Reducing False-Positive Screening MRI Rate in Women with Extremely Dense Breasts Using Prediction Models Based on Data from the DENSE Trial
Abstract
Background
High breast density increases breast cancer risk and lowers mammographic sensitivity. Supplemental MRI screening improves cancer detection but increases the number of false-positive screenings. Thus, methods to distinguish true-positive MRI screening results from false-positive ones are needed.
Purpose
To build prediction models based on clinical characteristics and MRI findings to reduce the rate of false-positive screening MRI findings in women with extremely dense breasts.
Materials and Methods
Clinical characteristics and MRI findings in Dutch breast cancer screening participants (age range, 50–75 years) with positive first-round MRI screening results (Breast Imaging Reporting and Data System 3, 4, or 5) after a normal screening mammography with extremely dense breasts (Volpara density category 4) were prospectively collected within the randomized controlled Dense Tissue and Early Breast Neoplasm Screening (DENSE) trial from December 2011 through November 2015. In this secondary analysis, prediction models were built using multivariable logistic regression analysis to distinguish true-positive MRI screening findings from false-positive ones.
Results
Among 454 women (median age, 52 years; interquartile range, 50–57 years) with a positive MRI result in a first supplemental MRI screening round, 79 were diagnosed with breast cancer (true-positive findings), and 375 had false-positive MRI results. The full prediction model (area under the receiver operating characteristics curve [AUC], 0.88; 95% CI: 0.84, 0.92), based on all collected clinical characteristics and MRI findings, could have prevented 45.5% (95% CI: 39.6, 51.5) of false-positive recalls and 21.3% (95% CI: 15.7, 28.3) of benign biopsies without missing any cancers. The model solely based on readily available MRI findings and age had a comparable performance (AUC, 0.84; 95% CI: 0.79, 0.88; P = .15) and could have prevented 35.5% (95% CI: 30.4, 41.1) of false-positive MRI screening results and 13.0% (95% CI: 8.8, 18.6) of benign biopsies.
Conclusion
Prediction models based on clinical characteristics and MRI findings may be useful to reduce the false-positive first-round screening MRI rate and benign biopsy rate in women with extremely dense breasts.
Clinical trial registration no. NCT01315015
© RSNA, 2021
Online supplemental material is available for this article.
See also the editorial by Imbriaco in this issue.
Summary
Prediction models based on clinical characteristics and MRI findings may be useful to reduce the false-positive rate of first-round supplemental screening MRI in women with extremely dense breasts.
Key Results
■ Using prospectively collected data of 454 women from the Dense Tissue and Early Breast Neoplasm Screening trial, a prediction model based on clinical characteristics and MRI findings (area under the receiver operating characteristics curve [AUC], 0.88) could prevent 45.5% of false-positive recalls and 21.3% of benign biopsies after first-round supplemental MRI screening in women with extremely dense breast tissue without missing any cancers.
■ A model using data from MRI findings plus age (AUC, 0.84) could help prevent 35.5% of false-positive recalls and 13.0% of benign biopsies without missing any cancers.
Introduction
Population-based mammographic screening has proven effective in terms of mortality reduction due to breast cancer detection and treatment at an early stage (1,2). However, in women with high breast density, the sensitivity of mammography is markedly reduced because of the masking effect of the fibroglandular tissue (3–6). This is particularly important, as the breast cancer risk in women with extremely dense breasts is twice as high as that in women with average breast density (7).
