Five Consecutive Years of Screening with Digital Breast Tomosynthesis: Outcomes by Screening Year and Round
Limited data exist beyond prevalence rounds of digital breast tomosynthesis (DBT) screening.
To compare DBT outcomes over multiple years and rounds to outcomes of digital mammography (DM) screening.
Materials and Methods
Retrospective analysis included 1 year of DM and 5 years of DBT screening (September 2011 to September 2016); 67 350 examinations were performed in 29 310 women. Recall rate (RR) percentage, cancer detection rate (CDR) per 1000 women screened, false-negative rate per 1000 women screened, positive predictive value of recall (PPV1) percentage, positive predictive value of biopsies performed percentage, sensitivity, and specificity were calculated. Cancers diagnosed within 1 year of screening were captured by means of linkage to state cancer registry, and biologic characteristics were grouped by prognostic factors. Performance trends across DBT rounds were compared with those from DM rounds by using logistic regression to account for examinations in the same woman. Analyses were adjusted for age, race, breast density, baseline examination, and reader.
There were 56 839 DBT and 10 511 DM examinations. The mean patient age (± standard deviation) was 56 years ±11 for the entire cohort, 55 years ±11 for the DBT group, and 57 years ±11 for the DM group. RRs were significantly lower for the DBT group (8.0%, 4522 of 56 839; 95% confidence interval [CI]: 7.7, 8.2) than for the DM group (10.4%, 1094 of 10 511; 95% CI: 9.8, 11.0) (P < .001). CDRs were higher with DBT (6.0 per 1000 women screened; 95% CI: 5.4, 6.7 per 1000 women screened; 340 of 56 839) than with DM (5.1 per 1000 women screened; 95% CI: 3.9, 6.6 per 1000 women screened; 54 of 10 511) (P = .25), but this difference was not statistically significant. Both RR and CDR remained improved compared with DM for 5 years of DBT at the population level. False-negative rates were slightly lower for DBT (0.6 per 1000 women screened; 95% CI: 0.4, 0.8 per 1000 women screened; 33 of 56 839) than DM (0.9 per 1000 women screened; 0.4, 1.6 per 1000 women screened; nine of 10 511) overall (P = .30), but the difference was not statistically significant. In adjusted analyses, RR, biopsy recommendation rates, and PPV1 were improved for DBT versus DM (P ≤ .001). Compared with DM, a higher proportion of DBT-detected cancers were invasive (70% [238 of 340] vs 68.5% [37 of 54]) and had poor prognoses characteristics (32.6% [76 of 233] vs 25.0% [nine of 36]).
Favorable outcomes with digital breast tomosynthesis screening were sustained over multiple years and rounds. Digital breast tomosynthesis screening was associated with detection of a higher proportion of poor-prognosis cancers than was digital mammography.
© RSNA, 2020
See also the editorial by Moy and Heller in this issue.
The sensitivity of digital breast tomosynthesis (DBT) was higher than that of digital mammography (DM) in each of the first 5 years after implementation of DBT. DBT also helped detect a higher proportion of poor-prognosis cancers than DM did.
■ At the population level, screening recall rates were lower (8.0% vs 10.4%) and cancer detection rates were higher (six of 1000 women screened vs 5.1 of 1000 women screened) for digital breast tomosynthesis (DBT) compared with digital mammography (DM) alone and remained so across all 5 years of tomosynthesis screening.
■ At the patient level, recall rate, positive predictive value of recall, and biopsy recommendation rate were significantly improved for all rounds of DBT compared with screening with DM alone (P ≤ .001 for all measures).
■ Cancers detected with DBT screening were more often invasive (70.0% vs 68.5%) and associated with a poorer prognosis (32.6% vs 25.0%) than those detected with DM alone.
