerning specific quality issues in the included studies: patient selection, index test, reference standard, and flow and timing. Domain 2 consisted of 3 items related to the external validity of the included studies in relation to the specific aim of this meta-analysis: patient selection, index test, and reference standard. Two investigators independently assessed both domains for each study, with any disagreements solved through discussion with a third investigator. Data Extraction Data extraction was conducted independently by 2 investigators. The ASL perfusion parameters composed the maximum or mean CBF, relative or equivalent, and the maximum or mean intensity signal (ITS), relative or equivalent, nCBFmeanTI (CBF derived from multi-inversion time [mTI] ASL). In addition, the following items were extracted: first author name, publication year, town/country of origin, inclusion years, study design, age (mean and standard deviation), WHO classification, specifications for the MR camera used, numbers of channel in head coil, ASL method, flip angle, repetition time, echo time, post labeling delay, acquisition time, region of interest (ROI) technique, reference region, tumor types included. Any inconsistencies were solved through discussion with a third investigator until consensus. Main Outcome Measure The main outcome measure was the diagnostic performance of the ASL parameters for glioma grading. The secondary outcome measure was the diagnostic performance of the different ASL techniques for glioma grading and to assess the heterogeneity of their diagnostic performance. Data Synthesis Absolute and relative CBF measures were calculated where possible with mean and SD. For each study, the sensitivity and specificity was calculated with corresponding confidence intervals. The true positive, true negative, false positive, false negative rates and the area under the curve (AUC) were calculated for each study. For studies that only reported individual patient data, receiver operating characteristics curves were calculated optimizing the cutoff for both specificity and sensitivity. The abovementioned calculations were performed independently by 2 investigators and assessed for congruency. Any discrepancies between the investigators were solved through discussion. Mean difference with 95% CI was calculated for HGG and LGG for absolute CBF and regional CBF max with a random effects model. The sensitivity and specificity heterogeneity between studies was visually assessed by plotting their distribution in a univariable analysis. Exploring heterogeneity by I2 is not useful for diagnostic meta-analyses that include studies with different thresholds, since this results in high heterogeneity.23 Instead, we used bivariate meta-regression and subgroup and sensitivity analyses to explore heterogeneity. The following factors were explored in the meta-regression: WHO classification system (year published), MR field strength (in tesla), ASL method (PASL, pCASL, or CASL), and perfusion measure (mean or max CBF [mL/min/100 g], mean or max signal intensity [SI]). We used the funnel plot asymmetry test as described by Egger to explore for potential publication bias and small study effect.24 Trial Sequential Analysis Results in meta-analysis might be caused by random errors rather than reflecting an actual effect. Furthermore, lack of statistical significance might be caused by a lack of statistical power. Random errors are more likely in meta-analysis with few studies or if the included studies have small populations.25 Trial sequential analysis (TSA) is similar to the interim analysis in clinical trials. Using TSA it is possible to calculate the sample size required to demonstrate a difference between LGG and HGG. We calculated the estimated sample size for the meta-analysis to be powered for a one-sided type 1 error of 2.5% using O’Brien-Fleming α-spending function at a power of 90%. We minimized the type 1 error in the TSA by using a highly stringent criterion of 2.5% type 1 error and 90% power. TSA was performed using TSA version 0.9. Bivariate Analysis For assessment of the overall diagnostic accuracy of ASL for glioma grading, all studies were included irrespective of the CBF estimate reported, since no threshold was calculated. However, the unit maximum regional CBF, rCBFmax, was used where possible. To obtain summary estimates with 95% CIs, we used a bivariate random effects meta-analysis with a restricted maximum likelihood estimation method to describe the summary receiver operating characteristics (SROC) curve with associated AUC. The following subgroup analyses were performed: studies where maximum and mean rCBF were reported, studies reporting on astrocytomas only, and studies using pulsed, continuous, and pseudocontinuous ASL, respectively. Furthermore, we performed sensitivity analysis by excluding studies with unclear or high risk of bias related to the domain of internal or external validity in QUADAS-2; at least 3 studies with low risk of bias were required to perform sensitivity analysis for any particular item. Statistical analysis was performed in R for Mac v1.1.383 using the mada and metafor package v0.5.8. Results Search Hits The electronic search yielded a total of 640 hits, with 326 remaining after removal of duplicates. All 326 hits were screened for potential inclusion using the criteria described above, and subsequently 48 studies were included in the full-text evaluation. After full-text evaluation, another 34 studies were excluded for the following reasons: absence of sensitivity, specificity, or individual patient data, n = 8; absence of quantitative CBF data n = 17; the paper was not in the English language, n = 4; the material included only glioblastomas, extra-axial tumors or metastases, n = 4. In total, 15 studies comprising 505 patients met the inclusion criteria and were subsequently included in the full meta-analysis (Figure 1).14–17,26–36 Fig. 1 View largeDownload slide PRISMA flow chart. Fig. 1 View largeDownload slide PRISMA flow chart. Study Characteristics and Risk of Bias Assessment A full description of the 15 included studies is presented in Table 1 with technical specifications in Supplementary Table S5. Eight studies (8/15, 53%) used the WHO classification system published in 200716,28,29,31,32,34,36,37 rather than the most recent classification system published in 2016. In 7 studies (47%), pCASL was used for CBF estimation,17,26,31–34,36 7 (47%) used PASL,14,16,27–30,35 and 1 (7%) used CASL.15 All studies reported on relative values; 9 (60%) used rCBFmax14–16,26–28,32,34,36; of the 4 remaining (27%),17,28,31,35 1 study reported both rCBFmax and rCBFmean,28 1 (7%) reported normalized intensity signal (nITSmax),30 1 nITSmean,29 and 1 nCBFmeanTI33 (CBF derived from mTI ASL). In 12 studies (80%),15–17,26,28,29,31–36 the MR field strength was 3T and in 3 studies (20%), 1.5T.14,27,30 In 13 studies (87%),14,16,17,26–34,36 the risk of bias for patient selection was low, and in 2 (13%) unclear.15,35 In all studies the risk of bias related to the reference standard was unclear, since there was no information as to whether the pathologist was blinded for ASL results or not. The time interval between imaging and surgery/biopsy was reported in only 2 (20%) studies.27,32,34 Under the domain of applicability, concerns and external validity related to patient selection: 11 (73%) studies showed low risk bias14–17,26–28,31,34,36 and 4 (27%) unclear risk of bias pertaining to inclusion of only one tumor type29,30,32,33 (Table 2). The risk of bias concerns were included in the sensitivity analyses. The distribution of sensitivity and specificity was depicted in forest plots and is suggestive of heterogeneity among the included studies related to the specificity for glioma grading among the included studies (Figure 2). Table 1 Characteristics of included studies First Author Year City / Country Years of Patient Inclusion Patient Inclusion (Retrospective or Prospective) Mean Age (SD) WHO Classification (Year of Publication) ROI Technique Reference Region (Location/Lobe, Gray or White Matter) Perfusion Parameters Tumors Included (WHO Grade) aCBF /rCBFmax In LGG Mean (SD), N Patients aCBF/rCBFmax HGG Mean (SD), N Patients Cebeci16 2014 Bursa / Turkey 2010–2013 Retrospective 47 (14) 2007 Manual ROI in maximal CBF Contralateral normal hemisphere rCBFmax / CBFmax / rSI GBM, AC (III), Gliosarcoma, OD (II), DNET, pilocystic AC 8.1 (28.1) / 0.96 (0.48), 13 23.65 (44.8) / 4.7 (1.38), 20 Fudaba28 2014 Oita / Japan 2010–2012 Retrospective 60 (17) 2007 Manual ROI (8 mm in diameter) in total tumor Mirrored normal white matter rCBFmin / max / mean AC (II–IV), OD (II–III), OA, GBM NA / NA, 9 NA / NA, 23 Furtner29 2014 Vienna / Austria 2009–2012 Prospective 54 (17) 2007 Manual ROI in whole tumor Mirrored normal contralateral healthy hemishpere nITS mean AC (II–IV) NA / 1.08 (0.39), 7 NA / 3.09 (2.57), 26 Kim30 2008 Suwon / South Korea 2005–2008 Prospective 43 (14) 2000 Manual ROIs in maximal tumor perfusion signal Contralateral white matter rITSmax AC (II–IV) NA / NA, 26 NA / NA, 35 Lehmann17 2010 Marseille / France 2007–2008 Prospective 58 (19) 2000 Manual ROI in whole tumor Contralateral white matter rCBFmean OD, AC, GBM, LGG NA / 1.09 (1.08), 4 NA / 2.26 (1.12), 4 Ma31 2017 Jiangsu / China 2014–2015 Prospective 46 (18) 2007 Manual ROIs in whole tumor (50–60 mm2) MIrrored contralateral normal matter CBFmean / rCBFmean GG, Pilocytic AC, AC gr I, AC (II–IV), AO, OD (gr III) 50.64 (35.89) / 2.13 (2.16), 27 88.03 (37.16) / 5.41 (3.74), 23 Shen36 2016 Wuhan / China 2014–2015 Prospective NA 2007 Manual ROIs in maximal CBF Contralateral normal appearing white matter CBFmax / rCBFmax AC (II–IV), OD (II–III), 38.97 (17.47) / 0.97 (0.27), 25 73.93 (23.74) / 2.02 (0.61), 27 Soni26 2017 Lucknow / India 2013–2014 Prospective 19–71 (range) NA Manual ROIs in maximal CBF Contralateral normal appearing frontal gray matter and periventricular white matter rCBFmax Glioma, LGG, Diffuse AC, GBM NA / 1.23 (0.46), 3 NA / 14.1, 1 Warmuth14 2003 Berlin / Germany NA Prospective 46 (15) 1993 Manual ROI in maximal CBF Contralateral mirrored tumor region of interest rCBFmax / CBFmean / CBFmax Gangliogliom, pleomorphic AC, AC, AOD, AA, GBM 186 (101) / 0.68 (0.08), 7 408 (267) / 1.64 (0.65), 9 Weber27 2006 Heidelberg / Germany NA Prospective 57 (14) NA Manual ROI (at least 20 voxel) whole tumor / maximal CBF Contralateral normalappearing gray and white matter rCBFmax AC (II–IV) NA / NA, 9 NA / NA, 35 Wolf15 2005 Philadelphia / USA NA Prospective 50 (12) NA Manual ROI marked by 3D mask, and 33 voxel ROI in maximal CBF and mean CBF Global CBF over 12 sections, excluding tumor and edema CBFmax / rCBFmax / rCBFmean GG, OD, OA, AC, AOA, AA, GBM 30.93 (9.44) / 0.87 (0.15), 4 95.68 (70.5)/ 2.97 (1.78), 16 Xiao32 2015 Bejing / China 2012–2014 Prospective 43 (17) 2007 Manual ROIs in maximal CBF Normalization in cerebellum CBFmax / rCBFmax Pilocytic AC, AC (II–IV) NA / 1.81 (0.98), 19 NA / 4.51 (2.27), 24 Yang33 2016 Jinan / China NA Prospective 51 (15) 2007 Manual ROI (8–10 voxel) in solid tumor Contralateral normal appearing frontal white matter rmTI-ASL / nCBFmeanTI AC (gr II–IV) NA / 1.62 (1.97), 15 NA / 6.7 (5.1), 28 Zeng34 2017 Zhejinag / China 2013–2015 Retrospective 50 (13) 2007 Manual ROI in maximal CBF Contralateral ROI in gray matter CBFmax / rCBFmax AC (II–IV), OD (gr II–III), OA, AOA 90.69 (36.95) / 1.12 (0.48), 13 168.82 (70.43) / 2.19 (0.87), 45 Zhang35 2014 Jȕlich / Germany NA Prospective 56 (14) NA Manual ROI in whole tumor Contralateral normalappearing gray and white matter rCBFmean AC (II–IV), OD (II–III) NA / 20.84 (16.87), 4 NA / 21.24 (9.61), 4 First Author Year City / Country Years of Patient Inclusion Patient Inclusion (Retrospective or Prospective) Mean Age (SD) WHO Classification (Year of Publication) ROI Technique Reference Region (Location/Lobe, Gray or White Matter) Perfusion Parameters Tumors Included (WHO Grade) aCBF /rCBFmax In LGG Mean (SD), N Patients aCBF/rCBFmax HGG Mean (SD), N Patients Cebeci16 2014 Bursa / Turkey 2010–2013 Retrospective 47 (14) 2007 Manual ROI in maximal CBF Contralateral normal hemisphere rCBFmax / CBFmax / rSI GBM, AC (III), Gliosarcoma, OD (II), DNET, pilocystic AC 8.1 (28.1) / 0.96 (0.48), 13 23.65 (44.8) / 4.7 (1.38), 20 Fudaba28 2014 Oita / Japan 2010–2012 Retrospective 60 (17) 2007 Manual ROI (8 mm in diameter) in total tumor Mirrored normal white matter rCBFmin / max / mean AC (II–IV), OD (II–III), OA, GBM NA / NA, 9 NA / NA, 23 Furtner29 2014 Vienna / Austria 2009–2012 Prospective 54 (17) 2007 Manual ROI in whole tumor Mirrored normal contralateral healthy hemishpere nITS mean AC (II–IV) NA / 1.08 (0.39), 7 NA / 3.09 (2.57), 26 Kim30 2008 Suwon / South Korea 2005–2008 Prospective 43 (14) 2000 Manual ROIs in maximal tumor perfusion signal Contralateral white matter rITSmax AC (II–IV) NA / NA, 26 NA / NA, 35 Lehmann17 2010 Marseille / France 2007–2008 Prospective 58 (19) 2000 Manual ROI in whole tumor Contralateral white matter rCBFmean OD, AC, GBM, LGG NA / 1.09 (1.08), 4 NA / 2.26 (1.12), 4 Ma31 2017 Jiangsu / China 2014–2015 Prospective 46 (18) 2007 Manual ROIs in whole tumor (50–60 mm2) MIrrored contralateral normal matter CBFmean / rCBFmean GG, Pilocytic AC, AC gr I, AC (II–IV), AO, OD (gr III) 50.64 (35.89) / 2.13 (2.16), 27 88.03 (37.16) / 5.41 (3.74), 23 Shen36 2016 Wuhan / China 2014–2015 Prospective NA 2007 Manual ROIs in maximal CBF Contralateral normal appearing white matter CBFmax / rCBFmax AC (II–IV), OD (II–III), 38.97 (17.47) / 0.97 (0.27), 25 73.93 (23.74) / 2.02 (0.61), 27 Soni26 2017 Lucknow / India 2013–2014 Prospective 19–71 (range) NA Manual ROIs in maximal CBF Contralateral normal appearing frontal gray matter and periventricular white matter rCBFmax Glioma, LGG, Diffuse AC, GBM NA / 1.23 (0.46), 3 NA / 14.1, 1 Warmuth14 2003 Berlin / Germany NA Prospective 46 (15) 1993 Manual ROI in maximal CBF Contralateral mirrored tumor region of interest rCBFmax / CBFmean / CBFmax Gangliogliom, pleomorphic AC, AC, AOD, AA, GBM 186 (101) / 0.68 (0.08), 7 408 (267) / 1.64 (0.65), 9 Weber27 2006 Heidelberg / Germany NA Prospective 57 (14) NA Manual ROI (at least 20 