Digital breast tomosynthesis(DBT) vs 2D mammography and impact of combined use: A meta-analysis

Background: Having breast cancer and receiving treatment for it is viewed as a very traumatic experience for women. From the recent advances in breast cancer diagnosis protocol; The triple assessment for a lump in the breast is the best practice and its application towards early disease detection is crucial. For which, 2D mammography is used. The limitation of this technique is a potential tissue overlap in dense breasts. Digital breast tomosynthesis (DBT) is a new imaging technology that can address the limitations of 2D mammography. This study aims to highlight the use of DBT.

The study population included all women aged from 18 to 80. For this analysis, the subdivision of the population is not created. A total of 28 RCTs with a total of 1,735,126 patients were selected for the DBT+FFDM vs FFDM study and 5 RCTs with a total of 554,419 patients were selected for the DBT vs FFDM study.

For the DBT+FFDM vs FFDM; The overall diagnostic odds ratio was 1.28 with a 95%CI [1.22-1.33], the Decrease in RR is 2.43% and the increase in CDR is 1.583. for DBT vs FFDM; the overall diagnostic odds ratio was 1.38 with a 95%CI [1.28-1.50], a decrease in RR is 1.70%, and an increase in CDR is 1.4.

DBT+FFDM and DBT do outweigh FFDM alone.

‘3D vs 2D mammography’, ‘DBT+FFDM vs FFDM’, ‘DBT vs FFDM’, ‘Digital Breast Tomosynthesis-meta analysis’, ‘DBT vs DM’, ‘DBT vs 2d mammography’, and ‘3d and 2d mammography’. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint

Breast cancer is the most common cancer observed in females overall. Having breast cancer and receiving treatment for it is viewed as a very traumatic experience for women because this reflects on their self-image, sexual relationship, and self- confidence and this may lead to other psychological problems such as anxiety, denial, depression, anger, etc.[1] This is why early breast cancer detection with minimum recalls is useful to avoid fears and anxiety toward the disease.[2] From the recent advances in breast cancer diagnosis protocol; The triple assessment for a lump in the breast is the best practice and its application towards early disease detection is crucial.[3] For screening of breast cancer, there are two main resources available; Mammography and a second clinical breast examination.[4] For mammography, there are few options available, Conventional two-dimensional (2D) mammography is widely acknowledged as the most effective method for detecting breast cancer. A meta-analysis of 11 randomized trials concluded that mammography screening can lead to a reduction of 20% mortality in breast cancer.[5] The limitation of this technique is a potential tissue overlap in dense breasts. This can consequently obscure the area of interest in the image, which can lead to false- negative and false-positive readings, following which recall of the patient for further imaging and biopsy will be needed. This explains 15-30% of undetected cancers by standard screening protocol. This can in turn be responsible for anxiety and non- attendance for the next routine breast screening tests.[6] Digital breast tomosynthesis (DBT) is a new imaging technology that can address the limitation caused by overlapping structures in 2D mammography. This technique provides a dual benefit for screening purposes. First, DBT increases the cancer detection rate mostly by highlighting architectural distortions [AD] and allowing a better assessment of the shape and margins of the mass. Second, it helps reduce the recall rate. However, DBT is not included in the majority of cancer screening programs worldwide.[5], [7], [8] The use of DBT with digital or synthetic mammography for breast cancer screenings increases the rates of overall and invasive breast cancer detection. however, no evidence shows that DBT , compared with digital or synthetic mammography may decrease recall rates, with high or moderate quality.[9] This study aims to combine different results and create the most accurate meta- analysis to date to experience clinical routine use of tomosynthesis, with an evaluation of the added value of this technique by comparing the diagnostic accuracy with the increase in cancer detection rate [CDR] and decreases in recall rate [RR] of Digital breast tomosynthesis vs 2D mammography and DBT combined with 2D mammography vs the accuracy of 2D mammography alone. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint From here onwards, Digital breast tomosynthesis [DBT] means 3D mammography and conventional mammography means 2D mammography; Full-Field Digital Mammography[FFDM]; Digital mammography[DM] or standard 2D mammography. These words are used interchangeably. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint Image 1: PRISMA Flow chart . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint METHODOLOGY: DATA COLLECTION For the data collection, Researchers reviewed all relevant literature, a search was done using PubMed, Google Scholar, and Cochrane Library databases. The used medical subject headings (MeSH) and keywords were the following: ‘3D vs 2D mammography’, ‘DBT+FFDM vs FFDM’, ‘DBT vs FFDM’, ‘Digital Breast Tomosynthesis-meta analysis’, ‘DBT vs DM’, ‘DBT vs 2d mammography’, and ‘3d and 2d mammography’. For further data collection from all the articles, a review of all the references and meta-analyses available, titles and abstracts were read methodically, and papers with detailed information about the study group and control group were selected. INCLUSION AND EXCLUSION CRITERIA In this paper, the data is compiled for the studies that provided Recall rates and cancer detection rates in the women who were not exposed to the risk factors. The main purpose of this study is to compare and provide the decrease in Recall rate and increase in cancer detection rate overall for the implementation of the DBT+FFDM method in routine screening protocol, The study population includes all women aged from 18 to up to 80. The subdivision of the population was not created as it is beyond the scope of this study. The following papers were excluded; Carbonaro et al 2016 [10], Michell et al 2012[11], and Chae et al 2016[12] as the study populations were exposed to risk factors thus hindering the result analysis. Inclusion criteria were the following: 1) only articles in the English language were selected, 2) Articles providing DBT vs FFDM comparison. 3) Articles with DBT+FFDM vs FFDM comparison 4) articles providing either of the following cancer detection rate [CDR] and recall rate [RR]. Exclusion criteria were the following: 1) non-English language articles, 2) articles where the study group was exposed to the risk factors 3) articles where provisional diagnoses for the patient were given as architectural distortion. 4) articles not providing CDR or RR. DATA EXTRACTION . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint According to PRISMA, a total of 28 RCTs with a total of 1,735,126 patients were selected for the DBT+FFDM vs FFDM study and 5 RCTs with a total of 554,419 patients were selected for the DBT vs FFDM study. Sumkin et al 2015 were only included in the RR table. Every appearing paper was independently studied by three different reviewers. Each article was analyzed for the number of patients, age, procedure modality, risk factors, BIRADs scoring of patients, recall rates, and cancer detection rate. Further discussion and consultation with the other authors and third- party reviewers were used to resolve conflicts. In the majority of the papers, the recall rate (RR) was calculated as the number of positive examinations (BI-RADS 0, 3, and 4) over the total number of screening examinations interpreted while Some papers considered only BI-RADS cat 0 for the calculation of RR. And some also include Cat 5. The modified Jadad score was employed to establish the quality of each paper for participation. ASSESSMENT OF STUDY QUALITY Three writers independently assessed the caliber of each included study. This test consists of 10 questions, each with a score between 0 and 2, with 20 being the maximum possible overall score. Two authors rated each article independently. For deciding the bias risk for RCTs, the Cochrane tool was applied. There were no assumptions made for any lacking or unclear material. There was no funding involved in collecting and reviewing the data analysis. STATISTICAL ANALYSIS For the determination of the odds ratio and other analyses, The statistical software packages RevMan (Review Manager, version 5.3), SPSS (Statistical Package for the Social Sciences, version 20), and Excel were used. Fixed- and random-effects models were used to estimate diagnostic odds ratios (DOR) with 95 percent confidence intervals to examine critical clinical outcomes. The odds ratio of the individual study was plotted on the funnel plot and forest chart. BIAS STUDY The risk of bias was evaluated by the QUADAS-2 analysis. which includes 4 Major parameters, as follows; 1. Patient selection, 2. Index test, 3. Reference standard, 4. The patient flow and Timing of the Index tests.