Breast cancer screening with digital breast tomosynthesis (DBT) is associated with increased cancer detection (1–5) and reduction in false-positive recalls compared with screening with digital mammography (DM) alone (1,2,6–11). However, the impact of DBT on these screening metrics varies according to the screening environment. In a recent meta-analysis of pooled data from prospective European trials and observational U.S. studies, DBT screening was associated with higher increases in cancer detection rates (CDRs) in Europe than in the United States, and greater reductions in recall in the United States than in Europe, where recall rates (RRs) are generally much lower (12). In all settings, the improved outcomes achieved with tomosynthesis are attributed to the increased conspicuity of both benign and malignant lesions and the decrease in tissue superimposition obtained with the quasi-three-dimensional format of DBT imaging (13–15).
Thus far, most of the published data on DBT screening, whether from prospective trials or observational studies, have been from first or prevalence round screening, where CDRs and RRs are expected to be higher than with subsequent, incidence rounds of screening (16). The few studies that have examined subsequent round screening have been limited, including only one to two additional rounds of screening; however, they have demonstrated lower recall rates and sustained improvements in CDRs for DBT compared with DM screening alone (17–19). Of note, few studies have had adequate follow-up to assess for false-negative findings beyond a single round, and the few studies with follow-up have not demonstrated any significant reduction in this important metric (17–20). Therefore, the longitudinal performance of DBT has not been well established at either the population level or at the individual patient level.
To optimize breast cancer screening strategies, understanding specific cancer biology–related outcomes is necessary (21). The National Cancer Institute–funded Tomosynthesis Mammographic Imaging Screening Trial (TMIST) will compare outcomes from approximately 165 000 women aged 45–74 years randomly assigned to either DM or DBT screening (22). Beyond routine screening outcomes, TMIST will also compare the number of advanced or poor- prognosis invasive cancers in each arm of the trial with the hypothesis that DBT screening will eventually decrease the number of advanced cancers compared with DM screening (23).
The aim of our study was to compare RRs, CDRs, biopsy recommendation rates, positive predictive values of recall (PPV1) and positive predictive values of biopsies performed (PPV3), false-negative rates and the biology, size, and nodal status of screen-detected and interval cancers across successive years and rounds of DBT screening versus DM screening.
Materials and Methods
This retrospective analysis was approved by the institutional review board and was compliant with the Health Insurance Portability and Accountability Act. The requirement to obtain informed consent was waived. The study population consisted of all women aged 40 years and older who underwent screening mammography at the Perelman Center for Advanced Medicine, Hospital of the University of Pennsylvania, Philadelphia, Pa, from September 1, 2010, to September 30, 2016. All patients had no prior history of breast cancer and no signs or symptoms of breast cancer. Examinations during the month of transition from DM to DBT, September 2011, were excluded. From September 1, 2010, to August 30, 2011, all patients were screened with DM alone; from October 1, 2011, to September 30, 2016, all patients were screened with DBT (Dimensions; Hologic, Bedford, Mass). From October 1, 2011, until January 6, 2015, DBT imaging was performed with DM (DM/DBT) (Dimensions); beginning on January 7, 2015, and until September 30, 2016, all screening was performed with synthetic DM and DBT (synthetic mammography/DBT) without DM. Portions of these data have been previously published (8,9,17,18,24).
In other publications, we have analyzed screening outcomes on subgroups of the larger population described herein (8,9,11,17,18,24–26). These prior publications included the comparison of RRs and CDRs of DBT and DM screening from 24 767 examinations (8), 26 299 examinations (9,25,26), 44 468 examinations (17), 23 206 examinations (18), 31 666 examinations (24), and 11 623 examinations (11), which are subgroups of patients included in our present analysis. These prior publications also include analyses that evaluated the impact of DBT on baseline examinations (25), the impact of DBT on the use of American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) category 3 (26), and the conspicuity of cancer at DBT compared with DM (15). These analyses are not specifically evaluated in the current analyses but are referenced. In the present analysis, we report on (a) an expanded patient population with additional new patients, (b) additional rounds of screening for previously reported patients and new patients, and (c) additional patient-level data regarding the pathologic characteristics of cancers detected based on TMIST criteria (22).