voxel) whole tumor / maximal CBF Contralateral normalappearing gray and white matter rCBFmax AC (II–IV) NA / NA, 9 NA / NA, 35 Wolf15 2005 Philadelphia / USA NA Prospective 50 (12) NA Manual ROI marked by 3D mask, and 33 voxel ROI in maximal CBF and mean CBF Global CBF over 12 sections, excluding tumor and edema CBFmax / rCBFmax / rCBFmean GG, OD, OA, AC, AOA, AA, GBM 30.93 (9.44) / 0.87 (0.15), 4 95.68 (70.5)/ 2.97 (1.78), 16 Xiao32 2015 Bejing / China 2012–2014 Prospective 43 (17) 2007 Manual ROIs in maximal CBF Normalization in cerebellum CBFmax / rCBFmax Pilocytic AC, AC (II–IV) NA / 1.81 (0.98), 19 NA / 4.51 (2.27), 24 Yang33 2016 Jinan / China NA Prospective 51 (15) 2007 Manual ROI (8–10 voxel) in solid tumor Contralateral normal appearing frontal white matter rmTI-ASL / nCBFmeanTI AC (gr II–IV) NA / 1.62 (1.97), 15 NA / 6.7 (5.1), 28 Zeng34 2017 Zhejinag / China 2013–2015 Retrospective 50 (13) 2007 Manual ROI in maximal CBF Contralateral ROI in gray matter CBFmax / rCBFmax AC (II–IV), OD (gr II–III), OA, AOA 90.69 (36.95) / 1.12 (0.48), 13 168.82 (70.43) / 2.19 (0.87), 45 Zhang35 2014 Jȕlich / Germany NA Prospective 56 (14) NA Manual ROI in whole tumor Contralateral normalappearing gray and white matter rCBFmean AC (II–IV), OD (II–III) NA / 20.84 (16.87), 4 NA / 21.24 (9.61), 4 Gr, grade, AC; astrocytoma, DNET, dysembryoplastic neuroepithelial tumor, OA, oligoastrocytoma, GG, ganglioglioma, OD, oligodendroglioma, AOD, (anaplastic OD), AOA, anaplastic OA, GBM, glioblastoma; NA; not available. View Large Table 1 Characteristics of included studies First Author Year City / Country Years of Patient Inclusion Patient Inclusion (Retrospective or Prospective) Mean Age (SD) WHO Classification (Year of Publication) ROI Technique Reference Region (Location/Lobe, Gray or White Matter) Perfusion Parameters Tumors Included (WHO Grade) aCBF /rCBFmax In LGG Mean (SD), N Patients aCBF/rCBFmax HGG Mean (SD), N Patients Cebeci16 2014 Bursa / Turkey 2010–2013 Retrospective 47 (14) 2007 Manual ROI in maximal CBF Contralateral normal hemisphere rCBFmax / CBFmax / rSI GBM, AC (III), Gliosarcoma, OD (II), DNET, pilocystic AC 8.1 (28.1) / 0.96 (0.48), 13 23.65 (44.8) / 4.7 (1.38), 20 Fudaba28 2014 Oita / Japan 2010–2012 Retrospective 60 (17) 2007 Manual ROI (8 mm in diameter) in total tumor Mirrored normal white matter rCBFmin / max / mean AC (II–IV), OD (II–III), OA, GBM NA / NA, 9 NA / NA, 23 Furtner29 2014 Vienna / Austria 2009–2012 Prospective 54 (17) 2007 Manual ROI in whole tumor Mirrored normal contralateral healthy hemishpere nITS mean AC (II–IV) NA / 1.08 (0.39), 7 NA / 3.09 (2.57), 26 Kim30 2008 Suwon / South Korea 2005–2008 Prospective 43 (14) 2000 Manual ROIs in maximal tumor perfusion signal Contralateral white matter rITSmax AC (II–IV) NA / NA, 26 NA / NA, 35 Lehmann17 2010 Marseille / France 2007–2008 Prospective 58 (19) 2000 Manual ROI in whole tumor Contralateral white matter rCBFmean OD, AC, GBM, LGG NA / 1.09 (1.08), 4 NA / 2.26 (1.12), 4 Ma31 2017 Jiangsu / China 2014–2015 Prospective 46 (18) 2007 Manual ROIs in whole tumor (50–60 mm2) MIrrored contralateral normal matter CBFmean / rCBFmean GG, Pilocytic AC, AC gr I, AC (II–IV), AO, OD (gr III) 50.64 (35.89) / 2.13 (2.16), 27 88.03 (37.16) / 5.41 (3.74), 23 Shen36 2016 Wuhan / China 2014–2015 Prospective NA 2007 Manual ROIs in maximal CBF Contralateral normal appearing white matter CBFmax / rCBFmax AC (II–IV), OD (II–III), 38.97 (17.47) / 0.97 (0.27), 25 73.93 (23.74) / 2.02 (0.61), 27 Soni26 2017 Lucknow / India 2013–2014 Prospective 19–71 (range) NA Manual ROIs in maximal CBF Contralateral normal appearing frontal gray matter and periventricular white matter rCBFmax Glioma, LGG, Diffuse AC, GBM NA / 1.23 (0.46), 3 NA / 14.1, 1 Warmuth14 2003 Berlin / Germany NA Prospective 46 (15) 1993 Manual ROI in maximal CBF Contralateral mirrored tumor region of interest rCBFmax / CBFmean / CBFmax Gangliogliom, pleomorphic AC, AC, AOD, AA, GBM 186 (101) / 0.68 (0.08), 7 408 (267) / 1.64 (0.65), 9 Weber27 2006 Heidelberg / Germany NA Prospective 57 (14) NA Manual ROI (at least 20 voxel) whole tumor / maximal CBF Contralateral normalappearing gray and white matter rCBFmax AC (II–IV) NA / NA, 9 NA / NA, 35 Wolf15 2005 Philadelphia / USA NA Prospective 50 (12) NA Manual ROI marked by 3D mask, and 33 voxel ROI in maximal CBF and mean CBF Global CBF over 12 sections, excluding tumor and edema CBFmax / rCBFmax / rCBFmean GG, OD, OA, AC, AOA, AA, GBM 30.93 (9.44) / 0.87 (0.15), 4 95.68 (70.5)/ 2.97 (1.78), 16 Xiao32 2015 Bejing / China 2012–2014 Prospective 43 (17) 2007 Manual ROIs in maximal CBF Normalization in cerebellum CBFmax / rCBFmax Pilocytic AC, AC (II–IV) NA / 1.81 (0.98), 19 NA / 4.51 (2.27), 24 Yang33 2016 Jinan / China NA Prospective 51 (15) 2007 Manual ROI (8–10 voxel) in solid tumor Contralateral normal appearing frontal white matter rmTI-ASL / nCBFmeanTI AC (gr II–IV) NA / 1.62 (1.97), 15 NA / 6.7 (5.1), 28 Zeng34 2017 Zhejinag / China 2013–2015 Retrospective 50 (13) 2007 Manual ROI in maximal CBF Contralateral ROI in gray matter CBFmax / rCBFmax AC (II–IV), OD (gr II–III), OA, AOA 90.69 (36.95) / 1.12 (0.48), 13 168.82 (70.43) / 2.19 (0.87), 45 Zhang35 2014 Jȕlich / Germany NA Prospective 56 (14) NA Manual ROI in whole tumor Contralateral normalappearing gray and white matter rCBFmean AC (II–IV), OD (II–III) NA / 20.84 (16.87), 4 NA / 21.24 (9.61), 4 First Author Year City / Country Years of Patient Inclusion Patient Inclusion (Retrospective or Prospective) Mean Age (SD) WHO Classification (Year of Publication) ROI Technique Reference Region (Location/Lobe, Gray or White Matter) Perfusion Parameters Tumors Included (WHO Grade) aCBF /rCBFmax In LGG Mean (SD), N Patients aCBF/rCBFmax HGG Mean (SD), N Patients Cebeci16 2014 Bursa / Turkey 2010–2013 Retrospective 47 (14) 2007 Manual ROI in maximal CBF Contralateral normal hemisphere rCBFmax / CBFmax / rSI GBM, AC (III), Gliosarcoma, OD (II), DNET, pilocystic AC 8.1 (28.1) / 0.96 (0.48), 13 23.65 (44.8) / 4.7 (1.38), 20 Fudaba28 2014 Oita / Japan 2010–2012 Retrospective 60 (17) 2007 Manual ROI (8 mm in diameter) in total tumor Mirrored normal white matter rCBFmin / max / mean AC (II–IV), OD (II–III), OA, GBM NA / NA, 9 NA / NA, 23 Furtner29 2014 Vienna / Austria 2009–2012 Prospective 54 (17) 2007 Manual ROI in whole tumor Mirrored normal contralateral healthy hemishpere nITS mean AC (II–IV) NA / 1.08 (0.39), 7 NA / 3.09 (2.57), 26 Kim30 2008 Suwon / South Korea 2005–2008 Prospective 43 (14) 2000 Manual ROIs in maximal tumor perfusion signal Contralateral white matter rITSmax AC (II–IV) NA / NA, 26 NA / NA, 35 Lehmann17 2010 Marseille / France 2007–2008 Prospective 58 (19) 2000 Manual ROI in whole tumor Contralateral white matter rCBFmean OD, AC, GBM, LGG NA / 1.09 (1.08), 4 NA / 2.26 (1.12), 4 Ma31 2017 Jiangsu / China 2014–2015 Prospective 46 (18) 2007 Manual ROIs in whole tumor (50–60 mm2) MIrrored contralateral normal matter CBFmean / rCBFmean GG, Pilocytic AC, AC gr I, AC (II–IV), AO, OD (gr III) 50.64 (35.89) / 2.13 (2.16), 27 88.03 (37.16) / 5.41 (3.74), 23 Shen36 2016 Wuhan / China 2014–2015 Prospective NA 2007 Manual ROIs in maximal CBF Contralateral normal appearing white matter CBFmax / rCBFmax AC (II–IV), OD (II–III), 38.97 (17.47) / 0.97 (0.27), 25 73.93 (23.74) / 2.02 (0.61), 27 Soni26 2017 Lucknow / India 2013–2014 Prospective 19–71 (range) NA Manual ROIs in maximal CBF Contralateral normal appearing frontal gray matter and periventricular white matter rCBFmax Glioma, LGG, Diffuse AC, GBM NA / 1.23 (0.46), 3 NA / 14.1, 1 Warmuth14 2003 Berlin / Germany NA Prospective 46 (15) 1993 Manual ROI in maximal CBF Contralateral mirrored tumor region of interest rCBFmax / CBFmean / CBFmax Gangliogliom, pleomorphic AC, AC, AOD, AA, GBM 186 (101) / 0.68 (0.08), 7 408 (267) / 1.64 (0.65), 9 Weber27 2006 Heidelberg / Germany NA Prospective 57 (14) NA Manual ROI (at least 20 voxel) whole tumor / maximal CBF Contralateral normalappearing gray and white matter rCBFmax AC (II–IV) NA / NA, 9 NA / NA, 35 Wolf15 2005 Philadelphia / USA NA Prospective 50 (12) NA Manual ROI marked by 3D mask, and 33 voxel ROI in maximal CBF and mean CBF Global CBF over 12 sections, excluding tumor and edema CBFmax / rCBFmax / rCBFmean GG, OD, OA, AC, AOA, AA, GBM 30.93 (9.44) / 0.87 (0.15), 4 95.68 (70.5)/ 2.97 (1.78), 16 Xiao32 2015 Bejing / China 2012–2014 Prospective 43 (17) 2007 Manual ROIs in maximal CBF Normalization in cerebellum CBFmax / rCBFmax Pilocytic AC, AC (II–IV) NA / 1.81 (0.98), 19 NA / 4.51 (2.27), 24 Yang33 2016 Jinan / China NA Prospective 51 (15) 2007 Manual ROI (8–10 voxel) in solid tumor Contralateral normal appearing frontal white matter rmTI-ASL / nCBFmeanTI AC (gr II–IV) NA / 1.62 (1.97), 15 NA / 6.7 (5.1), 28 Zeng34 2017 Zhejinag / China 2013–2015 Retrospective 50 (13) 2007 Manual ROI in maximal CBF Contralateral ROI in gray matter CBFmax / rCBFmax AC (II–IV), OD (gr II–III), OA, AOA 90.69 (36.95) / 1.12 (0.48), 13 168.82 (70.43) / 2.19 (0.87), 45 Zhang35 2014 Jȕlich / Germany NA Prospective 56 (14) NA Manual ROI in whole tumor Contralateral normalappearing gray and white matter rCBFmean AC (II–IV), OD (II–III) NA / 20.84 (16.87), 4 NA / 21.24 (9.61), 4 Gr, grade, AC; astrocytoma, DNET, dysembryoplastic neuroepithelial tumor, OA, oligoastrocytoma, GG, ganglioglioma, OD, oligodendroglioma, AOD, (anaplastic OD), AOA, anaplastic OA, GBM, glioblastoma; NA; not available. View Large Table 2 Quality assessment according to QUADAS-2 Risk of Bias Applicability Concerns Patient Selection Index Test Reference Standard Flow and Timing Patient Selection Index Test Reference Standard Cebeci 2014 + _ ? ? + _ + Fudaba 2014 + _ ? ? + _ + Furtner 2014 + + ? ? ? _ + Kim 2008 + + ? ? ? _ ? Lehmann 2010 + _ ? ? + + _ Ma 2017 + _ ? ? + + + Shen 2016 + + ? ? + + + Soni 2017 + _ ? ? + + _ Warmuth 2005 + _ ? ? + _ ? Weber 2006 + _ ? + + _ _ Wolf 2005 ? _ ? ? + _ ? Xiao 2015 + + ? + ? + + Yang 2016 + + ? ? ? _ + Zeng 2017 + + ? + + + + Zhang 2014 ? _ ? ? + _ ? Risk of Bias Applicability Concerns Patient Selection Index Test Reference Standard Flow and Timing Patient Selection Index Test Reference Standard Cebeci 2014 + _ ? ? + _ + Fudaba 2014 + _ ? ? + _ + Furtner 2014 + + ? ? ? _ + Kim 2008 + + ? ? ? _ ? Lehmann 2010 + _ ? ? + + _ Ma 2017 + _ ? ? + + + Shen 2016 + + ? ? + + + Soni 2017 + _ ? ? + + _ Warmuth 2005 + _ ? ? + _ ? Weber 2006 + _ ? + + _ _ Wolf 2005 ? _ ? ? + _ ? Xiao 2015 + + ? + ? + + Yang 2016 + + ? ? ? _ + Zeng 2017 + + ? + + + + Zhang 2014 ? _ ? ? + _ ? + indicates low risk. _ indicates high risk. ? indicates unclear risk. Risk of bias: Patient selection: Low if stated consecutive or reported years of inclusion together with clear inclusion criteria. Unclear if no mention on consecutive series of patients. High if reported a non-consecutive series. Index test: Low if ASL was interpreted blinded. Unclear if no information on blinding but pre-defined cut-off was specified for positive test. High if exploratory cut-off and no information on blinding or if reported unblinded. Reference standard: Low if reported on blinded evaluation and WHO adherence. Unclear if no information on blinding. High if reported on un-blinded evaluation or if pathology diagnosis reported only part of the tumors in the study. Flow and timing: Low if <30 days between ASL and histopathology. Unclear if not reported. High if reported more than 6 months. Applicability concerns: Patient selection: Low if mixed tumour types. Unclear if tumor types not reported or same tumor type reported. High if other comparison than between high- and low-grade. Index test: Low if presented relative CBF from 3D pseudocontinous ASL. Unclear if not normalized CBF but pseudocontinous ASL. High if presented other perfusion metric than CBF or if pulsed ASL or continuous ASL was used. Reference standard: Low if tumors were classified according to WHO 2007 or later. Unclear if WHO but unspecified, or before 2007. High if no report on histopathological diagnosis classification system or if part of the cohort did not have histopathology. View Large Table 2 Quality assessment according to QUADAS-2 Risk of Bias Applicability Concerns Patient Selection Index Test Reference Standard Flow and Timing Patient Selection Index Test Reference Standard Cebeci 2014 + _ ? ? + _ + Fudaba 2014 + _ ? ? + _ + Furtner 2014 + + ? ? ? _ + Kim 2008 + + ? ? ? _ ? Lehmann 2010 + _ ? ? + + _ Ma 2017 + _ ? ? + + + Shen 2016 + + ? ? + + + Soni 2017 + _ ? ? + + _ Warmuth 2005 + _ ? ? + _ ? Weber 2006 + _ ? + + _ _ Wolf 2005 ? _ ? ? + _ ? Xiao 2015 + + ? + ? + + Yang 2016 + + ? ? ? _ + Zeng 2017 + + ? + + + + Zhang 2014 ? _ ? ? + _ ? Risk of Bias Applicability Concerns Patient Selection Index Test Reference Standard Flow and Timing Patient Selection Index Test Reference Standard Cebeci 2014 + _ ? ? + _ + Fudaba 2014 + _ ? ? + _ + Furtner 2014 + + ? ? ? _ + Kim 2008 + + ? ? ? _ ? Lehmann 2010 + _ ? ? + + _ Ma 2017 + _ ? ? + + + Shen 2016 + + ? ? + + + Soni 2017 + _ ? ? + + _ Warmuth 2005 + _ ? ? + _ ? Weber 2006 + _ ? + + _ _ Wolf 2005 ? _ ? ? + _ ? Xiao 2015 + + ? + ? + + Yang 2016 + + ? ? ? _ + Zeng 2017 + + ? + + + + Zhang 2014 ? _ ? ? + _ ? + indicates low risk. _ indicates high risk. ? indicates unclear risk. Risk of bias: Patient selection: Low if stated consecutive or reported years of inclusion together with clear inclusion criteria. Unclear if no mention on consecutive series of patients. High if reported a non-consecutive series. Index test: Low if ASL was interpreted blinded. Unclear if no information on blinding but pre-defined cut-off was specified for positive test. High if exploratory cut-off and no information on blinding or if reported unblinded. Reference standard: Low if reported on blinded evaluation and WHO adherence. Unclear if no information on blinding. High if reported on un-blinded evaluation or if pathology diagnosis reported only part of the tumors in the study. Flow and timing: Low if <30 days between ASL and histopathology. Unclear if not reported. High if reported more than 6 months. Applicability concerns: Patient selection: Low if mixed tumour types. Unclear if tumor types not reported or same tumor type reported. High if other comparison than between high- and