Table 1: Table of the description of papers Author Name and Ye ar Location Duration of the study Population Study design Modalities Increase in CDR Decrease in Recall rate DBT DBT+DM DBT DBT+DM . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint v DM v DM v DM v DM Alsheik et al 2019[13] US Jun 2015- Sept 2017 99660 (DM) 95370 (DBT) Prospective study DBT vs DM 1 -2.15 Aujero et al 2017[14] US Oct 2011- Jun 2016 32076 (DM) 30561 (DBT +DM) Retrospective study DBT+DM vs DM 1.1 -2.9 Bahl et al 2018[15] US Jan 2009- Feb 2011 (DM) Jan 2013- Feb 2015 (DBT) 78 385 (DM) 76 896 (DBT +DM) Retrospective study DBT+DM vs DM ResultA: 0 ResultB: 0 Bernardi et al 2016[16] Italy May 2013- May 2015 9677 (DM) 9677 (DBT +DM) Prospective study DBT+DM vs DM 2.2 0.55 Chikarmane et al 2020[17] US Oct 2014- Sept 2016 (DM) Feb 2017- Dec 2018 (DBT) 5706 (DM) 4440 (DBT +DM) Retrospective study DBT+DM vs DM 0.1 -2.24 Ciatto et al 2013[18] Italy Aug 2011- Jun 2012 7294 (DM) 7294 (DBT +DM) Prospective comparison study DBT+DM vs DM 2.8 -1 Cohen et al 2018[19] US Feb 2011- Jun 2014 71656 (DM) 31414 (DBT +DM) Retrospective review DBT+DM vs DM 0.8 -1.8 Conant et al 2016[20] US Jan 2011- Dec 2014 113061 (DM) 25268 (DBT +DM) Retrospective analysis of prospective data DBT+DM vs DM 1.5 -1.7 Conant et al 2020[21] US Sept 2010- Sept 2016 10511 (DM) 56839 (DBT) Retrospective study DBT vs DM 1 -2.4 Destounis et al 2014[22] US Jun 2011- Dec 2011 524 (DM) 524 (DBT +DM) Retrospective review DBT+DM vs DM 1.9 -7.25 Durand et al 2015[23] US Aug 2011- Dec 2012 9364 (DM) 8591 (DBT +DM) Retrospective review DBT+DM vs DM 0.2 -4.5 Friedewald et al 2014[24] US Mar 2010- Dec 2012 2,80,698 (DM) 1,72,727 (DBT +DM) Retrospective multi-centre analysis DBT+DM vs DM 1.2 -1.63 Giess et al 2017[25] US Oct 2012- May 2015 14180 (DM) 9817 (DBT +DM) Retrospective study DBT +DM vs DM 2 0.4 Greenberg et al 2014[26] US Aug 2011- Nov 2012 38674 (DM) 20943 (DBT +DM) Retrospective multisite study DBT+DM vs DM 1.4 -2.6 Haas et al 2013[27] US Oct 2011- Sept 2012 7058 (DM) 6100 (DBT +DM) Retrospective study DBT+DM vs DM 0.5 -3.6 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint Heindel et al 2022[28] Germany July 2018- Dec 2020 49762 (DM) 49 715 (DBT +DM) Prospective RCT DBT+DM vs DM 2.3 Houssami et al 2014[29] Italy Aug 2011- Jun 2012 7292 (DM) 7292 (DBT +DM) Prospective study DBT+DM vs DM ResultA: 2.7 ResultB: 2.8 ResultA: -0.9 ResultB: -1 Houssami et al 2017[30] Italy May 2013- May 2015 9677 (DM) 9677 (DBT +DM) Secondary analysis Prospective study DBT+DM vs DM 1.9 Lang et al 2016[31] Sweden Jan 2010- Dec 2012 7500 (DM) 7500 (DBT +DM) Prospective study DBT +DM vs DM 2.6 1.2 Lourenco et al 2015[32] US Mar 2011- Feb 2012 (DM) Mar 2012- Feb 2013 (DBT) 12577 (DM) 12921 (DBT) Retrospective study DBT vs DM 0.4 -2.94 McCarthy et al 2014[33] US Sept 2010- Aug 2011 (DM) Oct 2011- Feb 2013 (DBT +DM) 10728 (DM) 15571 (DBT +DM) Retrospective – Prospective study DBT +DM vs DM 0.9 -1.6 McDonald et al 2015[34] US Sept 2010- Aug 2011 (DM) Oct 2011- Feb 2013 (DBT +DM) 10728 (DM) 15571 (DBT +DM) Retrospective study DBT +DM vs DM ResultA: 1.7 ResultB: 0.8 ResultA: -4.5 ResultB: -1.3 McDonald et al 2016[35] US Sept 2010- Sept 2014 10728 (DM) 33740 (DBT +DM) Retrospective study DBT +DM vs DM ResultA: 1.5 ResultB: 1.2 ResultC: 0.9 ResultA: -1.6 ResultB: -1.4 ResultC: -1.2 Pattacini et al 2018[36] Italy Mar 2015- Mar 2016 9783 (DM) 9777 (DBT +DM) Prospective RCT DBT+DM vs DM 4.1 0 Powell et al 2017[37] US Jun 2012- Aug 2014 10477 (DM) 2304 (DBT +DM) Retrospective study DBT +DM vs DM 2.6 -2.33 Rose et al 2013[38] US May 2011- Jan 2012 13856 (DM) 9499 (DBT +DM) Retrospective study DBT+DM vs DM 1.33 -3.2 Rose et al 2014[39] US May 2011- Jan 2012 10878 (DM) 10878 (DBT +DM) Prospective interpretations of DBT compared with Retrospective readings of DM DBT+DM vs DM 1.9 -2.8 Roth et al 2014[40] US Sept 2010- 10751 (DM) Interval analysis DBT+DM vs DM 1 -1.6 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint Aug 2011 (DM) Sept 2011- Aug 2012 (DBT+DM) 11115 (DBT +DM) Sharpe et al 2016[41] US Jan 2011- Mar 2014 80149 (DM) 5703 (DBT +DM) Prospective study with Retrospective cohort DBT +DM vs DM 1.9 -1.41 Skaane et al 2013[42] Norway Nov 2010- Dec 2011 12621 (DM) 12621 (DBT +DM) Prospective study DBT+DM vs DM 2.3 -0.77 Starikov et al 2016[43] US Jan 2013- Dec 2013 12157 (DM) 2070 (DBT +DM) Retrospective cohort DBT+DM vs DM 2.1 -7.3 Sumkin et al 2015 [44] US May 2010- Sept 2014 1074 (DM) 1074 (DBT +DM) Prospective single-site clinical study DBT+ DM vs DM -12.9 Winter et al 2020[45] US Jan 2012- Dec 2014 (DM) Jan 2016- Dec 2018 (DBT) 117099 (DM) 119746 (DBT) Retrospective study DBT vs DM 2.4 -2.1 Zackrisson et al 2018[46] Sweden Jan 2010- Feb 2015 14848 (DM) 14848 (DBT) Prospective survey DBT vs DM 2.2 1.1 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint DBT+FFDM vs FFDM A total of 28 RCTs with 1,735,126 patients were selected for the study of DBT+FFDM vs FFDM arm. In Image 2, the forest chart and in Image 3, funnel plots are described. This image suggests as the sample size increases, the odds ratio increases, which can be seen in Cohen et al 2019, Conant et al 2016, Friedewald et al 2014, Greenberg et al 2014, and Heindel et al 2022. The overall diagnostic odds ratio was 1.28 with a 95%CI [1.22 to 1.33] . . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint DBT vs FFDM A total of 5 RCTS with 554,419 patients were selected for this study, image 4 describes the forest charts and image 5 describes the funnel plot. Here overall diagnostic odds ratio was 1.38 with a 95%CI [1.28 to 1.50]. here, it is also notable that an increase in sample size leads to an increase in the odds ratio for DBT. This can be seen in Alsheik et al 2019 and Winter et al 2020. Table 2: Decrease in recall for DBT+FFDM vs FFDM intervention intervention Control Control DBT + DM DBT + DM DM DM Event (recalled people) Total recall rate Event (recalled people) Total recall rate Difference Chikarmane et al, 2020 349 4440 7.86% 575 5706 10.10% -2.24% Bernardi et al. 2016 381 9587 3.97% 328 9587 3.42% 0.55% Ciatto et al. 2013 73 72 35 1% 141 72 35 2% -1% Skaane et al 2013 365 12,621 2.89% 463 12,621 3.66% -0.77% haas et al. 2013 512 6100 8.40% 847 7058 12.00% -3.60% Durand et 671 8591 7.80% 1154 9364 12.30% -4.50% . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint al.2015 Conant et al. 2016 4856 55,998 8.70% 14884 142,883 10.40% -1.70% Destounis et al. 2014 22 524 4.20% 60 524 11.45% -7.25% Greenberg et al. 2014 2845 20,943 13.60% 6247 38,674 16.20% -2.60% Friedewald et al. 2014 15541 173663 8.94% 29726 281187 10.57% -1.63% Rose et al. 2013 518 9499 5.50% 1208 13,856 8.70% -3.20% Aujero et al 2017 1785 30,561 5.80% 2799 32,076 8.70% -2.90% Cohen et al 2018 1914 31,414 6.10% 5641 71,656 7.90% -1.80% Pattacini et al 2018 344 9777 3.50% 339 9783 3.50% 0.00% Rose et al 2014 588 10,878 5.40% 888 10,878 8.20% -2.80% Roth et al 2014 978 11,115 8.80% 1118 10,751 10.40% -1.60% Sumkin et al 2015 274 1074 25.50% 412 1074 38.40% -12.90% Starikov et al 2016 211 2070 10.20% 2128 12,157 17.50% -7.30% Houssami et al 2014 56 7292 0.80% 124 7292 1.70% -0.90% Houssami et al 2014 73 7292 1.00% 141 7292 2.00% -1.00% Lång et al. 2016 282 7500 3.80% 197 7500 2.60% 1.20% Powell et al. 2017 319 2,304 13.84% 1,694 10,477 16.17% -2.33% Sharpe et al 2016 341 5587 6.10% 5270 70173 7.51% -1.41% McCarthy et al 2014 1366 15571 8.80% 1112 10728 10.40% -1.60% Giess et al 2017 2254 21,074 10.70% 1683 16,264 10.30% 0.40% McDonald et al 2015 298 1859 16% 247 1204 20.50% -5% McDonald et al 2015 1068 13,712 7.80% 865 9524 9.10% -1.30% McDonald et al 2016 969 11,007 8.80% 1116 10728 10.40% -1.60% McDonald et al 2016 1004 11157 9.00% 1116 10728 10.40% -1.40% McDonald et al 2016 1065 11576 9.20% 1116 10728 10.40% -1.20% Average 7.80% 10.46% -2.43% P < 0.0001 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint Table 3: Decrease in Recall Rate for DBT vs FFDM intervention intervention Control Control DBT DBT dm dm Event (recalled people) Total Recall rate Event (recalled people) Total recall rate Difference winter et al. 2020 10347 119746 8.60% 12508 117099 10.70% -2.10% Lourenco et al 2015 827 12921 6.40% 1175 12577 9.34% -2.94% Alsheik et al 2019 17,165 194,437 8.83% 14,415 131,292 10.98% -2.15% Conant et al. 2020 4547 56839 8.00% 1093 10511 10.40% -2.40% Zackrisson et al 2018 535 14848 3.60% 371 14848 2.50% 1.10% Average 7.09% 8.78% -1.70% P < 0.0001 Decrease in the recall rate Patients within the screening with BI-RADS 0, 3, 4, or 5 were recalled for further investigation and the recall rate was calculated. Table 2 and Table 3 show a decrease in recall rate for DBT+FFDM and DBT respectively. For DBT+FFDM vs FFDM, the decrease in recall rate is 2.43% overall. For DBT vs FFDM, the average decrease in recall rate was 1.70%. Table 4: Cancer detection rate for DBT+FFDM vs FFDM CDR/1000 DBT + DM DM Difference Heindel et al, 2022 7.1 4.8 2.3 Chikarmane et al, 2020 6.1 6 0.1 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint Table 5: Cancer detection rate for DBT vs FFDM bahl et al. 2018 1.1 1.1 0 bahl et al. 2018 5 5 0 Bernardi et al. 2016 8.5 6.3 2.2 Ciatto et al. 2013 8.1 5.3 2.8 Skaane et al 2013 9.4 7.1 2.3 haas et al. 2013 5.7 5.2 0.5 Durand et al.2015 5.9 5.7 0.2 Conant et al. 2016 5.9 4.4 1.5 Destounis et al. 2014 5.7 3.8 1.9 Greenberg et al. 2014 6.3 4.9 1.4 Friedewald et