All examinations were evaluated by one of eight board-certified radiologists who specialized in breast imaging (experience range, 8–27 years) using BI-RADS (27,28). Demographics, breast density, and final assessment categories were categorized using the 4th edition of BI-RADS for most of the study period (27). The exact date of conversion to breast density categories used in the 5th edition of BI-RADS is unknown but occurred sometime after 2014 (28). For statistical analysis, breast density was classified into two groups: nondense (4th edition categories 1 or 2 or 5th edition categories a and b) and dense (4th edition categories 3 and 4 or 5th edition categories c and d) (27,28). Race and/or ethnicity was self-reported and captured from the electronic medical record. Cancer diagnoses were available from the state cancer registry with 1-year follow-up for all examinations.
A positive screening examination was defined as a BI-RADS final assessment category of 0, 3, 4, or 5. RR was defined as the proportion of all examinations that were positive. Overall cancer rate was defined as the number of examinations in which cancer was diagnosed within 1 year of screening. Biopsy recommendation rate was defined as the proportion of examinations with a recommendation for biopsy after recall. CDR was defined as the number of positive screening examinations in which cancer was subsequently diagnosed within 1 year. The false-negative rate was defined as the proportion of negative examinations in which cancer was diagnosed within 1 year. PPV1 was defined as the proportion of examinations with cancer among those with positive screening examinations and PPV3 as the number of examinations with cancer among those with biopsies performed. We evaluated sensitivity and specificity using standard definitions. The sensitivity of detection and CDR of invasive cancer were estimated analogously but restricted to examinations with a diagnosis of invasive disease at histologic examination within 1 year of screening. Because Pennsylvania cancer registry data were judged to be complete through December 31, 2016, performance measures requiring complete cancer capture (cancer rates, false-negative rates, and sensitivity) were estimated using only examinations performed prior to December 31, 2015, to ensure complete cancer capture with 1-year of follow-up.
Cancers were classified as invasive or ductal carcinoma in situ (DCIS) based on registry data. DCIS was classified as high grade or other. Invasive disease was classified according to receptor status, nodal status, and tumor size (<1 cm, 1–2 cm, >2 cm). Advanced cancers (based on TMIST criteria) (22) had one of the following characteristics: (a) metastatic disease, (b) node-positivity, (c) human epidermal growth factor receptor 2–positive and size of at least 1 cm, (d) hormone receptor–negative (both estrogen receptor–negative and progesterone receptor–negative) and size of at least 1 cm, and (e) hormone receptor–positive (estrogen receptor–positive or progesterone receptor–positive) and size of at least 2 cm.
Patient demographics, breast density, and baseline examinations versus subsequent examinations were stratified according to modality (DM or DBT) for counts and proportions. Screening metrics were computed overall for DM and DBT as well as year since DBT implementation and age group (40–49, 50–74, and ≥75 years). In age-based analyses, DBT was further stratified according to first or subsequent round. Each DBT examination was classified by screening round according to the number of prior DBT examinations an individual woman had received to estimate screening performance according to round. Cancer characteristics were summarized overall, by age group, and by first or subsequent DBT screening round.
Logistic regression estimated by means of generalized estimating equations was used to estimate the odds of each performance measure at each screening round in which a woman participated relative to DM. Use of generalized estimating equations accounted for correlation among examinations performed in the same woman. Unadjusted generalized estimating equation models were used to obtain P values and 95% confidence intervals (CIs) for performance measures for DM and DBT overall and stratified according to year after implementation of DBT. We additionally constructed adjusted models for all measures adjusting for within-woman screening round, age, race, breast density, and whether the examination was a baseline. Models for recall and biopsy recommendation were additionally adjusted for reader using fixed effects. These analyses were repeated for all measures, estimating performance as a function of DBT round relative to DM. For these adjusted models, we reported odds ratios and 95% CIs as well as P values for the composite hypothesis test of any difference in performance across screening rounds. Composite hypothesis tests were conducted using Wald tests of the null hypothesis of all parameters associated with screening rounds simultaneously equal to zero. Adjusted performance measures based on these models were obtained using indirect standardization to create a weighted average of estimated performance measures for each combination of covariates included in the model, with weights based on the marginal distribution of each covariate combination (29). Statistical significance was evaluated relative to a two-sided α threshold of .05. P values less than .05, based on two-sided hypothesis tests, were considered statistically significant. All analyses were performed in R software version 3.5.2 (The R Project for Statistical Computing, Vienna, Austria).
