Implementation of Breast Tomosynthesis in a Routine Screening Practice: An Observational Study

Transcription

1 Women s Imaging Original Research Rose et al. Tomosynthesis in Routine Screening Women s Imaging Original Research Stephen L. Rose 1 Andra L. Tidwell Louis J. Bujnoch Anne C. Kushwaha Amy S. Nordmann Russell Sexton, Jr. Rose SL, Tidwell AL, Bujnoch LJ, Kushwaha AC, Nordmann AS, Sexton R Jr Keywords: breast cancer, cancer detection, mammography, recall rate, tomosynthesis DOI: /AJR Received July 26, 2012; accepted after revision November 20, S. L. Rose is a speaker and consultant for Hologic. L. J. Bujnoch is a Hologic shareholder. A. L. Tidwell is a consultant to Rose Imaging Specialists. This work was supported in part by Hologic. Additional support was provided by the Karen T. Stall Research and Breast Institute. 1 All authors: TOPS Comprehensive Breast Center, Red Oak Dr, Houston, TX Address correspondence to S. L. Rose (slrosemd@hbimaging.onmicrosoft.com). AJR 2013; 200: X/13/ American Roentgen Ray Society Implementation of Breast Tomosynthesis in a Routine Screening Practice: An Observational Study OBJECTIVE. Digital mammography combined with tomosynthesis is gaining clinical acceptance, but data are limited that show its impact in the clinical environment. We assessed the changes in performance measures, if any, after the introduction of tomosynthesis systems into our clinical practice. MATERIALS AND METHODS. In this observational study, we used verified practice- and outcome-related databases to compute and compare recall rates, biopsy rates, cancer detection rates, and positive predictive values for six radiologists who interpreted screening mammography studies without (n = 13,856) and with (n = 9499) the use of tomosynthesis. Two-sided analyses (significance declared at p < 0.05) accounting for reader variability, age of participants, and whether the examination in question was a baseline were performed. RESULTS. For the group as a whole, the introduction and routine use of tomosynthesis resulted in significant observed changes in recall rates from 8.7% to 5.5% (p < 0.001), nonsignificant changes in biopsy rates from 15.2 to 13.5 per 1000 screenings (p = 0.59), and cancer detection rates from 4.0 to 5.4 per 1000 screenings (p = 0.18). The invasive cancer detection rate increased from 2.8 to 4.3 per 1000 screening examinations (p = 0.07). The positive predictive value for recalls increased from 4.7% to 10.1% (p < 0.001). CONCLUSION. The introduction of breast tomosynthesis into our practice was associated with a significant reduction in recall rates and a simultaneous increase in breast cancer detection rates. A n increasing body of evidence suggests that early detection of breast cancer through periodic mammography screening reduces the morbidity and mortality associated with this disease [1 7]. Mammography is accepted worldwide as a screening tool, and the number of mammography procedures performed is large and continues to increase [8]. However, mammography screening has a relatively low cancer detection rate, particularly after the first 2 years of screening [9]. Variability in the performance levels among radiologists who read and interpret mammograms is large and in part is attributed to the difficulty in distinguishing suspicious regions from the surrounding or overlapping tissue [10 15]. In recent years, a major effort has been expended to develop new approaches to breast imaging, one of which is the use of tomosynthesis that enables the reconstruction of cross-sectional images aimed to assist radiologists with the interpretation process [16 18]. The goal of this approach is to improve performance by lowering recall rates while increasing sensitivity. The Food and Drug Administration (FDA) recently approved the first tomosynthesis system for clinical use. The initial FDA approval process for these systems included retrospective interpretations of select groups of cases in a laboratory environment under conditions that may be different from prospective clinical readings [19]. However, there are only limited data on the impact of tomosynthesis when used prospectively in a clinical screening environment (Skaane P et al., presented at the 2011 annual meeting of the Radiological Society of North America). In this single-site study, we used procedure outcome-related databases to evaluate the recall, biopsy, and cancer detection rates and computed positive predictive values (PPVs) in our clinical breast imaging practice before and after the introduction of tomosynthesis for routine screening mammography. AJR:200, June