low-grade. Index test: Low if presented relative CBF from 3D pseudocontinous ASL. Unclear if not normalized CBF but pseudocontinous ASL. High if presented other perfusion metric than CBF or if pulsed ASL or continuous ASL was used. Reference standard: Low if tumors were classified according to WHO 2007 or later. Unclear if WHO but unspecified, or before 2007. High if no report on histopathological diagnosis classification system or if part of the cohort did not have histopathology. View Large Fig. 2 View largeDownload slide (A) Distribution of sensitivity of rCBF to discriminate between LGG and HGG. (B) Distribution of specificity of rCBF to discriminate between LGG and HGG. Fig. 2 View largeDownload slide (A) Distribution of sensitivity of rCBF to discriminate between LGG and HGG. (B) Distribution of specificity of rCBF to discriminate between LGG and HGG. The bivariate meta-regression assessing potential moderator variables for the primary outcome (diagnostic performance) sensitivity revealed no moderating effect for the following variables; ASL type (PASL, pCASL, or CASL), WHO classification (year published), tesla strength (1.5 or 3), and perfusion parameter (rCBFmax, rCBFmean, nCBFmTI, nITSmax, nITSmean) (P = 0.15–0.65). However, use of the perfusion parameters nITSmax and rCBFmax had a significant effect on the specificity in the bivariate meta-regression. The parameter nITSmax was associated with a lower specificity (P = 0.007) compared with rCBFmean/max and nITSmean. The usage of rCBFmax was associated with a higher specificity (P = 0.009) compared with rCBFmean and nITmean/max. The funnel plot assymetry test did not detect any evidence for publication bias (P = 0.53). Meta-analytic evaluation revealed a mean difference in rCBFmax between HGG and LGG of 3.23 (95% CI: 0.90–5.76, P = 0.007, I2 = 99%). The mean difference for absolute CBF between HGG and LGG was 45.50 (95% CI: 26.06–64.94, P < 0.001, I2 = 70%). Trial sequential analysis revealed that the required sample size to detect a statistical difference between LGG and HGG was n = 187. Hence, this meta-analysis was statistically powered including 505 patients (Figure 1). Diagnostic Performance In the main analysis, the pooled diagnostic performance for differentiation between high and low grade glioma irrespective of the perfusion measure (CBF or SI) is exemplified by the AUC. The AUC was 0.90 with a summary sensitivity ranging from 0.79 to 0.90, and a summary specificity ranging from 0.72 to 0.87 (Figure 3, Table 3). Table 3 Main and subgroup analysis Participants No. Studies AUC Summary Sensitivity (95% CI) Summary Specificity (95% CI) Main analysis 505 15 0.90 0.86 (0.79–0.90) 0.80 (0.72–0.89) Subgroup analysis Astrocytomas only 243 9 0.89 0.86 (0.79–0.91) 0.79 (0.63–0.89) rCBF max 302 9 0.90 0.87 (0.77–0.93) 0.85 (0.76–0.91) rCBF mean 118 5 0.78 0.74 (0.63–0.83) 0.74 (0.60–0.84) ASL method:  PASL 227 7 0.90 0.85 (0.71–0.91) 0.83 (0.69–0.92)  3D pCASL 258 7 0.88 0.86 (0.79–0.91) 0.80 (0.65–0.87) Sensitivity analysis from QUADAS-2 items Internal validity  Excluding studies with high or unclear risk of bias related to patient selection 477 13 0.90 0.85 (0.78–0.90) 0.81 (0.73–0.88)  Excluding studies with high or unclear risk of bias related to the index test 290 6 0.89 0.86 (0.79–0.91) 0.85 (0.69–0.94)  Excluding studies with high or unclear risk of bias related to the flow and timing 145 3 0.87 0.88 (0.77–0.95) 0.83 (0.68–0.92) External validity  Excluding studies with high risk of bias or unclear bias related to patient selection within the domain of external validity 325 11 0.88 0.86 (0.76–0.92) 0.81 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to index test within the domain of external validity* 215 6 0.84 0.85 (0.77–0.90) 0.82 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to reference standard within the domain of external validity** 344 8 0.89 0.84 (0.75–0.90) 0.80 (0.69–0.88) Participants No. Studies AUC Summary Sensitivity (95% CI) Summary Specificity (95% CI) Main analysis 505 15 0.90 0.86 (0.79–0.90) 0.80 (0.72–0.89) Subgroup analysis Astrocytomas only 243 9 0.89 0.86 (0.79–0.91) 0.79 (0.63–0.89) rCBF max 302 9 0.90 0.87 (0.77–0.93) 0.85 (0.76–0.91) rCBF mean 118 5 0.78 0.74 (0.63–0.83) 0.74 (0.60–0.84) ASL method:  PASL 227 7 0.90 0.85 (0.71–0.91) 0.83 (0.69–0.92)  3D pCASL 258 7 0.88 0.86 (0.79–0.91) 0.80 (0.65–0.87) Sensitivity analysis from QUADAS-2 items Internal validity  Excluding studies with high or unclear risk of bias related to patient selection 477 13 0.90 0.85 (0.78–0.90) 0.81 (0.73–0.88)  Excluding studies with high or unclear risk of bias related to the index test 290 6 0.89 0.86 (0.79–0.91) 0.85 (0.69–0.94)  Excluding studies with high or unclear risk of bias related to the flow and timing 145 3 0.87 0.88 (0.77–0.95) 0.83 (0.68–0.92) External validity  Excluding studies with high risk of bias or unclear bias related to patient selection within the domain of external validity 325 11 0.88 0.86 (0.76–0.92) 0.81 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to index test within the domain of external validity* 215 6 0.84 0.85 (0.77–0.90) 0.82 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to reference standard within the domain of external validity** 344 8 0.89 0.84 (0.75–0.90) 0.80 (0.69–0.88) *Related to the applicability of the index test in relation to the aim of the meta-analysis. **Related to the applicability of the reference standard in relation to the aim of the meta-analysis. AUC; Area under the receiver operating characteristics curve; QUADAS-2; a revised tool for the quality assessment of diagnostic accuracy studies, PASL; pulsed arterial spin labeling, pCASL; pseudo-continous arterial spin labeling, rCBF; relative cerebral blood flow. View Large Table 3 Main and subgroup analysis Participants No. Studies AUC Summary Sensitivity (95% CI) Summary Specificity (95% CI) Main analysis 505 15 0.90 0.86 (0.79–0.90) 0.80 (0.72–0.89) Subgroup analysis Astrocytomas only 243 9 0.89 0.86 (0.79–0.91) 0.79 (0.63–0.89) rCBF max 302 9 0.90 0.87 (0.77–0.93) 0.85 (0.76–0.91) rCBF mean 118 5 0.78 0.74 (0.63–0.83) 0.74 (0.60–0.84) ASL method:  PASL 227 7 0.90 0.85 (0.71–0.91) 0.83 (0.69–0.92)  3D pCASL 258 7 0.88 0.86 (0.79–0.91) 0.80 (0.65–0.87) Sensitivity analysis from QUADAS-2 items Internal validity  Excluding studies with high or unclear risk of bias related to patient selection 477 13 0.90 0.85 (0.78–0.90) 0.81 (0.73–0.88)  Excluding studies with high or unclear risk of bias related to the index test 290 6 0.89 0.86 (0.79–0.91) 0.85 (0.69–0.94)  Excluding studies with high or unclear risk of bias related to the flow and timing 145 3 0.87 0.88 (0.77–0.95) 0.83 (0.68–0.92) External validity  Excluding studies with high risk of bias or unclear bias related to patient selection within the domain of external validity 325 11 0.88 0.86 (0.76–0.92) 0.81 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to index test within the domain of external validity* 215 6 0.84 0.85 (0.77–0.90) 0.82 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to reference standard within the domain of external validity** 344 8 0.89 0.84 (0.75–0.90) 0.80 (0.69–0.88) Participants No. Studies AUC Summary Sensitivity (95% CI) Summary Specificity (95% CI) Main analysis 505 15 0.90 0.86 (0.79–0.90) 0.80 (0.72–0.89) Subgroup analysis Astrocytomas only 243 9 0.89 0.86 (0.79–0.91) 0.79 (0.63–0.89) rCBF max 302 9 0.90 0.87 (0.77–0.93) 0.85 (0.76–0.91) rCBF mean 118 5 0.78 0.74 (0.63–0.83) 0.74 (0.60–0.84) ASL method:  PASL 227 7 0.90 0.85 (0.71–0.91) 0.83 (0.69–0.92)  3D pCASL 258 7 0.88 0.86 (0.79–0.91) 0.80 (0.65–0.87) Sensitivity analysis from QUADAS-2 items Internal validity  Excluding studies with high or unclear risk of bias related to patient selection 477 13 0.90 0.85 (0.78–0.90) 0.81 (0.73–0.88)  Excluding studies with high or unclear risk of bias related to the index test 290 6 0.89 0.86 (0.79–0.91) 0.85 (0.69–0.94)  Excluding studies with high or unclear risk of bias related to the flow and timing 145 3 0.87 0.88 (0.77–0.95) 0.83 (0.68–0.92) External validity  Excluding studies with high risk of bias or unclear bias related to patient selection within the domain of external validity 325 11 0.88 0.86 (0.76–0.92) 0.81 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to index test within the domain of external validity* 215 6 0.84 0.85 (0.77–0.90) 0.82 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to reference standard within the domain of external validity** 344 8 0.89 0.84 (0.75–0.90) 0.80 (0.69–0.88) *Related to the applicability of the index test in relation to the aim of the meta-analysis. **Related to the applicability of the reference standard in relation to the aim of the meta-analysis. AUC; Area under the receiver operating characteristics curve; QUADAS-2; a revised tool for the quality assessment of diagnostic accuracy studies, PASL; pulsed arterial spin labeling, pCASL; pseudo-continous arterial spin labeling, rCBF; relative cerebral blood flow. View Large Fig. 3 View largeDownload slide Bivariate summary receiver operating characteristic curve (SROC). Fig. 3 View largeDownload slide Bivariate summary receiver operating characteristic curve (SROC). When we only included studies that used rCBFmax to discriminate between LGG and HGG, we found AUC of 0.90, with sensitivity ranging between 0.77 and 0.93 and specificity between 0.76 and 0.91. When we analyzed studies presenting data on mean rCBF, the AUC decreased to 0.78 with a summary sensitivity between 0.63 and 0.83 and summary specificity between 0.60 and 0.84. In the subgroup analyses on studies using the PASL technique, the diagnostic performance for differentiation between LGG and HGG, the AUC was 0.90 with a pooled sensitivity between 0.71 and 0.93, and pooled sensitivity between 0.69 and 0.92. In the subgroup analysis on studies using pCASL, the diagnostic performance AUC was 0.88 with a corresponding summary sensitivity between 0.79 and 0.91, and specificity between 0.65 and 0.87. In order to assess the degree of robustness, we conducted sensitivity analyses by excluding studies with a high or unclear risk of bias in the domain of external validity related to patient selection. This yielded an AUC of 0.88 and a summary sensitivity and specificity between 0.76 and 0.92 and between 0.72 and 0.87, respectively. When excluding studies with high or unclear risk of bias related to the reference standard, AUC was 0.89 with a summary sensitivity of 0.75 to 0.90 and specificity of 0.69 to 0.88. Discussion This study provides evidence that ASL MR imaging has an excellent diagnostic performance of 0.90 with pooled sensitivity of 0.86 and specificity of 0.80 for the differentiation of HGG from LGG. Furthermore, it is non-invasive and does not require a contrast agent. Subgroup analyses showed a similar high diagnostic performance for PASL as for PCASL for HGG detection. We showed that the parameter maximum relative CBF was associated with a higher diagnostic performance compared with the parameter mean relative CBF values, as also suggested by the meta-regression analyses. Sensitivity analyses based on internal and external validity assessment did not alter the results and strengthen the robustness of the conclusion. TSA revealed that the study was adequately powered statistically for detecting a difference between LGG and HGG. The strengths of this study are related to a broad search strategy, including several databases with a large number of studies, and an adequately powered meta-analysis including a large study cohort with aggregated data from 505 patients. Due to this sample size we were able to pool the data in subgroup analyses. Further, relevant meta-regression could be performed to support the findings. In accordance with the current high standards for diagnostic meta-analysis, this study was prospectively registered in PROSPERO38 and used updated guidelines for reporting meta-analysis20 and current recommendations for statistical methods in diagnostic meta-analysis.23 Furthermore, all steps in the meta-analysis were conducted by at least 2 researchers. The electronic search was conducted in well-established databases by one experienced librarian, who also searched unpublished gray literature. There were no signs of publication bias in the study. In relation to our study, a recent meta-analysis by Kong and colleagues analyzed CBF in LGG and HGG and found that HGG had a significantly higher CBF compared with LGG,19 which is in line with our data. However, diagnostic performance evaluation was not assessed in this study, necessitating clarification of the diagnostic value for ASL in identifying HGG. Dynamic susceptibility MR perfusion, an MR technique utilizing a gadolinium-based contrast agent, has shown an AUC of 0.77 to discriminate between grade II and grade III glioma.8 However, one strength of the ASL technique compared with other contrast agent–based perfusion techniques is that it does not require a gadolinium contrast agent, which has shown on MR images to accumulate in the brain9 and in autopsy studies.39,40 Furthermore, gadolinium-based contrast agents are associated with the considerable cost of approximately $200 per examination (Swedish retail price, standard dose—male 70 kg). An examination without the need of contrast agent also does not require testing for plasma creatinine and does not involve a risk for gadolinium-related toxicity.41 Our results in combination with the lesser performance of DSC perfusion can suggest that there is no added value for the purpose of glioma grading using a gadolinium contrast agent. However, most glioma patients require gadolinium to detect contrast enhancing lesions, which although imperfect, is commonly used to describe and characterize tumors, plan surgical resections, and measure Response Assessment in Neuro-Oncology or other standardized response to determine and evaluate treatment. Mean acquisition time for the ASL techniques in the present study was 300 seconds and thus does not prolong the MR scanning time substantially. Proposed drawbacks or limitations of the ASL technique are the signal-to-noise ratio, which is low in the cerebral white matter,42,43 as well as aging