TITLE: Digital Tomosynthesis for the Screening and Diagnosis of Breast Cancer: A Review of the Diagnostic Accuracy

Transcription

1 TITLE: Digital Tomosynthesis for the Screening and Diagnosis of Breast Cancer: A Review of the Diagnostic Accuracy DATE: 26 September 2013 CONTEXT AND POLICY ISSUES Breast cancer is the second leading cause of cancer death in women. 1 One in eight to 10 women are likely to develop breast cancer during their lifetime. 2 Breast cancer occurrence is estimated as 18% in women under the age 50 years, 52% in women 50 to 69 years, and 30% over the age of 69 years. 1 In Canada, it is estimated that in 2013 there will be 91,400 new cancer cases in women and 26% of these will be breast cancer. 1 The number of cancer deaths in women in 2013 is estimated to be 36,100 and 13.9% of these are expected to be due to breast cancer. 1 Early detection and treatment of breast cancer are expected to reduce morbidity and mortality due to the disease 3 and breast cancer screening and diagnosis play an important role. Mammography has been used for routine breast cancer screening as well as for diagnostic purposes. Initially the screen-film mammography method was developed and was later followed by the digital mammography. Mammography is a two dimensional (2D) X-ray imaging technique. As this involves 2D projection, overlapping tissues and normal dense structures in the breast may hamper interpretation. 4,5 Dense breast tissues may reduce or obscure conspicuity of abnormalities and could result in cancer being missed. 6,7 Also overlapping dense tissue may give the impression of a lesion and could result in unnecessary workup or biopsy. 6 Detection with a three dimensional (3D) technique may reduce the interference from overlapping tissues and dense tissues and may improve interpretations. 4 The digital tomosynthesis method is a 3D imaging technique in which the X ray tube moves along a limited angle arc and a finite number of 2D projections are collected. From these 2D projections a 3D volume of the breast is reconstructed. 8 This allows cross-sectional visualization of the breast tissues and reduces hindrance due to overlapping or superimposed tissues and has the potential to improve interpretation. 9 Introduction of tomosynthesis may have resource implications and if used as an adjunct with mammography could entail additional cost, hence the extent of improvement if any with the use of tomosynthesis needs to be assessed. Disclaimer: The Rapid Response Service is an information service for those involved in planning and providing health care in Canada. Rapid responses are based on a limited literature search and are not comprehensive, systematic reviews. The intent is to provide a list of sources of the best evidence on the topic that CADTH could identify using all reasonable efforts within the time allowed. Rapid responses should be considered along with other types of information and health care considerations. The information included in this response is not intended to replace professional medical advice, nor should it be construed as a recommendation for or against the use of a particular health technology. Readers are also cautioned that a lack of good quality evidence does not necessarily mean a lack of effectiveness particularly in the case of new and emerging health technologies, for which little information can be found, but which may in future prove to be effective. While CADTH has taken care in the preparation of the report to ensure that its contents are accurate, complete and up to date, CADTH does not make any guarantee to that effect. CADTH is not liable for any loss or damages resulting from use of the information in the report. Copyright: This report contains CADTH copyright material and may contain material in which a third party owns copyright. This report may be used for the purposes of research or private study only. It may not be copied, posted on a web site, redistributed by or stored on an electronic system without the prior written permission of CADTH or applicable copyright owner. Links: This report may contain links to other information available on the websites of third parties on the Internet. CADTH does not have control over the content of such sites. Use of third party sites is governed by the owners own terms and conditions.

2 The purpose of this report is to provide evidence on the clinical effectiveness of tomosynthesis used alone or as an adjunct with mammography compared to mammography alone for screening and diagnosis of breast cancer. RESEARCH QUESTIONS 1. What is the clinical effectiveness of digital tomosynthesis compared with mammography for breast cancer screening? 2. What is the clinical effectiveness of digital tomosynthesis as an adjunct to mammography compared with mammography alone for breast cancer screening? 3. What is the clinical effectiveness of digital tomosynthesis compared with mammography for breast cancer diagnosis? 4. What is the clinical effectiveness of digital tomosynthesis as an adjunct to mammography compared with mammography alone for breast cancer diagnosis? KEY FINDINGS No relevant studies were identified on the clinical effectiveness of tomosynthesis alone compared with mammography alone for breast cancer screening. In comparison to mammography alone, the combination of tomosynthesis and mammography appeared to have higher detection and lower recall rates for breast cancer screening, however statistical significance of the difference was not always clear. Limited evidence suggested that the greatest reduction in recall rates were for those below 50 years of age and for those with dense breast tissues. For diagnostic purposes, assessments with tomosynthesis alone or combination of tomosynthesis and mammography were no worse than mammography alone but superiority could not be established. METHODS Literature Search Strategy A limited literature search was conducted on key resources including PubMed, The Cochrane Library (2010, Issue 7), University of York Centre for Reviews and Dissemination (CRD) databases, ECRI (Health Devices Gold), Canadian and major international health technology agencies, as well as a focused Internet search. Methodological filters were applied to limit retrieval to health technology assessments, systematic reviews, meta-analyses, randomized controlled trials and non-randomized studies. Where possible, retrieval was limited to the human population. The search was also limited to English language documents published between January 1, 2008 and August 20, Digital Tomosynthesis for Breast cancer Screening and Diagnosis 2