al. 2014 5.5 4.3 1.2 Rose et al. 2013 5.37 4.04 1.33 Aujero et al 2017 6.4 5.3 1.1 Houssami et al 2017 8.2 6.3 1.9 Cohen et al 2018 4.8 4 0.8 Pattacini et al 2018 8.6 4.5 4.1 Rose et al 2014 5.4 3.5 1.9 Roth et al 2014 5.4 4.4 1 Starikov et al 2016 5.3 3.2 2.1 Houssami et al 2014 7.5 4.8 2.7 Houssami et al 2014 8.1 5.3 2.8 Lång et al. 2016 8.9 6.3 2.6 Powell et al. 2017 7.8 5.2 2.6 Sharpe et al 2016 5.4 3.5 1.9 McCarthy et al 2014 5.5 4.6 0.9 Giess et al 2017 3.8 1.8 2 McDonald et al 2015 5.9 4.2 1.7 McDonald et al 2015 5.4 4.6 0.8 McDonald et al 2016 5.5 4.6 0.9 McDonald et al 2016 5.8 4.6 1.2 McDonald et al 2016 6.1 4.6 1.5 Average 6.226 4.643 1.582 P < 0.0001 CDR/1000 DBT DM Difference winter et al. 2020 5.9 3.5 2.4 Conant et al. 2020 6 5 1 Zackrisson et al 2018 8.7 6.5 2.2 Alsheik et al 2019 4.8 3.8 1 Lourenco et al 2015 7.2 6.8 0.4 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint Increase in Cancer detection rate Table 4 and Table 5 show an increase in cancer detection rate for DBT+FFDM and DBT respectively. The average increase in cancer detection rate was 1.583 for DBT+FFDM and for DBT it was 1.4. Bias Study: Table 6: Risk of bias and applicability concern Average 6.52 5.12 1.4 P = 0.1776 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint Author Name and Year Patient Selection Index test Reference standard Flow and timing Patient selection Index test Reference standard 1 Alsheik et al 2019 low low High high high low low 2 Aujero et al 2017 Low low Low low Low Low low 3 Bahl et al 2018 low low low high low low low 4 Bernardi et al 2016 unclear low Low high low low low 5 Chikarman e et al 2020 low low low unclear unclear low low 6 Ciatto et al 2013 Low Unclear Low Low low low Low 7 Cohen et al 2018 Low Low Unclear Low Low Low Low 8 Conant et al 2016 Low Unclear Low Low Low unclear Low 9 Conant et al 2020 Low Low Low unclear Unclear Low low 10 Destounis et al 2014 Low High unclear Low Low Low unclear 11 Durand et al 2015 unclear Low Low unclear Low Low low 12 Friedewald et al 2014 Low Unclear Low Low Low unclear low 13 Giess et al 2017 low Low High high high low low 14 Greenberg et al 2014 low Low unclear unclear low low low . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint 15 Haas et al 2013 Low Low Low Low Unclear Low unclear 16 Heindel et al 2022 Low Low Low low Low Unclear low 17 Houssami et al 2014 Low Low Low Low Unclear Low unclear 18 Houssami et al 2017 unclear Low unclear low Low Low low 19 Lang et al 2016 Low low Low unclear Low Low unclear 20 Lourenco et al 2015 Low Unclear Low Low low high Low 21 McCarthy et al 2014 Low unclear high Low Low Low unclear 22 McDonald et al 2015 Low High Low unclear Low Low low 23 McDonald et al 2016 Low Low Low Low Unclear Low unclear 24 Pattacini et al 2018 unclear low Low Low Low high unclear 25 Powell et al 2017 low Low unclear high Low Low low 26 Rose et al 2013 Low Unclear Low Low low low Low 27 Rose et al 2014 Low Unclear Low Low low low Low 28 Roth et al 2014 unclear low Low low Low Low low 29 Sharpe et al 2016 Low low high Low Low Low low 30 Skaane et al 2013 Low low Low low Low unclear low 31 Starikov et al 2016 Low low unclear low Low Low low . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint 32 Sumkin et al 2015 low low high low high low low 33 Winter et al 2020 low low low high low unclear low 34 Zackrisson et al 2018 low low low low low unclear low Publication Bias: The summary of publication bias for DBT+FFDM vs FFDM and DBT vs FFDM is shown in (image 6). For the publication bias, In, patient selection, bias was low for 29 studies and unclear for 5. In the index test, bias was low in 25 studies, unclear in 7, and high in 2 studies. While for the reference standard, the bias was low in 23, unclear in 6, and high in 5 studies. The flow and timing bias was low in 22, unclear in 6, and high in 6. The applicability concerns bias in patient selection were low in 26, unclear in 5, and high in 3 studies. The index test bias was low in 26, unclear in 6, and high in 2. The reference standard bias was low in 27, and unclear in 7. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint

To present the best available evidence to help clinicians with decision-making, this study was conducted as a meta-analysis. The primary aim of this study was to evaluate the increase in cancer detection rate[CDR] and decrease in recall rate[RR] for 3D+2D vs 2D and 3D vs 2D mammography. The cancer detection rate was calculated per 1000 population screened, and every patient with BI-RADS category for highly suggestive cat 5 and cat 6. The results show, with a mild to moderate quality of evidence, that implementing DBT plus digital or synthetic mammography in population-based breast cancer screening increases overall breast cancer detection rates, and decreases recall rate. For Lourenco et al 2015, the screening examinations performed during this study are taken as the total population and not the actual population as the data is provided in such a way in the paper. Multiple studies have shown that DBT can substitute mammographic spot compression views for the evaluation of noncalcified lesions. The decrease in RR for asymmetries and focal asymmetries with DBT is related to improved characterization of overlapping breast tissue as a benign finding, which would reduce RR for summation of overlapping tissue. It may also be related to improved visualization of mass margins. According to Zuley et al 2013, By using tomosynthesis in the screening, radiologists were able to find fewer benign masses as BIRADS category 3, 4, or 5, without loss of much sensitivity, and malignant lesions were found as highly suggestive of malignancy (BI-RADS category 5), and with the help of tomosynthesis, radiologists were able to differentiate between malignant and benign lesions better. [47]–[50] In this study, The CDR is identified with an average of 1.28 for 3D+2D vs 2D and 1.39 for 3D vs 2D study design. It has been seen in the result that the incorporation of DBT into conventional 2D mammography does increase the CDR. However, in comparison to radiation exposure, 3D+2D is almost double the exposure than 2D alone. This may theoretically limit the application of DBT with DM.[5] As the usage of DBT in screening is limited, it may be very useful for the detection of benign Architectural distortions and Assessment of symptomatic women and work-up of screening-detected suspicious findings as lumps are identified by their shape, size margins, density, and content. The shape of a lesion is often better defined at DBT and DBT+FFDM than FFDM alone.[50], [51] However, DBT-guided FNAC, DBT-guided vacuum-assisted breast biopsy, Contrast- media-enhanced DBT, and Fusion imaging like DBT with other 3D modalities such as 3D-automated breast ultrasound, radionuclide imaging, or even MRI is a promising research approach.[5], [52] With some of these limitations of the radiation exposure from DBT, there is a new opportunity for using 2D synthesised-3D mammography, which may show less radiation exposure and almost the same or more CDR than DBT+FFDM. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint

This study has not divided the study population into sub-age groups and also some papers only involved BIRADS cat 0 in RR. For the diagnosis and incorporation into . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint the routine screening, DBT requires more time and study and the equipment is costlier.

As shown above, DBT+FFDM and DBT do outweigh FFDM alone and must be used with precaution and proper understanding of reading it as it can identify more cancers and can separate benign from malignancy better for BI-RADS 1,2 and highly suggestive of malignancy cat 5 and cat 6. In clinical practice, the use of . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint tomosynthesis is likely to provide a lower recall rate, an increase in cancer detection rate, fewer screening examinations, fewer short-interval follow-up studies, and fewer biopsies for patients with benign lesions. Thus, relieving anxiety and resources of the patients and hospitals. Using this technology, patients can benefit more than its side effects. For hospitals where the patient load is very high and facilities that provide mass population screening and community diagnosis, these modalities can be implemented for the betterment of the population. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint

[1] K. T sara s et al. , “ As s e ssme n t o f Dep r e s s io n and An x iety in Br e a s t Ca nc er Pati ents: P r e v ale nce a nd A s s oci at e d F ac t ors, ” A sia n Pacific Jo u rnal o f Cance r Pr e ve n ti on , v ol. 19, no. 6, pp. 1 661– 16 69, 201 8, doi: 10.22034 / A P JC P .201 8.19 .6.16 61. [2] J . A. Me r i no B onilla , M . T or r e s T a banera, and L . H. R o s Mend o z a, “ El cánce r de ma ma en el siglo XXI: de l a de tec ci ón pr e c oz a l o s nu ev os tr at ami en t o s, ” R a di ol og i a , vol. 59 , n o. 5, pp. 36 8–379, Sep. 2017, doi : 10.101 6/ j. rx .20 17.06 .003. [3] M. O . Ka r i m, K. A . Kh an, A . J . Kh an, A. Ja v ed, S . F azid, and M . I. A slam, “ T ripl e A sse s s me nt o f Br ea s t L ump: Shou ld W e P e rf orm Core Bi ops y f or E v e r y Pa ti en t ? , ” Cureu s , M ar . 