MRI is the most sensitive technique with which to screen women at high risk (8–14). More recently, MRI has also been considered as a screening tool in women at average risk with dense breasts (15–19). The Dense Tissue and Early Breast Neoplasm Screening (DENSE) trial was designed to evaluate the effectiveness of screening with mammography and MRI compared with mammography alone in Dutch breast cancer screening participants (age range, 50–75 years) with extremely dense breasts (20). A significant difference in interval cancers was found in the intention-to-screen analysis: the rate was 5.0 of 1000 screenings in the mammography-alone group versus 2.5 of 1000 screenings in the group invited for supplemental MRI (P < .001). This rate in the MRI-invitation group consisted of an interval cancer rate of 0.8 of 1000 screenings in the women who underwent MRI (59% of invited women) and 4.9 of 1000 in those who did not accept the MRI invitation. This study therefore provided strong evidence that supplemental MRI screening in women with extremely dense breasts improves detection of clinically relevant cancers (21). In contrast with its high sensitivity, the moderate specificity of breast MRI can result in false-positive findings that necessitate further imaging, tissue sampling, or additional follow-up examinations (16–18,21) (Fig 1). The recall rate among first-round DENSE participants was 94.9 of 1000 screenings, with a false-positive rate (FPR) of 79.8 of 1000. Other supplemental screening MRI studies have reported similar FPRs ranging from 52 of 1000 to 97 of 1000 (16,18). Recall and the subsequent diagnostic work-up often are associated with patient anxiety, increased health care costs, and sometimes biopsy-related morbidity (22–24). High recall rates may complicate implementation of supplemental MRI screening on a large scale, as sufficient capacity for timely diagnostic work-up is required. Reduction of the FPR is an important issue when considering the use of breast MRI as a screening tool (25). Therefore, the purpose of this study was to explore the potential of prediction models, based on clinical characteristics and MRI findings, to reduce the FPR.
Figure 1: Example of a false-positive MRI. In a 59-year-old participant, first-round screening MRI showed an 8-mm irregular mass in the right breast. The lesion shows rapid heterogeneous type 2 enhancement, low T2 signal intensity, and no evident diffusion restriction. Histopathologic findings showed sclerotic stroma and apocrine metaplasia.
Materials and Methods
Study Sample
The DENSE trial design was reported elsewhere (20,21). The study protocol is available through ClinicalTrials.gov (NCT01315015). DENSE is a multicenter randomized controlled trial that evaluates the effect of supplemental MRI screening in women with extremely dense breasts. Participants in the biennial Dutch population–based screening program (aged 50–75 years) with negative mammographic findings and extremely dense breasts (Volpara density category 4) were eligible for randomization. A total of 40 373 women were randomized for supplemental MRI screening for three consecutive rounds (intervention arm) or mammographic screening only (control arm). Results regarding the primary outcome, the difference in interval cancer rate during a 2-year screening period, were reported elsewhere (21) as well as studies focusing on other research questions (26–29). Ethical approval for the DENSE trial was obtained from the Dutch Minister of Health, Welfare, and Sport on November 11, 2011 (2011/19 WBO, the Hague, the Netherlands). All study participants provided written informed consent.
From December 2011 to November 2015, 4783 DENSE participants underwent first-round screening MRI. In our present analysis, all 454 participants recalled for diagnostic work-up after a Breast Imaging Reporting and Data System (BI-RADS) category 3, 4, or 5 result were included (Fig 2). Study data are available from the corresponding author on reasonable request.
Figure 2: Flowchart shows enrollment in the Dense Tissue and Early Breast Neoplasm Screening (DENSE) trial and selection of participants for the present analysis. Breast cancer screening participants (aged 50–75 years) with extremely dense breast tissue were assigned in a 1:4 ratio to a group that was invited to undergo supplemental MRI screening or to a group that underwent mammographic screening only. Approximately 59% of the participants in the MRI-invitation group actually underwent MRI. A total of 454 participants had a positive MRI screening result and constituted the study sample for the present analysis (dashed box).
Clinical Characteristics and MRI Findings
All participants were asked to complete a questionnaire on health status and breast cancer risk factors at the time of recruitment, including personal and family history of breast cancer, age at menarche, parity, age at first live birth, breastfeeding status, body mass index, hormone use after age 50 years (hormonal replacement therapy and contraceptive use), and menopausal status.
MRI Protocol
All participants underwent 3.0-T MRI (Philips Achieva or Siemens Ingenia) using a dedicated phased-array bilateral breast coil. The full MRI protocol is described in detail elsewhere (21). A gadolinium-based contrast agent (Gadovist, Bayer) was injected (1 mL/sec) for a total dose of 0.1 mmol per kilogram of body weight.