3 Selection Criteria and Methods One reviewer screened the titles and abstracts of the retrieved publications, selected potentially relevant articles for retrieval of full-text publications for further investigation and evaluated the full-text publications for final selection, according to the criteria listed in Table 1. Table 1: Selection Criteria Population Adult women, subgroups: Adult women ages 40 to 49 Adult women ages 50 to 74 Adult women with low breast density Adult women with high breast density Intervention Comparator Outcomes Study Designs Digital tomosynthesis with or without mammography Mammography alone Accuracy, sensitivity, specificity, detection rates Health technology assessment (HTA), systematic review (SR) and meta-analysis (MA), randomized controlled trial (RCT), and nonrandomized study Exclusion Criteria Studies were excluded if they did not satisfy the selection criteria in Table 1, or if they were published prior to Studies focusing on detection of calcification and phantom studies were excluded. Studies that were deemed to have incomplete reporting of outcomes, such as not reporting numerical values for outcomes, were excluded. Critical Appraisal of Individual Studies Critical appraisal of a study was conducted based on an assessment tool appropriate for the particular study design. The AMSTAR checklist 10 was used for systematic reviews and the QUADAS2 checklist 11 was used for diagnostic and screening studies. For the critical appraisal, a numeric score was not calculated. Instead, the strength and limitations of the study were described narratively. SUMMARY OF EVIDENCE Quantity of Research Available The literature search yielded 219 citations. Upon screening titles and abstracts, 178 articles were excluded and 41 potentially relevant articles were selected for full-text review. No potentially relevant article was identified from the grey literature. Of these 41 articles, 29 did not satisfy the inclusion criteria and were excluded. The 12 included articles comprised of one systematic review 12 and eleven non-randomized studies. 2,3,7,13-20 No health technology Digital Tomosynthesis for Breast cancer Screening and Diagnosis 3

4 assessment or randomized controlled trials were identified. Details of the study selection process are outlined in Appendix 1. Summary of Study Characteristics Characteristics of the included systematic review and non-randomized clinical studies are summarized below and details are provided in Appendix 2. Systematic review One systematic review 12 comparing tomosynthesis used alone or in combination with mammography with mammography alone was included. It was published in 2013 from Australia. It included 14 non-randomized diagnostic studies. Of the 14 studies, seven studies compared tomosynthesis with mammography, three studies compared combination of tomosynthesis and mammography with mammography alone, two studies compared tomography versus mammography versus combination of tomosynthesis and mammography, one study examined the combination of tomosynthesis and synthetically generated 2D images with the combination of tomosynthesis and mammography, and one study examined the combination of tomosynthesis and mammography and later follow-up with tomosynthesis. A conventional 2D mammography image may be simulated by generating a synthetic 2D image from each tomosynthesis slice. The studies had numbers of participants ranging between 50 and 738 with most studies having between 100 and 150 participants. Participant accrual periods were available in nine studies and were between 2006 and Mean ages of participants were available in eight studies and ranged between 51 and 57 years. Results of the individual studies were summarized separately. Outcomes reported included accuracy, sensitivity, and specificity. Clinical studies Screening studies Four studies 3,7,13,14 on screening were included. All four studies were published in 2013; two 3,13 from USA, one 7 from Italy and one 14 from Norway. All four studies compared the combination of tomosynthesis and mammography with mammography alone. Two studies 7,14 were prospective and two studies 3,13 were retrospective. The number of participants in the studies ranged between 7,292 and 23,335. All four studies reported on cancer detection rates and recall rates and one study 3 in addition reported on positive predictive value (PPV). Recall rates refer to persons recalled for further investigation. All studies reported subgroup data. Three studies 3,7,13 presented recall rates and detection rates stratified by age and breast density and one study 14 reported number of cancers detected stratified by breast density. Diagnostic studies Seven studies 2,15-20 on diagnostic imaging were included. Five studies 2,15,17,19,20 were observer performance studies where multiple readers interpreted the images. In one study 16 the assessment was based on one reader s interpretation and in one study 18 two readers interpreted the images and a third reader was involved when the two readers could not reach consensus. Four studies were published in 2013; one study each from France, 17 Italy, 19 Switzerland, 18 and the USA. 20 Two studies were published in 2012; one study each from Australia 15 and Sweden. 2 One study was published in 2010 from Sweden. 16 Two studies 2,16 compared tomosynthesis with mammography, three studies 15,19,20 compared the combination of Digital Tomosynthesis for Breast cancer Screening and Diagnosis 4