20 2 0, doi: 10 .7759/ cureu s.747 9. [4] J . G. E lmore, M. B. Ba r ton, V . M. M oc eri, S. P olk, P . J . Ar en a , and S . W . F le t che r , “ T en-Y ea r Ri s k of F als e P o sitiv e Sc r e ening M a mmogr am s and C l inical B r e a s t Examination s, ” New Engl and Jour nal o f Me dic ine , v ol. 3 38, no . 1 6, pp. 1 089–109 6, Apr . 19 98, doi : 10 .1056/NE JM 1 9980416 3381601 . [5] C. Man doul et a l. , “ Br e a s t t omo s yn t he s i s : What do we kn ow an d where d o w e s t a nd? , ” D i ag n In t e rv Im agi n g, v o l. 100, no. 10, pp . 537– 5 51, Oct. 201 9, doi : 1 0.1016/j.dii i .2019.0 7.012. [6] M. G. Ma r mo t, D . G. Altma n, D . A. Came r on, J. A . Dew a r , S. G . T hompso n, and M . W ilc o x , “ The be nefi t s a nd harm s o f br e a s t c a ncer sc r e enin g: an indep ende n t r e view , ” Br J Canc er , v ol. 108, no . 11, pp. 2205–2 240, Jun. 2013, doi : 10 .103 8/ bjc .2013 .177 . [7] N. Ti r a da et a l. , “Dig i t al B r ea st T o mos y nt hes i s: Phys i c s, Artifa ct s , and Qu ality Co nt r ol C ons i d er at io ns , ” R a dio G ra phi cs , v ol. 39, no. 2, pp. 413–426 , Mar . 20 19, do i: 10 .1148/r g.2019 180046 . [8] D . Be r n a r d i et al. , “ App li c a t i on o f brea s t tomos y nthe s i s i n s c r e ening: increme nt a l eff ec t on ma mmogr aph y ac quisit i o n and rea ding ti me , ” Br J Ra diol , vol. 85 , no . 102 0, pp. e 1 174–e 1178, Dec . 2012 , doi : 10 .1259/bjr / 19 385909. [9] R. R . Giampie tr o , M . V . G. Cabral, S. A. M . L ima, S. A. T . W e be r , and V . do s Sa nt o s N une s- Nogue ir a , “ Acc ur a cy and E f f e c tiv e n ess o f Mammo gr aph y v ersu s Mam mogr aph y an d T omos yn th esi s f or P o pulation -Ba sed Br e as t C a ncer S creeni ng : A S y s t em at i c R ev ie w and Met a - Ana lysi s , ” S ci Rep , vol. 10, no. 1, p. 7991, M ay 202 0, doi: 10.1038/s41598-02 0 -648 02-x . [10] L . A. C arbon a r o et al. , “Impac t on the r e c all r a t e o f digit a l br e a s t t omo s y nthe s i s a s an adjunc t t o d igi t al m ammogr aphy in the s c r e ening setting. A do uble r ea ding expe rienc e a n d r eview of the li t era tu r e, ” Eu r J R a d i o l , v ol. 85, no. 4 , pp. 808–814 , Apr . 2016 , doi: 10 .1016/j.ejr ad . 2016.01 .004 . [11] M. J. M i chella et al . , “ A c ompa r i son o f t h e acc ur acy of fil m - scr e en mammogr ap h y , full - field di git al m ammogr aphy , a nd digi t a l b r ea st t omo s y nthe s i s ” . [12] E . Y . Chae , H . H. K im, J . H. Cha, H. J. Shin, and W . J . C hoi, “ De t ec tion a nd cha r a cteri z a ti on of br ea s t l e s ion s in a selectiv e dia gno st i c po pula ti on: dia gno st i c ac cur ac y s tudy f o r co mparis on be t we en o ne -view digi t a l brea st t omo s y n th es i s a nd tw o -vi ew full - fi e ld dig it a l ma mmogr aph y ” . . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint [13] N. H. Alsh eik et al. , “Compa r i so n of Re so ur c e Ut i liz a t i on and C linical Outc ome s Fo llowin g S cr eeni ng wi t h D ig it a l Br e a s t T omo s y n th esi s V e r su s Digit al Ma mmogr aph y : Find in gs Fr om a L earning H e alth S y s t em” . [14] M. P . Auje r o, S. C . G avenoni s, R. B enjamin, Z. Z hang , a nd J . S . H olt , “Cl i n ic al Perfor mance of Syn the sized T wo -dimen sion al Mammography Combin ed with T omos y n th es is in a Lar g e S cr eeni ng Popula tion” . [15] M. B ahl, S . Ga ffney , A. M . McCa r th y , K. P . L owry , P . A. D a ng , and C. D . L ehm an, “ Br e ast C ance r Ch ar ac t e ri s tic s A ssoci a t ed wi t h 2 D D i gi t al Mam mogr aph y v er su s Dig ital Brea s t T o m os y n th e s i s f or S creenin g -de t ec t ed and Int er v a l Canc ers ” . [16] D . Be r n a r d i et al. , “Brea st c a nc er scr e eni ng wi t h t o mo s y n th e s i s (3 D mam mogr aph y ) with a cquir ed o r s y nt he t i c 2D ma mmogr aph y c omp ared with 2D mammogr ap h y alon e (ST O RM -2) : a popula t io n - b a se d pr o s pe c tiv e s tudy ” . [17] S . A. Chikarman e, L. R . C och on, R . Khor a s ani, S . Sah u , and C . S. Gi e s s , “Sc r een ing Mam mogr aph y P erformance Me tr i c s o f 2D Digital M a mmogr aph y V ersu s Dig i t al Br ea st T omos yn th esi s in W omen W ith a Person al Hist o r y of B r e a s t Canc er ” . [18] S . Ci a tt o et a l . , “ I n tegr ation o f 3 D digi t al mammogr aphy wi t h t omo s y n th e s i s f or p opul ation br ea s t -c a nc er scr e eni ng (ST O R M) : a pr o s pec tiv e c ompari son study ” . [19] E . O . C ohen , H. H. T s o, K. A. Pha lak, R. C. Ma yo , and J . W . T . Le ung , “ Scr ee ning Ma mm ogr aph y F indings F r om O ne S t anda r d P r oj ectio n Only in the E r a o f Full - F i eld Dig i t al M ammo gr aph y and D i gi t a l Br eas t T o mos yn t h es i s ” . [20] E . F . Co nant et a l . , “B r e a s t