Image Analysis
MRI scans were read by dedicated breast radiologists (including, among others, W.B.V., R.M.P., R.M.M., M.B.I.L., K.M.D., and J.V.) with 8.5 to more than 25 years of experience reading breast MRI scans. MRI scans were classified according to the BI-RADS MRI lexicon (fourth edition). Readers had access and were not blinded to current or previous breast imaging findings in accordance with clinical and screening practice. In our present analysis, all participants with a BI-RADS 3, 4, or 5 finding were included. In case of a BI-RADS 3 finding, independent double reading was performed by a second breast radiologist from the same reader group mentioned previously, and discrepancies were resolved by consensus. In case of a final BI-RADS 3 finding, MRI was repeated after 6 months. In case of a BI-RADS 4 or 5 finding, histologic sampling was indicated.
Prospectively collected MRI variables included BI-RADS classification, background parenchymal enhancement, symmetry of enhancement, lesion location, size, type, enhancement kinetics, shape, margin, distribution, and internal enhancement characteristics. In case of multiple lesions, the BI-RADS descriptors of the most suspicious lesion (based on the highest BI-RADS classification) were used for analysis.
Breast cancer diagnosis was defined as a histopathologic result of ductal carcinoma in situ or invasive carcinoma at histologic sampling or surgery; any other histopathologic result was considered a benign finding.
Statistical Analysis
Data were analyzed by using SPSS software (version 25, IBM) and RStudio software (version 3.6.1, RStudio). Clinical characteristics and MRI findings were summarized using descriptive statistics, and differences between true-positive and false-positive results were tested using the two-sample t test, Wilcoxon rank sum test, Pearson χ2 test, or Fisher exact test. P values less than .05 were considered indicative of a significant difference.
Multivariable logistic regression models were built, starting with a full model including all clinical characteristics and MRI findings. For the second model, backward elimination was used to remove variables. Akaike Information Criterion was used to compare models, with the preferred model being the one with the minimum Akaike Information Criterion value. Akaike Information Criterion rewards the goodness of fit of the model but penalizes increasing numbers of variables, resulting in a value that reflects the trade-off between goodness of fit and simplicity of the model (30). A third model, based on age and MRI findings only, was built to evaluate the potential of a model based solely on readily available data. Areas under the receiver operating characteristic curve (AUCs) and calibration plots were obtained. Internal validation was performed using 200 bootstrap samples to obtain optimism-corrected estimates (31). In the receiver operating characteristic curves of all models, the cutoff point where sensitivity falls below 100% was identified to calculate the maximum number of false-positive MRI screenings that could be identified with the model without missing any breast cancers.
Strategy of Not Recalling Women with BI-RADS 3 Results
In the DENSE trial, women with a final BI-RADS 3 result were scheduled for repeat MRI after 6 months. However, a simple and easy-to-implement strategy to reduce FPR could be to not recall women with a BI-RADS 3 result at all. This theoretical strategy was compared with the performance of the prediction models.