affecting cerebral perfusion. Based on the data from 505 patients reviewed here, there appears to be little evidence for this claim, and there was only a small difference in the diagnostic performance between PASL and pCASL. We acknowledge that there are a number of limitations in this study. Firstly, none of the included studies used the current WHO classification system from 2016.2 One study did report on molecular genetic data; however, there was no separation between metastases and glioma, and therefore this study was excluded.44 We did try to minimize this effect by excluding oligoastrocytomas when possible. The new WHO classification does not introduce, per definition, any change regarding glioma grading, therefore the comparison between HGG and LGG is still valid, despite the new classification of tumor subtypes. The included studies used several WHO classification systems; however, the meta-regression analyses did not detect any modulating effect on the outcome. Most of the studies used a 3T scanner; however, this did not seem to alter the outcome. A further limitation is the heterogeneous placement of ROI in the different studies and in the reference regions. Furthermore, there were substantial differences in acquisition and in postprocessing algorithms, if specified at all. In addition, given that grade II and grade III might be a further diagnostic challenge to discriminate between than HGG and LGG, this is a limitation of the study. Moreover, the bivariate meta-regression analyses lack a multivariable implementation that can take several potential moderating variables in the same analysis into consideration. Future studies would need to adopt the WHO classification in order to further assess the role of ASL for glioma grading. ASL MR imaging has an excellent diagnostic performance for differentiation between HGG and LGG. Given the low cost, absence of side effects, and efficacy, ASL MR imaging should be implemented in the routine workup of glioma. Supplementary Material Supplementary material is available online at Neuro-Oncology (http://neuro-oncology.oxfordjournals.org/). Funding None. Acknowledgments We acknowledge the assistance of Klas Moberg, Karolinska Insitutet University Library. Conflict of interest statement. None declared. Authorship Contribution. An.F.D., F.D.L., and Al.F.D. extracted data; An.F.D. and Al.F.D. conducted analyses; all authors contributed to critical assessment of the content and contributed to the preparation and the writing of the manuscript. All authors have read and approved the final version of the manuscript. References 1. Ostrom QT , Gittleman H , Liao P , et al. CBTRUS statistical report: primary brain and central nervous system tumors diagnosed in the United States in 2007–2011 . Neuro Oncol . 2014 ; 16 ( Suppl 4 ): iv1 – iv63 . Google Scholar CrossRef Search ADS PubMed 2. Louis DN , Perry A , Reifenberger G , et al. The 2016 World Health Organization classification of tumors of the central nervous system: a summary . Acta Neuropathol . 2016 ; 131 ( 6 ): 803 – 820 . Google Scholar CrossRef Search ADS PubMed 3. Smoll NR , Gautschi OP , Schatlo B , Schaller K , Weber DC . Relative survival of patients with supratentorial low-grade gliomas . Neuro Oncol . 2012 ; 14 ( 8 ): 1062 – 1069 . Google Scholar CrossRef Search ADS PubMed 4. Ohgaki H , Kleihues P . Population-based studies on incidence, survival rates, and genetic alterations in astrocytic and oligodendroglial gliomas . J Neuropathol Exp Neurol . 2005 ; 64 ( 6 ): 479 – 489 . Google Scholar CrossRef Search ADS PubMed 5. Stupp R , Mason WP , van den Bent MJ , et al. ; European Organisation for Research and Treatment of Cancer Brain Tumor and Radiotherapy Groups; National Cancer Institute of Canada Clinical Trials Group . Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma . N Engl J Med . 2005 ; 352 ( 10 ): 987 – 996 . Google Scholar CrossRef Search ADS PubMed 6. Tobin WO , Meyer FB , Keegan BM . Diagnostic yield and safety of cerebellar and brainstem parenchymal biopsy . World Neurosurg . 2015 ; 84 ( 6 ): 1973 – 1976 . Google Scholar CrossRef Search ADS PubMed 7. Jackson RJ , Fuller GN , Abi-Said D , et al. Limitations of stereotactic biopsy in the initial management of gliomas . Neuro Oncol . 2001 ; 3 ( 3 ): 193 – 200 . Google Scholar CrossRef Search ADS PubMed 8. Delgado AF , Delgado AF . Discrimination between glioma grades II and III using dynamic susceptibility perfusion MRI: a meta-analysis . AJNR Am J Neuroradiol . 2017 ; 38 ( 7 ): 1348 – 1355 . Google Scholar CrossRef Search ADS PubMed 9. Flood TF , Stence NV , Maloney JA , Mirsky DM . Pediatric brain: repeated exposure to linear gadolinium-based contrast material is associated with increased signal intensity at unenhanced T1-weighted MR imaging . Radiology . 2017 ; 282 ( 1 ): 222 – 228 . Google Scholar CrossRef Search ADS PubMed 10. Williams DS , Detre JA , Leigh JS , Koretsky AP . Magnetic resonance imaging of perfusion using spin inversion of arterial water . Proc Natl Acad Sci U S A . 1992 ; 89 ( 1 ): 212 – 216 . Google Scholar CrossRef Search ADS PubMed 11. Detre JA , Leigh JS , Williams DS , Koretsky AP . Perfusion imaging . Magn Reson Med . 1992 ; 23 ( 1 ): 37 – 45 . Google Scholar CrossRef Search ADS PubMed 12. Haller S , Zaharchuk G , Thomas DL , Lovblad KO , Barkhof F , Golay X . Arterial spin labeling perfusion of the brain: emerging clinical applications . Radiology . 2016 ; 281 ( 2 ): 337 – 356 . Google Scholar CrossRef Search ADS PubMed 13. Alsop DC , Detre JA , Golay X , et al. Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: a consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia . Magn Reson Med . 2015 ; 73 ( 1 ): 102 – 116 . Google Scholar CrossRef Search ADS PubMed 14. Warmuth C , Gunther M , Zimmer C . Quantification of blood flow in brain tumors: comparison of arterial spin labeling and dynamic susceptibility-weighted contrast-enhanced MR imaging . Radiology . 2003 ; 228 ( 2 ): 523 – 532 . Google Scholar CrossRef Search ADS PubMed 15. Wolf RL , Wang J , Wang S , et al. Grading of CNS neoplasms using continuous arterial spin labeled perfusion MR imaging at 3 Tesla . J Magn Reson Imaging . 2005 ; 22 ( 4 ): 475 – 482 . Google Scholar CrossRef Search ADS PubMed 16. Cebeci H , Aydin O , Ozturk-Isik E , et al. Assesment of perfusion in glial tumors with arterial spin labeling; comparison with dynamic susceptibility contrast method . Eur J Radiol . 2014 ; 83 ( 10 ): 1914 – 1919 . Google Scholar CrossRef Search ADS PubMed 17. Lehmann P , Monet P , de Marco G , et al. A comparative study of perfusion measurement in brain tumours at 3 Tesla MR: arterial spin labeling versus dynamic susceptibility contrast-enhanced MRI . Eur Neurol . 2010 ; 64 ( 1 ): 21 – 26 . Google Scholar CrossRef Search ADS PubMed 18. Roy B , Awasthi R , Bindal A , et al. Comparative evaluation of 3-dimensional pseudocontinuous arterial spin labeling with dynamic contrast-enhanced perfusion magnetic resonance imaging in grading of human glioma . J Comput Assist Tomogr . 2013 ; 37 ( 3 ): 321 – 326 . Google Scholar CrossRef Search ADS PubMed 19. Kong L , Chen H , Yang Y , Chen L . A meta-analysis of arterial spin labelling perfusion values for the prediction of glioma grade . Clin Radiol . 2017 ; 72 ( 3 ): 255 – 261 . Google Scholar CrossRef Search ADS PubMed 20. Liberati A , Altman DG , Tetzlaff J , et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration . J Clin Epidemiol . 2009 ; 62 ( 10 ): e1 – 34 . Google Scholar CrossRef Search ADS PubMed 21. Petra Macaskill CG , Deeks J , Harbord R , Takwoingi Y. Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy . Version 1.0.0. The Cochrane Collaboration; 2013. http://srdta.cochrane.org/. 22. Whiting PF , Rutjes AW , Westwood ME , et al. ; QUADAS-2 Group . QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies . Ann Intern Med . 2011 ; 155 ( 8 ): 529 – 536 . Google Scholar CrossRef Search ADS PubMed 23. McGrath TA , McInnes MD , Korevaar DA , Bossuyt PM . Meta-analyses of diagnostic accuracy in imaging journals: analysis of pooling techniques and their effect on summary estimates of diagnostic accuracy . Radiology . 2016 ; 281 ( 1 ): 78 – 85 . Google Scholar CrossRef Search ADS PubMed 24. Egger M , Davey Smith G , Schneider M , Minder C . Bias in meta-analysis detected by a simple, graphical test . BMJ . 1997 ; 315 ( 7109 ): 629 – 634 . Google Scholar CrossRef Search ADS PubMed 25. Pogue J , Yusuf S . Overcoming the limitations of current meta-analysis of randomised controlled trials . Lancet . 1998 ; 351 ( 9095 ): 47 – 52 . Google Scholar CrossRef Search ADS PubMed 26. Soni N , Dhanota DPS , Kumar S , Jaiswal AK , Srivastava AK . Perfusion MR imaging of enhancing brain tumors: comparison of arterial spin labeling technique with dynamic susceptibility contrast technique . Neurol India . 2017 ; 65 ( 5 ): 1046 – 1052 . Google Scholar CrossRef Search ADS PubMed 27. Weber MA , Zoubaa S , Schlieter M , et al. Diagnostic performance of spectroscopic and perfusion MRI for distinction of brain tumors . Neurology . 2006 ; 66 ( 12 ): 1899 – 1906 . Google Scholar CrossRef Search ADS PubMed 28. Fudaba H , Shimomura T , Abe T , et al. Comparison of multiple parameters obtained on 3T pulsed arterial spin-labeling, diffusion tensor imaging, and MRS and the Ki-67 labeling index in evaluating glioma grading . AJNR Am J Neuroradiol . 2014 ; 35 ( 11 ): 2091 – 2098 . Google Scholar CrossRef Search ADS PubMed 29. Furtner J , Schöpf V , Schewzow K , et al. Arterial spin-labeling assessment of normalized vascular intratumoral signal intensity as a predictor of histologic grade of astrocytic neoplasms . AJNR Am J Neuroradiol . 2014 ; 35 ( 3 ): 482 – 489 . Google Scholar CrossRef Search ADS PubMed 30. Kim MJ , Kim HS , Kim JH , Cho KG , Kim SY . Diagnostic accuracy and interobserver variability of pulsed arterial spin labeling for glioma grading . Acta Radiol . 2008 ; 49 ( 4 ): 450 – 457 . Google Scholar CrossRef Search ADS PubMed 31. Ma H , Wang Z , Xu K , et al. Three-dimensional arterial spin labeling imaging and dynamic susceptibility contrast perfusion-weighted imaging value in diagnosing glioma grade prior to surgery . Exp Ther Med . 2017 ; 13 ( 6 ): 2691 – 2698 . Google Scholar CrossRef Search ADS PubMed 32. Xiao HF , Chen ZY , Lou X , et al. Astrocytic tumour grading: a comparative study of three-dimensional pseudocontinuous arterial spin labelling, dynamic susceptibility contrast-enhanced perfusion-weighted imaging, and diffusion-weighted imaging . Eur Radiol . 2015 ; 25 ( 12 ): 3423 – 3430 . Google Scholar CrossRef Search ADS PubMed 33. Yang S , Zhao B , Wang G , et al. Improving the grading accuracy of astrocytic neoplasms noninvasively by combining timing information with cerebral blood flow: a multi-TI arterial spin-labeling MR Imaging study . AJNR Am J Neuroradiol . 2016 ; 37 ( 12 ): 2209 – 2216 . Google Scholar CrossRef Search ADS PubMed 34. Zeng Q , Jiang B , Shi F , Ling C , Dong F , Zhang J . 3D pseudocontinuous arterial spin-labeling MR imaging in the preoperative evaluation of gliomas . AJNR Am J Neuroradiol . 2017 ; 38 ( 10 ): 1876 – 1883 . Google Scholar CrossRef Search ADS PubMed 35. Zhang K , Langen KJ , Neuner I , et al. Relationship of regional cerebral blood flow and kinetic behaviour of O-(2-(18)F-fluoroethyl)-L-tyrosine uptake in cerebral gliomas . Nucl Med Commun . 2014 ; 35 ( 3 ): 245 – 251 . Google Scholar CrossRef Search ADS PubMed 36. Shen N , Zhao L , Jiang J , et al. Intravoxel incoherent motion diffusion-weighted imaging analysis of diffusion and microperfusion in grading gliomas and comparison with arterial spin labeling for evaluation of tumor perfusion . J Magn Reson Imaging . 2016 ; 44 ( 3 ): 620 – 632 . Google Scholar CrossRef Search ADS PubMed 37. Yang M , Wang S , Zhao Y , Zhao J . Management of intracranial aneurysm in children: clipped and coiled . Childs Nerv Syst . 2008 ; 24 ( 9 ): 1005 – 1012 . Google Scholar CrossRef Search ADS PubMed 38. Prospero International Prospective Register of Systematic Reviews . https://www.crd.york.ac.uk/PROSPERO/. 39. McDonald JS , McDonald RJ , Jentoft ME , et al. Intracranial gadolinium deposition following gadodiamide-enhanced magnetic resonance imaging in pediatric patients: a case-control study . JAMA Pediatr . 2017 ; 171 ( 7 ): 705 – 707 . Google Scholar CrossRef Search ADS PubMed 40. McDonald RJ , McDonald JS , Kallmes DF , et al. Gadolinium deposition in human brain tissues after contrast-enhanced MR imaging in adult patients without intracranial abnormalities . Radiology . 2017 ; 285 ( 2 ): 546 – 554 . Google Scholar CrossRef Search ADS PubMed 41. Marckmann P , Skov L , Rossen K , et al. Nephrogenic systemic fibrosis: suspected causative role of gadodiamide used for contrast-enhanced magnetic resonance imaging . J Am Soc Nephrol . 2006 ; 17 ( 9 ): 2359 – 2362 . Google Scholar CrossRef Search ADS PubMed 42. Chen Y , Wang DJ , Detre JA . Test-retest reliability of arterial spin labeling with common labeling strategies . J Magn Reson Imaging . 2011 ; 33 ( 4 ): 940 – 949 . Google Scholar CrossRef Search ADS PubMed 43. Spann SM , Kazimierski KS , Aigner CS , Kraiger M , Bredies K , Stollberger R . Spatio-temporal TGV denoising for ASL perfusion imaging . Neuroimage . 2017 ; 157 : 81 – 96 . Google Scholar CrossRef Search ADS PubMed 44. Brendle C , Hempel JM , Schittenhelm J , et al. Glioma grading and determination of IDH mutation status and ATRX loss by DCE and ASL perfusion . Clin Neuroradiol . 2017 ;doi: https://doi.org/10.1007/s00062-017-0590-z . © The Author(s) 2018. Published by Oxford University Press on behalf of the Society for Neuro-Oncology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 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) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Neuro-Oncology Oxford University Press

Arterial spin labeling MR imaging for differentiation between high- and low-grade glioma—a meta-analysis

Loading next page...
 
/lp/ou_press/arterial-spin-labeling-mr-imaging-for-differentiation-between-high-and-k98G2fxpvt
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Society for Neuro-Oncology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1522-8517
eISSN
1523-5866
D.O.I.
