Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Breast cancer mammographic diagnosis performance in a public health institution: a retrospective cohort study

Breast cancer mammographic diagnosis performance in a public health institution: a retrospective... Insights Imaging (2017) 8:581–588 DOI 10.1007/s13244-017-0573-2 ORIGINAL ARTICLE Breast cancer mammographic diagnosis performance in a public health institution: a retrospective cohort study 1 1 1 Juliana M.R.B. Mello & Fernando P. Bittelbrunn & Marcio A. B. C. Rockenbach & 1 1,2 1 Guilherme G. May & Leonardo M. Vedolin & Marilia S. Kruger & 1 1 1 Matheus D. Soldatelli & Guilherme Zwetsch & Gabriel T. F. de Miranda & Saone I. P. 1 1 Teixeira & Bruna S. Arruda Received: 27 April 2017 /Revised: 15 August 2017 /Accepted: 25 August 2017 /Published online: 4 October 2017 The Author(s) 2017. This article is an open access publication Abstract negative predictive value (NPV) was 98.32%. The abnormal Objectives To evaluate the quality assurance of mam- interpretation rate (recall rate) was 12.26%. mography results at a reference institution for the diag- Conclusions The results are appropriate when compared to nosis and treatment of breast cancer in southern Brazil, the values proposed by the BIRADS 5th edition. based on the BIRADS (Breast Imaging Reporting and Additionally, the study provided self-reflection considering Data System) 5th edition recommendations for auditing our radiological practice, which is essential for improvements purposes. and collaboration regarding breast cancer detection. It may Materials and methods Retrospective cohort and cross- stimulate better radiological practice performance and con- sectional study with 4502 patients (9668 mammogra- tinuing education, despite possible infrastructure and facility phies)) who underwent at least one or both breast mam- limitations. mographies throughout 2013 at a regional public hospi- Main Messages tal, linked to a federal public university. The results � Accurate quality performance rates are possible despite fi- were followed until 31 December 2014, including true nancial and governmental limitations. positives (TPs), true negatives (TNs), false positives � Low-income institutions should develop standardised team- (FPs), false negatives (FNs), positive predictive values work to improve radiological practice. (PPVs), negative predictive value (NPV), sensitivity and � Regular mammography audits may help to increase the specificity, with a confidence interval of 95%. quality of public health systems. Results The study showed high quality assurance, particularly . . regarding sensitivity (90.22%) and specificity (92.31%). The Keywords Mammography auditing Breast cancer Quality . . overall positive predictive value (PPV) was 65.35%, and the assurance SISMAMA Public health Abbreviations ACR American College of Radiology BIRADS Breast Imaging Reporting and Data System * Juliana M.R.B. Mello jmrmello@hcpa.edu.br; julianamrbm@gmail.com; BCSC Breast Cancer Surveillance Consortium juliana.mello@ppcr.org CI Confidence interval FN False negative Fernando P. Bittelbrunn FP False positive fbittelbrunn@hcpa.edu.br HCPA Hospital de Clínicas de Porto Alegre (Porto 1 Alegre’s Clinical Hospital) Radiological Department, Hospital de Clínicas de Porto Alegre INCA Instituto Nacional do Câncer (Cancer National (HCPA), 2350, Ramiro Barcelos St. Second floor, Porto Alegre 90035-903, Brazil Institute) PPV Positive predictive value Neuroradiology Department, Hospital Moinhos de Vento, Porto Alegre, Brazil NPV Negative predictive value 582 Insights Imaging (2017) 8:581–588 SISMAMA Sistema de Informação do Câncer de Mama the need for specific attention to the mammography audit (Breast Cancer Information System) performance and breast imaging education worldwide. TN True negative However, in Brazil, most imaging centres do not perform TP True positive an internal audit to compare their rates with the BIRADS recommendations. There is a Brazilian government task force to solve this problem, and it has created a national standardised programme, called SISMAMA (since 2009). Introduction This programme will help institutions and hospitals all over the country collect more accurate data and facilitate perfor- The main goal of breast cancer screening is to reduce mortality mance rate evaluations [22]. rates through early detection and proper treatment [1–6]. Great To encourage good radiological practice in our service ei- effort has been made in the last years by the international ther individually or as a group of radiologists and to determine scientific community, particularly in the radiology field, to whether our practice is in accordance with the international achieve this goal [7–9]. Recently, a systematic review showed performance standards, we performed a retrospective cohort a decrease of at least 20% for the mortality rate due to the use observational and cross-sectional study for the diagnostic of mammography for breast cancer screening [10]. Another mammography results in our hospital from 2013. study attributes most of the decrease in the breast cancer mor- tality rate of at least 38% since 1990 to early mammography detection [11]. Nevertheless, uncertainties remain about the Materials and methods magnitude of overdiagnosis associated with different possible screening strategies [10]. Meanwhile, the main imaging mo- We present a retrospective cohort observational and dality for breast cancer screening is still mammography [12, cross-sectional study, with all the BIRADS mammogra- 13]. phy results from 1 January 2013 until 31 December, As a method of standardising the process of reporting 2013, in a regional reference public hospital, which mammograms and also with the objective of facilitating data were submitted to an internal audit. The BIRADS mam- collection, the American College of Radiology (ACR) created mography results were analysed using a national soft- the lexicon published in the atlas BBreast Imaging Reporting ware, provided by the Brazilian National Cancer and Data System (BIRADS)^, which is used worldwide as a Institute (INCA), called SISMAMA, which aims to na- single and united system for the breast imaging radiology tionalise all mammography results as a way to congre- subspecialty [13]. It provides a guide to mammography audits gate more data. and performance measures [14]. A total of 9668 mammographies were performed in 4502 Mammography auditing assumes a relevant role when patients between the ages 28 and 92 years in one or both discussing the quality of breast imaging, whether regarding breasts in 2013 in our hospital. The mammography results, the radiologist’s interpretation or the quality of the images under approval from our regional and hospital ethics commit- taken [1, 15, 6]. The majority of developed countries, such tee, were submitted to an internal audit. We considered all the as the USA, have a federal standardisation for evaluating the mammographies as Bdiagnostic^, because we always have at personal performance of a radiologist as well as the imaging least one radiologist to evaluate the images before releasing centres [9, 16, 17]. In Europe, for example, there is the Dutch the patient in our service, and also because as we are a refer- Reference Centre of Screening, which conducts triennial au- ence centre for breast imaging in public health, almost all our dits of 17 mammography performing centres [18]. In this patients have been previously referred by many smaller re- European programme, they showed increased detection rates gional health centres as symptomatic patients. and increased sensitivity (up to 71.6%) over the last audit All the mammographies that had positive results (classified series performed from 1996 to 2013 [18]. as BIRADS 0, 4 and 5) were followed to verify further exam- In Latin America, some underdeveloped countries are try- inations, surgical procedures and biopsies performed for 1 ing to implement policies to address the growing incidence of year after the mammogram had been taken, until 31 breast cancer, especially in Brazil and Mexico, and also to December 2014. Mammographies classified as BIRADS 1, promote an early breast cancer detection