1. Tomosynthesis vs Digital Mammography What do the published studies show? Why do we need TMIST? Etta D. Pisano, MD Harvard Medical School Beth Israel Deaconess Medical Center

2. NY TIMES HEALTH Vast Study Casts Doubts on Value of Mammograms By GINA KOLATAGINA KOLATA FEB. 11, 2014 Komen Mammogram Ads Misleading: Professors THEIR 'STATISTICS ARE MEANINGLESS,' ONE EXPLAINS

3. SCREENING IS UNDER ATTACK We KNOW screening saves lives based on 7 randomized controlled trials- -Support is stronger for mortality reduction in OLDER women. - Mortality reduction estimates range from 15- 49%. -Controversy centers on women in their 40s and how often to screen.

4. SCREENING IS UNDER ATTACK - There are many published statistical estimates of the rate of OVERDIAGNOSIS, ranging up to 50%. -What is the rate and role of OVERTREATMENT?

5. Should Tomosynthesis replace Digital Mammography For Breast Cancer Screening?

6. Literature on Screening Tomosynthesis  The majority of DBT published studies are small reader studies. Limited subjects, single site, single DBT vendor.  Diagnostic Performance (AUC, Sensitivity, Specificity, recall rate, Cancer Detection Rate, PPV)  Assessment of DBT alone (TWO VIEW OR ONE VIEW) and DBT as ADJUNCT to one and two-view mammography.  LIMITED CLINICAL DATA published to date on SYNTHETIC 2-D PLUS TOMO.

7. Tomosynthesis Literature  There are several papers that provide insight into use of DBT in clinical practice (single vendor studies)  Clinical Performance (AUC, Sensitivity, Specificity, recall rate, cancer detection rate, etc.)  Screening Population  Radiation Exposure  AGD of DBT compared to Screening FFDM  Interpretation Time

8.  in reader studies published to date (* stat significant) Diagnostic Accuracy (ROC) BETTER for TM+DM vs DM CountryMachine# Sites (#Readers) Δ in AUC (Z Tomo, Z FFDM) Svahn (2010) Svahn (2012) SwedenSiemens (1vDBT+D)v DM 1vDBT vs DM 1 (5) 0.114* 0.094* (0.86, 0.766) Alakrhas (2014) (?pure screening) AustraliaHologic (2vDBT+DM) vs DM 1 (26) 0.107* (0.788,0.681) Wallis (2012)England & Sweden Philips (2vDBT+DM) vs DM 2 (20) 0.079* (0.851, 0.772) Rafferty (2013)USHologic (2vDBT+DM) vs DM 5 (15) 0.068* (0.895, 0.828)

9. CountryMachine# Sites #Women Δ FP rate (TM+DM, Rate DM) ΔCancer Detection Rates/1000 (Rate TM+DM, rate DM) Ciatto 2013 ItalyHologic2 7292 Reduced FPs from 322 to 254* 2.8* (8.1, 5.3) Bigger cancers Skaane 2013 NorwayHologic1 12631 8.0/1000 (53.1/61.1)* 1.9* (8.0, 6.1) More invasive CAs Lower FPs and Higher Cancer Detection Rates for TM+DM vs DM in 2 European Prospective Studies

10. 8 US retrospective studies-  recalls,  CDR CountryMachine# SitesΔ % recalls (rate TM+DM, rate DM) Δ Cancer Detection Rates/1000 (Rate TM+DM, rate DM) Rose 2013USHologic1 2.8%* (5.5%, 8.7%) 1.4 (5.4,4.0) Friedewald 2014USHologic13 1.5%* (9.1%, 10.7%) 1.2 (5.4, 4.2) Haas 2013USHologic4 3.6%* (8.4%, 12%) 0.5 (5.7,5.2) McCarthy 2014USHologic1 1.6%* (8.8%,10.4%) 0.9 (5.5,4.6) Lourenco 2014USHologic1 2.9% (6.4%,9.3%) 0.8 (5.4,4.6) Greenberg 2014USHologic1 2.6% (13.6%,16.2%) 1.4 (6.3,4.9) Destounis 2014USHologic1 7.25% (4.2%,11.45%) 1.9 (5.7,3.8) Durand 2014USHologic1 12.3% (8.4%, 12.0%) 0.2 (5.9,5.7)