Results
Clinical Characteristics and MRI Findings
Among 40 373 women in the DENSE trial, 454 (median age, 52 years; interquartile range, 50–57 years) had positive results at MRI screening (Fig 2). A total of 79 participants (median age, 54 years; interquartile range, 51–61 years) were diagnosed with breast cancer (true-positive findings) and 375 (median age, 52 years; interquartile range, 50–56 years) had false-positive MRI findings (Table 1). A total of 432 women (95.2%) completed the questionnaire on health status and breast cancer risk factors. Most were postmenopausal, were 13 years old at menarche (mean age, 13.4 years ± 1.4 [standard deviation]), had a healthy weight (mean body mass index, 22.1 kg/m2 ± 2.8), had given birth two or more times, had zero first-degree relatives with breast cancer, had never used hormones after age 50 years, and had no history of breast biopsy. Median Volpara breast density was 18.8% (interquartile range, 16.7%–22.0%). Among the 454 recalled participants, 150 had BI-RADS 3 lesions (33%), 286 had BI-RADS 4 lesions (63%), and 18 had BI-RADS 5 lesions (4%) (Table 2). Most women had symmetric minimal background parenchymal enhancement. Lesions detected were predominantly masses, and half were smaller than 1 cm (median size, 1.0 cm; interquartile range 0.7–1.0 cm). Most lesions showed rapid initial uptake followed by washout in the delayed phase (type 3 enhancement curve). Of the 279 mass lesions detected, most had an irregular shape, irregular margin, and a heterogeneous internal enhancement pattern (Table E1 [online]). Of the 143 nonmass enhancement lesions detected, most had a focal distribution and heterogeneous internal enhancement pattern (Table E2 [online]). A total of 300 women underwent biopsy—276 after a BI-RADS 4 or 5 lesion was seen at screening MRI and 24 after a BI-RADS 4 or 5 lesion was seen at 6-month follow-up MRI after an initial BI-RADS 3 finding on screening MRI. There were 31 women with an indication for biopsy who did not undergo the procedure because the lesion was no longer visible at additional imaging (n = 18), the lesion was an intramammary lymph node or cyst (n = 8), biopsy of the lesion was technically impossible (n = 2), or the lesion was already histologically proven benign in the past (n = 3). A total of 79 women were diagnosed with breast cancer, of whom 64 had invasive carcinoma and 15 had ductal carcinoma in situ. Women with a breast cancer diagnosis, and thus a true-positive MRI screening examination, were older, had a higher body mass index, and more often had two or more first-degree relatives with breast cancer compared with women with a false-positive screening MRI result (Table 1). MRI findings associated with a true-positive screening MRI result were a higher BI-RADS classification (P < .001), a mass lesion type (P = .02), rapid initial contrast material uptake (P = .01), and wash-out kinetics (P < .001) (Table 2). For the 279 mass lesions seen at screening MRI, an irregular shape (P < .001) and an irregular or spiculated margin (P < .001) were associated with breast cancer diagnosis and thus a true-positive MRI screening result (Table E1 [online]).
 |
 |
Prediction Models
The full model (model A) (Table 3), including all clinical characteristics and MRI findings, based on 328 participants with complete data regarding these variables, yielded an AUC of 0.88 (95% CI: 0.84, 0.92) (Fig 3). The optimism-corrected AUC was 0.80. Among these 328 participants, 64 had breast cancer and 264 had false-positive findings. With the full prediction model, a total of 120 of 264 (45.5%, 95% CI: 39.6, 51.5) false-positive findings could have been identified as false-positive cases without missing any cancers. Among the participants identified as having false-positive findings were 35 of 164 (21%, 95% CI: 16, 28) who underwent biopsy for a benign lesion. Backward selection resulted in a reduced prediction model (model B, Table 3) with the optimal trade-off between goodness of fit and simplicity of the model. The reduced model included the variables age (odds ratio [OR], 1.14; 95% CI: 1.05, 1.24), menopausal status (perimenopausal vs postmenopausal OR, 4.1; 95% CI: 1.7, 10.8), number of first-degree relatives with breast cancer (two or more relatives vs zero relatives: OR, 11; 95% CI: 1, 250), history of breast biopsy (OR, 2.5; 95% CI: 1.1, 5.3), MRI BI-RADS classification (BI-RADS 4 vs BI-RADS 3: OR, 33; 95% CI: 7, 590; BI-RADS 5 vs BI-RADS 3: OR, 710; 95% CI: 83, 17 724), and initial phase kinetics (rapid vs slow: OR, 6.9; 95% CI: 1.4, 63.7). The AUC of the reduced prediction model was 0.85 (95% CI: 0.80, 0.90) (Fig 3). The optimism-corrected AUC was 0.82. With the reduced prediction model, a total of 85 of 264 (32.2%; 95% CI: 26.8, 38.1) false-positive MRI screenings could have been identified without missing any cancers. Among the identified false-positive findings were 17 of 164 (10.4%; 95% CI: 6.5, 16.1) participants who underwent biopsy for a benign lesion. A third model (model C, Table 3), including all MRI findings as presented in Table 2 plus age, based on 369 participants for whom data were complete, yielded a comparable performance with AUC of 0.84 ([95% CI: 0.79, 0.88] versus full-model AUC of 0.88; P = .15) (Fig 3). The optimism-corrected AUC was 0.78. Among these 369 participants, 68 had breast cancer and 301 had false-positive findings. With the MRI plus age model, 107 of 301 (35.5%; 95% CI: 30.4, 41.1) false-positive MRI screenings could have been identified as false-positive without missing any cancers. Among the identified false-positive cases were 24 of 185 (13%, 95% CI: 9, 19) participants who underwent biopsy for a benign lesion.