strategy [19]. In 2 and 3 were considered negative and all results of BIRADS Asia, in Taiwan, there is a study showing improvement in their 6 were also not considered for audit purposes, because the performances, with a sensitivity rate varying from 79.6% to BIRADS 5theditionrecommends doingsoindiagnostic 87% and a specificity rate varying from 90.5% to 91.1% [20]. mammography audits. Meanwhile, even in developed countries, like the USA, many After selecting all considered BIRADS-positive results radiologists report their accuracy goals are below the pub- (BIRADS 0, 4 and 5), which added up to 664 mammogra- lished desirable benchmarks [21]. This information enhances phies, we reviewed the mammographic reports, breast Insights Imaging (2017) 8:581–588 583 ultrasound reports, anatomopathologic biopsy results (guided or further investigations that were needed because of the by either ultrasound or stereotactic imaging-guided proce- mammography results. dures to investigate suspicious cluster microcalcifications) No patient had any known direct or prompt adverse event and anatomopathologic results from surgery procedures. due to this study, besides undergoing a mammogram and be- The formulas and calculus used were from the last 5th ing exposed to the examination radiation. This radiation mean editionof the BIRADS atlas for auditing purposes. The perfor- (entrance surface dose) was 2.6 mGy during 2013 in our ser- mance measures were calculated based on the numbers of vice, depending on the thickness and composition of the true-positive (TP), false-positive (FP), true-negative (TN), breast. false-negative (FN) and total mammographic examinations STATA software (StataCorp 2013; Stata: Release 13 (N) summed over the year of 2013. Statistical Software) was chosen for statistical analysis. We The performance measures were calculated as follows: calculated the overall positive predictive value (PPV), nega- tive predictive value (NPV), sensitivity and specificity of our Sensitivity = TP/(TP + FN); mammography results with 95% of confidence intervals (CI). Specificity = TN/(TN + FP); Positive predictive value (PPV) = TP/(TP + FP); Negative predictive value (NPV) = TN/(TN + FN); Results Positive predictive value 2 (PPV2 or biopsy recommended) = TP/(number of diagnostic examinations The BIRADS distribution of the 9668 mammography recommended for tissue diagnosis); examinations (accounting a mammography for each Positive predictive value 3 (PPV 3 or biopsy breast separately) during 2013 in our institution is performed) = TP/(number of biopsies performed). shown in Table 1. The distribution of the anatomy composition of the breast, True positive (TP) is defined as a tissue diagnosis of cancer following the BIRADS recommendations, is shown in within 1 year after a positive examination. True negative (TN) Table 2. The anatomy of the breast (divided into extremely is defined as no known tissue diagnosis of cancer within 1 year dense, heterogeneously dense, scattered fibroglandular tissue after a negative examination. False negative (FN) is defined as and entirely fat) was divided into each BIRADS category, 0, 1, tissue diagnosis of cancer within 1 year after a negative exam- 2, 3, 4, 5 and 6, during 2013 at our institution. ination. False positive (FP) is defined as no known tissue We selected the categories 0, 4 and 5, which totaled 664 diagnosis of cancer within 1 year after a positive examination. mammographies in 2013, as recommended by the BIRADS Our standard comparisons were the values proposed by the 5th edition, to perform an internal audit of diagnostic BIRADS 5th edition recommendations for diagnostic auditing mammograms. purposes (table presented on page 593) and also by the Breast As the BIRADS 5th edition says, category 0 means the Cancer Surveillance Consortium (BCSC) Benchmarks results radiologist who did the report believes the mammogram find- available in the last BIRADS 5th edition (table presented on ing will need an additional evaluation, such as an additional page 591) to check if the radiological practice was accurate for examination or breast ultrasound. Categories 4 and 5 mean the a breast imaging reference centre. The BCSC Benchmarks radiologist who did the report believes the mammogram find- results we used to compare with our data were collected from ing will need further histopathological evaluation, such as a a table presented in the 5th BIRADS Atlas Edition (page 591), comprising 401,572 diagnostic mammography examinations Table 1 Number of mammographies and their distribution percentages taken between 1996 and 2005, collected from 153 mammog- in the BIRADS categories raphy facilities and 741 interpreting physicians that demon- Mammography distribution into the BIRADS categories in our institution strate a geographically and ethnically representative sample of the US population [13]. BIRADS categories No. of mammographies % of mammographies Our mammographic results included reports from four different radiologists (with 5 to 20 years of experience Category 0 478 4.94% in breast imaging), which were retrospectively accessed. Category 1 1641 16.97% This was a single-centre study with two digital mam- Category 2 6927 71.65% mography machines (Siemens Mammomat Inspiration), Category 3 235 2.43% whichwererecentlyacquiredbythehospital,andfive Category 4 137 1.42% ultrasound machines (Philips and Aloka). These ma- Category 5 49 0.51% chines were used to perform all mammographies and Category 6 201 2.08% any complementary examinations that needed to be car- Total 9668 100.00% ried out, such as ultrasounds, ultrasound-guided biopsies 584 Insights Imaging (2017) 8:581–588 Table 2 Breast anatomy composition distribution into the BIRADS categories during 2013 at our institution Breast anatomy composition into the BIRADS categories at our institution BIRADS category Breast anatomy composition Total % 0 Entirely fat 13 2.72% Extremely dense 134 28.03% Scattered areas of figroglandular tissue 155 32.43% Heterogeneously dense 176 36.82% BIRADS 0 total 478 1 Entirely fat 118 7.19% Extremely dense 374 22.79% Scattered areas of figroglandular tissue 598 36.44% Heterogeneously dense 551 33.58% BIRADS 1 total 1641 2 Entirely fat 571 8.27% Extremely dense 1342 19.43% Scattered areas of figroglandular tissue 2783 40.29% Heterogeneously dense 2211 32.01% BIRADS 2 total 6907 3 Entirely fat 11 4.68% Extremely dense 72 30.64% Scattered areas of figroglandular tissue 76 32.34% Heterogeneously dense 76 32.34% BIRADS 3 total 235 4 Entirely fat 5 3.65% Extremely dense 46 33.58% Scattered areas of figroglandular tissue 33 24.09% Heterogeneously dense 53 38.69% BIRADS 4 total 137 5 Entirely fat 3 6.12% Extremely dense 12 24.49% Scattered areas of figroglandular tissue 24 48.98% Heterogeneously dense 10 20.41% BIRADS 5 total 49 6 Entirely fat 9 4.48% Extremely dense 62 30.85% Scattered areas of figroglandular tissue 65 32.34% Heterogeneously dense 65 32.34% BIRADS 6 total 201 core biopsy. The radiologist decision between categories 4 or Our cancer detection rate was 19.34% and the abnormal 5 depends on his/her analysis of the mammogram finding and interpretation rate (or recall rate) was 12.26%, as shown in how likely it is to be a breast cancer or not. How the radiolo- Fig. 1. The BCSC Benchmarks (for either palpable or non- gist makes his/her decision depends on training, education, palpable mammographic findings) show similar values to our experience and the BIRADS 5th edition recommendations institution, although smaller. Nevertheless, our institution re- [13]. call rate value (for both palpable and non-palpable findings) is There were 83 true positives (TP), 44 false positives (FP), 9 smaller than the recommended BIRADS (for either palpable false negatives (FN) and 528 true negatives (TN) when con- and non-palpable findings) standards. sidering the sample of 664 mammographies in the categories Our positive predictive value 2 (recommendation for tissue 0, 4 and 5. diagnosis or PPV2) was 43%, as illustrated in Fig. 2. This is an Insights Imaging (2017) 8:581–588 585 Fig. 1 Percentages of abnormal 30.00% 25.00% 25.00% interpretation/recall rate 25.00% 20.00% 15.00% 12.26% 11.30% 9.60% 10.00% 5.00% 0.00% HPCA for both