Our study population consisted of 67 350 screening examinations (DM = 10 511 and DBT = 56 839) performed in 29 310 women. The median age was 54 years (interquartile range, 46–63 years). Patient characteristics, including age, race, breast density, and whether the screening study was a baseline study or not, were similar between the DM and DBT cohorts (Table 1).
At the population level, the screening outcomes overall, as well as stratified according to screening year, are presented in Table 2. The overall RR was significantly lower for DBT compared with DM (8.0% [95% CI: 7.7%, 8.2%] vs 10.4% [95% CI: 9.8%, 11.0%], respectively; P < .001). The overall CDR was higher for DBT (6.0 per 1000 women screened; 95% CI: 5.5, 6.7 per 1000 women screened) than DM (5.0 per 1000 women screened; 95% CI: 3.9, 6.6 per 1000 women screened) as well as the individual CDRs for DBT years 1–5 compared with DM, although this difference was not statistically significant (P = .25). The overall sensitivity of DBT was higher than that of DM in DBT years 1–4, and in DBT year 5, the overall sensitivity was equal to that of DM at 85.7% (P = .17). Of note, when sensitivity was calculated for invasive cancers only, the overall invasive cancer sensitivity was 89.8% (95% CI: 85.6%, 92.9%) for DBT compared with 80.4% (95% CI: 66.5%, 89.5%) for DM (P = .07). The false-negative rates overall for DBT and for DBT years 1–4 (0.6, 0.6, 0.3, 0.4, and 0.7 per 1000 women screened, respectively) were lower than that for DM (0.9 per 1000 women screened; P = .30); however, this difference was not statistically significant. In DBT year 5, the false-negative rate was equal to that with DM.
At the population level, DBT specificity was higher across DBT years 1–5 relative to DM (P < .001). However, the rate of biopsy recommendations was similar for DBT years 1–3 and DM. The biopsy recommendation rates for DBT years 4 and 5 (1.5% and 1.3%, respectively) were slightly lower than those for DM and DBT years 1–3 (2.0%, 2.2%, 2.0%, and 2.0%, respectively). Overall, the rate of biopsy recommendations was lower with DBT than with DM (1.8% vs 2.0%, P = .23); however, this difference was not statistically significant. The PPV1 for all years of DBT was higher than that for DM (P = .002).
During the 5 years of DBT screening, 27 287 unique women underwent one round of imaging, 12 320 (45%) contributed one round of screening to the analysis, 6476 (24%) contributed two rounds, 4065 (15%) contributed three rounds, 2762 (10%) contributed four rounds, and 1664 (6%) contributed five or six rounds of DBT screening (Table 3). Odds ratios for screening outcomes from individual patient rounds 1–5 of DBT screening compared with DM are presented in Table E1 (online). The overall CDR per 1000 women screened (DBT round 1–5 = 6.89, 5.48, 5.61, 4.16, 5.72, respectively; DM = 5.03 per 1000 women; P = .17), invasive CDR per 1000 women screened (DBT round 1–5 = 4.78, 4.10, 3.17, 3.65, 4.42; DM = 3.57; P = .38), and sensitivity (DBT round 1–5 = 94.31%, 92.07%, 83.14%, 89.29%, 80.42%; DM = 84.82%; P = .10) at each DBT round were not different for DBT compared with DM. However, PPV1 (DBT round 1–5 = 6.92%, 7.49%, 10.40%, 9.41%, 14.09%, respectively; DM = 4.92%), biopsy recommendation rate (DBT round 1–5 = 2.27%, 1.45%, 1.41%, 0.86%, 0.84%; DM = 1.84%), and specificity (DBT round 1–5 = 90.93%, 93.26%, 95.13%, 95.94%, 96.07%; DM = 90.71) were significantly different between the two modalities (P ≤ .001). Figure 1 illustrates the trends of these screening outcomes compared with DM according to round of DBT. In the 4th month of year 4 of DBT screening, we converted to synthetic mammography/DBT screening, without the use of DM. Data on the early implementation of synthetic mammography with DBT have previously been published on this same patient population (24).