c anc er s c r e eni ng u s ing t omo s y nthe si s i n combi nation w ith digit a l ma mmogr aph y c om par ed t o dig i t al ma m mogr aph y alone: a co hort s tudy w ithin t he P RO S P R con sortium” . [21] E . F . Co nant et a l . , “F iv e Con se cutive Y ea r s o f S cr e enin g with D ig i t al Brea st T omos y n th es i s: Outc ome s by Sc r eening Y ea r a nd R ound” . [22] S . De s toun i s , A. A ri e no , an d R. M or g an, “ Initial e xpe r ie nc e w ith combin a ti on d igital br e a s t t omo s ynthe si s plu s full fi eld dig i t al mam m ogr aph y or full fi eld digit a l mammog r a ph y a lone in the s c r e ening e n vir onm en t ” . [23] M. A . D ur a n d et al. , “E a r ly cl inic a l e x peri enc e with d igital b r e a s t t omo s ynthe si s f o r s c reening ma mmogr aph y ” . [24] S . M. Fri ed ew ald et al. , “Brea st c a nce r sc r eeni n g using t omo s yn th e si s in combin atio n with di git al m ammogr aphy ” . [25] C. S. Gie s s , S. P ou r ja bb ar , I . K. Ip, R . Lac s o n, E. Alpe r , a nd R. Khora s a ni, “ Comp aring Dia gno s tic Per forma nce o f D ig i t al B r ea st T omos y n th es i s a n d Full - F i eld Dig i t al Mammo gr aph y in a Hybrid S cr eeni ng En v ir onme nt ” . [26] J . S . G r een ber g , M . C. Ja vitt, J . Katz en , S . Mic hael , and A. E . H olland, “Cli ni c al pe rfor m ance m et r i cs of 3 D d i g it a l br e as t t o m o s yn t he s is c o m p ar e d wi th 2 D d i g it al m a m m o gr aph y f o r b r e a s t canc er s c r e ening in communi ty pr ac tice” . . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint [27] B. M. H a as , V . K a lra, J . G ei sel, M . Raghu, M. D ur a n d, and L . E. Philpo t t s , “Comp a r i son o f t omo s ynthe si s plu s dig it a l mammo gr aph y and dig it al ma mmogr aph y alone f o r br e a s t canc er scree ning ” . [28] W . H e ind el et a l ., “ D i git al b r ea s t t omo s ynthe si s plu s s y n the si sed mammogr a ph y vers u s digi t al scree ning ma mmogr aph y f or the de t ec t i on of in va siv e b r ea s t canc e r ( TO SYM A): a multic en t r e , op en -label , r andomi sed , co n tr o lled, sup e riority t rial” . [29] N. Hou s s a mi et al . , “Brea s t sc r eeni ng us i ng 2D-mammog r a ph y o r i nt e gr ating d igital br e a s t t omo s ynthe si s (3 D -mammogr ap h y) f or si ngle-read ing or d oubl e -r e ading – evide nc e t o guid e futu r e screen ing str a t e gie s” . [30] N. Hou s s a mi et al . , “Brea s t c a nce r de t e ct ion using sing le -reading o f br e a s t t o mo s y n th es i s (3D - ma mmogr aph y ) c omp ared to double -r ea ding of 2D-ma mmogr aph y : E v idence fr o m a po pulation -ba se d trial” . [31] K. Lång , I. Ande r s s o n , A . R o sso , P . T imberg , a nd S. Z ack ris s on, “ P er f ormanc e of o n e-vi ew br ea s t t omo s yn th e si s a s a s tand -alon e brea s t canc er s c reening modali t y : re s ul ts from the Malmö Brea s t T omo s ynt h e s i s Sc r e ening T rial , a popul a ti on -ba sed st u dy ” , doi: 10 .1007/s00330-015 - 3 803-3 . [32] A. P . L ourenc o , M. Ba rry -Brooks, G . L. Bai r d, A . T ut tle , a nd M. B . Maini er o , “Ch an ges in r ec all typ e and p a tie n t treatme n t f ollo wing im plementat i o n of screening digi t a l brea st t omo s ynthe si s” . [33] A. M. M c Ca rth y et a l. , “ Scr ee n ing outco mes follow ing impl ement ation o f digital br ea st t omo s ynthe si s in a g en e r a l-population sc r ee ni ng pr og r a m” . [34] E . S. Mc D o n ald, A. M. M c Car th y , A. L. Ak h tar , M. B . S ynne stv ed t , M. Schn all, a nd E. F . Co n ant, “ Baseli ne Screen ing Mammo gr aph y: Perfor m ance o f Full-F i eld Dig i t al M ammog r aph y V ers u s D i gi t a l Br eas t T o mos yn t h es i s, ” Americ an Jour nal o f R oe nt ge no lo gy , v ol. 2 05, no. 5, pp. 11 43– 11 48, Nov . 2015, doi: 10. 2214/ A J R . 15.14 406 . [35] M. D . E. S. M c Don ald, M . A. O u s timov , S . P . W ei n s t ein , M. B. S y nne s tvedt , P . M. S . MD , and M. E . F . Co nant, “ E ff ec t i vene s s o f Digi t al Br e a s t T omos ynthe si s Compared Wi th Digi t a l Mam mogr aph y O utcome s Analy si s F r om 3 Y ea r s o f Brea st Canc er Sc r e ening ” . [36] P . Pa t ta cini et al. , “ Digit al Ma mmogr aph y v er su s D i gital Ma mmogr aph y Plu s T omo s y n th e s i s f or Br ea s t C an cer Sc r ee ning: The R e g gio E milia T omos yn th e sis R andom iz ed T ri al” . [37] J . L. P o well et al ., “ Impac t o f th e Addi tion of D ig i t al Brea st T o mos y nthe s i s ( DB T) t o S t a nd ar d 2D Dig ital Sc r e ening Mammogr ap h y on t he Rat e s o f Pa ti ent Recal l, C anc er Detecti on, and R e c omme nd a ti on s f