BCSC Benchmark BCSC for palpable BIRADS for non BIRADS for palpable and non for non palpable mammographic palpable palpable palpable mammographic findings mammographic mammographic mammographic findings findings findings findings Our Institution and the BCSC and BIRADS standard proposed values important subanalysis because it shows the accuracy radiolo- The negative predictive value (NPV) was estimated in gists expect for recommending a histopathological correlation 98.32% (95% CI 96.93%–99.09%). for a suspicious mammographic finding. Positive predictive value 3 (biopsy performed or PPV3) was 60%, as shown in Fig. 3. Discussion Usually the PPV3 is higher than the PPV2, since the biop- sies that will actually be done will probably show positive As a limitation of this study, the comparison between our rates results for cancer. Regarding this topic, our results showed and the BCSC rates regarding Bminimally invasive cancers or the same outcome. As BIRADS recommends, PPV3 is a more DCIS^ and Binvasive cancers with negative axillary lymph accurate indicator of cancer status, and PPV2 is more accurate nodes^ could not be done using proper additional statistical for interpretive performance [13]. analysis. The main reasons were limited access to BCSC data Our rate of negative axillary lymph nodes was and also because their sample size for the table presented on 75.93%, as shown in Fig. 4. page 591 of the last 5th edition of BIRADS (401,572 diag- The minimum cancer rate (invasive cancer with 1.0 cm or nostic mammograms between 1996 to 2005) had a distinct less or DCIS, ductal carcinoma in situ) was 36%, as illustrated difference from our sample size (9668 mammograms during in Fig. 5. 2013). This distinct difference between the sample sizes com- The sensibility of our diagnostic mammograms in pared could have led us to false results. However, if we 2013 was 90.22% (95% CI 82.24%–95.43%) and the consider a confidence interval of 95%, our sample size specificity was 92.31% (95% CI 89.91% to 94.36%). should be accurate to compare our proportions and The positive predictive value (PPV) was estimated in BCSC proportions, but we could not calculate a specific 65.35% (95% CI 58.49%–71.63%). p-value for these comparisons. 60.00% Fig. 2 PPV2 showing how 50.00% frequently the performed biopsies 50.00% 43.00% 43.00% indicate cancer 40.00% 40.00% 31.00% 30.00% 20.00% 10.00% 0.00% HPCA for both BCSC Benchmark for BCSC for palpable BIRADS for non BIRADS for palpable palpable and non non palpable mammographic palpable mammographic palpable mammographic findings mammographic findings mammographic findings findings findings Our Institution and the BCSC and BIRADS Standard proposed values Recall Rate 586 Insights Imaging (2017) 8:581–588 70.00% Fig. 3 PPV3 showing how 60.00% 55.00% frequently performed biopsies 60.00% 49.00% indicated cancers 45.00% 50.00% 36.00% 40.00% 30.00% 20.00% 10.00% 0.00% HPCA for both BCSC Benchmark BCSC for palpable BIRADS for non BIRADS for palpable and non for non palpable mammographic palpable palpable palpable mammographic findings mammographic mammographic mammographic findings findings findings findings Our Institution and the BCSC and BIRADS Standard proposed values We had a total of 58 losses, whose data records were not taken, allowing performing additional mammography exami- available in the hospital electronic system, mainly because nations when the patient is still inside the service. Other rea- these 58 patients (8.7% of our sample of 664 positive mam- sons are the experience and training of the radiologist's team. mography results) had private health care systems, which did Our positive predictive value 2 (recommendation for tissue not allow them to be classified into the hospital's public health diagnosis or PPV2) was 43%, which is exactly the same value care electronic system. as in BCSC for palpable findings (43%), above the value of Our cancer detection rate was 19.34%, close to the mini- BCSC Benchmarks for non-palpable findings (31%), and near mum 20% required for diagnostic mammograms according to the value requested by BIRADS for non-palpable findings BIRADS. This result is possibly related to a minor skewed (15–40%) and by BIRADS for palpable findings (25–50%). distribution of cases mainly due to the few screening mam- This PPV2 shows that our recommendations for biopsies are mograms in the total number of 9668 mammographies select- accurate. ed and considered as diagnostic in our study. Additionally, it Our positive predictive value 3 (biopsy performed or may be secondary to the fact that our SISMAMA nationalised PPV3) was 60%, which was above all the standard compari- software does not have a digital option to separate systematic sons: BCSC Benchmarks for non-palpable findings (36%), screening from diagnostic mammograms. This may result in BCSC for palpable findings (49%), BIRADS for non- some minor skewed distribution of the data when considering palpable findings (20–45%) and BIRADS for palpable find- the cancer detection rate, which would certainly not invalidate ings (30–55%). We believe the reason for this high PPV3 is our results. the fact that most of symptomatic patients were referred by The abnormal interpretation rate (or recall rate) was previous smaller health care centres to our hospital, which 12.26%, below the BCSC Benchmarks for diagnostic mam- centralise most of the patients who will need to be treated mograms (13.3%) and adequate for the value recommended for breast cancer using the national public health system. by BIRADS (8–25% for non-palpable findings or 10–25% for Moreover, we have patients who already had undergone some palpable findings), as shown in Fig. 1. There are many reasons previous screening before having a mammography in our hos- for this lower recall rate, such as the prompt review of the pital. Another reason to consider is that many patients are images by a breast radiologist immediately after they are already symptomatic at the time of the breast cancer diagnosis 75.93% Fig. 4 Percentages of invasive 80.00% 68.20% cancers with negative axillary 70.00% lymph nodes 56.50% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% HPCA for both palpable and non BCSC Benchmark for non BCSC for palpable palpable mammographic findings palpable mammographic findings mammographic findings Our Institution and the BCSC values Insights Imaging (2017) 8:581–588 587 45.00% Fig. 5 Percentages of minimum 39.00% 40.00% 36.00% invasive cancers (<1.0 cm) or DCIS 35.00% 30.00% 25.00% 20.00% 15.00% 15.00% 10.00% 5.00% 0.00% HPCA for both palpable and non BCSC Benchmark for non BCSC for palpable palpable mammographic findings palpable mammographic findings mammographic findings Our Institution and BCSC values in south Brazil, as well as in other low-income regions world- institutions worldwide, even with poorer general infrastruc- wide. This situation reflects the fact that some patients do not ture conditions when compared to the available facilities in have ideal social and economic possibilities for having a reference institutions in developed countries. mammography performed in an earlier stage of the disease. Our secondary goal in doing the study was to encourage Our rate of negative axillary lymph nodes was 75.93%, radiologists to be curious and try to look for their overall above the BCSC Benchmarks for non-palpable findings performance and self-evaluation (either individually or as a (68.2%) and BCSC in palpable findings (56.5%). This is an- team). As discussed in the literature, the feedback to the team other positive result, which may contribute to lower morbidity of radiologists as well as the moments of self-reflection and and mortality rates for the patients’ outcomes. insights in recall behaviour on mammograms is extremely The minimum cancer rate (invasive cancer with 1.0 cm or valuable. By doing this, other institutions also may improve less or DCIS, ductal carcinoma in situ) was 36%, near the their internal collaboration process towards meeting the pa- BCSC Benchmarks for non-palpable findings (39%) and tients’ needs. Additionally, we had more understanding about above the BCSC for palpable findings (15%). For the mini- how we can achieve