10.1093/neuonc/noy095
Publisher site
See Article on Publisher Site

Abstract

Abstract Background Arterial spin labeling is an MR imaging technique that measures cerebral blood flow (CBF) non-invasively. The aim of the study is to assess the diagnostic performance of arterial spin labeling (ASL) MR imaging for differentiation between high-grade glioma and low-grade glioma. Methods Cochrane Library, Embase, Medline, and Web of Science Core Collection were searched. Study selection ended November 2017. This study was prospectively registered in PROSPERO (CRD42017080885). Two authors screened all titles and abstracts for possible inclusion. Data were extracted independently by 2 authors. Bivariate random effects meta-analysis was used to describe summary receiver operating characteristics. Trial sequential analysis (TSA) was performed. Results In total, 15 studies with 505 patients were included. The diagnostic performance of ASL CBF for glioma grading was 0.90 with summary sensitivity 0.89 (0.79–0.90) and specificity 0.80 (0.72–0.89). The diagnostic performance was similar between pulsed ASL (AUC 0.90) with a sensitivity 0.85 (0.71–0.91) and specificity 0.83 (0.69–0.92) and pseudocontinuous ASL (AUC 0.88) with a sensitivity 0.86 (0.79–0.91) and specificity 0.80 (0.65–0.87). In astrocytomas, the diagnostic performance was 0.89 with sensitivity 0.86 (0.79 to 0.91) and specificity 0.79 (0.63 to 0.89). Sensitivity analysis confirmed the robustness of the findings. TSA revealed that the meta-analysis was adequately powered. Conclusion Arterial spin labeling MR imaging had an excellent diagnostic accuracy for differentiation between high-grade and low-grade glioma. Given its low cost, non-invasiveness, and efficacy, ASL MR imaging should be considered for implementation in the routine workup of patients with glioma. arterial spin labeling, brain tumors, CNS, glioma, imaging Importance of the study Arterial spin labeling is an MR imaging technique for measurement of CBF without administration of contrast agent. The summary diagnostic performance of ASL to discriminate between glioma grades is unknown, the individual trials assessing this having been limited and hampered by a small number of participants. Accurate diagnosis is important for prognostication, treatment planning, and assessment of treatment response for glioma. Biopsy is highly invasive and carries the risk of undersampling. MRI plays an important role in this regard. This meta-analysis provides evidence that ASL MR imaging has an excellent diagnostic accuracy for differentiation between high-grade and low-grade glioma, in addition to its low cost and non-invasiveness. Its high diagnostic performance was independent of the specific ASL technique used. ASL can be considered for implementation in the routine workup of patients with glioma when gadolinium administration is unwanted or unnecessary. Gliomas are the most common primary brain tumors and comprise 80% of all malignant tumors in the brain. The average incidence of glioma was 6.61 (6.57% to 6.6%) per 100000 in 2007–2011 in the United States.1 Gliomas originate from the cerebral glial cells and are classified according to their cell type based on histological appearance and biomolecular status, most commonly into astrocytoma or oligodendroglioma, and according to their grade, low-grade (World Health Organization [WHO] grades I–II) or high-grade (III–IV).2 The 10-year survival for low-grade gliomas in adults is approximately 43%3; oligodendrogliomas have been reported associated with a relatively more favorable outcome compared with astrocytomas.4 Grade III glioma has a significantly worse prognosis compared with glioma grades I and II. The most common intracranial neoplasm, WHO grade IV glioma, also named glioblastoma, has a median survival of 15 months.5 Accurate assessment of tumor grade and type is crucial for optimal treatment planning. Tumor diagnosis, classification, and grading is established by biopsy or excision, which is associated with considerable risk, including death, especially in central areas of the brain.6 In addition, diagnosis from biopsy sampling or subtotal resection is potentially limited by the sample not being representative, leading to understaging of tumor grade7; in turn, this may mislead treatment planning toward a more conservative strategy. Furthermore, patients who are inoperable or unwilling to undergo surgery, or in whom tumors are located in eloquent areas, might not be suitable for surgical intervention even though they qualify for adjuvant therapy, including chemotherapy and or radiotherapy. Thus, there is a strong need for alternative methods for grading brain tumors preoperatively with imaging as a reasonable alternative even if there is no ideal imaging technique as of today. MRI, which is done routinely in the presurgical workup of patients with brain cancer, is a reasonable candidate for assessment of tumor grade in gliomas. Potential advantages of MR are its ability to non-invasively map the entire tumor and quantify cerebral blood flow (CBF), a marker of angiogenesis and thus of tumor grade. The diagnostic performance of dynamic susceptibility contrast (DSC) perfusion to discriminate between WHO grades II and III is limited8; in addition the technique requires administration of a gadolinium contrast agent with considerable costs; and more importantly, recent evidence of gadolinium deposition in the brain parenchyma has been reported.9 Arterial spin labeling (ASL) is a completely non-invasive MR imaging method mainly for measurement of CBF.10,11 The ASL technique is based on magnetic labeling of inflowing protons, which subsequently are measured in the brain parenchyma; roughly, CBF maps are acquired by subtraction of background non-labeled tissue. Based on the labeling method, the ASL technique is categorized into continuous ASL (CASL), pulsed ASL (PASL), or pseudocontinuous ASL (pCASL). At present mainly multiphase PASL and 3D pCASL are recommended12,13; however, there is no consensus regarding the optimal ASL technique for brain tumor evaluation. A large number of publications have suggested a promising role for ASL for discrimination between high-grade glioma (HGG) and low-grade glioma (LGG),14–16 with a strong correlation between DSC-derived CBF and ASL-derived CBF.17 However, concerns have been raised regarding the interpatient variability in the measurements,18 as well as the use of divergent parameters and techniques, WHO classification system, and postprocessing schemes. This lack of consensus and guidelines has probably contributed to the reluctance to implement ASL in clinical routine for the management of glioma patients. One meta-analysis concluded that there is a significant difference in blood flow between glioma grades,19 but the diagnostic performance of ASL to discriminate between tumor grades was not assessed in the study. Therefore the aim of this meta-analysis is to fully assess the diagnostic performance of ASL for differentiation of HGG from LGG, including potential sources of heterogeneity of its diagnostic performance. Materials and Methods This meta-analysis was conducted with adherence to the the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines20 and the Cochrane handbook for systematic reviews of diagnostic test accuracy.21 The study protocol was prospectively registered in PROSPERO (CRD42017080885). Search Strategy and Selection Criteria Literature search A literature search was performed in the following databases: Medline (OVID), Embase.com (Elsevier), Web of Science Core Collection (Clarivate), and Cochrane Library (Wiley). The MeSH (Medical Subject Headings) terms identified for searching Medline (OVID) were adapted in accordance to the corresponding vocabulary in Embase. Each search concept was also complemented with relevant free-text terms like: brain tumor, astrocytoma, arterial spin labeling, and ASL. The free-text terms were, if appropriate, truncated and/or combined with proximity operators. Conference abstracts were excluded in the search strategy. To include all eligible studies in the search, no language restriction was applied. Databases were searched from inception. All searches were performed by a librarian at the library of Karolinska Institutet in Stockholm, Sweden, in November 2017; the search strategies are available in Supplementary Tables S1–S4. Study eligibility criteria We included retrospective and prospective studies that evaluated patients with primary brain tumors, who received diagnoses according to the WHO classification, who were preoperatively examined using ASL, and for whom subsequent neuropathological diagnosis was available. Inclusion criteria were availability of (i) absolute CBF values and normalized CBF values stratified for tumor, and (ii) individual patient data or sensitivity and specificity for discrimination between high- and low-grade glioma. All types of ASL techniques were eligible for inclusion: PASL, pCASL, and CASL were eligible for inclusion. All tumor types as listed in the WHO classification were eligible, since the latest update of the WHO 2016 classification mainly added a strengthened biomolecular classification of tumor types with none or only slight changes in tumor grade classification.2 Furthermore, only untreated gliomas were eligible for inclusion. The following exclusion criteria were applied in the meta-analysis: studies with a strict pediatric cohort; studies grouping metastases or meningioma with intracranial tumors, studies reporting on recurrent or treated gliomas only, meningiomas, studies in other languages than English, duplicate or overlapping cohorts, abstracts or editorials. Oligoastrocytomas were excluded if possible from the analysis, since this diagnosis has been removed from the WHO classification 2016 and these tumors are now classified as either oligodendrogliomas or astrocytomas based on biomolecular mutational status. In the case of a duplicate cohort, the largest was included and the overlapping study was excluded from the meta-analysis. Study selection After the electronic search was conducted, 2 investigators independently screened and reviewed all titles and abstracts for possible inclusion in the qualitative assessment. Any disagreement between the investigators was solved through discussion with a third investigator, who cast the deciding vote. In addition, the third investigator hand-searched reference lists to identify additional studies. Quality Assessment Quality assessment was performed according to the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool,22 which was adjusted and developed to be used for this particular meta-analysis. Each study was graded according to prespecified criteria in 2 domains. Domain 1 included 4 internal bias items concerning specific quality issues in the included studies: patient selection, index test, reference standard, and flow and timing. Domain 2 consisted of 3 items related to the external validity of the included studies in relation to the specific aim of this meta-analysis: patient selection, index test, and reference standard. Two investigators independently assessed both domains for each study, with any disagreements solved through discussion with a third investigator. Data Extraction Data extraction was conducted independently by 2 investigators. The ASL perfusion parameters composed the maximum or mean CBF, relative or equivalent, and the maximum or mean intensity signal (ITS), relative or equivalent, nCBFmeanTI (CBF derived from multi-inversion time [mTI] ASL). In addition, the following items were extracted: first author name, publication year, town/country of origin, inclusion years, study design, age (mean and standard deviation), WHO classification, specifications for the MR camera used, numbers of channel in head coil, ASL method, flip angle, repetition time, echo time, post labeling delay, acquisition time, region of interest (ROI) technique, reference region, tumor types included. Any inconsistencies were solved through discussion with a third investigator until consensus. Main Outcome Measure The main outcome measure was the diagnostic performance of the ASL parameters for glioma grading. The secondary outcome measure was the diagnostic performance of the different ASL techniques for glioma grading and to assess the heterogeneity of their diagnostic performance. Data Synthesis Absolute and relative CBF measures were calculated where possible with mean and SD. For each study, the sensitivity and specificity was calculated with corresponding confidence intervals. The true positive, true negative, false positive, false negative rates and the area under the curve (AUC) were calculated for each study. For studies that only reported individual patient data, receiver operating characteristics curves were calculated optimizing the cutoff for both specificity and sensitivity. The abovementioned calculations were performed independently by 2 investigators and assessed for congruency. Any discrepancies between the investigators were solved through discussion. Mean difference with 95% CI was calculated for HGG and LGG for absolute CBF and regional CBF max with a random effects model. The sensitivity and specificity heterogeneity between studies was visually assessed by plotting their distribution in a univariable analysis. Exploring heterogeneity by I2 is not useful for diagnostic meta-analyses that include studies with different thresholds, since this results in high heterogeneity.23 Instead, we used bivariate meta-regression and subgroup and sensitivity analyses to explore heterogeneity. The following factors were explored in the meta-regression: WHO classification system (year published), MR field strength (in tesla), ASL method (PASL, pCASL, or CASL), and perfusion measure (mean or max CBF [mL/min/100 g], mean or max signal intensity [SI]). We used the funnel plot asymmetry test as described by Egger to explore for potential publication bias and small study effect.24 Trial Sequential Analysis Results in meta-analysis might be caused by random errors rather than reflecting an actual effect. Furthermore, lack of statistical significance might be caused by a lack of statistical power. Random errors are more likely in meta-analysis with few studies or if the included studies have small populations.25 Trial sequential analysis (TSA) is similar to the interim analysis in clinical trials. Using TSA it is possible to calculate the sample size required to demonstrate a difference between LGG and HGG. We calculated the estimated sample size for the meta-analysis to be powered for a one-sided type 1 error of 2.5% using O’Brien-Fleming α-spending function at a power of 90%. We minimized the type 1 error in the TSA by using a highly stringent criterion of 2.5% type 1 error and 90% power. TSA was performed using TSA version 0.9. Bivariate Analysis For assessment of the overall diagnostic accuracy of ASL for glioma grading, all studies were included irrespective of the CBF estimate reported, since no threshold was calculated. However, the unit maximum regional CBF, rCBFmax, was used where possible. To obtain summary estimates with 95% CIs, we used a bivariate random effects meta-analysis with a restricted maximum likelihood estimation method to describe the summary receiver operating characteristics (SROC) curve with associated AUC. The following subgroup analyses were performed: studies where maximum and mean rCBF were reported, studies reporting on astrocytomas only, and studies using pulsed, continuous, and pseudocontinuous ASL, respectively. Furthermore, we performed sensitivity analysis by excluding studies with unclear or high risk of bias related to the domain of internal or external validity in QUADAS-2; at least 3 studies with low risk of bias were required to perform sensitivity analysis for any particular item. Statistical analysis was performed in R for Mac v1.1.383 using the mada and metafor package v0.5.8. Results Search Hits The electronic search yielded a total of 640 hits, with 326 remaining after removal of duplicates. All 326 hits were screened for potential inclusion using the criteria described above, and subsequently 48 studies were included in the full-text evaluation. After full-text evaluation, another 34 studies were excluded for the following reasons: absence of sensitivity, specificity, or individual patient data, n = 8; absence of quantitative CBF data n = 17; the paper was not in the English language, n = 4; the material included only glioblastomas, extra-axial tumors or metastases, n = 4. In total, 15 studies comprising 505 patients met the inclusion criteria and were subsequently included in the full meta-analysis (Figure 1).14–17,26–36 Fig. 1 View largeDownload slide PRISMA flow chart. Fig. 1 View largeDownload slide