2 Rose et al. Materials and Methods Subjects This study compares performance measures from screening mammography at a multisite communitybased comprehensive breast center. The analysis compares the outcomes of screening mammography studies interpreted in 2010 before the introduction of tomosynthesis to results acquired after the introduction of tomosynthesis during the period of May 2011 to the end of January 2012 that were interpreted by radiologists who read at least 500 mammography studies during each period. The time period for imaging with tomosynthesis was chosen to allow evaluation of performance measures after approximately 10,000 screening examinations. Figure 1 shows the numbers of cases interpreted by these radiologists with digital mammography alone (n = 13,856) and digital mammography plus tomosynthesis (n = 9499). This dataset was chosen to allow comparison of both individual- and practice-based performance measures. The breast center has a relatively stable screening population, with new patients representing approximately 23% and 17% of screening examinations in the digital mammography alone and digital mammography plus tomosynthesis time periods, respectively. We note that during the digital mammography plus tomosynthesis time period, the percentage of patients agreeing to have the combined digital mammography plus tomosynthesis procedure was 88%. For the digital mammography plus tomosynthesis time period, only those cases imaged with digital mammography plus tomosynthesis are included in this analysis. Cases were included if the patient was asymptomatic and 18 years of age and elected (self-selected) to have tomosynthesis as the standard screening examination. Approval with waiver of written informed consent was obtained from Western Institutional Review Board. Materials and Methods The sources for our analysis were databases (version rev 4.0, PenRad) that contained information on procedure scheduling, procedure completion, radiology reporting, and procedure-related outcomes determined from relevant biopsy and surgical pathology reports. The databases allowed determination of the imaging method (digital mammography alone or digital mammography plus tomosynthesis) that was used. These databases were assembled from the original reports for quality assurance purposes as required by the Mammography Quality Standards Act (MQSA), and for other reasons [20]. Validation of the database information was performed by reviewing imaging and pathology reports. All recall, biopsy, and cancer cases included in the analysis were verified by clinical record review and determined to be the result of a follow-up of findings of a screening mammography. Digital mammography 18,202 Screening examinations (4346 excluded; low-volume readers) 13,856 Screening examinations 1208 (87/1000 examinations) BI-RADS (16/1000 examinations) Biopsies recommended 211 (15/1000 examinations) Biopsies performed 56 Cancers (4.0/1000 examinations) 39 Invasive (2.8/1000 examinations) 17 In situ (1.2/1000 examinations) Fig 1. Flowchart shows study process. Digital mammography plus tomosynthesis 10,878 Screening examinations (1379 excluded; low-volume readers) 9499 Screening examinations 518 (55/1000 examinations) BI-RADS (14/1000 examinations) Biopsies recommended 128 (13/1000 examinations) Biopsies performed 51 Cancers (5.4/1000 examinations) 41 Invasive (4.3/1000 examinations) 10 In situ (1.1/1000 examinations) Before the introduction of tomosynthesis, images were obtained with a digital mammography system (Selenia, Hologic). After the introduction of tomosynthesis, the digital mammography and tomosynthesis images for each view were obtained under a single compression (Dimensions, Hologic). The radiation doses used for digital mammography and tomosynthesis are approximately the same. Thus, the radiation dose when digital mammography plus tomosynthesis images were obtained was approximately twice that used for digital mammography alone. The radiation dose for the digital mammography plus tomosynthesis combined modality is lower than the MQSA limit for a single screening mammography examination. During the two study periods, 13 radiologists interpreted a total of 18,202 digital mammography alone screening examinations, whereas 11 radiologists interpreted 10,878 digital mammography plus tomosynthesis screening examinations. However, because several of the radiologists had interpreted few mammograms during one or both periods in question, no satisfactory reference data were available for these radiologists for both periods. Therefore, we chose to focus the primary analysis on the six radiologists who interpreted at least 500 cases during each of the study time periods before and after the introduction of tomosynthesis. These radiologists are all MQSA-qualified and had an average of 12 years (range, 2 32 years) experience after training in mammography interpretation. The six radiologists interpret an average of more than 6000 screening examinations per year (range 4000 to 10,000). Recall Rates Recall rates for each radiologist and for the group of six radiologists were computed directly