higher scores and solve some problems mum cancer rate, BIRADS does not determine a fixed num- concerning data collection. ber, because it is implicit that the higher it is, the better it will Knowing the teamwork performance is essential to discov- be for both patients and professionals related to the treatment ering different means of improvement and to improving the of these patients. radiological medical practice. By pointing out the limitations The sensitivity of our diagnostic mammograms in 2013 and also our good results, the current study had a positive was 90.22%, above all the standard comparisons: the impact on our daily practice, especially regarding our focus BIRADS recommendation for non-palpable findings (80% on enhancing the breast cancer diagnosis detection when or higher), BIRADS for palpable findings (85% or higher), performing mammograms. In the future, we intend to imple- BCSC Benchmarks for non-palpable findings (83.1%) and ment arrangements for collecting data, analysing results and BCSC in palpable findings (87.8%). also concentrating more efforts on continuing medical educa- The specificity of our diagnostic mammograms in 2013 tion for all radiologists. was 92.31%, adequate for all the standard comparisons: BIRADS recommendations for non-palpable findings (80– 95%), BIRADS recommendations for palpable findings (83– 95%), BCSC Benchmarks for non-palpable findings (92.2%) Conclusion and BCSC for palpable findings (93.2%). As shown in the results, most of our performance rates This study is an internal audit in a regional reference institu- regarding mammography reports are similar to (sometimes tion for breast cancer diagnosis, treatment and follow-up in an better than) the values proposed by BIRADS as good radio- underdeveloped country. We performed a retrospective obser- logical practice. The sensitivity and specificity of our diagnos- vational and cross-sectional cohort study to analyse our radio- tic mammographies are adequate for all the standard compar- logical practice accuracy considering mammography reports. isons, moving the values of our institution to the international Despite possible limitations, our results showed high-quality aim of reducing the mortality and morbidity related to breast rates, demonstrating that standardised teamwork can promote cancer. the quality of public health systems worldwide. Additionally, Nevertheless, this study has a higher internal than external it may encourage radiologists to seek performance feedback validity mainly because it is a single-centre study. Despite this, and continuing medical education, thus resulting in better out- it can stimulate good radiological practice in public health comes for the patients. 588 Insights Imaging (2017) 8:581–588 Acknowledgements We specially thank the Hospital de Clínicas de versus experienced radiographers in the Dutch breast cancer screen- Porto Alegre (HCPA), Radiology Department, the medical physicists ing programme. Eur Radiol 25(11):3322–3327 who work with us and the physicist Rafael Zimmer from the 8. Vernacchia FS, Pena ZG (2009) Digital mammography: its impact Postgraduate and Research Department. on recall rates and cancer detection rates in a small community- We kindly thank Prof. Wolnei Caumo, MD, PhD, Prof. José Roberto based radiology practice. Am J Roentgenol 193(2):582–585 Goldim, PhD, Prof. Jorge Villanova Biazús, MD, PhD, and Prof. Andrea 9. Sickles EA (2007) Auditing your breast imaging practice: an Pires Souto Damin, MD, PhD, from the Federal University of Rio Grande evidence-based approach. Semin Roentgenol 42(4):211–217 do Sul (UFRGS), Prof. Maria Helena Pitombeira Rigatto, MD, PhD, from 10. Myers ER, Moorman P, Gierisch JM, Havrilesky LJ, Grimm LJ, PUCRS, and Prof. Felipe Fregni, MD, PhD, MPH, MEd, from the PPCR Ghate S et al (2015) Benefits and harms of breast cancer screening. Course offered by the Harvard T.H. Chan School of Public Health. JAMA 314(15):1615 11. Monticciolo DL, Newell MS, Hendrick RE, Helvie MA, Moy L, Monsees B et al (2017) Breast cancer screening for average-risk Compliance with ethical standards All procedures performed in this women: recommendations from the ACR Commission on breast study were in accordance to the ethical standards of the institutional and imaging. J Am Coll Radiol. https://doi.org/10.1016/j.jacr.2017.06. national research committee. 12. Duarte DL (2006) In: Guanabara Koogan (ed) A Mama Em Conflicts of interest The authors have no conflicts of interest to Imagens. Guanabara Koogan, Rio de Janeiro 327 p declare. 13. D’Orsi CJ, Sickles EA, Mendelson EB, Morris EA (2013) Breast imaging reporting and data system: ACR BI-RADS—breast imag- Open Access This article is distributed under the terms of the Creative ing atlas 5th edition, 5th edn. American College of Radiology Commons Attribution 4.0 International License (http:// (ACR), Reston creativecommons.org/licenses/by/4.0/), which permits unrestricted use, 14. Wu Y, Rubin DL, Woods RW, Elezaby M, Burnside ES (2014) distribution, and reproduction in any medium, provided you give appro- Developing a comprehensive database management system. priate credit to the original author(s) and the source, provide a link to the Cancer Informat 13:53–62 Creative Commons license, and indicate if changes were made. 15. Geller BM, Ichikawa L, Miglioretti DL, Eastman D (2012) Web- based mammography audit feedback. Am J Roentgenol 198(6): 562–567 16. Dee KE, Sickles EA (2001) Medical audit of diagnostic mammog- raphy examinations: comparison with screening outcomes obtained References concurrently. Am J Roentgenol 176:729–733 17. Tabar LVBCH et al (2000) The Swedish Two-County trial twenty years later: updated mortality results an new insights from long- 1. Geertse TD, Holland R, Timmers JMH, Paap E, Pijnappel RM, term follow-up. Radiol Clin N Am 38:625–651 Broeders MJM et al (2015) Value of audits in breast cancer screen- 18. Geertse TD, Holland R, Timmers JM, Paap E, Pijnappel RM, ing quality assurance programmes. Eur Radiol 25 (11):3338–3347 Broeders MJ et al (2015) Value of audits in breast cancer screening 2. Howard JH, Bland KI (2012) Current management and treatment quality assurance programmes. Eur Radiol 25(11):3338–3347 strategies for breast cancer. Curr Opin Obstet Gynecol 24(1):44–48 19. Nigenda G, Gonzalez-Robledo MC, Gonzalez-Robledo LM, 3. Kopans D (2008) In: Koogan G (ed) Diagnóstico Por Imagem da Bejarano-Arias RM (2016) Breast cancer policy in Latin Mama, 3 edição edn. Guanabara Koogan, Rio de Janeiro 1061 p America: account of achievements and challenges in five countries. 4. Lee CH, Dershaw DD, Kopans D, Evans P, Monsees B, Glob Health 12(1):39 Monticciolo D et al (2010) Breast cancer screening with imaging: 20. Pan HB, Wong KF, Yang TL, Hsu GC, Chou CP, Huang JS et al recommendations from the Society of Breast Imaging and the ACR (2014) The outcome of a quality-controlled mammography screen- on the use of mammography, breast MRI, breast ultrasound, and ing program: experience from a population-based study in Taiwan. other technologies for the detection of clinically occult breast can- J Chin Med Assoc 77(10):531–534 cer. J Am Coll Radiol 7(1):18–27 21. Jackson SL, Cook AJ, Miglioretti DL, Carney PA, Geller BM, 5. Oeffinger KC, Fontham ETH, Etzioni R, Herzig A, Michaelson JS, Onega T et al (2012) Are radiologists' goals for mammography Shih Y-CT et al (2015) Breast cancer screening for women at aver- accuracy consistent with published recommendations? Acad age risk. JAMA 314(15):1599 Radiol 19(3):289–295 6. Elmore JG et al (2010) Radiologists attitudes and use of mammog- 22. Passman L, Ramalho Ortigão Farias A, Tomazelli J, Franco de raphy audit reports. Acad Radiol 17(6):752–760 Abreu D, Dias M, de Assis M et al (2011) SISMAMA- 7. van Landsveld-Verhoeven C, den Heeten GJ, Timmers J, Broeders implementation of an information system for breast cancer early MJM (2015) Mammographic positioning quality of newly trained detection programs in Brazil. Breast 20:35–39 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Insights into Imaging Springer Journals

Breast cancer mammographic diagnosis performance in a public health institution: a retrospective cohort study

Loading next page...