5 mammography and tomosynthesis with mammography alone and two studies 17,18 compared three modalities: combination of mammography and tomosynthesis, mammography alone and tomosynthesis alone. The number of participants in the studies ranged between 62 and 622. Four studies 2,17,19,20 reported on accuracy determined by the receiver operating characteristic (ROC) analysis. Five studies 2,17-20 reported on sensitivity and specificity. Three studies 17,18,20 reported on PPV and negative predictive value (NPV). Two studies 15,20 reported on recall rates and two studies 15,16 reported on detection rates. Two studies 16,18 reported subgroup data. One study 18 reported data stratified by breast density for sensitivity, specificity, PPV and NPV. One study 16 reported data stratified by breast density for number of cancers detected. Summary of Critical Appraisal Systematic review One systematic review 12 was included. The objectives, and the inclusion and exclusion criteria were clearly stated. The authors mentioned that a systematic review of the literature was conducted and provided the search dates, however a search strategy was not provided and a single database was searched. One reviewer selected articles for inclusion and the article selection process was not described. Characteristics of the individual studies were described. It was not mentioned if data extraction was done in duplicate, if quality assessment of studies was conducted, or if publication bias was explored. Conflict of interest was declared and it was stated that both authors received support from the manufacturer but that the support had no role in the conceptual or reporting aspects of the review. Clinical studies Screening studies Four studies 3,7,13,14 on screening were included. Two studies 7,14 were prospective and two studies were retrospective. 3,13 In all studies, it was not mentioned if a consecutive sequence or random selection of participants were considered and it was unclear if the participants were representative of participants generally seen in clinical practices. All participants appeared to be included in the analyses in three studies 3,7,14 and it was unclear in one study as it did not explicitly state the number of patients when reporting results. 13 Image sets from the two modalities were interpreted independently in three studies 3,13,14 and not independently in one study. 7 Diagnostic studies Seven studies 2,15-20 on diagnostic imaging were included. Consecutive patients were considered in two studies, 17,18 four studies 2,16,19,20 did not mention if patients considered were consecutive or a random selection and in one study 15 images were randomly selected from an image archive. In all studies, the image sets obtained using tomosynthesis, mammography or their combination were interpreted independently. Readers were blinded to the truth status (i.e. confirmed results using various other methods such as ultrasonography and biopsy) in five studies. 2,17-20 In two studies it was not explicitly mentioned if the readers were blinded to the truth status. 15,16 In the studies, the sample sets used were enriched with disease cases and may not be representative of that typically seen in clinical practices. Strengths and limitations of individual studies are provided in Appendix 3. Digital Tomosynthesis for Breast cancer Screening and Diagnosis 5

6 Summary of Findings The overall findings from the systematic review and individual clinical studies are summarized below and details are available in Appendix 4. What is the clinical effectiveness of digital tomosynthesis compared with mammography for breast cancer screening? No relevant studies were identified on the clinical effectiveness of tomosynthesis compared with mammography alone for breast cancer screening. What is the clinical effectiveness of digital tomosynthesis as an adjunct to mammography compared with mammography alone for breast cancer screening? Four relevant studies 3,7,13,14 on the clinical effectiveness of the combination of tomosynthesis and mammography compared with mammography alone for breast cancer screening, were identified. Four studies 3,7,13,14 showed that the cancer detection rates were higher with the combination of tomosynthesis and mammography compared with mammography alone and of the these four studies, one study 7 reported the difference to be statistically significant and the others did not report on statistical significance. Two studies 3,13 showed that the recall rates were lower with the combination of tomosynthesis and mammography compared with mammography alone and of the these two studies, one study 13 reported the difference to be statistically significant and the other did not report on statistical significance. One study 7 showed that the false positive recall rate was lower with the combination of tomosynthesis and mammography compared with mammography alone but did not report on statistical significance. One study 3 showed that PPV was higher with the combination of tomosynthesis and mammography compared with mammography alone, however statistical significance of the difference was not reported. Subgroup data stratified according to age and breast density were reported in three studies 3,7,13 and subgroup data stratified by breast density were reported in one study. 14 Stratification by age In comparison to mammography alone, the combination of tomosynthesis and mammography showed statistically significantly higher detection rates in age groups of <60 years and 60 years, in one study, 7 but whether the effect was statistically significantly higher in one group compared to another was not reported. In comparison to mammography alone, the combination of tomosynthesis and mammography showed statistically significantly lower recall rates in age groups of <40 years, 40 to 49 years, 50 to 59 years, and 60 to 69 years, and no statistically significant difference in recall rates in age group >70 years, in one study. 13 In comparison to mammography alone, the combination of tomosynthesis and mammography showed lower recall rates in age groups of <60 years and 60 years, in one study 7 and in age groups of <50 years, 50 to 64 years and > 64 years in one study 3 however, statistical significance of the differences were not reported and whether the effect was statistically significantly higher in one group compared to another was also not Digital Tomosynthesis for Breast cancer Screening and Diagnosis 6