from mammographic interpretation records (BI-RADS rating 0) [21]. We excluded recommendations for recall due to technical reasons, such as image artifacts. A total of six and two cases recalled during the two periods, respectively, were given a diagnostic BI-RADS 3 rating as a result of the diagnostic workup that followed the screening examination. These patients all underwent an additional mammography examination before any management decisions on whether to biopsy and were considered negative (or false-positive recalls) for the purpose of all analyses. Biopsy and Cancer Detection Rates All breast biopsies that resulted from a recalled case were included in this analysis. If the biopsy was positive for cancer, the cancer detection was considered a screen-detected cancer. Pathology reports were used to verify all biopsy results. Only the interpreter of the original screening mammography that eventually led to the diagnosis of breast cancer was credited with the finding. Cases were excluded from the cancer detection rate analysis if the most recent screening mammography before biopsy had been performed more than 180 days earlier or if the original interpreter had not recommended a recall (i.e., false-negative results) AJR:200, June 2013

3 Tomosynthesis in Routine Screening TABLE 1: Comparison of Recalls, Biopsies, Cancer Detection, and Positive Predictive Value (PPV) for Digital Mammography Alone and Digital Mammography Plus Tomosynthesis Recalled Cases Lost to Follow-Up PPV1 a (%) PPV3 b (%) Invasive Cancers per 1000 Cases Invasive Cancers Cancers per 1000 Cases Biopsies per 1000 Cases Cancers Recalls per 1000 Cases Biopsies Recall Rate (%) Recalled Cases Reader Cases Read Digital mammography alone All c 13, Digital mammography plus tomosynthesis All c a PPV values corrected for cases lost to follow-up. b Positive biopsy rate. c All represents pooled nonadjusted data. Statistical Analysis The main outcome measures included the recall rate, cancer detection rate, and invasive cancer detection rate. For each of the main outcome measures, a generalized linear mixed model (PROC GLIMMIX procedure, SAS Institute) was used to analyze the difference between digital mammography alone and digital mammography plus tomosynthesis. The model accounted for age, new patients (first time imaged at the breast center), and each radiologist s individual performance levels in a random-reader model. The correlation between interpretations by the same radiologist was accounted for by specifying the radiologist as a G-side random effect. Data tabulation, statistical modeling, data analyses, and inference were performed using statistical software (version 9.3, SAS Institute). In a secondary analysis using the same generalized mixed model that was used for the primary analysis, we assessed whether the exclusion of the lowvolume radiologists (and the examinations they interpreted during one or both periods in question) could affect any of the study conclusions. To assess the similarity of the types of invasive cancer and nodal status, a chi-square test was used. The similarity of invasive cancer size was assessed using a Wilcoxon rank sum test. All analyses were two-sided with statistical significance declared at p < Results Table 1 summarizes results for the six radiologists who interpreted at least 500 screening mammography studies without and with the use of tomosynthesis. The table includes results for recall, biopsy, cancer detection, and PPV. Recall rates for the group of six radiologists were 8.7% for mammography interpreted without tomosynthesis and 5.5% for mammography interpreted with tomosynthesis (p < 0.001). Four of the six radiologists (readers 2, 4, 5, and 6) significantly reduced their individual recall rates (p < 0.05). Table 1 also summarizes biopsy recommendation by radiologist and verified findings. There were 211 and 128 biopsies, 15.2/1000 and 13.5/1000 screenings (p = 0.59), performed as a result of diagnostic workup of screening-suspected abnormalities with digital mammography alone and digital mammography plus tomosynthesis, respectively. Breast cancer detection rates for the group were 4.04 per 1000 screening examinations for mammography interpreted without tomosynthesis and 5.37 per 1000 screening examinations for mammography interpreted with tomosynthesis (p = 0.18). We note that, during the tomosynthesis period, there were a total of 1408 digital mammography alone examinations that resulted in the detection of six cancers or 4.3/1000 AJR:200, June