PRISMA flow chart. Study Characteristics and Risk of Bias Assessment A full description of the 15 included studies is presented in Table 1 with technical specifications in Supplementary Table S5. Eight studies (8/15, 53%) used the WHO classification system published in 200716,28,29,31,32,34,36,37 rather than the most recent classification system published in 2016. In 7 studies (47%), pCASL was used for CBF estimation,17,26,31–34,36 7 (47%) used PASL,14,16,27–30,35 and 1 (7%) used CASL.15 All studies reported on relative values; 9 (60%) used rCBFmax14–16,26–28,32,34,36; of the 4 remaining (27%),17,28,31,35 1 study reported both rCBFmax and rCBFmean,28 1 (7%) reported normalized intensity signal (nITSmax),30 1 nITSmean,29 and 1 nCBFmeanTI33 (CBF derived from mTI ASL). In 12 studies (80%),15–17,26,28,29,31–36 the MR field strength was 3T and in 3 studies (20%), 1.5T.14,27,30 In 13 studies (87%),14,16,17,26–34,36 the risk of bias for patient selection was low, and in 2 (13%) unclear.15,35 In all studies the risk of bias related to the reference standard was unclear, since there was no information as to whether the pathologist was blinded for ASL results or not. The time interval between imaging and surgery/biopsy was reported in only 2 (20%) studies.27,32,34 Under the domain of applicability, concerns and external validity related to patient selection: 11 (73%) studies showed low risk bias14–17,26–28,31,34,36 and 4 (27%) unclear risk of bias pertaining to inclusion of only one tumor type29,30,32,33 (Table 2). The risk of bias concerns were included in the sensitivity analyses. The distribution of sensitivity and specificity was depicted in forest plots and is suggestive of heterogeneity among the included studies related to the specificity for glioma grading among the included studies (Figure 2). Table 1 Characteristics of included studies First Author Year City / Country Years of Patient Inclusion Patient Inclusion (Retrospective or Prospective) Mean Age (SD) WHO Classification (Year of Publication) ROI Technique Reference Region (Location/Lobe, Gray or White Matter) Perfusion Parameters Tumors Included (WHO Grade) aCBF /rCBFmax In LGG Mean (SD), N Patients aCBF/rCBFmax HGG Mean (SD), N Patients Cebeci16 2014 Bursa / Turkey 2010–2013 Retrospective 47 (14) 2007 Manual ROI in maximal CBF Contralateral normal hemisphere rCBFmax / CBFmax / rSI GBM, AC (III), Gliosarcoma, OD (II), DNET, pilocystic AC 8.1 (28.1) / 0.96 (0.48), 13 23.65 (44.8) / 4.7 (1.38), 20 Fudaba28 2014 Oita / Japan 2010–2012 Retrospective 60 (17) 2007 Manual ROI (8 mm in diameter) in total tumor Mirrored normal white matter rCBFmin / max / mean AC (II–IV), OD (II–III), OA, GBM NA / NA, 9 NA / NA, 23 Furtner29 2014 Vienna / Austria 2009–2012 Prospective 54 (17) 2007 Manual ROI in whole tumor Mirrored normal contralateral healthy hemishpere nITS mean AC (II–IV) NA / 1.08 (0.39), 7 NA / 3.09 (2.57), 26 Kim30 2008 Suwon / South Korea 2005–2008 Prospective 43 (14) 2000 Manual ROIs in maximal tumor perfusion signal Contralateral white matter rITSmax AC (II–IV) NA / NA, 26 NA / NA, 35 Lehmann17 2010 Marseille / France 2007–2008 Prospective 58 (19) 2000 Manual ROI in whole tumor Contralateral white matter rCBFmean OD, AC, GBM, LGG NA / 1.09 (1.08), 4 NA / 2.26 (1.12), 4 Ma31 2017 Jiangsu / China 2014–2015 Prospective 46 (18) 2007 Manual ROIs in whole tumor (50–60 mm2) MIrrored contralateral normal matter CBFmean / rCBFmean GG, Pilocytic AC, AC gr I, AC (II–IV), AO, OD (gr III) 50.64 (35.89) / 2.13 (2.16), 27 88.03 (37.16) / 5.41 (3.74), 23 Shen36 2016 Wuhan / China 2014–2015 Prospective NA 2007 Manual ROIs in maximal CBF Contralateral normal appearing white matter CBFmax / rCBFmax AC (II–IV), OD (II–III), 38.97 (17.47) / 0.97 (0.27), 25 73.93 (23.74) / 2.02 (0.61), 27 Soni26 2017 Lucknow / India 2013–2014 Prospective 19–71 (range) NA Manual ROIs in maximal CBF Contralateral normal appearing frontal gray matter and periventricular white matter rCBFmax Glioma, LGG, Diffuse AC, GBM NA / 1.23 (0.46), 3 NA / 14.1, 1 Warmuth14 2003 Berlin / Germany NA Prospective 46 (15) 1993 Manual ROI in maximal CBF Contralateral mirrored tumor region of interest rCBFmax / CBFmean / CBFmax Gangliogliom, pleomorphic AC, AC, AOD, AA, GBM 186 (101) / 0.68 (0.08), 7 408 (267) / 1.64 (0.65), 9 Weber27 2006 Heidelberg / Germany NA Prospective 57 (14) NA Manual ROI (at least 20 voxel) whole tumor / maximal CBF Contralateral normalappearing gray and white matter rCBFmax AC (II–IV) NA / NA, 9 NA / NA, 35 Wolf15 2005 Philadelphia / USA NA Prospective 50 (12) NA Manual ROI marked by 3D mask, and 33 voxel ROI in maximal CBF and mean CBF Global CBF over 12 sections, excluding tumor and edema CBFmax / rCBFmax / rCBFmean GG, OD, OA, AC, AOA, AA, GBM 30.93 (9.44) / 0.87 (0.15), 4 95.68 (70.5)/ 2.97 (1.78), 16 Xiao32 2015 Bejing / China 2012–2014 Prospective 43 (17) 2007 Manual ROIs in maximal CBF Normalization in cerebellum CBFmax / rCBFmax Pilocytic AC, AC (II–IV) NA / 1.81 (0.98), 19 NA / 4.51 (2.27), 24 Yang33 2016 Jinan / China NA Prospective 51 (15) 2007 Manual ROI (8–10 voxel) in solid tumor Contralateral normal appearing frontal white matter rmTI-ASL / nCBFmeanTI AC (gr II–IV) NA / 1.62 (1.97), 15 NA / 6.7 (5.1), 28 Zeng34 2017 Zhejinag / China 2013–2015 Retrospective 50 (13) 2007 Manual ROI in maximal CBF Contralateral ROI in gray matter CBFmax / rCBFmax AC (II–IV), OD (gr II–III), OA, AOA 90.69 (36.95) / 1.12 (0.48), 13 168.82 (70.43) / 2.19 (0.87), 45 Zhang35 2014 Jȕlich / Germany NA Prospective 56 (14) NA Manual ROI in whole tumor Contralateral normalappearing gray and white matter rCBFmean AC (II–IV), OD (II–III) NA / 20.84 (16.87), 4 NA / 21.24 (9.61), 4 First Author Year City / Country Years of Patient Inclusion Patient Inclusion (Retrospective or Prospective) Mean Age (SD) WHO Classification (Year of Publication) ROI Technique Reference Region (Location/Lobe, Gray or White Matter) Perfusion Parameters Tumors Included (WHO Grade) aCBF /rCBFmax In LGG Mean (SD), N Patients aCBF/rCBFmax HGG Mean (SD), N Patients Cebeci16 2014 Bursa / Turkey 2010–2013 Retrospective 47 (14) 2007 Manual ROI in maximal CBF Contralateral normal hemisphere rCBFmax / CBFmax / rSI GBM, AC (III), Gliosarcoma, OD (II), DNET, pilocystic AC 8.1 (28.1) / 0.96 (0.48), 13 23.65 (44.8) / 4.7 (1.38), 20 Fudaba28 2014 Oita / Japan 2010–2012 Retrospective 60 (17) 2007 Manual ROI (8 mm in diameter) in total tumor Mirrored normal white matter rCBFmin / max / mean AC (II–IV), OD (II–III), OA, GBM NA / NA, 9 NA / NA, 23 Furtner29 2014 Vienna / Austria 2009–2012 Prospective 54 (17) 2007 Manual ROI in whole tumor Mirrored normal contralateral healthy hemishpere nITS mean AC (II–IV) NA / 1.08 (0.39), 7 NA / 3.09 (2.57), 26 Kim30 2008 Suwon / South Korea 2005–2008 Prospective 43 (14) 2000 Manual ROIs in maximal tumor perfusion signal Contralateral white matter rITSmax AC (II–IV) NA / NA, 26 NA / NA, 35 Lehmann17 2010 Marseille / France 2007–2008 Prospective 58 (19) 2000 Manual ROI in whole tumor Contralateral white matter rCBFmean OD, AC, GBM, LGG NA / 1.09 (1.08), 4 NA / 2.26 (1.12), 4 Ma31 2017 Jiangsu / China 2014–2015 Prospective 46 (18) 2007 Manual ROIs in whole tumor (50–60 mm2) MIrrored contralateral normal matter CBFmean / rCBFmean GG, Pilocytic AC, AC gr I, AC (II–IV), AO, OD (gr III) 50.64 (35.89) / 2.13 (2.16), 27 88.03 (37.16) / 5.41 (3.74), 23 Shen36 2016 Wuhan / China 2014–2015 Prospective NA 2007 Manual ROIs in maximal CBF Contralateral normal appearing white matter CBFmax / rCBFmax AC (II–IV), OD (II–III), 38.97 (17.47) / 0.97 (0.27), 25 73.93 (23.74) / 2.02 (0.61), 27 Soni26 2017 Lucknow / India 2013–2014 Prospective 19–71 (range) NA Manual ROIs in maximal CBF Contralateral normal appearing frontal gray matter and periventricular white matter rCBFmax Glioma, LGG, Diffuse AC, GBM NA / 1.23 (0.46), 3 NA / 14.1, 1 Warmuth14 2003 Berlin / Germany NA Prospective 46 (15) 1993 Manual ROI in maximal CBF Contralateral mirrored tumor region of interest rCBFmax / CBFmean / CBFmax Gangliogliom, pleomorphic AC, AC, AOD, AA, GBM 186 (101) / 0.68 (0.08), 7 408 (267) / 1.64 (0.65), 9 Weber27 2006 Heidelberg / Germany NA Prospective 57 (14) NA Manual ROI (at least 20 voxel) whole tumor / maximal CBF Contralateral normalappearing gray and white matter rCBFmax AC (II–IV) NA / NA, 9 NA / NA, 35 Wolf15 2005 Philadelphia / USA NA Prospective 50 (12) NA Manual ROI marked by 3D mask, and 33 voxel ROI in maximal CBF and mean CBF Global CBF over 12 sections, excluding tumor and edema CBFmax / rCBFmax / rCBFmean GG, OD, OA, AC, AOA, AA, GBM 30.93 (9.44) / 0.87 (0.15), 4 95.68 (70.5)/ 2.97 (1.78), 16 Xiao32 2015 Bejing / China 2012–2014 Prospective 43 (17) 2007 Manual ROIs in maximal CBF Normalization in cerebellum CBFmax / rCBFmax Pilocytic AC, AC (II–IV) NA / 1.81 (0.98), 19 NA / 4.51 (2.27), 24 Yang33 2016 Jinan / China NA Prospective 51 (15) 2007 Manual ROI (8–10 voxel) in solid tumor Contralateral normal appearing frontal white matter rmTI-ASL / nCBFmeanTI AC (gr II–IV) NA / 1.62 (1.97), 15 NA / 6.7 (5.1), 28 Zeng34 2017 Zhejinag / China 2013–2015 Retrospective 50 (13) 2007 Manual ROI in maximal CBF Contralateral ROI in gray matter CBFmax / rCBFmax AC (II–IV), OD (gr II–III), OA, AOA 90.69 (36.95) / 1.12 (0.48), 13 168.82 (70.43) / 2.19 (0.87), 45 Zhang35 2014 Jȕlich / Germany NA Prospective 56 (14) NA Manual ROI in whole tumor Contralateral normalappearing gray and white matter rCBFmean AC (II–IV), OD (II–III) NA / 20.84 (16.87), 4 NA / 21.24 (9.61), 4 Gr, grade, AC; astrocytoma, DNET, dysembryoplastic neuroepithelial tumor, OA, oligoastrocytoma, GG, ganglioglioma, OD, oligodendroglioma, AOD, (anaplastic OD), AOA, anaplastic OA, GBM, glioblastoma; NA; not available. View Large Table 1 Characteristics of included studies First Author Year City / Country Years of Patient Inclusion Patient Inclusion (Retrospective or Prospective) Mean Age (SD) WHO Classification (Year of Publication) ROI Technique Reference Region (Location/Lobe, Gray or White Matter) Perfusion Parameters Tumors Included (WHO Grade) aCBF /rCBFmax In LGG Mean (SD), N Patients aCBF/rCBFmax HGG Mean (SD), N Patients Cebeci16 2014 Bursa / Turkey 2010–2013 Retrospective 47 (14) 2007 Manual ROI in maximal CBF Contralateral normal hemisphere rCBFmax / CBFmax / rSI GBM, AC (III), Gliosarcoma, OD (II), DNET, pilocystic AC 8.1 (28.1) / 0.96 (0.48), 13 23.65 (44.8) / 4.7 (1.38), 20 Fudaba28 2014 Oita / Japan 2010–2012 Retrospective 60 (17) 2007 Manual ROI (8 mm in diameter) in total tumor Mirrored normal white matter rCBFmin / max / mean AC (II–IV), OD (II–III), OA, GBM NA / NA, 9 NA / NA, 23 Furtner29 2014 Vienna / Austria 2009–2012 Prospective 54 (17) 2007 Manual ROI in whole tumor Mirrored normal contralateral healthy hemishpere nITS mean AC (II–IV) NA / 1.08 (0.39), 7 NA / 3.09 (2.57), 26 Kim30 2008 Suwon / South Korea 2005–2008 Prospective 43 (14) 2000 Manual ROIs in maximal tumor perfusion signal Contralateral white matter rITSmax AC (II–IV) NA / NA, 26 NA / NA, 35 Lehmann17 2010 Marseille / France 2007–2008 Prospective 58 (19) 2000 Manual ROI in whole tumor Contralateral white matter rCBFmean OD, AC, GBM, LGG NA / 1.09 (1.08), 4 NA / 2.26 (1.12), 4 Ma31 2017 Jiangsu / China 2014–2015 Prospective 46 (18) 2007 Manual ROIs in whole tumor (50–60 mm2) MIrrored contralateral normal matter CBFmean / rCBFmean GG, Pilocytic AC, AC gr I, AC (II–IV), AO, OD (gr III) 50.64 (35.89) / 2.13 (2.16), 27 88.03 (37.16) / 5.41 (3.74), 23 Shen36 2016 Wuhan / China 2014–2015 Prospective NA 2007 Manual ROIs in maximal CBF Contralateral normal appearing white matter CBFmax / rCBFmax AC (II–IV), OD (II–III), 38.97 (17.47) / 0.97 (0.27), 25 73.93 (23.74) / 2.02 (0.61), 27 Soni26 2017 Lucknow / India 2013–2014 Prospective 19–71 (range) NA Manual ROIs in maximal CBF Contralateral normal appearing frontal gray matter and periventricular white matter rCBFmax Glioma, LGG, Diffuse AC, GBM NA / 1.23 (0.46), 3 NA / 14.1, 1 Warmuth14 2003 Berlin / Germany NA Prospective 46 (15) 1993 Manual ROI in maximal CBF Contralateral mirrored tumor region of interest rCBFmax / CBFmean / CBFmax Gangliogliom, pleomorphic AC, AC, AOD, AA, GBM 186 (101) / 0.68 (0.08), 7 408 (267) / 1.64 (0.65), 9 Weber27 2006 Heidelberg / Germany NA Prospective 57 (14) NA Manual ROI (at least 20 voxel) whole tumor / maximal CBF Contralateral normalappearing gray and white matter rCBFmax AC (II–IV) NA / NA, 9 NA / NA, 35 Wolf15 2005 Philadelphia / USA NA Prospective 50 (12) NA Manual ROI marked by 3D mask, and 33 voxel ROI in maximal CBF and mean CBF Global CBF over 12 sections, excluding tumor and edema CBFmax / rCBFmax / rCBFmean GG, OD, OA, AC, AOA, AA, GBM 30.93 (9.44) / 0.87 (0.15), 4 95.68 (70.5)/ 2.97 (1.78), 16 Xiao32 2015 Bejing / China 2012–2014 Prospective 43 (17) 2007 Manual ROIs in maximal CBF Normalization in cerebellum CBFmax / rCBFmax Pilocytic AC, AC (II–IV) NA / 1.81 (0.98), 19 NA / 4.51 (2.27), 24 Yang33 2016 Jinan / China NA Prospective 51 (15) 2007 Manual ROI (8–10 voxel) in solid tumor Contralateral normal appearing frontal white matter rmTI-ASL / nCBFmeanTI AC (gr II–IV) NA / 1.62 (1.97), 15 NA / 6.7 (5.1), 28 Zeng34 2017 Zhejinag / China 2013–2015 Retrospective 50 (13) 2007 Manual ROI in maximal CBF Contralateral ROI in gray matter CBFmax / rCBFmax AC (II–IV), OD (gr II–III), OA, AOA 90.69 (36.95) / 1.12 (0.48), 13 168.82 (70.43) / 2.19 (0.87), 45 Zhang35 2014 Jȕlich / Germany NA Prospective 56 (14) NA Manual ROI in whole tumor Contralateral normalappearing gray and white matter rCBFmean AC (II–IV), OD (II–III) NA / 20.84 (16.87), 4 NA / 21.24 (9.61), 4 First Author Year City / Country Years of Patient Inclusion Patient Inclusion (Retrospective or Prospective) Mean Age (SD) WHO Classification (Year of Publication) ROI Technique Reference Region (Location/Lobe, Gray or White Matter) Perfusion Parameters Tumors Included (WHO Grade) aCBF /rCBFmax In LGG Mean (SD), N Patients aCBF/rCBFmax HGG Mean (SD), N Patients Cebeci16 2014 Bursa / Turkey 2010–2013 Retrospective 47 (14) 2007 Manual ROI in maximal CBF Contralateral normal hemisphere rCBFmax / CBFmax / rSI GBM, AC (III), Gliosarcoma, OD (II), DNET, pilocystic AC 8.1 (28.1) / 0.96 (0.48), 13 23.65 (44.8) / 4.7 (1.38), 20 Fudaba28 2014 Oita / Japan 2010–2012 Retrospective 60 (17) 2007 Manual ROI (8 mm in diameter) in total tumor Mirrored normal white matter rCBFmin / max / mean AC (II–IV), OD (II–III), OA, GBM NA / NA, 9 NA / NA, 23 Furtner29 2014 Vienna / Austria 2009–2012 Prospective 54 (17) 2007 Manual ROI in whole tumor Mirrored normal contralateral healthy hemishpere nITS mean AC (II–IV) NA / 1.08 (0.39), 7 NA / 3.09 (2.57), 26 Kim30 2008 Suwon / South Korea 2005–2008 Prospective 43 (14) 2000 Manual ROIs in maximal tumor perfusion signal Contralateral white matter rITSmax AC (II–IV) NA / NA, 26 NA / NA, 35 Lehmann17 2010 Marseille / France 2007–2008 Prospective 58 (19) 2000 Manual ROI in whole tumor Contralateral white matter rCBFmean OD, AC, GBM, LGG NA / 1.09 (1.08), 4 NA / 2.26 (1.12), 4 Ma31 2017 Jiangsu / China 2014–2015 Prospective 46 (18) 2007 Manual ROIs in whole tumor (50–60 mm2) MIrrored contralateral normal matter CBFmean / rCBFmean GG, Pilocytic AC, AC gr I, AC (II–IV), AO, OD (gr III) 50.64 (35.89) / 2.13 (2.16), 27 88.03 (37.16) / 5.41 (3.74), 23 Shen36 2016 Wuhan / China 2014–2015 Prospective NA 2007 Manual ROIs in maximal CBF Contralateral normal appearing white matter CBFmax / rCBFmax AC (II–IV), OD (II–III), 38.97 (17.47) / 0.97 (0.27), 25 73.93 (23.74) / 2.02 (0.61), 27 Soni26 2017 Lucknow / India 2013–2014 Prospective 19–71 (range) NA Manual ROIs in maximal CBF Contralateral normal appearing frontal gray matter and periventricular white matter rCBFmax Glioma, LGG, Diffuse AC, GBM NA / 1.23 (0.46), 3 NA / 14.1, 1 Warmuth14 2003 Berlin / Germany NA Prospective 46 (15) 1993 Manual ROI in maximal CBF Contralateral mirrored tumor region of interest rCBFmax / CBFmean / CBFmax Gangliogliom, pleomorphic AC, AC, AOD, AA, GBM 186 (101) / 0.68 (0.08), 7 408 (267) / 1.64 (0.65), 9 Weber27 2006 Heidelberg / Germany NA Prospective 57 (14) NA Manual ROI (at least 20 voxel) whole tumor / maximal CBF Contralateral normalappearing gray and white matter rCBFmax AC (II–IV) NA / NA, 9 NA / NA, 35 Wolf15 2005 Philadelphia / USA NA Prospective 50 (12) NA Manual ROI marked by 3D mask, and 33 voxel ROI in maximal CBF and mean CBF Global CBF over 12 sections, excluding tumor and edema CBFmax / rCBFmax / rCBFmean GG, OD, OA, AC, AOA, AA, GBM 30.93 (9.44) / 0.87 (0.15), 4 95.68 (70.5)/ 2.97 (1.78), 16 Xiao32 2015 Bejing / China 2012–2014 Prospective 43 (17) 2007 Manual ROIs in maximal CBF Normalization in cerebellum CBFmax / rCBFmax Pilocytic AC, AC (II–IV) NA / 1.81 (0.98), 19 NA / 4.51 (2.27), 24 Yang33 2016 Jinan / China NA Prospective 51 (15) 2007 Manual ROI (8–10 voxel) in solid tumor Contralateral normal appearing