7 reported. In one study 13 the greatest reduction in recall rates was for those of age less than 50 years. Stratification by breast density In comparison to mammography alone, the combination of tomosynthesis and mammography showed higher detection rates in one study, 7 the difference being statistically significant for the low breast density group and not statistically significant for the high breast density group. In one study, 14 in comparison to mammography alone, the combination of tomosynthesis and mammography showed that the numbers of cancers detected in the patient group with invasive cancers were higher for the different breast density sub-groups, however the statistical significance of the differences were not reported. In one study, 13 in comparison to mammography alone, the combination of tomosynthesis and mammography showed statistically significant lower recall rates in cases of breasts with scattered fibroglandular density, heterogeneously dense breasts, and extremely dense breasts and no statistically significant difference in cases of breasts of low density. In comparison to mammography alone, the combination of tomosynthesis and mammography showed lower recall rates in both low and high breast density groups in two studies 3,7 but the statistical significance of the differences were not reported. In one study 13 the greatest reduction in recall rates was for those with dense breasts. What is the clinical effectiveness of digital tomosynthesis compared with mammography for breast cancer diagnosis? One relevant systematic review 12 was identified. In this systematic review, the individual studies were summarized separately and no pooling was undertaken. Four studies included in the systematic review showed that tomosynthesis (one view) appears to be comparable with mammography (two view) based on sensitivity, specificity or accuracy. Images may be acquired using different views such as craniocaudal, mediolateral, and mediolateral-oblique and the terminology one view or two view is used to indicate whether a single view or a dual view was obtained. Three studies included in the systematic review showed that tomosynthesis (two view) appeared to have at least equal or better accuracy than mammography (two view). Five studies included in the systematic review showed that cancers were equally conspicuous or more conspicuous with tomosynthesis compared to mammography when qualitatively examined or quantitatively scored. Four relevant studies 2,16-18 on the clinical effectiveness of tomosynthesis compared with mammography alone for breast cancer diagnosis were identified. Of these four studies, two studies 17,18 also compared the combination of tomosynthesis and mammography with mammography alone and results for this particular comparison were included in the summary of findings section for question 4 (below). Accuracy was determined by ROC analyses in two studies. 2,17 In comparison to mammography (two view), accuracy obtained with tomosynthesis (one view) appeared to be numerically higher in the two studies. 2,17 In comparison to mammography (two view), sensitivity obtained with tomosynthesis (one view) appeared to be numerically higher in two studies 2,18 and numerically lower in one study. 17 In comparison to mammography (two view), specificity obtained with tomosynthesis (one view) appeared to be numerically higher in two studies 17,18 and lower in one study. 2 In comparison to mammography (two view), PPV with tomosynthesis (one view) appeared to be numerically higher in two Digital Tomosynthesis for Breast cancer Screening and Diagnosis 7

8 studies. 17,18 In comparison to mammography (two view), NPV with tomosynthesis (one view) appeared to be numerically higher in one study 18 and lower in one study. 17 In most instances statistical significance or non-significance of the difference was not reported. One study 16 showed that the detection rate was statistically significantly higher with tomosynthesis compared with mammography. The results are summarized in Table 2. and details are available in Appendix 4. Table 2: Diagnostic studies comparing tomosynthesis with mammography Study Accuracy (ROC analysis) Sensitivity Specificity PPV NPV Cancer detection rate Thibault, NR Waldherr, NR NR Svahn, (S) (S) (NS) NR NR NR Fornvik, NR NR NR NR NR (S) NR = not reported, NS = not statistically significant, S = statistically significant, indicates numerical increase and indicates numerical decrease with tomosynthesis compared to mammography. however these need to be interpreted with caution as differences were small in some cases. The actual values for the two modalities are available in Appendix 4. Statistical significance if reported is indicated in parenthesis. Subgroup data stratified according to breast density were reported in two studies. 16,18 In one study 18 for both high and low breast density, the sensitivity, specificity, PPV and NPV appeared to be numerically higher with tomosynthesis compared to mammography, with the differences being greater for sensitivity and NPV and lesser for specificity and PPV. In one study 16 for intermediate and dense parenchyma, the number of tumors detected was statistically significantly higher with tomosynthesis compared with mammography and the difference in detection between the two modalities was not statistically significant for fatty parenchyma. What is the clinical effectiveness of digital tomosynthesis as an adjunct to mammography compared with mammography alone for breast cancer diagnosis? One relevant systematic review 12 was identified. In this systematic review, the individual studies were summarized separately and no pooling was undertaken. Six studies included in the systematic review showed that addition of tomosynthesis to standard mammography (two view) appeared to increase accuracy and this was partly or mostly due to reduction in false positive interpretations. Five relevant studies 15,17-20 on the clinical effectiveness of the combination of tomosynthesis and mammography compared with mammography alone for breast cancer diagnosis were identified. Of these five studies, two studies 17,18 also compared tomosynthesis alone with mammography alone and results for this particular comparison were included in the summary of findings section for question 3 (above). In comparison to mammography alone, accuracy obtained with combination of tomosynthesis and mammography appeared to be numerically higher in two studies 17,19 and statistically significantly higher in one study. 20 In comparison to mammography alone, sensitivity obtained with combination of tomosynthesis and mammography appeared to be numerically higher in three studies and numerically lower in one study. 17 In comparison to mammography alone, specificity obtained with the combination of tomosynthesis and mammography appeared to be numerically higher in three studies 17,19,20 and numerically lower Digital Tomosynthesis for Breast cancer Screening and Diagnosis 8