4 Rose et al. TABLE 2: Comparison of Recall Rates for Digital Mammography Alone and Digital Mammography Plus Tomosynthesis for Different Breast Density BI-RADS Ratings Breast Digital Mammography Alone Digital Mammography Plus Tomosynthesis Density (BI-RADS) No. of Cases Cases (%) No. of Recalls Recall Rate (%) No. of Cases Cases (%) No. of Recalls Recall Rate (%) Relative Change (%) NA All 13, Note NA indicates not available. Dash indicates not applicable. TABLE 3: Comparison of Recall Rates for Digital Mammography Alone and Digital Mammography Plus Tomosynthesis by Subject Age Digital Mammography Alone Digital Mammography Plus Tomosynthesis Age (y) No. of Cases Cases (%) No. of Recalls Recall Rate (%) No. of Cases Cases (%) No. of Recalls Recall Rate (%) Relative Change (%) < > Total 13, screening examinations, suggesting that the digital mammography only practice remained quite stable. Invasive cancer detection rates increased from 2.81 to 4.32 per 1000 screening examinations with the addition of tomosynthesis (p = 0.07). The PPV for recalls increased from 4.7% with digital mammography alone to 10.1% for digital mammography plus tomosynthesis (p < 0.001). The PPV for biopsies increased from 26.5% for digital mammography alone to 39.8% for digital mammography plus tomosynthesis (p = 0.06). Recall rates are shown in Table 2 as a function of subjectively rated breast density (BI- RADS density score) by the interpreting radiologist. Recall rates decreased for all breast density categories, and the improvement with the addition of tomosynthesis was similar for each BI-RADS breast density category and ranged from 32.4% to 41.0%. Recall rates as a function of age are shown in Table 3. The recall rate reduction with the addition of tomosynthesis was greater than 30% for all age groups. The distribution of ages in the two time periods was similar. The mean age of the screened population was 53.8 years for women screened with digital mammography alone and 54.5 years for women screened with digital mammography plus tomosynthesis. Figure 2 shows an example of a cancer detected on digital mammography plus tomosynthesis that was more clearly visible on the tomosynthesis images. Figure 3 demonstrates the potential of tomosynthesis to reduce recall rate and shows a benign case that was retrospectively determined to be a recall by three different radiologists when using digital mammography alone but was actually rated a benign (no recall) during the clinical interpretation using digital mammography plus tomosynthesis (Fig. 3). A description of the screening-detected cancers is shown in Table 4 and includes grade, size, and stage for invasive and in situ cancers. The types of invasive cancer, cancer size from imaging, and node status shown by the N stage were similar for those detected with each of the two imaging methods (p = 0.69; p = 0.91; p = 0.84, respectively). The cancer grade distribution was also similar, but more grade 3 cancers were detected using digital mammography alone and more grade 2 cancers were detected using digital mammography plus tomosynthesis. However, if we compare the number of higher-grade invasive cancers (grade 2 or 3) detected per 1000 screening examinations, there were 2.6 high-grade invasive cancers detected per 1000 screening examinations with tomosynthesis compared with 1.9 without tomosynthesis. The radiologic signs reported by the interpreting radiologist of the screening study showed large differences. In particular, the cancers detected by digital mammography plus tomosynthesis were more likely to be described as spiculated masses or distortions, compared with cancers detected by digital mammography alone. For digital mammography alone, cancers were more likely to be described as asymmetries, densities, or indistinct masses. These results suggest an improved visualization of the abnormalities in terms of shape and margins with the addition of tomosynthesis. The inclusion of the eight low-volume radiologists and the cases they interpreted during one or both periods in question did not affect any of the study conclusions. Recall rates in the subgroup of women undergoing the first screening examination were 13.9% and 9.6% during the digital mammography alone and digital mammography plus tomosynthesis periods, respectively, and cancer detection rates were 4.1/1000 and 7.7/1000 in these groups, respectively. The difference in the fraction of baseline examinations during the two periods being compared was accounted for in the model, and it did not change any of the study conclusions. Discussion Our results suggest that the introduction of tomosynthesis imaging into our screening practice was associated with statistically significant decreases in recall rates, a statistically nonsignificant decrease in biopsy rates, and large gains in PPV. A substantial, albeit not 1404 AJR:200, June 2013