or Sh ort -t e rm Follow -up” . [38] S . L. R ose , A . L. Tidw e ll, L . J . Bujnoc h, A . C . K us hwaha , A . S. No r dmann, and R. S . Jr , “ Impleme nt ation o f brea st tomos y n th e si s in a r ou tine s c r e ening pr a c tice : a n obs e r v at i onal study ” . [39] S . L. R ose , A . L. Tidw e ll, M . F . Ic e, A. S . N o r dma nn, R . S e x t on, an d D . O. R. S ong , “ A R ea de r St ud y C o m p ar in g P r os pe c t i v e T om os yn t he s is In t e rp r et a t i o ns w it h R et r os pe ct i v e R e a d i n g s o f the C or re sponding FF DM E x amin a ti on s ” . [40] R. G. R oth, A. D. A . Maidm e n t , S. P . W ein s t ein , S. O. R oth, an d E. F . Cona n t, “Digi t a l Br ea st T omos yn th esi s : Le sso n s L earn ed from Earl y Clinic a l Imple menta ti on” . . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint [41] R. E. S. J r et a l. , “ Increa sed Ca nce r De t ec ti on Rat e a n d V ari a ti on s in t he Recall R a t e R esul ting fr om Im plementa tio n of 3 D Digi t al Br ea s t T omos yn the si s i n to a P o pulation -ba sed S cr eeni ng Pr og r a m” . [42] P . S k aan e et al ., “Pr o sp ectiv e trial compa ring f u ll -fi eld di git al ma mmogr aph y (FF D M) v e r su s combine d FF DM and t omo s y nthe s i s in a population -ba se d sc r een in g pr ogr amme us i ng i ndepend en t d ouble r ea d ing w ith arbi tr ation, ” E ur Ra di ol , v o l. 23, n o. 8 , pp. 2061 –20 71, Aug. 20 13, doi: 10.1007 / s00330 - 0 13-2820 -3. [43] A. S t arik ov , M . Drotman, K. He nt e l, J. K a t z en, R. J. Min, an d E. K. A rl e o, “2D ma mmogr aph y , di git al b r ea s t tomos y n th esi s , and u ltr a so und: whic h s h ould b e u sed f o r th e di ff e r e n t brea st de n s itie s in b r ea st c a nc er s c reening ? ” . [44] J . H. S umkin et al. , “R e call R at e R educ t i o n with T omos yn th esi s Dur i ng Ba s eline Sc r ee ning Ex a m i na t i o ns , ” Ac ad Rad i ol , vol. 22, n o. 1 2, p p. 1477– 1482, D ec . 201 5, doi: 10 .1016/j.ac r a .2015. 08.015 . [45] A. M. W i nt e r , S. Kaz mi , A. K. Ha r dy , and D . L. Be nne t t , “C ompa r i s on o f interval b r e a s t canc er s w ith 2D digi t al mamm ogr aph y v ersu s 3 D di git al b r e a s t tomos y n th esi s in a l ar ge c o mmunity - ba sed pr actic e” . [46] S . Zac kris s on et al . , “On e-view brea st t o mos y n t he s i s vers u s t w o-v iew ma mmogr a ph y in the Malmö Brea s t T omo s ynt h e s i s Sc r e ening T rial ( MB T ST): a pr o s p ec tiv e , popula tio n-b a sed, di agno s tic a cc ur a cy s tudy , ” Lanc et Onc o l , v ol. 19, no. 11, pp . 1493–15 03, 201 8, do i: h ttps ://doi .or g /10.10 16 / S 1470-204 5 (18) 30521-7. [47] M. Noroo zi an et a l ., “ D i git al Br e a s t T omo s y n t he si s I s Comp a r abl e t o M a mmogr ap hic Spot Vie ws f o r Ma ss Cha r a cteriz ati on, ” Radi ol o gy, vol. 262 , no. 1, pp . 61–6 8, Jan. 2 012, doi: 10 .1148/r adio l.1 1101763 . [48] K. R. Br and t et al ., “Ca n Digital Br e a s t T o mos yn t he si s R eplac e Con v entional Diagn ostic Mam mogr aph y Views f o r Sc r e ening R ec a lls Wi t h out Calc i fica tio n s? A C ompari son Study in a S imulat ed Cli nical S e t t i ng , ” America n J ou rnal o f Roe n tgen ol ogy , vol. 20 0, no. 2 , pp . 291 –298, Feb. 20 13, doi : 10.2214/ A JR .12.88 81. [49] C. M. Ha kim, D . M. Chough, M. A . Gan ott, J . H . Sumk in, M. L . Zule y , a nd D . Gur , “ D i g it al Brea s t T omos yn th esi s in the D iag no s tic En v ir onmen t : A Su bjectiv e Si de -by-Sid e R evi ew , ” Ame r i can Jour nal o f R oe nt ge nolo gy , v ol. 195, n o. 2, pp. W 1 72–W176, Aug . 2010, doi: 10 .2214/ A J R.09.324 4 . [50] M. L . Zu ley et a l . , “ D i gital Brea s t T omos ynthe si s vers u s S uppleme n tal Diagno stic Mam mogr aphic View s for Ev aluati on of Nonc a lci fie d B r e a s t Le sion s, ” Ra diol og y , v o l. 266, no. 1, pp. 89– 95, J an. 2013, do i : 10.11 48/ rad i ol.1212 0552. [51] T . I. A ls ha f e iy , J . V . Nguyen , C . M. R ochma n, B. T . Nichol son , J. T . P a tr i e, and J. A. Ha rvey , “O u t come o f Ar c hi t ec tur al Di st or t io n De tect e d Only at Brea s t T omo s ynt h e si s v e r s us 2D Mam mogr aph y , ” Radiolo gy , vol. 288 , no. 1, pp. 38– 46, Jul. 2 018, doi : 10 .1148/r adio l.2 0181711 59. [52] A. V ou r t si s, “ Thr ee -di m en s i onal a u t o m ated br e a s t ul tr a sound : T echnical a s pe c t s a nd first re s u lt s , ” Di ag n I n ter v Im a gin g , vol. 10 0, n o. 1 0, pp. 57 9–592, O c t . 2019, doi : 10 .1016/j.dii i.2019 .0 3.012 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 14, 2023. ; https://doi.org/10.1101/2023.12.07.23299674doi: medRxiv preprint