11. 8 US retrospective studies-  recalls,  CDR CountryMachine# SitesΔ % recalls (rate TM+DM, rate DM) Δ Cancer Detection Rates/1000 (Rate TM+DM, rate DM) Rose 2013US Hologic 1 2.8%* (5.5%, 8.7%) 1.4 (5.4,4.0) Friedewald 2014US Hologic 13 1.5%* (9.1%, 10.7%) 1.2 (5.4, 4.2) Haas 2013US Hologic 4 3.6%* (8.4%, 12%) 0.5 (5.7,5.2) McCarthy 2014US Hologic 1 1.6%* (8.8%,10.4%) 0.9 (5.5,4.6) Lourenco 2014US Hologic 1 2.9% (6.4%,9.3%) 0.8 (5.4,4.6) Greenberg 2014US Hologic 1 2.6% (13.6%,16.2%) 1.4 (6.3,4.9) Destounis 2014US Hologic 1 7.25% (4.2%,11.45%) 1.9 (5.7,3.8) Durand 2014US Hologic 1 12.3% (8.4%, 12.0%) 0.2 (5.9,5.7)

12. 8 US retrospective studies-  recalls,  CDR CountryMachine# SitesΔ % recalls (rate TM+DM, rate DM) Δ Cancer Detection Rates/1000 (Rate TM+DM, rate DM) Rose 2013 USHologic1 2.8%* (5.5%, 8.7%) 1.4 (5.4,4.0) Friedewald 2014USHologic 131.5%* (9.1%, 10.7%) 1.2 (5.4, 4.2) Haas 2013 USHologic4 3.6%* (8.4%, 12%) 0.5 (5.7,5.2) McCarthy 2014 USHologic1 1.6%* (8.8%,10.4%) 0.9 (5.5,4.6) Lourenco 2014USHologic1 2.9% (6.4%,9.3%) 0.8 (5.4,4.6) Greenberg 2014 USHologic1 2.6% (13.6%,16.2%) 1.4 (6.3,4.9) Destounis 2014USHologic1 7.25% (4.2%,11.45%) 1.9 (5.7,3.8) Durand 2014 USHologic1 12.3% (8.4%, 12.0%) 0.2 (5.9,5.7)

13. All of these studies were on DM+TM vs DM, none on Synth DM+TM vs DM 3 US retrospective studies-  recalls,  CDR CountryMachine# SitesΔ % recalls (rate TM+DM, rate DM) Δ Cancer Detection Rates/1000 (Rate TM+DM, rate DM) Friedewald 2014USHologic 131.5%* (9.1%, 10.7%) 1.2 (5.4, 4.2) Lourenco 2014USHologic1 2.9% (6.4%,9.3%) 0.8 (5.4,4.6) Destounis 2014USHologic1 7.25% (4.2%,11.45%) 1.9 (5.7,3.8)

14. DBT Average Glandular Dose  Shin et al. Comparative evaluation of average glandular dose and breast cancer detection between single-view digital breast tomosynthesis (DBT) plus single-view digital mammography (DM) and two-view DM: correlation with breast thickness and density. Eur Radiol. 2014 Sep 3. [Epub]  Retrospective comparison of average glandular dose of Hologic Dimensions DBT and Hologic Dimensions FFDM for 149 subjects (102 cancer, 35 benign, 12 normal)  Korea, Single Site, Hologic Dimensions MetricFFDM mean[range] DBT mean[range] Difference P-Value Mean AGD1.63 mGy (0.68-7.41 mGy) 1.7 mGy (0.93-5.02 mGy) <0.001*

15. Reading Times Trend Longer for (DBT+FFDM) vs. FFDM CountryMachine# SitesReading Time Difference Skaane 2013NorwayHologic1 45 vs. 91 sec* Bernardi 2012ItalyHologic1 33 vs. 77 sec*

16. TOMO COSTS MORE THAN DM Should all centers replace ALL DM units? Would some women be equally served with DM instead of Tomo at a lower cost? Initial Cost-Effectiveness Analysis (Lee 2014) suggests TOMO is cost- effective in women with dense breasts, if specificity gains are realized outside of early adopters/research studies.