5 Tomosynthesis in Routine Screening statistically significant, increase in breast cancer detection rates was observed. Recall rates were reduced for all radiologists in the study. The magnitudes of recall rate reductions we observed were in general agreement with the ranges reported in retrospective studies [22, C A 23]. The improvements we observed may be attributed to improved detectability as well as clearer definitions of shape and margins leading to more accurate interpretations. Our data were not adjusted for any learning effect; the interpretations without tomosynthesis occurred D Fig. 2 Comparison of digital mammography and tomosynthesis images in 42-year-old woman with 10- mm invasive ductal carcinoma in left breast. A D, Craniocaudal images for digital mammography (A) and tomosynthesis (B) and mediolateral oblique images for digital mammography (C) and tomosynthesis (D). Tomosynthesis images show improved visibility of cancer (arrow) and cancer margins compared with digital mammography. chronologically before those with tomosynthesis. Also, we did not account for any effect that may result from the continuous effort to improve performance (e.g., assessments of false-negative findings or continuous efforts to reduce recall rates). However, because experience with tomosynthesis to date is generally limited, we believe our observations are likely to be conservative and represent a lower level of possible improvement in performance. A comprehensive assessment as to whether the improvement we observed was limited by lack of substantial experience with tomosynthesis is beyond the scope of this investigation. At the same time, we note that the interradiologist variability in recall and attributable cancer detection rates were well within the range of published observer variability values [24 26]. Despite the consistent improvements in all performance measures of interest, the relatively small sample size (in particular in terms of the total number of detected cancers during each period) combined with the fact that detection rates of noninvasive cancers decreased during the digital mammography plus tomosynthesis period resulted in our inability to generate inferences about sta- B AJR:200, June

6 Rose et al. A E B Fig year-old woman. Examination was retrospectively reviewed by three radiologists who stated they would have recalled woman without availability of tomosynthesis, and she was actually not recalled for diagnostic workup during clinical interpretation. She remained negative during subsequent examination 1 year later. Digital mammography images (A and B) demonstrate a suspicious area as shown by the arrow (A). Tomosynthesis images (C E) show three widely spaced images from the tomosynthesis image set. Tomosynthesis images show that structures at different levels in breast can summate to create suspicious region on digital mammography image that may be identified as negative or superimposed tissue on tomosynthesis images. tistically significant improvements in overall cancer detection rates. As expected, the improvement in performance levels was larger in terms of increasing cancer detection and in decreasing recall rates in the group of women visiting our practice for the first time. We note that we observed a statistically insignificant decrease in the detection of ductal carcinoma in situ (DCIS). The underlying reason for this finding is not clear, but it could easily be attributed to the small number of actually detected DCIS in both periods, 17 (1.23/1000) and 10 (1.05/1000), respectively. For a comparable DCIS detection rate, the C expected number of DCIS would have been 12 rather than the observed 10. This difference is too small for any reliable inferences. It could be the result of lower prevalence during the period with tomosynthesis or the number of actual DCIS cases seen by the six radiologists included in this study during the peri- D 1406 AJR:200, June 2013

7 Tomosynthesis in Routine Screening od with tomosynthesis or a combination of both. It could also possibly be explained as follows: When provided both digital mammography and tomosynthesis imaging, radiologists TABLE 4: Staging and Description of Screening-Detected Cancers: Digital Mammography Alone and Digital Mammography Plus Tomosynthesis Description Finding Digital Mammography Alone Digital Mammography Plus Tomosynthesis Invasive cancer 39 (2.81) 41 (4.32) Histology IDC ILC 3 6 ILC and IDC 4 1 Mucinous 1 0 Grade NA 1 0 N stage Nx 2 1 N N1 2 6 N2 2 0 Imaging size (mm) Mean Median Radiologic sign Spiculated mass 1 14 Architectural distortion 1 4 Lobulated mass 0 2 Asymmetry, density, indistinct mass Calcifications 4 3 DCIS only 17 (1.22) 10 (1.05) Grade Low 3 0 Medium 6 3 High 7 7 NA 1 0 All cancer cases 56 (4.04) 51 (5.37) T stage No stage 4 1 Tis T1mic 1 1 T1a 4 3 T1b 10 6 T1c T T3 2 1 Breast density Note Data in parentheses are per IDC = invasive ductal carcinoma, ILC = invasive lobular carcinoma, DCIS = ductal carcinoma in situ, NA = not available. tend to focus more on the latter (definitely spending most of the interpretation time on the tomosynthesis image set) and, as a result, are perhaps not as meticulous in terms of identifying calcifications or