frontal white matter rmTI-ASL / nCBFmeanTI AC (gr II–IV) NA / 1.62 (1.97), 15 NA / 6.7 (5.1), 28 Zeng34 2017 Zhejinag / China 2013–2015 Retrospective 50 (13) 2007 Manual ROI in maximal CBF Contralateral ROI in gray matter CBFmax / rCBFmax AC (II–IV), OD (gr II–III), OA, AOA 90.69 (36.95) / 1.12 (0.48), 13 168.82 (70.43) / 2.19 (0.87), 45 Zhang35 2014 Jȕlich / Germany NA Prospective 56 (14) NA Manual ROI in whole tumor Contralateral normalappearing gray and white matter rCBFmean AC (II–IV), OD (II–III) NA / 20.84 (16.87), 4 NA / 21.24 (9.61), 4 Gr, grade, AC; astrocytoma, DNET, dysembryoplastic neuroepithelial tumor, OA, oligoastrocytoma, GG, ganglioglioma, OD, oligodendroglioma, AOD, (anaplastic OD), AOA, anaplastic OA, GBM, glioblastoma; NA; not available. View Large Table 2 Quality assessment according to QUADAS-2 Risk of Bias Applicability Concerns Patient Selection Index Test Reference Standard Flow and Timing Patient Selection Index Test Reference Standard Cebeci 2014 + _ ? ? + _ + Fudaba 2014 + _ ? ? + _ + Furtner 2014 + + ? ? ? _ + Kim 2008 + + ? ? ? _ ? Lehmann 2010 + _ ? ? + + _ Ma 2017 + _ ? ? + + + Shen 2016 + + ? ? + + + Soni 2017 + _ ? ? + + _ Warmuth 2005 + _ ? ? + _ ? Weber 2006 + _ ? + + _ _ Wolf 2005 ? _ ? ? + _ ? Xiao 2015 + + ? + ? + + Yang 2016 + + ? ? ? _ + Zeng 2017 + + ? + + + + Zhang 2014 ? _ ? ? + _ ? Risk of Bias Applicability Concerns Patient Selection Index Test Reference Standard Flow and Timing Patient Selection Index Test Reference Standard Cebeci 2014 + _ ? ? + _ + Fudaba 2014 + _ ? ? + _ + Furtner 2014 + + ? ? ? _ + Kim 2008 + + ? ? ? _ ? Lehmann 2010 + _ ? ? + + _ Ma 2017 + _ ? ? + + + Shen 2016 + + ? ? + + + Soni 2017 + _ ? ? + + _ Warmuth 2005 + _ ? ? + _ ? Weber 2006 + _ ? + + _ _ Wolf 2005 ? _ ? ? + _ ? Xiao 2015 + + ? + ? + + Yang 2016 + + ? ? ? _ + Zeng 2017 + + ? + + + + Zhang 2014 ? _ ? ? + _ ? + indicates low risk. _ indicates high risk. ? indicates unclear risk. Risk of bias: Patient selection: Low if stated consecutive or reported years of inclusion together with clear inclusion criteria. Unclear if no mention on consecutive series of patients. High if reported a non-consecutive series. Index test: Low if ASL was interpreted blinded. Unclear if no information on blinding but pre-defined cut-off was specified for positive test. High if exploratory cut-off and no information on blinding or if reported unblinded. Reference standard: Low if reported on blinded evaluation and WHO adherence. Unclear if no information on blinding. High if reported on un-blinded evaluation or if pathology diagnosis reported only part of the tumors in the study. Flow and timing: Low if <30 days between ASL and histopathology. Unclear if not reported. High if reported more than 6 months. Applicability concerns: Patient selection: Low if mixed tumour types. Unclear if tumor types not reported or same tumor type reported. High if other comparison than between high- and low-grade. Index test: Low if presented relative CBF from 3D pseudocontinous ASL. Unclear if not normalized CBF but pseudocontinous ASL. High if presented other perfusion metric than CBF or if pulsed ASL or continuous ASL was used. Reference standard: Low if tumors were classified according to WHO 2007 or later. Unclear if WHO but unspecified, or before 2007. High if no report on histopathological diagnosis classification system or if part of the cohort did not have histopathology. View Large Table 2 Quality assessment according to QUADAS-2 Risk of Bias Applicability Concerns Patient Selection Index Test Reference Standard Flow and Timing Patient Selection Index Test Reference Standard Cebeci 2014 + _ ? ? + _ + Fudaba 2014 + _ ? ? + _ + Furtner 2014 + + ? ? ? _ + Kim 2008 + + ? ? ? _ ? Lehmann 2010 + _ ? ? + + _ Ma 2017 + _ ? ? + + + Shen 2016 + + ? ? + + + Soni 2017 + _ ? ? + + _ Warmuth 2005 + _ ? ? + _ ? Weber 2006 + _ ? + + _ _ Wolf 2005 ? _ ? ? + _ ? Xiao 2015 + + ? + ? + + Yang 2016 + + ? ? ? _ + Zeng 2017 + + ? + + + + Zhang 2014 ? _ ? ? + _ ? Risk of Bias Applicability Concerns Patient Selection Index Test Reference Standard Flow and Timing Patient Selection Index Test Reference Standard Cebeci 2014 + _ ? ? + _ + Fudaba 2014 + _ ? ? + _ + Furtner 2014 + + ? ? ? _ + Kim 2008 + + ? ? ? _ ? Lehmann 2010 + _ ? ? + + _ Ma 2017 + _ ? ? + + + Shen 2016 + + ? ? + + + Soni 2017 + _ ? ? + + _ Warmuth 2005 + _ ? ? + _ ? Weber 2006 + _ ? + + _ _ Wolf 2005 ? _ ? ? + _ ? Xiao 2015 + + ? + ? + + Yang 2016 + + ? ? ? _ + Zeng 2017 + + ? + + + + Zhang 2014 ? _ ? ? + _ ? + indicates low risk. _ indicates high risk. ? indicates unclear risk. Risk of bias: Patient selection: Low if stated consecutive or reported years of inclusion together with clear inclusion criteria. Unclear if no mention on consecutive series of patients. High if reported a non-consecutive series. Index test: Low if ASL was interpreted blinded. Unclear if no information on blinding but pre-defined cut-off was specified for positive test. High if exploratory cut-off and no information on blinding or if reported unblinded. Reference standard: Low if reported on blinded evaluation and WHO adherence. Unclear if no information on blinding. High if reported on un-blinded evaluation or if pathology diagnosis reported only part of the tumors in the study. Flow and timing: Low if <30 days between ASL and histopathology. Unclear if not reported. High if reported more than 6 months. Applicability concerns: Patient selection: Low if mixed tumour types. Unclear if tumor types not reported or same tumor type reported. High if other comparison than between high- and low-grade. Index test: Low if presented relative CBF from 3D pseudocontinous ASL. Unclear if not normalized CBF but pseudocontinous ASL. High if presented other perfusion metric than CBF or if pulsed ASL or continuous ASL was used. Reference standard: Low if tumors were classified according to WHO 2007 or later. Unclear if WHO but unspecified, or before 2007. High if no report on histopathological diagnosis classification system or if part of the cohort did not have histopathology. View Large Fig. 2 View largeDownload slide (A) Distribution of sensitivity of rCBF to discriminate between LGG and HGG. (B) Distribution of specificity of rCBF to discriminate between LGG and HGG. Fig. 2 View largeDownload slide (A) Distribution of sensitivity of rCBF to discriminate between LGG and HGG. (B) Distribution of specificity of rCBF to discriminate between LGG and HGG. The bivariate meta-regression assessing potential moderator variables for the primary outcome (diagnostic performance) sensitivity revealed no moderating effect for the following variables; ASL type (PASL, pCASL, or CASL), WHO classification (year published), tesla strength (1.5 or 3), and perfusion parameter (rCBFmax, rCBFmean, nCBFmTI, nITSmax, nITSmean) (P = 0.15–0.65). However, use of the perfusion parameters nITSmax and rCBFmax had a significant effect on the specificity in the bivariate meta-regression. The parameter nITSmax was associated with a lower specificity (P = 0.007) compared with rCBFmean/max and nITSmean. The usage of rCBFmax was associated with a higher specificity (P = 0.009) compared with rCBFmean and nITmean/max. The funnel plot assymetry test did not detect any evidence for publication bias (P = 0.53). Meta-analytic evaluation revealed a mean difference in rCBFmax between HGG and LGG of 3.23 (95% CI: 0.90–5.76, P = 0.007, I2 = 99%). The mean difference for absolute CBF between HGG and LGG was 45.50 (95% CI: 26.06–64.94, P < 0.001, I2 = 70%). Trial sequential analysis revealed that the required sample size to detect a statistical difference between LGG and HGG was n = 187. Hence, this meta-analysis was statistically powered including 505 patients (Figure 1). Diagnostic Performance In the main analysis, the pooled diagnostic performance for differentiation between high and low grade glioma irrespective of the perfusion measure (CBF or SI) is exemplified by the AUC. The AUC was 0.90 with a summary sensitivity ranging from 0.79 to 0.90, and a summary specificity ranging from 0.72 to 0.87 (Figure 3, Table 3). Table 3 Main and subgroup analysis Participants No. Studies AUC Summary Sensitivity (95% CI) Summary Specificity (95% CI) Main analysis 505 15 0.90 0.86 (0.79–0.90) 0.80 (0.72–0.89) Subgroup analysis Astrocytomas only 243 9 0.89 0.86 (0.79–0.91) 0.79 (0.63–0.89) rCBF max 302 9 0.90 0.87 (0.77–0.93) 0.85 (0.76–0.91) rCBF mean 118 5 0.78 0.74 (0.63–0.83) 0.74 (0.60–0.84) ASL method:  PASL 227 7 0.90 0.85 (0.71–0.91) 0.83 (0.69–0.92)  3D pCASL 258 7 0.88 0.86 (0.79–0.91) 0.80 (0.65–0.87) Sensitivity analysis from QUADAS-2 items Internal validity  Excluding studies with high or unclear risk of bias related to patient selection 477 13 0.90 0.85 (0.78–0.90) 0.81 (0.73–0.88)  Excluding studies with high or unclear risk of bias related to the index test 290 6 0.89 0.86 (0.79–0.91) 0.85 (0.69–0.94)  Excluding studies with high or unclear risk of bias related to the flow and timing 145 3 0.87 0.88 (0.77–0.95) 0.83 (0.68–0.92) External validity  Excluding studies with high risk of bias or unclear bias related to patient selection within the domain of external validity 325 11 0.88 0.86 (0.76–0.92) 0.81 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to index test within the domain of external validity* 215 6 0.84 0.85 (0.77–0.90) 0.82 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to reference standard within the domain of external validity** 344 8 0.89 0.84 (0.75–0.90) 0.80 (0.69–0.88) Participants No. Studies AUC Summary Sensitivity (95% CI) Summary Specificity (95% CI) Main analysis 505 15 0.90 0.86 (0.79–0.90) 0.80 (0.72–0.89) Subgroup analysis Astrocytomas only 243 9 0.89 0.86 (0.79–0.91) 0.79 (0.63–0.89) rCBF max 302 9 0.90 0.87 (0.77–0.93) 0.85 (0.76–0.91) rCBF mean 118 5 0.78 0.74 (0.63–0.83) 0.74 (0.60–0.84) ASL method:  PASL 227 7 0.90 0.85 (0.71–0.91) 0.83 (0.69–0.92)  3D pCASL 258 7 0.88 0.86 (0.79–0.91) 0.80 (0.65–0.87) Sensitivity analysis from QUADAS-2 items Internal validity  Excluding studies with high or unclear risk of bias related to patient selection 477 13 0.90 0.85 (0.78–0.90) 0.81 (0.73–0.88)  Excluding studies with high or unclear risk of bias related to the index test 290 6 0.89 0.86 (0.79–0.91) 0.85 (0.69–0.94)  Excluding studies with high or unclear risk of bias related to the flow and timing 145 3 0.87 0.88 (0.77–0.95) 0.83 (0.68–0.92) External validity  Excluding studies with high risk of bias or unclear bias related to patient selection within the domain of external validity 325 11 0.88 0.86 (0.76–0.92) 0.81 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to index test within the domain of external validity* 215 6 0.84 0.85 (0.77–0.90) 0.82 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to reference standard within the domain of external validity** 344 8 0.89 0.84 (0.75–0.90) 0.80 (0.69–0.88) *Related to the applicability of the index test in relation to the aim of the meta-analysis. **Related to the applicability of the reference standard in relation to the aim of the meta-analysis. AUC; Area under the receiver operating characteristics curve; QUADAS-2; a revised tool for the quality assessment of diagnostic accuracy studies, PASL; pulsed arterial spin labeling, pCASL; pseudo-continous arterial spin labeling, rCBF; relative cerebral blood flow. View Large Table 3 Main and subgroup analysis Participants No. Studies AUC Summary Sensitivity (95% CI) Summary Specificity (95% CI) Main analysis 505 15 0.90 0.86 (0.79–0.90) 0.80 (0.72–0.89) Subgroup analysis Astrocytomas only 243 9 0.89 0.86 (0.79–0.91) 0.79 (0.63–0.89) rCBF max 302 9 0.90 0.87 (0.77–0.93) 0.85 (0.76–0.91) rCBF mean 118 5 0.78 0.74 (0.63–0.83) 0.74 (0.60–0.84) ASL method:  PASL 227 7 0.90 0.85 (0.71–0.91) 0.83 (0.69–0.92)  3D pCASL 258 7 0.88 0.86 (0.79–0.91) 0.80 (0.65–0.87) Sensitivity analysis from QUADAS-2 items Internal validity  Excluding studies with high or unclear risk of bias related to patient selection 477 13 0.90 0.85 (0.78–0.90) 0.81 (0.73–0.88)  Excluding studies with high or unclear risk of bias related to the index test 290 6 0.89 0.86 (0.79–0.91) 0.85 (0.69–0.94)  Excluding studies with high or unclear risk of bias related to the flow and timing 145 3 0.87 0.88 (0.77–0.95) 0.83 (0.68–0.92) External validity  Excluding studies with high risk of bias or unclear bias related to patient selection within the domain of external validity 325 11 0.88 0.86 (0.76–0.92) 0.81 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to index test within the domain of external validity* 215 6 0.84 0.85 (0.77–0.90) 0.82 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to reference standard within the domain of external validity** 344 8 0.89 0.84 (0.75–0.90) 0.80 (0.69–0.88) Participants No. Studies AUC Summary Sensitivity (95% CI) Summary Specificity (95% CI) Main analysis 505 15 0.90 0.86 (0.79–0.90) 0.80 (0.72–0.89) Subgroup analysis Astrocytomas only 243 9 0.89 0.86 (0.79–0.91) 0.79 (0.63–0.89) rCBF max 302 9 0.90 0.87 (0.77–0.93) 0.85 (0.76–0.91) rCBF mean 118 5 0.78 0.74 (0.63–0.83) 0.74 (0.60–0.84) ASL method:  PASL 227 7 0.90 0.85 (0.71–0.91) 0.83 (0.69–0.92)  3D pCASL 258 7 0.88 0.86 (0.79–0.91) 0.80 (0.65–0.87) Sensitivity analysis from QUADAS-2 items Internal validity  Excluding studies with high or unclear risk of bias related to patient selection 477 13 0.90 0.85 (0.78–0.90) 0.81 (0.73–0.88)  Excluding studies with high or unclear risk of bias related to the index test 290 6 0.89 0.86 (0.79–0.91) 0.85 (0.69–0.94)  Excluding studies with high or unclear risk of bias related to the flow and timing 145 3 0.87 0.88 (0.77–0.95) 0.83 (0.68–0.92) External validity  Excluding studies with high risk of bias or unclear bias related to patient selection within the domain of external validity 325 11 0.88 0.86 (0.76–0.92) 0.81 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to index test within the domain of external validity* 215 6 0.84 0.85 (0.77–0.90) 0.82 (0.72–0.87)  Excluding studies with high risk of bias or unclear bias related to reference standard within the domain of external validity** 344 8 0.89 0.84 (0.75–0.90) 0.80 (0.69–0.88) *Related to the applicability of the index test in relation to the aim of the meta-analysis. **Related to the applicability of the reference standard in relation to the aim of the meta-analysis. AUC; Area under the receiver operating characteristics curve; QUADAS-2; a revised tool for the quality assessment of diagnostic accuracy studies, PASL; pulsed arterial spin labeling, pCASL; pseudo-continous arterial spin labeling, rCBF; relative cerebral blood flow. View Large Fig. 3 View largeDownload slide Bivariate summary receiver operating characteristic curve (SROC). Fig. 3 View largeDownload slide Bivariate summary receiver operating characteristic curve (SROC). When we only included studies that used rCBFmax to discriminate between LGG and HGG, we found AUC of 0.90, with sensitivity ranging between 0.77 and 0.93 and specificity between 0.76 and 0.91. When we analyzed studies presenting data on mean rCBF, the AUC decreased to 0.78 with a summary sensitivity between 0.63 and 0.83 and summary specificity between 0.60 and 0.84. In the subgroup analyses on studies using the PASL technique, the diagnostic performance for differentiation between LGG and HGG, the AUC was 0.90 with a pooled sensitivity between 0.71 and 0.93, and pooled sensitivity between 0.69 and 0.92. In the subgroup analysis on studies using pCASL, the diagnostic performance AUC was 0.88 with a corresponding summary sensitivity between 0.79 and 0.91, and specificity between 0.65 and 0.87. In order to assess the degree of robustness, we conducted sensitivity analyses by excluding studies with a high or unclear risk of bias in the domain of external validity related to patient selection. This yielded an AUC of 0.88 and a summary sensitivity and specificity between 0.76 and 0.92 and between 0.72 and 0.87, respectively. When excluding studies with high or unclear risk of bias related to the reference standard, AUC was 0.89 with a summary sensitivity of 0.75 to 0.90 and specificity of 0.69 to 0.88. Discussion This study provides evidence that ASL MR imaging has an excellent diagnostic performance of 0.90 with pooled sensitivity of 0.86 and specificity of 0.80 for the differentiation of HGG from LGG. Furthermore, it is non-invasive and does not require a contrast agent. Subgroup analyses showed a similar high diagnostic performance for PASL as for PCASL for HGG detection. We showed that the parameter maximum relative CBF was associated with a higher diagnostic performance compared with the parameter mean relative CBF values, as also suggested by the meta-regression analyses. Sensitivity analyses based on internal and external validity assessment did not alter the results and strengthen the robustness of the