Table 4 presents the detected invasive cancers stratified according to histopathologic findings, TMIST advanced criteria (22), positive nodal status, receptor status, and size. The rate of cancers with a maximum histologic classification of high-grade DCIS is also included. Cancer cases are stratified according to study type (DM vs DBT) and, within DBT, as first or subsequent rounds of screening as well as by age group. Overall, the proportion of detected cancers classified as advanced cancers and cancers manifesting with positive nodes was higher for DBT than for DM. When evaluated according to age group, for women aged 40–49 years and 50–74 years, DBT-detected cancers tended to be more often advanced; however, the reverse was found for women older than 74 years (Table E2 [online]). An example of a “good prognosis” (estrogen and progesterone receptor–positive) invasive cancer (node-negative, subcentimeter, invasive) detected at screening and better seen with DBT is shown in Figure 2. An example of an advanced triple-negative cancer detected with DBT is shown in Figure 3.
Digital breast tomosynthesis (DBT) is rapidly being implemented as the new and better mammographic technique despite limited data on long-term outcomes. Although DBT screening has been shown to help detect more cancers thus far, the additional cancers have been mostly luminal A subtypes, which are associated with better prognoses. Such lower-grade cancers most often manifest at mammographic screening as areas of architectural distortion that, even when subtle, are more conspicuous at DBT compared with imaging with digital mammography (DM) alone (14,15). However, studies evaluating subtypes of cancers detected at DBT versus DM screening are mostly based on first-round screening. These analyses may be biased because of a DBT prevalence effect resulting from screening a population with a more sensitive method (ie, DBT) for the first time (30). In addition, these prior studies have had less diverse patients and may not reflect the diversity of cancer subtypes manifesting in different racial populations at different ages.
This report builds on previously published data from subgroups of this population (8,9,11,17,18,24–26). As shown before and again in this analysis with more patients and years of follow-up, the overall RR was lower (8.0% vs 10.4%) and the overall CDR was higher (6.0 vs 5.1 per 1000 women screened) with DBT compared with DM screening alone and continued to improve with each successive year of DBT screening.
In our population, the overall cancer and invasive cancer rates (cancers detected at screening and cancers manifesting within a year after a negative screening examination) increased from DBT year 1 to DBT year 5 compared with DM, likely due to the increase in supplemental screening during the time period. However, despite the increasing overall rate of cancers, the overall false-negative rate remained lower for DBT compared with DM. In addition, across all years of DBT screening, the sensitivity for invasive cancer detection was higher than that with DM. PPV1 and PPV3 also remained higher for all DBT years compared with DM. These data support that the improved outcomes achieved with DBT at the population level are sustainable beyond prevalence round screening.