diagnosing clusters. Because digital mammography is available to the interpreter, this decrease is likely to be addressed with greater familiarity regarding the advantages and disadvantages of the different imaging presentations: digital mammography versus tomosynthesis. In addition, the current acquisition protocol for digital mammography plus tomosynthesis acquires the digital mammography images after the completion of the 15 low-dose tomosynthesis projection images (dataset) that are used in the reconstruction of the tomosynthesis examination. As a result, the digital mammography images may be more prone to patient motion than digital mammography images acquired without tomosynthesis imaging. This issue is likely to be addressed in future tomosynthesis systems. One could argue that some or all of the reduction in recall rates and improvement in detection rates we observed for the high-volume radiologists may be attributable simply to the introduction of an additional modality (tomosynthesis) and therefore may not persist over time. However, the magnitude of the simultaneous changes we observed for both recall and cancer detection rates, particularly invasive cancers, suggests that the underlying technology is indeed aiding the radiologists. Interestingly, unlike the authors of several retrospective studies, we observed simultaneous improvements in all performance measures and, in particular, in the detection of invasive cancers. These improvements were distributed over all breast density categories. The increased PPV for recalls and biopsies suggests that tomosynthesis may limit the harms of mammography [6, 27]. A large reduction in recalls was seen for women in all age groups, including women under 50 years old. Our study has several limitations. First, we recognize that this was an observational study that could be affected by a changing environment or changes in adherence to practice guidelines. However, we do not believe this was a large factor because we chose to review and analyze consecutive periods during which the primary change in practice was the introduction of tomosynthesis. Second, with the introduction of a new technology, particularly case selection with the new technology in this study, we cannot rule out self-selection bias because women elected whether they would undergo the tomosynthesis procedure. Third, assessing absolute sensitivity, specificity, and outcome measures in terms of person-years gained or lost and morbidity or mortality is clearly beyond the scope of this short-duration observational study. How- AJR:200, June

8 Rose et al. ever, this type of a study adds important observations in terms of changes in the number and types of cancers detected when tomosynthesis is incorporated into screening practice. If one believes that earlier detection of the additional cancers identified using digital mammography plus tomosynthesis is important, in particular the earlier detection of invasive cancers, the observations made as a result of this study may ultimately prove to be important with regard to other summary performance measures directly related to the cost-benefit of screening. Fourth, there is a potential risk of the additional radiation on the basis of commonly used models that use a linear extrapolation from the risk of high radiation dose, but this risk is extremely small when compared with the potential benefits shown in this study. Substantial investigative work is under way to reduce the radiation dose to the breast during tomosynthesis procedures to comparable levels of digital mammography; however, this topic is beyond the scope of this article [28]. Finally, the results presented here are from a single practice and therefore may or may not be generalized to other practices. Despite these limitations and the large interradiologist variability in performance measures, we observed consistent significant improvements in recall rates and notable, albeit not statistically significant, improvements in cancer detection related performance measures that applied to individuals as well as the group of radiologists as a whole, and we believe that similar improvements are possible, if not likely, in other practices. Clearly, clinical validation of these findings as well as the reproducibility of our own results in other studies when tomosynthesis is used during interpretation will require additional supporting information from other studies; however, preliminary results suggest that indeed this is likely to be the case (Haas B et al. and Butler R et al., presented at the 2012 annual meeting of the Radiological Society of North America). In conclusion, implementation of tomosynthesis in our screening practice resulted in a consistent significant improvement in performance. The PPV of recalls doubled with the addition of tomosynthesis. Improved performance resulted in significant