17.  What is the impact of using 2D mammogram synthesized from a DBT exam on diagnostic accuracy (in lieu of DM+TM) in a screening population in the US?  ARE the new cancers found the kind that kill women?  Do ALL DBT systems perform with the same diagnostic accuracy and reduce recalls?  What is the interval cancer rate for TM? What we still need to learn about TOMO for SCREENING

18. TMIST Purpose Should breast tomosynthesis (TM) replace digital mammography (DM) for breast cancer screening in North America? Study designed by a team including reps from ACRIN/ECOG, NCI, BCSC, CISNET, lay advisory committee and ACS, plus screening skeptics!

19. TMIST Study Design  Study Type: Randomized Clinical Trial.  Time Frame: 5 years plus possibly another 5 years of f/up  Patient Population: Asymptomatic women presenting for screening mammography who are at least 40 years of age.  Estimated Number of Women: 88,312 (44,156 per arm)  Randomization Assignments: Breast Tomosynthesis (TM) or Digital Mammography (DM). All women will undergo one test or the other for 3 consecutive years.

20. Primary Aim  Compare the number of advanced or aggressive cancers detected using Tomo vs. Digital mammography.  With advanced or aggressive cancers defined as- 1) All invasive cancers over 1.0 cm in size, regardless of nodal status 2) All invasive tumors that are over 6 mm in size and which have prognostic markers that suggest aggressive behavior, (ie triple negative or her 2+).

21. Secondary Study Aims  Patient-Centered Aims  To compare the recall and biopsy rates due to abnormal screening examinations for TM versus DM  To compare quality of life (QOL) and patient preferences for breast cancer screening regimens among women receiving TM versus DM  To assess treatment decision making among women diagnosed with breast cancer  Imaging Aims  To compare diagnostic accuracy of TM versus DM  To compare interval cancer rates

22. Secondary Study Aims  Biological Aims  To determine correlates of TM/DM findings, pathology and genetic analysis, and other patient characteristics with long-term patient outcomes, including interval cancer rate.  To ascertain potential biomarkers for aggressive tumors and/or poor outcomes via blood DNA/RNA/protein analysis.  To assess and compare the characteristics (e.g. stage, grade, cell subtype) of cancers detected from screening by TM and DM.

23. Eligibility  Inclusion Criteria  Women age 40 and over  Presenting for screening mammogram with no symptoms of breast disease  Exclusion Criteria  Previous history of breast cancer  Breast implants  Known Pregnancy, or lactating

24. Study Procedures  Enrolled subjects would be randomly assigned to one of two imaging arms: TM or DM.  Subjects would undergo three consecutive (1 year) screening rounds under the same modality.  One additional year of follow-up information would be recorded 12 months after 3 rd screening round. (Could be 5 years of follow-up, if funded for this).  All women who undergo breast biopsy during the trial will have sample of breast tissue go to the study central bio- repository for genomic testing.  Genomic data on all subjects will be collected from a blood sample.  Quality of Life and Treatment Decision Surveys for subsets.

25. Study Timeline  Funding from the National Breast Cancer Foundation of Canada has allowed the study to open in Vancouver, Ottawa, Toronto and Boston, with over 1000 women enrolled to date. Plan to enroll 6300 women through that funding in 5 centers (4 in Canada, 1 in Boston).  NCI funding decision pending. Total planned enrollment is 88312.

26. NY TIMES HEALTH Vast Study Casts Doubts on Value of Mammograms By GINA KOLATAGINA KOLATA FEB. 11, 2014 Komen Mammogram Ads Misleading: Professors THEIR 'STATISTICS ARE MEANINGLESS,' ONE EXPLAINS

27. We need to re-prove the efficacy of modern breast cancer screening. TMIST

28. Questions? After today, email me at episano@bidmc.harvard.edu.