1
 
/lp/springer_journal/breast-cancer-mammographic-diagnosis-performance-in-a-public-health-sMm8fWjOx6

References (31)

Publisher
Springer Journals
Copyright
Copyright © 2017 by The Author(s)
Subject
Medicine & Public Health; Imaging / Radiology; Diagnostic Radiology; Interventional Radiology; Neuroradiology; Ultrasound; Internal Medicine
eISSN
1869-4101
DOI
10.1007/s13244-017-0573-2
Publisher site
See Article on Publisher Site

Abstract

Insights Imaging (2017) 8:581–588 DOI 10.1007/s13244-017-0573-2 ORIGINAL ARTICLE Breast cancer mammographic diagnosis performance in a public health institution: a retrospective cohort study 1 1 1 Juliana M.R.B. Mello & Fernando P. Bittelbrunn & Marcio A. B. C. Rockenbach & 1 1,2 1 Guilherme G. May & Leonardo M. Vedolin & Marilia S. Kruger & 1 1 1 Matheus D. Soldatelli & Guilherme Zwetsch & Gabriel T. F. de Miranda & Saone I. P. 1 1 Teixeira & Bruna S. Arruda Received: 27 April 2017 /Revised: 15 August 2017 /Accepted: 25 August 2017 /Published online: 4 October 2017 The Author(s) 2017. This article is an open access publication Abstract negative predictive value (NPV) was 98.32%. The abnormal Objectives To evaluate the quality assurance of mam- interpretation rate (recall rate) was 12.26%. mography results at a reference institution for the diag- Conclusions The results are appropriate when compared to nosis and treatment of breast cancer in southern Brazil, the values proposed by the BIRADS 5th edition. based on the BIRADS (Breast Imaging Reporting and Additionally, the study provided self-reflection considering Data System) 5th edition recommendations for auditing our radiological practice, which is essential for improvements purposes. and collaboration regarding breast cancer detection. It may Materials and methods Retrospective cohort and cross- stimulate better radiological practice performance and con- sectional study with 4502 patients (9668 mammogra- tinuing education, despite possible infrastructure and facility phies)) who underwent at least one or both breast mam- limitations. mographies throughout 2013 at a regional public hospi- Main Messages tal, linked to a federal public university. The results � Accurate quality performance rates are possible despite fi- were followed until 31 December 2014, including true nancial and governmental limitations. positives (TPs), true negatives (TNs), false positives � Low-income institutions should develop standardised team- (FPs), false negatives (FNs), positive predictive values work to improve radiological practice. (PPVs), negative predictive value (NPV), sensitivity and � Regular mammography audits may help to increase the specificity, with a confidence interval of 95%. quality of public health systems. Results The study showed high quality assurance, particularly . . regarding sensitivity (90.22%) and specificity (92.31%). The Keywords Mammography auditing Breast cancer Quality . . overall positive predictive value (PPV) was 65.35%, and the assurance SISMAMA Public health Abbreviations ACR American College of Radiology BIRADS Breast Imaging Reporting and Data System * Juliana M.R.B. Mello jmrmello@hcpa.edu.br; julianamrbm@gmail.com; BCSC Breast Cancer Surveillance Consortium juliana.mello@ppcr.org CI Confidence interval FN False negative Fernando P. Bittelbrunn FP False positive fbittelbrunn@hcpa.edu.br HCPA Hospital de Clínicas de Porto Alegre (Porto 1 Alegre’s Clinical Hospital) Radiological Department, Hospital de Clínicas de Porto Alegre INCA Instituto Nacional do Câncer (Cancer National (HCPA), 2350, Ramiro Barcelos St. Second floor, Porto Alegre 90035-903, Brazil Institute) PPV Positive predictive value Neuroradiology Department, Hospital Moinhos de Vento, Porto Alegre, Brazil NPV Negative predictive value 582 Insights Imaging (2017) 8:581–588 SISMAMA Sistema de Informação do Câncer de Mama the need for specific attention to the mammography audit (Breast Cancer Information System) performance and breast imaging education worldwide. TN True negative However, in Brazil, most imaging centres do not perform TP True positive an internal audit to compare their rates with the BIRADS recommendations. There is a Brazilian government task force to solve this problem, and it has created a national standardised programme, called SISMAMA (since 2009). Introduction This programme will help institutions and hospitals all over the country collect more accurate data and facilitate perfor- The main goal of breast cancer screening is to reduce mortality mance rate evaluations [22]. rates through early detection and proper treatment [1–6]. Great To encourage good radiological practice in our service ei- effort has been made in the last years by the international ther individually or as a group of radiologists and to determine scientific community, particularly in the radiology field, to whether our practice is in accordance with the international achieve this goal [7–9]. Recently, a systematic review showed performance standards, we performed a retrospective cohort a decrease of at least 20% for the mortality rate due to the use observational and cross-sectional study for the diagnostic of mammography for breast cancer screening [10]. Another mammography results in our hospital from 2013. study attributes most of the decrease in the breast cancer mor- tality rate of at least 38% since 1990 to early mammography detection [11]. Nevertheless, uncertainties remain about the Materials and methods magnitude of overdiagnosis associated with different possible screening strategies [10]. Meanwhile, the main imaging mo- We present a retrospective cohort observational and dality for breast cancer screening is still mammography [12, cross-sectional study, with all the BIRADS mammogra- 13]. phy results from 1 January 2013 until 31 December, As a method of standardising the process of reporting 2013, in a regional reference public hospital, which mammograms and also with the objective of facilitating data were submitted to an internal audit. The BIRADS mam- collection, the American College of Radiology (ACR) created mography results were analysed using a national soft- the lexicon published in the atlas BBreast Imaging Reporting ware, provided by the Brazilian National Cancer and Data System (BIRADS)^, which is used worldwide as a Institute (INCA), called SISMAMA, which aims to na- single and united system for the breast imaging radiology tionalise all mammography results as a way to congre- subspecialty [13]. It provides a guide to mammography audits gate more data. and performance measures [14]. A total of 9668 mammographies were performed in 4502 Mammography auditing assumes a relevant role when patients between the ages 28 and 92 years in one or both discussing the quality of breast imaging, whether regarding breasts in 2013 in our hospital. The mammography results, the radiologist’s interpretation or the quality of the images under approval from our regional and hospital ethics commit- taken [1, 15, 6]. The majority of developed countries, such tee, were submitted to an internal audit. We considered all the as the USA, have a federal standardisation for evaluating the mammographies as Bdiagnostic^, because we always have at personal performance of a radiologist as well as the imaging least one radiologist to evaluate the images before releasing centres [9, 16, 17]. In Europe, for example, there is the Dutch the patient in our service, and also because as we are a refer- Reference Centre of Screening, which conducts triennial au- ence centre for breast imaging in public health, almost all our dits of 17 mammography performing centres [18]. In this patients have been previously referred by many smaller re- European programme, they showed increased detection rates gional health centres as symptomatic patients. and increased sensitivity (up to 71.6%) over the last audit All the mammographies that had positive results (classified series performed from 1996 to 2013 [18]. as BIRADS 0, 4 and 5) were followed to verify further exam- In Latin America, some underdeveloped countries are try- inations, surgical procedures and biopsies performed for 1 ing to implement policies to address the growing incidence of year after the mammogram had been taken, until 