decreases in recall rates concurrent with increases in cancer detection rates, in particular those of invasive cancers. script. We also acknowledge PenRad for support with data collection. References 1. Tabár L, Vitak B, Chen HH, Yen MF, Duffy SW, Smith RA. Beyond randomized controlled trials: organized mammographic screening substantially reduces breast carcinoma mortality. Cancer 2001; 91: Tabár L, Vitak B, Chen TH, et al. Swedish twocounty trial: impact of mammographic screening on breast cancer mortality during 3 decades. Radiology 2011; 260: Kerlikowske K, Grady D, Rubin SM, Sandrock C, Ernster VL. Efficacy of screening mammography: a meta-analysis. JAMA 1995; 273: Kerlikowske K, Grady D, Ernster V. Benefit of mammography screening in women ages years: current evidence from randomized controlled trials. Cancer 1995; 76: Lehtimäki T, Lundin M, Linder N, et al. Long-term prognosis of breast cancer detected by mammography screening or other methods. Breast Cancer Res 2011; 13:R Humphrey LL, Helfand M, Chan BK, Woolf SH. Breast cancer screening: a summary of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med 2002; 137: Duffy SW, Tabár L, Chen HH, et al. The impact of organized mammography service screening on breast carcinoma mortality in seven Swedish counties. Cancer 2002; 95: Youlden DR, Cramb SM, Dunn NA, Muller JM, Pyke CM, Baade PD. The descriptive epidemiology of female breast cancer: an international comparison of screening, incidence, survival and mortality. Cancer Epidemiol 2012; 36: Ng EH, Ng FC, Tan PH, et al. Results of intermediate measures from a population-based, randomized trial of mammographic screening prevalence and detection of breast carcinoma among Asian women: the Singapore Breast Screening Project. Cancer 1998; 82: Sickles EA, Wolverton DE, Dee KE. Performance parameters for screening and diagnostic mammography: specialist and general radiologists. Radiology 2002; 224: Jiang Y, Miglioretti DL, Metz CE, Schmidt RA. Breast cancer detection rate: designing imaging trials to demonstrate improvements. Radiology 2007; 243: Elmore JG, Wells CK, Lee CH, Howard DH, Fein- 14. Bird RE, Wallace TW, Yankaskas BC. Analysis of cancers missed at screening mammography. Radiology 1992; 184: Carney PA, Miglioretti DL, Yankaskas BC, et al. Individual and combined effects of age, breast density, and hormone replacement therapy use on the accuracy of screening mammography. Ann Intern Med 2003; 138: Niklason LT, Christian BT, Niklason LE, et al. Digital tomosynthesis in breast imaging. Radiology 1997; 205: Rafferty EA. Digital mammography: novel applications. Radiol Clin North Am 2007; 45: Baker JA, Lo JY. Breast tomosynthesis: state-ofthe-art and review of the literature. Acad Radiol 2011; 18: U.S. Food and Drug Administration website. Summary of safety and effectiveness data. gov/downloads/advisorycommittees/committees- MeetingMaterials/MedicalDevices/MedicalDevicesAdvisoryCommittee/RadiologicalDevicesPanel/ UCM pdf. Published Accessed February 7, U.S. Food and Drug Administration website. Radiation-emitting products: Mammography Quality Standards Act Regulations. EmittingProducts/MammographyQualityStandards- ActandProgram/Regulations/default.htm. Updated December 4, Accessed February 7, American College of Radiology. Breast imaging reporting and data system atlas. Reston, VA: American College of Radiology, Gur D, Abrams GS, Chough DM, et al. Digital breast tomosynthesis: observer performance study. AJR 2009; 193: Poplack SP, Tosteson TD, Kogel CA, Nagy HM. Digital breast tomosynthesis: initial experience in 98 women with abnormal digital screening mammography. AJR 2007; 189: Gur D, Sumkin JH, Hardesty LA, et al. Recall and detection rates in screening mammography. Cancer 2004; 100: Beam CA, Layde PM, Sullivan DC. Variability in the interpretation of screening mammograms by US radiologists. Arch Intern Med 1996; 156: Skaane P, Diekmann F, Balleyguier C, et al. Observer variability in screen-film mammography versus full-field digital mammography with softcopy reading. Eur Radiol 2008; 18: U.S. Preventive Services Task Force. Screening for Breast Cancer: U.S. Preventive Services Task Force recommendation statement. Ann Intern Acknowledgments The authors thank Kathleen Willison for her participation in the data collection and analysis and Loren Niklason for providing editorial contributions and review of the manu- stein AR. Variability in radiologists interpretations of mammograms. N Engl J Med 1994; 331: Beam CA, Conant EF, Sickles EA, Weinstein SP. Evaluation of proscriptive health care policy implementation in screening mammography. Radiology 2003; 229: Med 2009; 151: Gur D, Zuley ML, Anello MI, et al. Dose reduction in digital breast tomosynthesis (DBT) screening using synthetically reconstructed projection images: an observer performance study. Acad Radiol 2012; 19: AJR:200, June 2013