 |
Figure 3: Graph shows areas under the receiver operating characteristic curves (AUCs) of three prediction models for true-positive MRI screening. Model A (green): The full model, including all clinical characteristics and MRI findings, as presented in Tables 1 and 2, based on 328 participants. Model B (blue): The reduced model, including the variables age, menopausal status, number of first-degree relatives with breast cancer, history of breast biopsy, MRI Breast Imaging Reporting and Data System classification, and initial phase kinetics. Model C (purple): The model based on readily available data, including all MRI findings, as presented in Table 2 plus age, based on 369 participants.
Strategy of Not Recalling Women with BI-RADS 3 Findings
Not recalling women with a BI-RADS 3 finding at supplemental screening MRI (n = 150) would reduce the recall rate by 33% (from 94.9 of 1000 to 63.6 of 1000) and the FPR by 39.0% (from 79.8 of 1000 to 48.7 of 1000) at the cost of missing four breast cancers. These cases included one invasive lobular carcinoma (Bloom and Richardson grade 2) in a 62-year-old woman, one papillary carcinoma with ductal carcinoma in situ in a 50-year-old woman, one ductal carcinoma in situ (grade unknown) in a 71-year-old woman, and one triple-negative invasive carcinoma (grade 3) in a 52-year-old woman. The triple-negative cancer was in fact not visible at screening MRI but was identified as a new lesion at 6-month follow-up MRI performed for a BI-RADS 3 lesion in the contralateral breast. Therefore, realistically, three cancers would have been missed if all 150 participants with a BI-RADS 3 screening MRI result were not recalled. This is in line with the no more than 2% likelihood of a malignant neoplasm for a BI-RADS 3 score (32).
Discussion
Prediction models based on clinical characteristics and MRI findings may enable identification of a substantial part of false-positive first-round supplemental MRI screenings, reducing false-positive rate and benign biopsy rate without missing any cancers. The full model (area under the receiver operating characteristics curve [AUC], 0.88; 95% CI: 0.84, 0.92), based on all clinical characteristics and MRI findings, could help prevent 45.5% of false-positive recalls and 21.3% of benign biopsies after first-round supplemental MRI screening in women with extremely dense breast tissue without missing any cancers. However, if such a model were implemented, this would require the investment of a substantial amount of time, money, and effort to collect all these clinical characteristics in screening practice. The amount of missing data, even within the context of a prospective randomized-controlled trial, reflects the difficulty in collecting these characteristics and raises questions about the feasibility of such a model in daily practice. The model solely based on readily available MRI findings and age had a comparable performance (AUC, 0.84; 95% CI: 0.79, 0.88; P = .15) and could have prevented 35.5% of false-positive MRI screenings and 13.0% of benign biopsies. As this model is based on readily available data, its implementation could be relatively easy.
Not recalling women with a BI-RADS 3 score at screening MRI (after double reading and consensus in case of discrepancies) would reduce the FPR in our first-round screened study population from 79.8 of 1000 to 48.7 of 1000, at the cost of missing three breast cancers (two invasive carcinomas and one ductal carcinoma in situ) and one incidental detection of a highly aggressive fourth cancer. This may seem unfavorable compared with using our prediction model, which in theory does not miss any breast cancers. However, the performance of the developed model will probably be poorer when used in daily practice because of overfitting and might also result in missed breast cancer cases.