9 in one study. 18 In comparison to mammography alone, PPV with the combination of tomosynthesis and mammography appeared to be numerically higher in three studies. 17,18,20 In comparison to mammography alone, NPV with the combination of tomosynthesis and mammography appeared to be numerically higher in two studies 18,20 and numerically lower in one study. 17 In comparison to mammography alone, the combination of tomosynthesis and mammography showed higher detection rate in one study 15 and reduced false recall rates in two studies. 15,20 The results are summarized in Table 3 and details are available in Appendix 4. Table 3: Diagnostic studies comparing combination of tomosynthesis and mammography with mammography alone Study Accuracy (ROC analysis Sensitivity Specificity PPV NPV Cancer detection rate Recall rate Gennaro, NR NR NR NR Rafferty, ; Study 1 Rafferty, ; Study 2 (S) NR (S) NR Thibault, NR NR Waldherr, NR NR NR Bernardi, NR NR NR NR NR NR = not reported, NS = not statistically significant, S = statistically significant, indicates numerical increase and indicates numerical decrease with the combination of tomosynthesis and mammography compared with mammography alone, however these need to be interpreted with caution as differences were small in some cases. The actual values obtained with the two modalities are available in Appendix 4. Statistical significance if reported is indicated in parenthesis. Limitations Not all studies reported all outcomes. For the diagnostic studies there were some inconsistencies in the findings between studies, hence results need to be interpreted with caution. P-values and 95% confidence intervals for the difference between modalities were not always reported, hence statistical significance of the differences could not be ascertained. Estimates of sensitivity and specificity varied. This could be due to differences in research methods used and differences between readers. Results may be impacted by extent of experience with breast imaging and experience with tomosynthesis. There were inconsistencies in the reporting of breast density categories. The terminology used varied among the studies. In the diagnostic studies, the case sets used were enriched with disease cases and is unlikely to be representative of the case mix typically encountered in clinical practice. None of the studies were conducted in Canada hence it is unclear if the findings are applicable to the Canadian setting. CONCLUSIONS AND IMPLICATIONS FOR DECISION OR POLICY MAKING No relevant studies were identified on the clinical effectiveness of tomosynthesis compared with mammography alone for breast cancer screening. In comparison to mammography alone, the combination of tomosynthesis and mammography appeared to have higher cancer detection Digital Tomosynthesis for Breast cancer Screening and Diagnosis 9

10 rates and lower recall rates for breast cancer screening. However, it should be noted that it was not always clear if the differences were statistically significant. Limited evidence showed that the reduction in recall rates were greatest for those of age less than 50 years and those with dense breast tissues. Ongoing screening studies such as TOMMY 21 in the UK may provide further insights. For diagnostic purposes, assessments with tomosynthesis or combination of tomosynthesis and mammography were no worse than mammography alone but superiority could not be established. Some authors mentioned that more prospective studies are required to determine the role of tomosynthesis for application in diagnostics. When considering the incorporation of tomosynthesis in clinical practice, various factors such as acquisition time, operation and training would also need to be taken into account. PREPARED BY: Canadian Agency for Drugs and Technologies in Health Tel: Digital Tomosynthesis for Breast cancer Screening and Diagnosis 10

11 REFERENCES 1. Canadian cancer statistics 2013: special topic: liver cancer [Internet]. Toronto: Canadian Cancer Society; [cited 2013 Sep 12]. Available from: adian%20cancer%20statistics/canadian-cancer-statistics-2013-en.pdf 2. Svahn TM, Chakraborty DP, Ikeda D, Zackrisson S, Do Y, Mattsson S, et al. Breast tomosynthesis and digital mammography: a comparison of diagnostic accuracy. Br J Radiol Nov;85(1019):e1074-e Rose SL, Tidwell AL, Bujnoch LJ, Kushwaha AC, Nordmann AS, Sexton R Jr. Implementation of breast tomosynthesis in a routine screening practice: an observational study. AJR Am J Roentgenol Jun;200(6): Timberg P, Bath M, Andersson I, Mattsson S, Tingberg A, Ruschin M. In-plane visibility of lesions using breast tomosynthesis and digital mammography. Med Phys Nov;37(11): Svahn T, Andersson I, Chakraborty D, Svensson S, Ikeda D, Fornvik D, et al. The diagnostic accuracy of dual-view digital mammography, single-view breast tomosynthesis and a dual-view combination of breast tomosynthesis and digital mammography in a free-response observer performance study. Radiat Prot Dosimetry [Internet] Apr [cited 2013 Aug 27];139(1-3): Available from: 6. Chan HP, Wei J, Zhang Y, Helvie MA, Moore RH, Sahiner B, et al. Computer-aided detection of masses in digital tomosynthesis mammography: comparison of three approaches. Med Phys [Internet] Sep [cited 2013 Aug 27];35(9): Available from: 7. Ciatto S, Houssami N, Bernardi D, Caumo F, Pellegrini M, Brunelli S, et al. Integration of 3D digital mammography with tomosynthesis for population breast-cancer screening (STORM): a prospective comparison study. Lancet Oncol Jun;14(7): Gennaro G, Toledano A, di Maggio C, Baldan E, Bezzon E, La Grassa M, et al. Digital breast tomosynthesis versus digital mammography: a clinical performance study. Eur Radiol Jul;20(7): Gur D, Abrams GS, Chough DM, Ganott MA, Hakim CM, Perrin RL, et al. Digital breast tomosynthesis: observer performance study. AJR Am J Roentgenol Aug;193(2): Shea BJ, Grimshaw JM, Wells GA, Boers M, Andersson N, Hamel C, et al. Development of AMSTAR: a measurement tool to assess the methodological quality of systematic reviews. BMC Med Res Methodol [Internet] [cited 2013 Sep 6];7:10. Available from: Methodology checklist 5: studies of diagnostic accuracy. Edinburgh: Scottish Intercollegiate Guidelines Network (SIGN); 2006 Jan. Digital Tomosynthesis for Breast cancer Screening and Diagnosis 11