31 breast cancer, especially in Brazil and Mexico, and also to December 2014. Mammographies classified as BIRADS 1, promote an early breast cancer detection strategy [19]. In 2 and 3 were considered negative and all results of BIRADS Asia, in Taiwan, there is a study showing improvement in their 6 were also not considered for audit purposes, because the performances, with a sensitivity rate varying from 79.6% to BIRADS 5theditionrecommends doingsoindiagnostic 87% and a specificity rate varying from 90.5% to 91.1% [20]. mammography audits. Meanwhile, even in developed countries, like the USA, many After selecting all considered BIRADS-positive results radiologists report their accuracy goals are below the pub- (BIRADS 0, 4 and 5), which added up to 664 mammogra- lished desirable benchmarks [21]. This information enhances phies, we reviewed the mammographic reports, breast Insights Imaging (2017) 8:581–588 583 ultrasound reports, anatomopathologic biopsy results (guided or further investigations that were needed because of the by either ultrasound or stereotactic imaging-guided proce- mammography results. dures to investigate suspicious cluster microcalcifications) No patient had any known direct or prompt adverse event and anatomopathologic results from surgery procedures. due to this study, besides undergoing a mammogram and be- The formulas and calculus used were from the last 5th ing exposed to the examination radiation. This radiation mean editionof the BIRADS atlas for auditing purposes. The perfor- (entrance surface dose) was 2.6 mGy during 2013 in our ser- mance measures were calculated based on the numbers of vice, depending on the thickness and composition of the true-positive (TP), false-positive (FP), true-negative (TN), breast. false-negative (FN) and total mammographic examinations STATA software (StataCorp 2013; Stata: Release 13 (N) summed over the year of 2013. Statistical Software) was chosen for statistical analysis. We The performance measures were calculated as follows: calculated the overall positive predictive value (PPV), nega- tive predictive value (NPV), sensitivity and specificity of our Sensitivity = TP/(TP + FN); mammography results with 95% of confidence intervals (CI). Specificity = TN/(TN + FP); Positive predictive value (PPV) = TP/(TP + FP); Negative predictive value (NPV) = TN/(TN + FN); Results Positive predictive value 2 (PPV2 or biopsy recommended) = TP/(number of diagnostic examinations The BIRADS distribution of the 9668 mammography recommended for tissue diagnosis); examinations (accounting a mammography for each Positive predictive value 3 (PPV 3 or biopsy breast separately) during 2013 in our institution is performed) = TP/(number of biopsies performed). shown in Table 1. The distribution of the anatomy composition of the breast, True positive (TP) is defined as a tissue diagnosis of cancer following the BIRADS recommendations, is shown in within 1 year after a positive examination. True negative (TN) Table 2. The anatomy of the breast (divided into extremely is defined as no known tissue diagnosis of cancer within 1 year dense, heterogeneously dense, scattered fibroglandular tissue after a negative examination. False negative (FN) is defined as and entirely fat) was divided into each BIRADS category, 0, 1, tissue diagnosis of cancer within 1 year after a negative exam- 2, 3, 4, 5 and 6, during 2013 at our institution. ination. False positive (FP) is defined as no known tissue We selected the categories 0, 4 and 5, which totaled 664 diagnosis of cancer within 1 year after a positive examination. mammographies in 2013, as recommended by the BIRADS Our standard comparisons were the values proposed by the 5th edition, to perform an internal audit of diagnostic BIRADS 5th edition recommendations for diagnostic auditing mammograms. purposes (table presented on page 593) and also by the Breast As the BIRADS 5th edition says, category 0 means the Cancer Surveillance Consortium (BCSC) Benchmarks results radiologist who did the report believes the mammogram find- available in the last BIRADS 5th edition (table presented on ing will need an additional evaluation, such as an additional page 591) to check if the radiological practice was accurate for examination or breast ultrasound. Categories 4 and 5 mean the a breast imaging reference centre. The BCSC Benchmarks radiologist who did the report believes the mammogram find- results we used to compare with our data were collected from ing will need further histopathological evaluation, such as a a table presented in the 5th BIRADS Atlas Edition (page 591), comprising 401,572 diagnostic mammography examinations Table 1 Number of mammographies and their distribution percentages taken between 1996 and 2005, collected from 153 mammog- in the BIRADS categories raphy facilities and 741 interpreting physicians that demon- Mammography distribution into the BIRADS categories in our institution strate a geographically and ethnically representative sample of the US population [13]. BIRADS categories No. of mammographies % of mammographies Our mammographic results included reports from four different radiologists (with 5 to 20 years of experience Category 0 478 4.94% in breast imaging), which were retrospectively accessed. Category 1 1641 16.97% This was a single-centre study with two digital mam- Category 2 6927 71.65% mography machines (Siemens Mammomat Inspiration), Category 3 235 2.43% whichwererecentlyacquiredbythehospital,andfive Category 4 137 1.42% ultrasound machines (Philips and Aloka). These ma- Category 5 49 0.51% chines were used to perform all mammographies and Category 6 201 2.08% any complementary examinations that needed to be car- Total 9668 100.00% ried out, such as ultrasounds, ultrasound-guided biopsies 584 Insights Imaging (2017) 8:581–588 Table 2 Breast anatomy composition distribution into the BIRADS categories during 2013 at our institution Breast anatomy composition into the BIRADS categories at our institution BIRADS category Breast anatomy composition Total % 0 Entirely fat 13 2.72% Extremely dense 134 28.03% Scattered areas of figroglandular tissue 155 32.43% Heterogeneously dense 176 36.82% BIRADS 0 total 478 1 Entirely fat 118 7.19% Extremely dense 374 22.79% Scattered areas of figroglandular tissue 598 36.44% Heterogeneously dense 551 33.58% BIRADS 1 total 1641 2 Entirely fat 571 8.27% Extremely dense 1342 19.43% Scattered areas of figroglandular tissue 2783 40.29% Heterogeneously dense 2211 32.01% BIRADS 2 total 6907 3 Entirely fat 11 4.68% Extremely dense 72 30.64% Scattered areas of figroglandular tissue 76 32.34% Heterogeneously dense 76 32.34% BIRADS 3 total 235 4 Entirely fat 5 3.65% Extremely dense 46 33.58% Scattered areas of figroglandular tissue 33 24.09% Heterogeneously dense 53 38.69% BIRADS 4 total 137 5 Entirely fat 3 6.12% Extremely dense 12 24.49% Scattered areas of figroglandular tissue 24 48.98% Heterogeneously dense 10 20.41% BIRADS 5 total 49 6 Entirely fat 9 4.48% Extremely dense 62 30.85% Scattered areas of figroglandular tissue 65 32.34% Heterogeneously dense 65 32.34% BIRADS 6 total 201 core biopsy. The radiologist decision between categories 4 or Our cancer detection rate was 19.34% and the abnormal 5 depends on his/her analysis of the mammogram finding and interpretation rate (or recall rate) was 12.26%, as shown in how likely it is to be a breast cancer or not. How the radiolo- Fig. 1. The BCSC Benchmarks (for either palpable or non- gist makes his/her decision depends on training, education, palpable mammographic findings) show similar values to our experience and the BIRADS 5th edition recommendations institution, although smaller. Nevertheless, our institution re- [13]. call rate value (for both palpable and non-palpable findings) is There were 83 true positives (TP), 44 false positives (FP), 9 smaller than the recommended BIRADS (for either palpable false negatives (FN) and 528 true negatives (TN) when con- and non-palpable findings) standards. sidering the sample of 664 mammographies in the categories Our positive predictive value 2 (recommendation for tissue 0, 4 and 5. diagnosis or PPV2) was 43%, as illustrated in Fig. 2. This is an Insights Imaging (2017) 8:581–588 585 Fig. 1 