The association of certain clinical characteristics with breast cancer has been well established and has led to numerous breast cancer risk models (33). However, literature on clinical characteristics and MRI findings in relation to false-positive MRI screening is scarce. Vreeman et al (34) reported an increased FPR in first-round MRI screening in women with a high amount of fibroglandular tissue (OR, 1.26; P = .05) and high background parenchymal enhancement (OR, 1.48; P = .003). In our study sample of women with extremely dense breasts, the amount of fibroglandular tissue (Volpara automated breast density measurements) and background parenchymal enhancement were not clearly associated with a higher FPR. This difference in findings may be explained in part by different study populations (extremely dense breasts vs all breast densities) and different methods of background parenchymal enhancement measurements (visual inspection as used in Dense Tissue and Early Breast Neoplasm Screening [DENSE] study vs automated tool).
Baltzer et al (35) reported that nonmass lesions were the major cause of false-positive breast MRI findings and that BI-RADS descriptors are insufficient to differentiate benign from malignant nonmass lesions. In our study, the ratio of mass to nonmass lesions differed between benign (1.7:1) and malignant (4.1:1) findings, indicating that mass lesions are more often malignant (P = .02). However, in our multivariable model, lesion type was not of considerable help in discrimination of false-positive from true-positive screening MRI findings. It should be noted that in the DENSE trial, specific attention was paid to not overcalling nonmass enhancement lesions, including repeated training every 6 months focused on nonmass enhancement recall rate. Although our approach was to combine clinical characteristics and MRI findings to lower FPR at MRI, McCarthy et al (36) reported that using breast cancer risk factors, including body mass index and genetic markers to guide follow-up of BI-RADS 4 mammograms, could reduce the burden of false-positive mammograms.
Our study had some limitations. First, the estimates of the expected reduction in false-positive findings and benign biopsies were obtained from the set of cases that was used to generate the regression models. As such, they are likely to be optimistic. Second, subsequent incident screening rounds had considerably fewer false-positive findings (mostly associated with the lower rate of BI-RADS 3 results) than the first prevalent screening round (29). In line with the DENSE policy of an additional evaluation for BI-RADS 3 results, the presence of prior imaging allows a much clearer differentiation between suspicious and nonsuspicious lesions. Therefore, the results cannot be generalized to incident screening.
In conclusion, prediction models based on clinical characteristics and MRI findings may reduce the false-positive first-round screening MRI rate and the number of benign biopsies, bringing supplemental screening MRI for women with dense breasts one step closer to implementation. Validation studies using data from different populations and incident screening rounds are warranted. As incident screening rounds have a much lower false-positive rate, separate models may have to be created.
Acknowledgments
We thank the trial participants for their contributions; the regional screening organizations, Volpara Health Technologies, the Dutch Expert Center for Screening, and the National Institute for Public Health and the Environment for their advice and in-kind contributions; and the data managers and project support personnel for their assistance during the trial.
Author Contributions
Author contributions: Guarantor of integrity of entire study, B.M.d.D., C.E.L., C.H.v.G.; 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, B.M.d.D., W.B.V., C.E.L., R.M.P.; clinical studies, B.M.d.D., M.F.B., S.V.d.L., W.B.V., K.M.D., M.B.I.L., C.E.L., R.M.M., E.M.M., J.V., R.M.P., C.H.v.G.; statistical analysis, B.M.d.D., M.F.B., C.H.v.G.; and manuscript editing, all authors
Supported by the University Medical Center Utrecht (project number: UMCU DENSE), the Netherlands Organization for Health Research and Development (project numbers: ZonMW-200320002-UMCU and ZonMW Preventie 50-53125-98-014), the Dutch Cancer Society (project numbers: DCS-UU-2009-4348, UU-2014-6859, and UU-2014-7151), the Dutch Pink Ribbon—A Sister’s Hope (project number: Pink Ribbon-10074), Bayer Pharmaceuticals (project number: BSP-DENSE), and Stichting Kankerpreventie Midden-West. For research purposes, Volpara Health Technologies provided Volpara Imaging Software, version 1.5, for installation on servers in the screening units.