At the individual patient level across multiple rounds of DBT screening, RR, PPV1, and biopsy recommendation rate continued to improve compared with DM, reflecting sustained outcomes of these additional important screening metrics. When outcomes were analyzed according to age groups and according to whether a first or subsequent screening examination was performed, the overall sensitivity of DBT was higher for all age groups at the first round and remained higher for subsequent rounds for patients aged 40–49 years and those older than 74 years compared with DM. For the 50–74-year-old cohort, subsequent-round DBT sensitivity was slightly lower than that of DM. However, the sensitivity for invasive cancer detection with DBT, whether at the first round of screening or a subsequent round, was higher than that with DM for all age groups. These patient-level data, which included age and breast density over multiple screening rounds, build on data previously reported that demonstrate sustained and improved outcomes at successive rounds of DBT screening (17,18). Also, as additional data are accrued across multiple sites of diverse populations, such patient-level data may help inform new recommendations for the age of initiation of screening and the interval to screening with DBT.
Our results on cancer subtypes detected with each modality differ from the multisite Population-based Research to Optimize the Screening Process (PROSPR) report in which our data were included as a single site. In the PROSPR study, DBT-detected invasive cancers tended to be smaller, less often node positive, and less often human epidermal growth factor receptor 2 enriched compared with DM-detected cancers (18). However, in the PROSPR data set, more than twice the number of women were screened with DM than with DBT. Although the reverse is true in this analysis, significantly more DBT screening examinations were performed than DM screening examinations. Other smaller studies have also reported that DBT-detected cancers tend to be lower grade and more often manifest as subtle spiculations or distortions not evident with two-dimensional mammography (15,30–36). Such low-grade cancers could potentially be found at later DM screening examinations without any significant change in long-term outcomes (eg, morbidity and mortality). It should be noted, however, that many of those earlier studies were based on first or prevalence screening when detection of indolent cancers may be more likely with improved conspicuity of DBT (36). In addition, the other smaller studies did not include as diverse a patient population as our study, where approximately 50% of the patient population screened were African Americans, who are known to develop more aggressive breast cancer subtypes at an earlier age. In the present data set, the invasive cancers detected with DBT at both first and subsequent screenings tended to be more often cancers with poor prognoses than those detected with DM. This suggests that DBT screening not only helps detect these clinically significant cancers, but also that this detection is sustained over multiple rounds of screening. Clearly, large data sets with diverse populations and long-term follow-up are needed to support this finding.
Several limitations must be considered when interpreting our results. All data came from a single academic institution, and the results may not translate to other practices. In addition, we did not have detailed risk information on all patients during the extended time period. However, we previously published outcomes from DM and from the first 18 months of DBT screening where risk data were available, and no difference existed between the risk profiles of the DM and DBT cohorts (9). Finally, midway through DBT year 4, we replaced two-dimensional DM with synthetic mammography in DBT screening. However, previously published screening outcomes from this time period showed no significant change in CDR with synthetic mammography and/or DBT compared with DM and/or DBT (24).
In conclusion, in our study of multiple years and rounds of digital breast tomosynthesis (DBT) screening, we demonstrated sustained and improved screening performance with DBT compared with digital mammography (DM) screening. Specifically, we showed that DBT screening had a higher sensitivity for invasive cancers and, overall, a lower-false negative rate. In addition, the invasive cancers detected with DBT tended to be larger and more often advanced compared with those detected with DM. Such poor-prognosis cancers are generally not considered to contribute to “overdiagnosis” (37) but rather are key targets for early detection when more treatment options are available and the potential exists for long-term benefits, including mortality reductions. Although additional data and longer-term follow-up are needed to substantiate such claims, our results support the continued use of DBT in routine breast cancer screening.
Author contributions: Guarantors of integrity of entire study, E.F.C., R.A.H.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, E.F.C., S.P.Z., K.E.K.; clinical studies, E.F.C., E.S.M., S.P.W., K.E.K., J.A.B., J.D.T.; experimental studies, E.S.M., M.D.S.; statistical analysis, R.A.H.; and manuscript editing, E.F.C., S.P.Z., E.S.M., S.P.W., K.E.K., J.A.B., M.D.S., R.A.H.
E.F.C. and M.D.S. supported by the National Cancer Institute grant U54CA163313.