12 12. Houssami N, Skaane P. Overview of the evidence on digital breast tomosynthesis in breast cancer detection. Breast Apr;22(2): Haas BM, Kalra V, Geisel J, Raghu M, Durand M, Philpotts LE. Comparison of tomosynthesis plus digital mammography and digital mammography alone for breast cancer screening. Radiology Jul 30. Epub ahead of print. 14. Skaane P, Bandos AI, Gullien R, Eben EB, Ekseth U, Haakenaasen U, et al. Comparison of digital mammography alone and digital mammography plus tomosynthesis in a population-based screening program. Radiology Apr;267(1): Bernardi D, Ciatto S, Pellegrini M, Anesi V, Burlon S, Cauli E, et al. Application of breast tomosynthesis in screening: incremental effect on mammography acquisition and reading time. Br J Radiol Dec;85(1020):e1174-e Fornvik D, Zackrisson S, Ljungberg O, Svahn T, Timberg P, Tingberg A, et al. Breast tomosynthesis: accuracy of tumor measurement compared with digital mammography and ultrasonography. Acta Radiol Apr;51(3): Thibault F, Dromain C, Breucq C, Balleyguier CS, Malhaire C, Steyaert L, et al. Digital breast tomosynthesis versus mammography and breast ultrasound: a multireader performance study. Eur Radiol Sep;23(9): Waldherr C, Cerny P, Altermatt HJ, Berclaz G, Ciriolo M, Buser K, et al. Value of oneview breast tomosynthesis versus two-view mammography in diagnostic workup of women with clinical signs and symptoms and in women recalled from screening. AJR Am J Roentgenol Jan;200(1): Gennaro G, Hendrick RE, Ruppel P, Chersevani R, di Maggio C, La Grassa M, et al. Performance comparison of single-view digital breast tomosynthesis plus single-view digital mammography with two-view digital mammography. Eur Radiol Mar;23(3): Rafferty EA, Park JM, Philpotts LE, Poplack SP, Sumkin JH, Halpern EF, et al. Assessing radiologist performance using combined digital mammography and breast tomosynthesis compared with digital mammography alone: results of a multicenter, multireader trial. Radiology Jan;266(1): Health Technology Assessment Programme [Internet]. Southampton (UK): National Institute for Health Research. Details of HTA project in progress: TOMMY trial: a comparison of TOMosynthesis with digital MammographY in the UK NHS breast screening programme; 2013 Sep 18 [cited 2013 Sep 19]. Available from: Digital Tomosynthesis for Breast cancer Screening and Diagnosis 12

13 ABBREVIATIONS ACR AUC BI-RADS CI FN FP FPF FROC JAFROC M NPV NR NS PPV ROC SD SR T TN TP TPF American College of Radiology area under the curve Breast Imaging-Reporting and Data System confidence interval false negative false positive false positive fraction free response receiver operating characteristic Jackknife alternative free response receiver operating characteristic mammography negative predictive value not reported not significant positive predictive value receiver operating characteristic standard deviation systematic review tomosynthesis true negative true positive true positive fraction Digital Tomosynthesis for Breast cancer Screening and Diagnosis 13

14 APPENDIX 1: Selection of Included Studies 219 citations identified from electronic literature search and screened 178 citations excluded 41 potentially relevant articles retrieved for scrutiny (full text, if available) No potentially relevant report retrieved from other sources (grey literature, hand search) 41 potentially relevant reports 29 reports excluded: -irrelevant outcome (10) - irrelevant comparison (1) -irrelevant design (2) -irelevant intervention (1) -already included in the selected systematic reviews (13) -other (review articles, editorials)(2) 12 reports included in review Digital Tomosynthesis for Breast cancer Screening and Diagnosis 14

15 APPENDIX 2: Characteristics of Included Studies First Author, Year, Country Systematic reviews Houssami, , Australia Study Design, Duration SR with 12 studies comprising of 7 on T vs M; 3 on (M+T) vs M; 2 on T vs M vs (M+T); in addition there were 2 other studies:1 on (M syn +T) vs (M actual +T); 1 on (M+T) followed later by T) Patient Characteristics, Sample Size (N) Recalled cases, cases with symptoms or cases used as a test set by radiologists. Age (year): means 51 to 57 for 8 studies and NR for 6 studies. N: 50 to 738 (with most studies between 100 and 150) Interventio n Compara tors Outcomes Measured T or (M+T) M Accuracy, sensitivity, specificity Subject accrual periods between 2006 and 2011 for 9 studies; NR for 5 Clinical Studies - Screening Ciatto, , Italy (STORM) Prospective population screening study (at 2 screening services). Asymptomatic women Age (years) (median [interquartile range]: 58 [54 63] (M+T) i.e (2D+3D) M i.e. (2D) Detection rate, recall rate Device: A Selenia Dimensions unit with integrated 2D and 3D mammograph y done in the COMBO mode N = 7292 (7294 screens analyzed as two subjects had bilateral cancer [ detected with different screenreading techniques]) Subject accrual period: Aug 2011 to Digital Tomosynthesis for Breast cancer Screening and Diagnosis 15