1 Members of the DENSE trial study group are listed at the end of this article.
Members of the DENSE Trial Study Group: University Medical Center Utrecht, Utrecht University, Utrecht, the Netherlands: C. H. van Gils, PhD; M. F. Bakker, PhD; S. V. de Lange, MD; S. G. A. Veenhuizen, MSc; W. B. Veldhuis, MD, PhD; R. M. Pijnappel, MD, PhD; M. J. Emaus, PhD; P. H. M. Peeters, MD, PhD; E. M. Monninkhof, PhD; M. A. Fernandez-Gallardo, MD; W. P. T. M. Mali, MD, PhD; M. A. A. J. van den Bosch, MD, PhD; and P. J. van Diest, MD, PhD. Radboud University Nijmegen Medical Center, Nijmegen, the Netherlands: R. M. Mann, MD, PhD; R. Mus, MD; M. W. Imhof-Tas, MD; and N. Karssemeijer, PhD. The Netherlands Cancer Institute – Antoni van Leeuwenhoek Hospital, Amsterdam, the Netherlands: C. E. Loo, MD, PhD; P. K. de Koekkoek-Doll, MD; and H. A. O. Winter-Warnars, MD, PhD. Albert Schweitzer Hospital, Dordrecht, the Netherlands: R. H. C. Bisschops, MD, PhD; M. C. J. M. Kock, MD, PhD; R. K. Storm, MD; and P. H. M. van der Valk, MD. Maastricht University Medical Centre, Maastricht, the Netherlands: M. B. I. Lobbes, MD, PhD, and S. Gommers, MD. Jeroen Bosch Hospital, ’s-Hertogenbosch, the Netherlands: M. D. F. de Jong, MD, and M. J. C. M. Rutten, MD, PhD. Amsterdam UMC, Vrije Universiteit Amsterdam, Amsterdam, the Netherlands: K. M. Duvivier, MD, and P. de Graaf, MD, PhD. Hospital Group Twente (ZGT), Almelo, the Netherlands: J. Veltman, MD, PhD, and R. L. J. H. Bourez, MD. Erasmus Medical Center, Rotterdam, the Netherlands: H. J. de Koning, MD, PhD.
References
- 1. . The effect of population-based mammography screening in Dutch municipalities on breast cancer mortality: 20 years of follow-up. Int J Cancer 2017;141(4):671–677.
- 2. National Evaluation Team for Breast Cancer Screening (NETB). National evaluation of breast cancer screening in the Netherlands 1990 – 2011/2012. Rotterdam, the Netherlands: Optima Grafische Communicatie,2014.
- 3. . Volumetric breast density affects performance of digital screening mammography. Breast Cancer Res Treat 2017;162(1):95–103.
- 4. . Volumetric breast density and risk of advanced cancers after a negative screening episode: a cohort study. Breast Cancer Res 2018;20(1):95.
- 5. . Automated and clinical breast imaging reporting and data system density measures predict risk for screen-detected and interval cancers: a case-control study. Ann Intern Med 2018;168(11):757–765.
- 6. . Digital mammography screening: sensitivity of the programme dependent on breast density. Eur Radiol 2017;27(7):2744–2751.
- 7. . Mammographic density and the risk and detection of breast cancer. N Engl J Med 2007;356(3):227–236.
- 8. . Efficacy of MRI and mammography for breast-cancer screening in women with a familial or genetic predisposition. N Engl J Med 2004;351(5):427–437.
- 9. . Prospective multicenter cohort study to refine management recommendations for women at elevated familial risk of breast cancer: the EVA trial. J Clin Oncol 2010;28(9):1450–1457.
- 10. . Cancer yield of mammography, MR, and US in high-risk women: prospective multi-institution breast cancer screening study. Radiology 2007;244(2):381–388.
- 11. . Accuracy of screening women at familial risk of breast cancer without a known gene mutation: Individual patient data meta-analysis. Eur J Cancer 2017;85(31):38.
- 12. . Cost-effectiveness of screening women with familial risk for breast cancer with magnetic resonance imaging. J Natl Cancer Inst 2013;105(17):1314–1321.