conclusion. TSA revealed that the study was adequately powered statistically for detecting a difference between LGG and HGG. The strengths of this study are related to a broad search strategy, including several databases with a large number of studies, and an adequately powered meta-analysis including a large study cohort with aggregated data from 505 patients. Due to this sample size we were able to pool the data in subgroup analyses. Further, relevant meta-regression could be performed to support the findings. In accordance with the current high standards for diagnostic meta-analysis, this study was prospectively registered in PROSPERO38 and used updated guidelines for reporting meta-analysis20 and current recommendations for statistical methods in diagnostic meta-analysis.23 Furthermore, all steps in the meta-analysis were conducted by at least 2 researchers. The electronic search was conducted in well-established databases by one experienced librarian, who also searched unpublished gray literature. There were no signs of publication bias in the study. In relation to our study, a recent meta-analysis by Kong and colleagues analyzed CBF in LGG and HGG and found that HGG had a significantly higher CBF compared with LGG,19 which is in line with our data. However, diagnostic performance evaluation was not assessed in this study, necessitating clarification of the diagnostic value for ASL in identifying HGG. Dynamic susceptibility MR perfusion, an MR technique utilizing a gadolinium-based contrast agent, has shown an AUC of 0.77 to discriminate between grade II and grade III glioma.8 However, one strength of the ASL technique compared with other contrast agent–based perfusion techniques is that it does not require a gadolinium contrast agent, which has shown on MR images to accumulate in the brain9 and in autopsy studies.39,40 Furthermore, gadolinium-based contrast agents are associated with the considerable cost of approximately $200 per examination (Swedish retail price, standard dose—male 70 kg). An examination without the need of contrast agent also does not require testing for plasma creatinine and does not involve a risk for gadolinium-related toxicity.41 Our results in combination with the lesser performance of DSC perfusion can suggest that there is no added value for the purpose of glioma grading using a gadolinium contrast agent. However, most glioma patients require gadolinium to detect contrast enhancing lesions, which although imperfect, is commonly used to describe and characterize tumors, plan surgical resections, and measure Response Assessment in Neuro-Oncology or other standardized response to determine and evaluate treatment. Mean acquisition time for the ASL techniques in the present study was 300 seconds and thus does not prolong the MR scanning time substantially. Proposed drawbacks or limitations of the ASL technique are the signal-to-noise ratio, which is low in the cerebral white matter,42,43 as well as aging affecting cerebral perfusion. Based on the data from 505 patients reviewed here, there appears to be little evidence for this claim, and there was only a small difference in the diagnostic performance between PASL and pCASL. We acknowledge that there are a number of limitations in this study. Firstly, none of the included studies used the current WHO classification system from 2016.2 One study did report on molecular genetic data; however, there was no separation between metastases and glioma, and therefore this study was excluded.44 We did try to minimize this effect by excluding oligoastrocytomas when possible. The new WHO classification does not introduce, per definition, any change regarding glioma grading, therefore the comparison between HGG and LGG is still valid, despite the new classification of tumor subtypes. The included studies used several WHO classification systems; however, the meta-regression analyses did not detect any modulating effect on the outcome. Most of the studies used a 3T scanner; however, this did not seem to alter the outcome. A further limitation is the heterogeneous placement of ROI in the different studies and in the reference regions. Furthermore, there were substantial differences in acquisition and in postprocessing algorithms, if specified at all. In addition, given that grade II and grade III might be a further diagnostic challenge to discriminate between than HGG and LGG, this is a limitation of the study. Moreover, the bivariate meta-regression analyses lack a multivariable implementation that can take several potential moderating variables in the same analysis into consideration. Future studies would need to adopt the WHO classification in order to further assess the role of ASL for glioma grading. ASL MR imaging has an excellent diagnostic performance for differentiation between HGG and LGG. Given the low cost, absence of side effects, and efficacy, ASL MR imaging should be implemented in the routine workup of glioma. Supplementary Material Supplementary material is available online at Neuro-Oncology (http://neuro-oncology.oxfordjournals.org/). Funding None. Acknowledgments We acknowledge the assistance of Klas Moberg, Karolinska Insitutet University Library. Conflict of interest statement. None declared. Authorship Contribution. An.F.D., F.D.L., and Al.F.D. extracted data; An.F.D. and Al.F.D. conducted analyses; all authors contributed to critical assessment of the content and contributed to the preparation and the writing of the manuscript. All authors have read and approved the final version of the manuscript. References 1. Ostrom QT , Gittleman H , Liao P , et al. CBTRUS statistical report: primary brain and central nervous system tumors diagnosed in the United States in 2007–2011 . Neuro Oncol . 2014 ; 16 ( Suppl 4 ): iv1 – iv63 . Google Scholar CrossRef Search ADS PubMed 2. Louis DN , Perry A , Reifenberger G , et al. The 2016 World Health Organization classification of tumors of the central nervous system: a summary . Acta Neuropathol . 2016 ; 131 ( 6 ): 803 – 820 . Google Scholar CrossRef Search ADS PubMed 3. Smoll NR , Gautschi OP , Schatlo B , Schaller K , Weber DC . Relative survival of patients with supratentorial low-grade gliomas . Neuro Oncol . 2012 ; 14 ( 8 ): 1062 – 1069 . Google Scholar CrossRef Search ADS PubMed 4. Ohgaki H , Kleihues P . Population-based studies on incidence, survival rates, and genetic alterations in astrocytic and oligodendroglial gliomas . J Neuropathol Exp Neurol . 2005 ; 64 ( 6 ): 479 – 489 . Google Scholar CrossRef Search ADS PubMed 5. Stupp R , Mason WP , van den Bent MJ , et al. ; European Organisation for Research and Treatment of Cancer Brain Tumor and Radiotherapy Groups; National Cancer Institute of Canada Clinical Trials Group . Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma . N Engl J Med . 2005 ; 352 ( 10 ): 987 – 996 . Google Scholar CrossRef Search ADS PubMed 6. Tobin WO , Meyer FB , Keegan BM . Diagnostic yield and safety of cerebellar and brainstem parenchymal biopsy . World Neurosurg . 2015 ; 84 ( 6 ): 1973 – 1976 . Google Scholar CrossRef Search ADS PubMed 7. Jackson RJ , Fuller GN , Abi-Said D , et al. Limitations of stereotactic biopsy in the initial management of gliomas . Neuro Oncol . 2001 ; 3 ( 3 ): 193 – 200 . Google Scholar CrossRef Search ADS PubMed 8. Delgado AF , Delgado AF . Discrimination between glioma grades II and III using dynamic susceptibility perfusion MRI: a meta-analysis . AJNR Am J Neuroradiol . 2017 ; 38 ( 7 ): 1348 – 1355 . Google Scholar CrossRef Search ADS PubMed 9. Flood TF , Stence NV , Maloney JA , Mirsky DM . Pediatric brain: repeated exposure to linear gadolinium-based contrast material is associated with increased signal intensity at unenhanced T1-weighted MR imaging . Radiology . 2017 ; 282 ( 1 ): 222 – 228 . Google Scholar CrossRef Search ADS PubMed 10. Williams DS , Detre JA , Leigh JS , Koretsky AP . Magnetic resonance imaging of perfusion using spin inversion of arterial water . Proc Natl Acad Sci U S A . 1992 ; 89 ( 1 ): 212 – 216 . Google Scholar CrossRef Search ADS PubMed 11. Detre JA , Leigh JS , Williams DS , Koretsky AP . Perfusion imaging . Magn Reson Med . 1992 ; 23 ( 1 ): 37 – 45 . Google Scholar CrossRef Search ADS PubMed 12. Haller S , Zaharchuk G , Thomas DL , Lovblad KO , Barkhof F , Golay X . Arterial spin labeling perfusion of the brain: emerging clinical applications . Radiology . 2016 ; 281 ( 2 ): 337 – 356 . Google Scholar CrossRef Search ADS PubMed 13. Alsop DC , Detre JA , Golay X , et al. Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: a consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia . Magn Reson Med . 2015 ; 73 ( 1 ): 102 – 116 . Google Scholar CrossRef Search ADS PubMed 14. Warmuth C , Gunther M , Zimmer C . Quantification of blood flow in brain tumors: comparison of arterial spin labeling and dynamic susceptibility-weighted contrast-enhanced MR imaging . Radiology . 2003 ; 228 ( 2 ): 523 – 532 . Google Scholar CrossRef Search ADS PubMed 15. Wolf RL , Wang J , Wang S , et al. Grading of CNS neoplasms using continuous arterial spin labeled perfusion MR imaging at 3 Tesla . J Magn Reson Imaging . 2005 ; 22 ( 4 ): 475 – 482 . Google Scholar CrossRef Search ADS PubMed 16. Cebeci H , Aydin O , Ozturk-Isik E , et al. Assesment of perfusion in glial tumors with arterial spin labeling; comparison with dynamic susceptibility contrast method . Eur J Radiol . 2014 ; 83 ( 10 ): 1914 – 1919 . Google Scholar CrossRef Search ADS PubMed 17. Lehmann P , Monet P , de Marco G , et al. A comparative study of perfusion measurement in brain tumours at 3 Tesla MR: arterial spin labeling versus dynamic susceptibility contrast-enhanced MRI . Eur Neurol . 2010 ; 64 ( 1 ): 21 – 26 . Google Scholar CrossRef Search ADS PubMed 18. Roy B , Awasthi R , Bindal A , et al. Comparative evaluation of 3-dimensional pseudocontinuous arterial spin labeling with dynamic contrast-enhanced perfusion magnetic resonance imaging in grading of human glioma . J Comput Assist Tomogr . 2013 ; 37 ( 3 ): 321 – 326 . Google Scholar CrossRef Search ADS PubMed 19. Kong L , Chen H , Yang Y , Chen L . A meta-analysis of arterial spin labelling perfusion values for the prediction of glioma grade . Clin Radiol . 2017 ; 72 ( 3 ): 255 – 261 . Google Scholar CrossRef Search ADS PubMed 20. Liberati A , Altman DG , Tetzlaff J , et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration . J Clin Epidemiol . 2009 ; 62 ( 10 ): e1 – 34 . Google Scholar CrossRef Search ADS PubMed 21. Petra Macaskill CG , Deeks J , Harbord R , Takwoingi Y. Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy . Version 1.0.0. The Cochrane Collaboration; 2013. http://srdta.cochrane.org/. 22. Whiting PF , Rutjes AW , Westwood ME , et al. ; QUADAS-2 Group . QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies . Ann Intern Med . 2011 ; 155 ( 8 ): 529 – 536 . Google Scholar CrossRef Search ADS PubMed 23. McGrath TA , McInnes MD , Korevaar DA , Bossuyt PM . Meta-analyses of diagnostic accuracy in imaging journals: analysis of pooling techniques and their effect on summary estimates of diagnostic accuracy . Radiology . 2016 ; 281 ( 1 ): 78 – 85 . Google Scholar CrossRef Search ADS PubMed 24. Egger M , Davey Smith G , Schneider M , Minder C . Bias in meta-analysis detected by a simple, graphical test . BMJ . 1997 ; 315 ( 7109 ): 629 – 634 . Google Scholar CrossRef Search ADS PubMed 25. Pogue J , Yusuf S . Overcoming the limitations of current meta-analysis of randomised controlled trials . Lancet . 1998 ; 351 ( 9095 ): 47 – 52 . Google Scholar CrossRef Search ADS PubMed 26. Soni N , Dhanota DPS , Kumar S , Jaiswal AK , Srivastava AK . Perfusion MR imaging of enhancing brain tumors: comparison of arterial spin labeling technique with dynamic susceptibility contrast technique . Neurol India . 2017 ; 65 ( 5 ): 1046 – 1052 . Google Scholar CrossRef Search ADS PubMed 27. Weber MA , Zoubaa S , Schlieter M , et al. Diagnostic performance of spectroscopic and perfusion MRI for distinction of brain tumors . Neurology . 2006 ; 66 ( 12 ): 1899 – 1906 . Google Scholar CrossRef Search ADS PubMed 28. Fudaba H , Shimomura T , Abe T , et al. Comparison of multiple parameters obtained on 3T pulsed arterial spin-labeling, diffusion tensor imaging, and MRS and the Ki-67 labeling index in evaluating glioma grading . AJNR Am J Neuroradiol . 2014 ; 35 ( 11 ): 2091 – 2098 . Google Scholar CrossRef Search ADS PubMed 29. Furtner J , Schöpf V , Schewzow K , et al. Arterial spin-labeling assessment of normalized vascular intratumoral signal intensity as a predictor of histologic grade of astrocytic neoplasms . AJNR Am J Neuroradiol . 2014 ; 35 ( 3 ): 482 – 489 . Google Scholar CrossRef Search ADS PubMed 30. Kim MJ , Kim HS , Kim JH , Cho KG , Kim SY . Diagnostic accuracy and interobserver variability of pulsed arterial spin labeling for glioma grading . Acta Radiol . 2008 ; 49 ( 4 ): 450 – 457 . Google Scholar CrossRef Search ADS PubMed 31. Ma H , Wang Z , Xu K , et al. Three-dimensional arterial spin labeling imaging and dynamic susceptibility contrast perfusion-weighted imaging value in diagnosing glioma grade prior to surgery . Exp Ther Med . 2017 ; 13 ( 6 ): 2691 – 2698 . Google Scholar CrossRef Search ADS PubMed 32. Xiao HF , Chen ZY , Lou X , et al. Astrocytic tumour grading: a comparative study of three-dimensional pseudocontinuous arterial spin labelling, dynamic susceptibility contrast-enhanced perfusion-weighted imaging, and diffusion-weighted imaging . Eur Radiol . 2015 ; 25 ( 12 ): 3423 – 3430 . Google Scholar CrossRef Search ADS PubMed 33. Yang S , Zhao B , Wang G , et al. Improving the grading accuracy of astrocytic neoplasms noninvasively by combining timing information with cerebral blood flow: a multi-TI arterial spin-labeling MR Imaging study . AJNR Am J Neuroradiol . 2016 ; 37 ( 12 ): 2209 – 2216 . Google Scholar CrossRef Search ADS PubMed 34. Zeng Q , Jiang B , Shi F , Ling C , Dong F , Zhang J . 3D pseudocontinuous arterial spin-labeling MR imaging in the preoperative evaluation of gliomas . AJNR Am J Neuroradiol . 2017 ; 38 ( 10 ): 1876 – 1883 . Google Scholar CrossRef Search ADS PubMed 35. Zhang K , Langen KJ , Neuner I , et al. Relationship of regional cerebral blood flow and kinetic behaviour of O-(2-(18)F-fluoroethyl)-L-tyrosine uptake in cerebral gliomas . Nucl Med Commun . 2014 ; 35 ( 3 ): 245 – 251 . Google Scholar CrossRef Search ADS PubMed 36. Shen N , Zhao L , Jiang J , et al. Intravoxel incoherent motion diffusion-weighted imaging analysis of diffusion and microperfusion in grading gliomas and comparison with arterial spin labeling for evaluation of tumor perfusion . J Magn Reson Imaging . 2016 ; 44 ( 3 ): 620 – 632 . Google Scholar CrossRef Search ADS PubMed 37. Yang M , Wang S , Zhao Y , Zhao J . Management of intracranial aneurysm in children: clipped and coiled . Childs Nerv Syst . 2008 ; 24 ( 9 ): 1005 – 1012 . Google Scholar CrossRef Search ADS PubMed 38. Prospero International Prospective Register of Systematic Reviews . https://www.crd.york.ac.uk/PROSPERO/. 39. McDonald JS , McDonald RJ , Jentoft ME , et al. Intracranial gadolinium deposition following gadodiamide-enhanced magnetic resonance imaging in pediatric patients: a case-control study . JAMA Pediatr . 2017 ; 171 ( 7 ): 705 – 707 . Google Scholar CrossRef Search ADS PubMed 40. McDonald RJ , McDonald JS , Kallmes DF , et al. Gadolinium deposition in human brain tissues after contrast-enhanced MR imaging in adult patients without intracranial abnormalities . Radiology . 2017 ; 285 ( 2 ): 546 – 554 . Google Scholar CrossRef Search ADS PubMed 41. Marckmann P , Skov L , Rossen K , et al. Nephrogenic systemic fibrosis: suspected causative role of gadodiamide used for contrast-enhanced magnetic resonance imaging . J Am Soc Nephrol . 2006 ; 17 ( 9 ): 2359 – 2362 . Google Scholar CrossRef Search ADS PubMed 42. Chen Y , Wang DJ , Detre JA . Test-retest reliability of arterial spin labeling with common labeling strategies . J Magn Reson Imaging . 2011 ; 33 ( 4 ): 940 – 949 . Google Scholar CrossRef Search ADS PubMed 43. Spann SM , Kazimierski KS , Aigner CS , Kraiger M , Bredies K , Stollberger R . Spatio-temporal TGV denoising for ASL perfusion imaging . Neuroimage . 2017 ; 157 : 81 – 96 . Google Scholar CrossRef Search ADS PubMed 44. Brendle C , Hempel JM , Schittenhelm J , et al. Glioma grading and determination of IDH mutation status and ATRX loss by DCE and ASL perfusion . Clin Neuroradiol . 2017 ;doi: https://doi.org/10.1007/s00062-017-0590-z . © The Author(s) 2018. Published by Oxford University Press on behalf of the Society for Neuro-Oncology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 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)

Journal

Neuro-OncologyOxford University Press

Published: Jun 2, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off