- 1. . Comparison of digital mammography alone and digital mammography plus tomosynthesis in a population-based screening program. Radiology 2013;267(1):47–56. Link, Google Scholar
- 2. . Integration of 3D digital mammography with tomosynthesis for population breast cancer screening (STORM): A prospective comparison study. Lancet Oncol 2013;14(7):583–589. Crossref, Medline, Google Scholar
- 3. . Performance of one-view breast tomosynthesis as a stand-alone breast cancer screening modality: Results from the Malmö Breast Tomosynthesis Screening Trial, a population-based study. Eur Radiol 2016;26(1):184–190. Crossref, Medline, Google Scholar
- 4. . Digital Breast Tomosynthesis with Synthesized Two-Dimensional Images versus Full-Field Digital Mammography for Population Screening: Outcomes from the Verona Screening Program. Radiology 2018;287(1):37–46. Link, Google Scholar
- 5. . Digital Breast Tomosynthesis and Synthetic 2D Mammography versus Digital Mammography: Evaluation in a Population-based Screening Program. Radiology 2018;287(3):787–794. Link, Google Scholar
- 6. . Comparison of tomosynthesis plus digital mammography and digital mammography alone for breast cancer screening. Radiology 2013;269(3):694–700. Link, Google Scholar
- 7. . Implementation of breast tomosynthesis in a routine screening practice: An observational study. AJR Am J Roentgenol 2013;200(6):1401–1408. Crossref, Medline, Google Scholar
- 8. . Breast cancer screening using tomosynthesis in combination with digital mammography. JAMA 2014;311(24):2499–2507. Crossref, Medline, Google Scholar
- 9. . Screening outcomes following implementation of digital breast tomosynthesis in a general-population screening program. J Natl Cancer Inst 2014;106(11):dju316. Crossref, Medline, Google Scholar
- 10. . Early clinical experience with digital breast tomosynthesis for screening mammography. Radiology 2015;274(1):85–92. Link, Google Scholar
- 11. . Breast cancer screening using tomosynthesis in combination with digital mammography compared to digital mammography alone: A cohort study within the PROSPR consortium. Breast Cancer Res Treat 2016;156(1):109–116. Crossref, Medline, Google Scholar
- 12. . Breast Cancer Screening Using Tomosynthesis or Mammography: A Meta-analysis of Cancer Detection and Recall. J Natl Cancer Inst 2018;110(9):942–949. Crossref, Medline, Google Scholar
- 13. . Clinical implementation of digital breast tomosynthesis. Radiol Clin North Am 2014;52(3):499–518. Crossref, Medline, Google Scholar
- 14. . Strategies to Increase Cancer Detection: Review of True-Positive and False-Negative Results at Digital Breast Tomosynthesis Screening. RadioGraphics 2016;36(7):1954–1965. Link, Google Scholar
- 15. . Breast Cancer Conspicuity on Simultaneously Acquired Digital Mammographic Images versus Digital Breast Tomosynthesis Images. Radiology 2019;292(1):69–76. Link, Google Scholar
- 16. . Update of the Swedish two-county program of mammographic screening for breast cancer. Radiol Clin North Am 1992;30(1):187–210. Crossref, Medline, Google Scholar
- 17. . Effectiveness of Digital Breast Tomosynthesis Compared with Digital Mammography: Outcomes Analysis From 3 Years of Breast Cancer Screening. JAMA Oncol 2016;2(6):737–743. Crossref, Medline, Google Scholar
- 18. . Association of Digital Breast Tomosynthesis versus Digital Mammography with Cancer Detection and Recall Rates by Age and Breast Density. JAMA Oncol 2019;5(5):635–642. Crossref, Medline, Google Scholar
- 19. . Screening outcome for consecutive examinations with digital breast tomosynthesis versus standard digital mammography in a population-based screening program. Eur Radiol 2019;29(12):6991–6999. Crossref, Medline, Google Scholar