- 13. . Systematic review: using magnetic resonance imaging to screen women at high risk for breast cancer. Ann Intern Med 2008;148(9):671–679.
- 14. . Multimodality screening of high-risk women: a prospective cohort study. J Clin Oncol 2009;27(36):6124–6128.
- 15. . Comparison of abbreviated breast MRI vs digital breast tomosynthesis for breast cancer detection among women with dense breasts undergoing screening. JAMA 2020;323(8):746–756.
- 16. . Supplemental breast MR imaging screening of women with average risk of breast cancer. Radiology 2017;283(2):361–370.
- 17. . Detection of breast cancer with addition of annual screening ultrasound or a single screening MRI to mammography in women with elevated breast cancer risk. JAMA 2012;307(13):1394–1404.
- 18. . Application of abbreviated protocol of magnetic resonance imaging for breast cancer screening in dense breast tissue. Acad Radiol 2017;24(3):316–320.
- 19. . MRI versus mammography for breast cancer screening in women with familial risk (FaMRIsc): a multicentre, randomised, controlled trial. Lancet Oncol 2019;20(8):1136–1147.
- 20. . MR imaging as an additional screening modality for the detection of breast cancer in women aged 50–75 years with extremely dense breasts: the DENSE trial study design. Radiology 2015;277(2):527–537.
- 21. . Supplemental MRI screening for women with extremely dense breast tissue. N Engl J Med 2019;381(22):2091–2102.
- 22. . Anxiety after an abnormal screening mammogram is a serious problem. Breast 2012;21(1):83–88.
- 23. . Consequences of false-positive screening mammograms. JAMA Intern Med 2014;174(6):954–961.
- 24. . Effect of abnormal screening mammogram on quality of life. Br J Surg 2011;98(4):537–542.
- 25. . Magnetic resonance imaging in breast cancer: A literature review and future perspectives. World J Clin Oncol 2014;5(2):61–70.
- 26. . Reasons for (non)participation in supplemental population-based MRI breast screening for women with extremely dense breasts. Clin Radiol 2018;73(8):759.e1–759.e9.e1–e9.
- 27. . Knowledge-based and deep learning-based automated chest wall segmentation in magnetic resonance images of extremely dense breasts. Med Phys 2019;46(10):4405–4416.
- 28. . Computer-aided diagnosis in multiparametric magnetic resonance imaging screening of women with extremely dense breasts to reduce false-positive diagnoses. Invest Radiol 2020;55(7):438–444.
- 29. . Supplemental breast MRI for women with extremely dense breasts: results of the second screening round of the DENSE trial. Radiology 2021;299(2):278–286.
- 30. . Selection of important variables and determination of functional form for continuous predictors in multivariable model building. Stat Med 2007;26(30):5512–5528.
- 31. . Clinical prediction models: a practical approach to development, validation, and updating. New York, NY:Springer,2009.
- 32.
American College of Radiology . ACR BI-RADS Atlas: breast imaging reporting and data system. 5th ed. Reston, Va:American College of Radiology,2013. - 33. . Breast cancer risk models: a comprehensive overview of existing models, validation, and clinical applications. Breast Cancer Res Treat 2017;164(2):263–28.[Published correction appears in Breast Cancer Res Treat 2017;164(3):745.].
- 34. . Amount of fibroglandular tissue FGT and background parenchymal enhancement BPE in relation to breast cancer risk and false positives in a breast MRI screening program: a retrospective cohort study. Eur Radiol 2019;29(9):4678–4690.
- 35. . False-positive findings at contrast-enhanced breast MRI: a BI-RADS descriptor study. AJR Am J Roentgenol 2010;194(6):1658–1663.
- 36. . The use of the Gail model, body mass index and SNPs to predict breast cancer among women with abnormal (BI-RADS 4) mammograms. Breast Cancer Res 2015;17(1):1.
Article History
Received: Feb 9 2021Revision requested: Mar 19 2021
Revision received: May 18 2021
Accepted: June 4 2021
Published online: Aug 17 2021
Published in print: Nov 2021