16 First Author, Year, Country Study Design, Duration June 2012 Patient Characteristics, Sample Size (N) Interventio n Compara tors Outcomes Measured Haas, , USA Retrospective study ( subjects screened at 4 sites) All sites used digital mammograph y (Selenia). The tertiary care hospital site also had digital tomosynthesis (Dimensions) Women undergoing screening mammography Age(years): most between 40 and 69; <40 y (n= 307), y (n=3748), y (n=4341), y (n=3091), 70 y (n=1671) N = 13,158 (6100 in [M+T] group and 7058 in M group) (M+T) M Detection rate, recall rate Rose, , USA Subject accrual period: 1 October 2011 to 30 September 2012 Retrospective study using database information of a multisite communitybased comprehensiv e breast centre (single practice). Women undergoing screening mammography Age (years) 18 N= 23,355 (9499 in [M+T] group and13,856 in M group) (M+T) M Detection rate, recall rate Device; NR Mammograph y screening results interpreted in 2010 before introduction of tomosynthesis was compared to Digital Tomosynthesis for Breast cancer Screening and Diagnosis 16

17 First Author, Year, Country Skaane, , Norway Study Design, Duration results after introduction of tomosynthesis during May 2011 to January Prospective screening study ( single institution) Oslo Tomosynthesi s Screening Trial Patient Characteristics, Sample Size (N) Women undergoing screening and consenting to participate in the study Age range (years): 50 to 69 Interventio n Compara tors Outcomes Measured (M+T) M Detection rate, false positive rate Device: Digital mammograph y and tomosynthesis using a commercially available system (Dimensions) N = 12,621 Pre-planned interim analysis of 2 arms of a 4- arm study with subjects participating from 22 November 2010 to 31 December, 2011 Clinical Studies Diagnosis (T vs M) Svahn, , Sweden Observer performance study. Device: M was performed using a Mammomat Novation DR unit and T was Symptomatic as well as asymptomatic subjects. Age (years) (mean [range]): 60 (42 to 79) N = 185 (89 with cancer) T(1 view); For T[1 view] approx. twice the ma needed for a single M view was used resulting in same absorbed M (2 view) JAFROC, ROC AUC, TPF, FPF Digital Tomosynthesis for Breast cancer Screening and Diagnosis 17

18 First Author, Year, Country Study Design, Duration performed using a prototype device based on the same type of unit. Patient Characteristics, Sample Size (N) Interventio n dose as M (2view) Compara tors Outcomes Measured Fornvik, , Sweden Subject accrual period: 19 June 2006 to 21 May 2008 Comparative study. Device: M was performed using a Mammomat Novation unit and T was performed using a prototype unit adapted from the Mammomat Novation. Accrual period: June 2006 to December 2007 Women with findings suspicious of malignancy recalled from screening and women with clinical symptoms with subtle or negative findings of malignancy with M. Age (years) (mean [range]): 60 [42 to 79] N= 62 (with 73 malignant breast tumors) T M (1 view) or M (2 view) Measurable tumor, tumor size, tumor staging, Clinical Studies Diagnosis (T vs M vs [M+T]) Thibault, ,France Observe performance study Device: M was performed using a Senographe 2000D or a Senograph DS unit. T was performed Women with symptoms or referred to the clinic due to screening recalls with unresolved mammographic or ultrasound work-up. Age (years): 40 N = 131 (55 T (1 view) or (M+T) M (2 view) ROC AUC, sensitivity, specificity, PPV, NPV Digital Tomosynthesis for Breast cancer Screening and Diagnosis 18

19 First Author, Year, Country Study Design, Duration using a prototype unit adapted from the Senograph DS platform. Patient Characteristics, Sample Size (N) malignant and 76 benign or normal) Interventio n Compara tors Outcomes Measured Waldherr, , Switzerland Subject accrual period: NR Comparative study Device: NR Subject accrual period: December 2008 to June 2009 Women with signs and symptoms or recalled from screening for diagnostic work-up and had an abnormal two view mammography. Age: NR T (1 view) or (M+T) M (2 view) Sensitivity, specificity, PPV, NPV, TP, TN, FP, FN Clinical Studies Diagnosis ( M vs [M+T]) Gennaro, , Italy Observer performance study Device: Standard examination with a GE Senographe 2000D mammograph y unit and tomosynthesis (1 view) with a prototype system developed by GE Healthcare Subject accrual period: April 2007 to N = 144 ( 78 with cancer, 66 recalled from screening and had abnormal mammogram) Women with at least one breast lesion identified by mammography and/or ultrasound and considered to be doubtful, suspicious or probably malignant. Age (years): 40 N = 235 ( 469 breasts which included 68 with cancer, 200 with benign lesions and 201 with no lesion [normal]) M(1 view) + T (1 view) M ( 2 view) ROC AUC, sensitivity, specificity Digital Tomosynthesis for Breast cancer Screening and Diagnosis 19

20 First Author, Year, Country Rafferty, , USA Study Design, Duration July 2008 Observer performance study (2 studies: Study 1 and Study 2). Device: Digital mammograph y with a commercially available system (Selenia) and tomosynthesis with an investigational tomosynthesis system Patient Characteristics, Sample Size (N) Women presenting for screening or biopsy. Age (years) (mean [range]): Study1: 51.7 [25 to 80] Study2: 53.5 [25 to 87] N: Study 1: 312 (48 cancers, 264 noncancers); Study 2: 310 (51 cancer cases, 259 non-cancer cases) Interventio n Compara tors Outcomes Measured (M+T) M ROC-AUC, sensitivity, specificity, PPV, NPV Accrual period: July 2006 to May 2007 Bernardi, , Australia Observer performance study. 100 cases were randomly selected from the 2D+3D image archive. 100 cases comprising of 10 cancers and 90 noncancers were randomly selected from the 2D+3D image archive. Age: NR (M+T) i.e (2D+3D) M i.e. (2D) Detection rate, recall rate (primary focus was acquisition time and reading time) Device: Standard 2D mammograhy and the COMBO procedure which acquires both 2D and 3D images were used. N = 100 Digital Tomosynthesis for Breast cancer Screening and Diagnosis 20