Percentages of abnormal 30.00% 25.00% 25.00% interpretation/recall rate 25.00% 20.00% 15.00% 12.26% 11.30% 9.60% 10.00% 5.00% 0.00% HPCA for both BCSC Benchmark BCSC for palpable BIRADS for non BIRADS for palpable and non for non palpable mammographic palpable palpable palpable mammographic findings mammographic mammographic mammographic findings findings findings findings Our Institution and the BCSC and BIRADS standard proposed values important subanalysis because it shows the accuracy radiolo- The negative predictive value (NPV) was estimated in gists expect for recommending a histopathological correlation 98.32% (95% CI 96.93%–99.09%). for a suspicious mammographic finding. Positive predictive value 3 (biopsy performed or PPV3) was 60%, as shown in Fig. 3. Discussion Usually the PPV3 is higher than the PPV2, since the biop- sies that will actually be done will probably show positive As a limitation of this study, the comparison between our rates results for cancer. Regarding this topic, our results showed and the BCSC rates regarding Bminimally invasive cancers or the same outcome. As BIRADS recommends, PPV3 is a more DCIS^ and Binvasive cancers with negative axillary lymph accurate indicator of cancer status, and PPV2 is more accurate nodes^ could not be done using proper additional statistical for interpretive performance [13]. analysis. The main reasons were limited access to BCSC data Our rate of negative axillary lymph nodes was and also because their sample size for the table presented on 75.93%, as shown in Fig. 4. page 591 of the last 5th edition of BIRADS (401,572 diag- The minimum cancer rate (invasive cancer with 1.0 cm or nostic mammograms between 1996 to 2005) had a distinct less or DCIS, ductal carcinoma in situ) was 36%, as illustrated difference from our sample size (9668 mammograms during in Fig. 5. 2013). This distinct difference between the sample sizes com- The sensibility of our diagnostic mammograms in pared could have led us to false results. However, if we 2013 was 90.22% (95% CI 82.24%–95.43%) and the consider a confidence interval of 95%, our sample size specificity was 92.31% (95% CI 89.91% to 94.36%). should be accurate to compare our proportions and The positive predictive value (PPV) was estimated in BCSC proportions, but we could not calculate a specific 65.35% (95% CI 58.49%–71.63%). p-value for these comparisons. 60.00% Fig. 2 PPV2 showing how 50.00% frequently the performed biopsies 50.00% 43.00% 43.00% indicate cancer 40.00% 40.00% 31.00% 30.00% 20.00% 10.00% 0.00% HPCA for both BCSC Benchmark for BCSC for palpable BIRADS for non BIRADS for palpable palpable and non non palpable mammographic palpable mammographic palpable mammographic findings mammographic findings mammographic findings findings findings Our Institution and the BCSC and BIRADS Standard proposed values Recall Rate 586 Insights Imaging (2017) 8:581–588 70.00% Fig. 3 PPV3 showing how 60.00% 55.00% frequently performed biopsies 60.00% 49.00% indicated cancers 45.00% 50.00% 36.00% 40.00% 30.00% 20.00% 10.00% 0.00% HPCA for both BCSC Benchmark BCSC for palpable BIRADS for non BIRADS for palpable and non for non palpable mammographic palpable palpable palpable mammographic findings mammographic mammographic mammographic findings findings findings findings Our Institution and the BCSC and BIRADS Standard proposed values We had a total of 58 losses, whose data records were not taken, allowing performing additional mammography exami- available in the hospital electronic system, mainly because nations when the patient is still inside the service. Other rea- these 58 patients (8.7% of our sample of 664 positive mam- sons are the experience and training of the radiologist's team. mography results) had private health care systems, which did Our positive predictive value 2 (recommendation for tissue not allow them to be classified into the hospital's public health diagnosis or PPV2) was 43%, which is exactly the same value care electronic system. as in BCSC for palpable findings (43%), above the value of Our cancer detection rate was 19.34%, close to the mini- BCSC Benchmarks for non-palpable findings (31%), and near mum 20% required for diagnostic mammograms according to the value requested by BIRADS for non-palpable findings BIRADS. This result is possibly related to a minor skewed (15–40%) and by BIRADS for palpable findings (25–50%). distribution of cases mainly due to the few screening mam- This PPV2 shows that our recommendations for biopsies are mograms in the total number of 9668 mammographies select- accurate. ed and considered as diagnostic in our study. Additionally, it Our positive predictive value 3 (biopsy performed or may be secondary to the fact that our SISMAMA nationalised PPV3) was 60%, which was above all the standard compari- software does not have a digital option to separate systematic sons: BCSC Benchmarks for non-palpable findings (36%), screening from diagnostic mammograms. This may result in BCSC for palpable findings (49%), BIRADS for non- some minor skewed distribution of the data when considering palpable findings (20–45%) and BIRADS for palpable find- the cancer detection rate, which would certainly not invalidate ings (30–55%). We believe the reason for this high PPV3 is our results. the fact that most of symptomatic patients were referred by The abnormal interpretation rate (or recall rate) was previous smaller health care centres to our hospital, which 12.26%, below the BCSC Benchmarks for diagnostic mam- centralise most of the patients who will need to be treated mograms (13.3%) and adequate for the value recommended for breast cancer using the national public health system. by BIRADS (8–25% for non-palpable findings or 10–25% for Moreover, we have patients who already had undergone some palpable findings), as shown in Fig. 1. There are many reasons previous screening before having a mammography in our hos- for this lower recall rate, such as the prompt review of the pital. Another reason to consider is that many patients are images by a breast radiologist immediately after they are already symptomatic at the time of the breast cancer diagnosis 75.93% Fig. 4 Percentages of invasive 80.00% 68.20% cancers with negative axillary 70.00% lymph nodes 56.50% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% HPCA for both palpable and non BCSC Benchmark for non BCSC for palpable palpable mammographic findings palpable mammographic findings mammographic findings Our Institution and the BCSC values Insights Imaging (2017) 8:581–588 587 45.00% Fig. 5 Percentages of minimum 39.00% 40.00% 36.00% invasive cancers (<1.0 cm) or DCIS 35.00% 30.00% 25.00% 20.00% 15.00% 15.00% 10.00% 5.00% 0.00% HPCA for both palpable and non BCSC Benchmark for non BCSC for palpable palpable mammographic findings palpable mammographic findings mammographic findings Our Institution and BCSC values in south Brazil, as well as in other low-income regions world- institutions worldwide, even with poorer general infrastruc- wide. This situation reflects the fact that some patients do not ture conditions when compared to the available facilities in have ideal social and economic possibilities for having a reference institutions in developed countries. mammography performed in an earlier stage of the disease. Our secondary goal in doing the study was to encourage Our rate of negative axillary lymph nodes was 75.93%, radiologists to be curious and try to look for their overall above the BCSC Benchmarks for non-palpable findings performance and self-evaluation (either individually or as a (68.2%) and BCSC in palpable findings (56.5%). This is an- team). As discussed in the literature, the feedback to the team other positive result, which may contribute to lower morbidity of radiologists as well as the moments of self-reflection and and mortality rates for the patients’ outcomes. insights in recall behaviour on mammograms is extremely The minimum cancer rate (invasive cancer with 1.0 cm or valuable. By doing this, other institutions also may improve less or DCIS, ductal carcinoma in situ) was 36%, near the their internal collaboration process towards meeting the pa- BCSC Benchmarks for non-palpable findings (39%) and tients’ needs. Additionally, we had more understanding about above the BCSC for palpable findings (15%). For the mini- how we can achieve higher scores and solve some problems mum cancer rate, BIRADS does not determine a fixed num- concerning data collection. ber, because it is implicit that the higher it is, the better it will Knowing the teamwork performance is essential to discov- be for both patients and professionals related to the treatment ering different means of improvement and to improving the of these patients. radiological medical practice. By pointing out the limitations The sensitivity of our diagnostic mammograms in 2013 and also our good results, the current study had a positive was 90.22%, above all the standard comparisons: the impact on our daily practice, especially regarding our focus BIRADS recommendation for non-palpable findings (80% on enhancing the breast cancer diagnosis detection when or higher), BIRADS for palpable findings (85% or higher), performing mammograms. In the future, we intend to imple- BCSC Benchmarks for non-palpable findings (83.1%) and ment arrangements for collecting data, analysing results and BCSC in palpable findings (87.8%). also concentrating more efforts on continuing medical educa- The specificity of our diagnostic mammograms in 2013 tion for all radiologists. was 92.31%, adequate for all the standard comparisons: BIRADS recommendations for non-palpable findings (80– 95%), BIRADS recommendations for palpable findings (83– 95%), BCSC Benchmarks for non-palpable findings (92.2%) Conclusion and BCSC for palpable findings (93.2%). As shown in the results, most of our performance rates This study is an internal audit in a regional reference institu- regarding mammography reports are similar to (sometimes tion for breast cancer diagnosis, treatment and follow-up in an better than) the values proposed by BIRADS as good radio- underdeveloped country. We performed a retrospective obser- logical practice. The sensitivity and specificity of our diagnos- vational and cross-sectional cohort study to analyse our radio- tic mammographies are adequate for all the standard compar- logical practice accuracy considering mammography reports. isons, moving the values of our institution to the international Despite possible limitations, our results showed high-quality aim of reducing the mortality and morbidity related to breast rates, demonstrating that standardised teamwork can promote cancer. the quality of public health systems worldwide. Additionally, Nevertheless, this study has a higher internal than external it may encourage radiologists to seek performance feedback validity mainly because it is a single-centre study. Despite this, and continuing medical education, thus resulting in better out- it can stimulate good radiological practice in public health comes for the patients. 588 Insights Imaging (2017) 8:581–588 Acknowledgements We specially thank the Hospital de Clínicas de versus experienced radiographers in the Dutch breast cancer screen- Porto Alegre (HCPA), Radiology Department, the medical physicists ing programme. Eur Radiol 25(11):3322–3327 who work with us and the physicist Rafael Zimmer from the 8. Vernacchia FS, Pena ZG (2009) Digital mammography: its impact Postgraduate and Research Department. on recall rates and cancer detection rates in a small community- We kindly thank Prof. Wolnei Caumo, MD, PhD, Prof. José Roberto based radiology practice. Am J Roentgenol 193(2):582–585 Goldim, PhD, Prof. Jorge Villanova Biazús, MD, PhD, and Prof. Andrea 9. Sickles EA (2007) Auditing your breast imaging practice: an Pires Souto Damin, MD, PhD, from the Federal University of Rio Grande evidence-based approach. Semin Roentgenol 42(4):211–217 do Sul (UFRGS), Prof. Maria Helena Pitombeira Rigatto, MD, PhD, from 10. Myers ER, Moorman P, Gierisch JM, Havrilesky LJ, Grimm LJ, PUCRS, and Prof. Felipe Fregni, MD, PhD, MPH, MEd, from the PPCR Ghate S et al (2015) Benefits and harms of breast cancer screening. Course offered by the Harvard T.H. Chan School of Public Health. JAMA 314(15):1615 11. Monticciolo DL, Newell MS, Hendrick RE, Helvie MA, Moy L, Monsees B et al (2017) Breast cancer screening for average-risk Compliance with ethical standards All procedures performed in this women: recommendations from the ACR Commission on breast study were in accordance to the ethical standards of the institutional and imaging. J Am Coll Radiol. https://doi.org/10.1016/j.jacr.2017.06. national research committee. 12. Duarte DL (2006) In: Guanabara Koogan (ed) A Mama Em Conflicts of interest The authors have no conflicts of interest to Imagens. Guanabara Koogan, Rio de Janeiro 327 p declare. 13. D’Orsi CJ, Sickles EA, Mendelson EB, Morris EA (2013) Breast imaging reporting and data system: ACR BI-RADS—breast imag- Open Access This article is distributed under the terms of the Creative ing atlas 5th edition, 5th edn. American College of Radiology Commons Attribution 4.0 International License (http:// (ACR), Reston creativecommons.org/licenses/by/4.0/), which permits unrestricted use, 14. Wu Y, Rubin DL, Woods RW, Elezaby M, Burnside ES (2014) distribution, and reproduction in any medium, provided you give appro- Developing a comprehensive database management system. priate credit to the original author(s) and the source, provide a link to the Cancer Informat 13:53–62 Creative Commons license, and indicate if changes were made. 15. Geller BM, Ichikawa L, Miglioretti DL, Eastman D (2012) Web- based mammography audit feedback. Am J Roentgenol 198(6): 562–567 16. Dee KE, Sickles EA (2001) Medical audit of diagnostic mammog- raphy examinations: comparison with screening outcomes obtained References concurrently. Am J Roentgenol 176:729–733 17. Tabar LVBCH et al (2000) The Swedish Two-County trial twenty years later: updated mortality results an new insights from long- 1. Geertse TD, Holland R, Timmers JMH, Paap E, Pijnappel RM, term follow-up. Radiol Clin N Am 38:625–651 Broeders MJM et al (2015) Value of audits in breast cancer screen- 18. Geertse TD, Holland R, Timmers JM, Paap E, Pijnappel RM, ing quality assurance programmes. Eur Radiol 25 (11):3338–3347 Broeders MJ et al (2015) Value of audits in breast cancer screening 2. Howard JH, Bland KI (2012) Current management and treatment quality assurance programmes. Eur Radiol 25(11):3338–3347 strategies for breast cancer. Curr Opin Obstet Gynecol 24(1):44–48 19. Nigenda G, Gonzalez-Robledo MC, Gonzalez-Robledo LM, 3. Kopans D (2008) In: Koogan G (ed) Diagnóstico Por Imagem da Bejarano-Arias RM (2016) Breast cancer policy in Latin Mama, 3 edição edn. Guanabara Koogan, Rio de Janeiro 1061 p America: account of achievements and challenges in five countries. 4. Lee CH, Dershaw DD, Kopans D, Evans P, Monsees B, Glob Health 12(1):39 Monticciolo D et al (2010) Breast cancer screening with imaging: 20. Pan HB, Wong KF, Yang TL, Hsu GC, Chou CP, Huang JS et al recommendations from the Society of Breast Imaging and the ACR (2014) The outcome of a quality-controlled mammography screen- on the use of mammography, breast MRI, breast ultrasound, and ing program: experience from a population-based study in Taiwan. other technologies for the detection of clinically occult breast can- J Chin Med Assoc 77(10):531–534 cer. J Am Coll Radiol 7(1):18–27 21. Jackson SL, Cook AJ, Miglioretti DL, Carney PA, Geller BM, 5. Oeffinger KC, Fontham ETH, Etzioni R, Herzig A, Michaelson JS, Onega T et al (2012) Are radiologists' goals for mammography Shih Y-CT et al (2015) Breast cancer screening for women at aver- accuracy consistent with published recommendations? Acad age risk. JAMA 314(15):1599 Radiol 19(3):289–295 6. Elmore JG et al (2010) Radiologists attitudes and use of mammog- 22. Passman L, Ramalho Ortigão Farias A, Tomazelli J, Franco de raphy audit reports. Acad Radiol 17(6):752–760 Abreu D, Dias M, de Assis M et al (2011) SISMAMA- 7. van Landsveld-Verhoeven C, den Heeten GJ, Timmers J, Broeders implementation of an information system for breast cancer early MJM (2015) Mammographic positioning quality of newly trained detection programs in Brazil. Breast 20:35–39

Journal

Insights into ImagingSpringer Journals

Published: Oct 4, 2017

There are no references for this article.