- 20. . Interval breast cancers in the ‘screening with tomosynthesis or standard mammography’ (STORM) population-based trial. Breast 2018;38:150–153. Crossref, Medline, Google Scholar
- 21. . Association of Screening and Treatment with Breast Cancer Mortality by Molecular Subtype in US Women, 2000-2012. JAMA 2018;319(2):154–164 [Published correction appears in JAMA 2018;319(7):724.]. Crossref, Medline, Google Scholar
- 22. . ECOG-ACRIN Cancer Research Group. http://ecog-acrin.org/tmist. Accessed July 7, 2019. Google Scholar
- 23. . Is Tomosynthesis the Future of Breast Cancer Screening? Radiology 2018;287(1):47–48. Link, Google Scholar
- 24. . Implementation of Synthesized Two-dimensional Mammography in a Population-based Digital Breast Tomosynthesis Screening Program. Radiology 2016;281(3):730–736. Link, Google Scholar
- 25. . Baseline Screening Mammography: Performance of Full-Field Digital Mammography Versus Digital Breast Tomosynthesis. AJR Am J Roentgenol 2015;205(5):1143–1148. Crossref, Medline, Google Scholar
- 26. . BI-RADS category 3 comparison: Probably benign category after recall from screening before and after implementation of digital breast tomosynthesis. Radiology 2017;285(3):778–787. Link, Google Scholar
- 27. . BI-RADS: Mammography. In: D’Orsi CJ, Mendelson EB, Ikeda DM, . Breast Imaging Reporting and Data System: ACR BI-RADS – Breast Imaging Atlas. 4th ed. Reston, Va: American College of Radiology, 2003. Google Scholar
- 28. . ACR BI-RADS Mammography. In: D’Orsi CJ, Sickles EA, Mendelson EB, . ACR BI-RADS Atlas, Breast Imaging Reporting and Data Systems. 5th ed. Reston, Va: American College of Radiology, 2013. Google Scholar
- 29. . Standardization and control for confounding in observational studies: A historical perspective. Stat Sci 2014;29(4):529–558. Crossref, Google Scholar
- 30. . One-view breast tomosynthesis versus two-view mammography in the Malmö Breast Tomosynthesis Screening Trial (MBTST): A prospective, population-based, diagnostic accuracy study. Lancet Oncol 2018;19(11):1493–1503. Crossref, Medline, Google Scholar
- 31. . Biologic Profiles of Invasive Breast Cancers Detected Only with Digital Breast Tomosynthesis. AJR Am J Roentgenol 2017;209(6):1411–1418. Crossref, Medline, Google Scholar
- 32. . Breast cancers found with digital breast tomosynthesis: A comparison of pathology and histologic grade. Breast J 2016;22(6):651–656. Crossref, Medline, Google Scholar
- 33. . Performance of breast cancer screening using digital breast tomosynthesis: Results from the prospective population-based Oslo Tomosynthesis Screening Trial. Breast Cancer Res Treat 2018;169(3):489–496. Crossref, Medline, Google Scholar
- 34. . Breast cancer characteristics associated with 2D digital mammography versus digital breast tomosynthesis for screening-detected and interval cancers. Radiology 2018;287(1):49–57. Link, Google Scholar
- 35. . Comparison of breast cancers detected in the Verona screening program following transition to digital breast tomosynthesis screening with cancers detected at digital mammography screening. Breast Cancer Res Treat 2018;170(2):391–397. Crossref, Medline, Google Scholar
- 36. . The Coming of Age of Breast Tomosynthesis in Screening. Radiology 2019;291(1):31–33. Link, Google Scholar
- 37. . Overdiagnosis in mammography screening: A 45-year journey from shadowy idea to acknowledged reality. BMJ 2015;350:h867. Crossref, Medline, Google Scholar
Article HistoryReceived: Aug 9 2019
Revision requested: Sept 24 2019
Revision received: Nov 27 2019
Accepted: Dec 12 2019
Published online: Mar 10 2020
Published in print: May 2020