21 First Author, Year, Country Study Design, Duration Patient Characteristics, Sample Size (N) Interventio n Compara tors Outcomes Measured Subject accrual period: NR AUC = area under the curve, FN = false negative, FP = false positive, FPF = false positive fraction, JAFROC = Jackknife alternative free response receiver operating characteristic, M = mammography, NPV = negative predictive value, NR = not reported, T = tomosynthesis, PPV = positive predictive value, ROC = receiver operating, characteristic, SR = systematic review, syn = synthetic, TN = true negative, TP = true positive, TPF = true positive fraction, Digital Tomosynthesis for Breast cancer Screening and Diagnosis 21

22 APPENDIX 3: Summary of Study Strengths and Limitations First Author, Year, Country Systematic reviews Houssami, , Australia Strengths The objective was stated. The inclusion and exclusion criteria were stated. A systematic review of the literature was conducted however a single database (Medline) searched, 2000 to Oct List of included studies provided Characteristics of the individual studies were provided Conflict of interest was stated. Both authors received support from Hologic Inc and stated that the support received had no role in the conceptual or reporting aspects of the review. Clinical Studies - Screening Ciatto, , Most participants received both Italy modalities (5% declined 2D and 3D (STORM) and received only 2D). Prospective study All participants appear to be included in the analysis Limitations Study selection was not described in detail List of excluded studies not provided Article selection was done by single reviewer Not mentioned if data extraction was done in duplicate Not mentioned if quality assessments of studies were conducted Not mentioned if publication bias was explored Unclear if a consecutive sequence or a random selection of participants were considered Mammograms were interpreted first on the basis of 2D alone and then by the same radiologist on the basis of 2D and 3D, so not independent. Unclear if participants representative of those generally seen in clinical practices Haas, , USA Image sets for the two modalities were from independent population groups, hence interpretations were independent Unclear if a consecutive sequence of participants were considered. Not randomized All participants did not receive both modalities. One group received M and the other group received M+T (M+T was only available at one of the four sites). Participants receiving (M+T) were more likely to have had breast cancer or have a first-degree relative with breast cancer Retrospective study Unclear if participants representative of those generally seen in clinical practices Digital Tomosynthesis for Breast cancer Screening and Diagnosis 22

23 First Author, Year, Country Rose, , USA Strengths All participants appear to be included in the analysis Image sets for the two modalities were from independent population groups, hence interpretations were independent Limitations Unclear if all participants were included in the analysis Unclear if a consecutive sequence of participants were considered. Not a random selection. All participants did not receive both modalities. One group received M and the other group received M+T The M group considered was of a time period before the introduction of T Retrospective study Participants were from a single practice, so unclear if representative of those seen in other clinical practices. Skaane, , Norway Prospective study All participants received both modalities Images from examinations performed with each modality were interpreted independently All participants appear to be included in the analysis Unclear if a consecutive sequence or a random selection of participants were considered Participants were from a single institution, so unclear if representative of those generally seen in clinical practice Clinical Studies Diagnosis (T vs M) Svahn, , All participants received both Sweden modalities Image sets from the two modalities were interpreted independently The truth status( positive/ negative) of the images were determined by two radiologists not participating in the study and the readers participating in the study were blinded to the truth status All image sets were analyzed Unclear if a consecutive sequence or a random selection of images were considered. The selected sample may not be representative of that typically seen in practice (sample enriched with difficult cases) Fornvik, , Sweden All participants received both modalities Image sets from the two modalities were interpreted independently Readers were likely blinded to the truth status but it was not explicitly Unclear if a consecutive sequence or a random selection of images were considered. The selected sample may not be representative of that typically seen in practice (sample consisted of Digital Tomosynthesis for Breast cancer Screening and Diagnosis 23

24 First Author, Year, Country Strengths mentioned All image sets were analyzed Limitations difficult cases) Clinical Studies Diagnosis (T vs M vs [M+T]) Thibault, 17 Consecutive participants were 2013,France enrolled All participants received all the modalities Image sets from the different modalities were interpreted independently Readers were blinded to the truth status All image sets were analyzed excepting those with incomplete data The selected sample may not be representative of that typically seen in practice (disease enriched sample) Waldherr, , Switzerland Consecutive participants were enrolled All participants received all the modalities Image sets from the different modalities were interpreted independently Readers were blinded to the truth status All image sets were analyzed. The selected sample may not be representative of that typically seen in practice Clinical Studies Diagnosis (M vs [M+T]) Gennaro, , All participants received both Italy modalities Image sets from the two modalities were interpreted independently Readers were blinded to the truth status All image sets were analyzed excepting those with incomplete data Unclear if consecutive participants were enrolled The selected sample may not be representative of that typically seen in practice Rafferty, , USA All participants received both modalities Image sets from the two modalities were interpreted independently Readers were blinded to the truth status All image sets were analyzed Unclear if consecutive participants were enrolled The selected sample may not be representative of that typically seen in practice Digital Tomosynthesis for Breast cancer Screening and Diagnosis 24