Collaborative Modeling to Compare Different Breast Cancer Screening Strategies: A Decision Analysis for the US Preventive Services Task Force.

Importance: The effects of breast cancer incidence changes and advances in screening and treatment on outcomes of different screening strategies are not well known. Objective: To estimate outcomes of various mammography screening strategies. Design, Setting, and Population: Comparison of outcomes using 6 Cancer Intervention and Surveillance Modeling Network (CISNET) models and national data on breast cancer incidence, mammography performance, treatment effects, and other-cause mortality in US women without previous cancer diagnoses. Exposures: Thirty-six screening strategies with varying start ages (40, 45, 50 years) and stop ages (74, 79 years) with digital mammography or digital breast tomosynthesis (DBT) annually, biennially, or a combination of intervals. Strategies were evaluated for all women and for Black women, assuming 100% screening adherence and "real-world" treatment. Main Outcomes and Measures: Estimated lifetime benefits (breast cancer deaths averted, percent reduction in breast cancer mortality, life-years gained), harms (false-positive recalls, benign biopsies, overdiagnosis), and number of mammograms per 1000 women. Results: Biennial screening with DBT starting at age 40, 45, or 50 years until age 74 years averted a median of 8.2, 7.5, or 6.7 breast cancer deaths per 1000 women screened, respectively, vs no screening. Biennial DBT screening at age 40 to 74 years (vs no screening) was associated with a 30.0% breast cancer mortality reduction, 1376 false-positive recalls, and 14 overdiagnosed cases per 1000 women screened. Digital mammography screening benefits were similar to those for DBT but had more false-positive recalls. Annual screening increased benefits but resulted in more false-positive recalls and overdiagnosed cases. Benefit-to-harm ratios of continuing screening until age 79 years were similar or superior to stopping at age 74. In all strategies, women with higher-than-average breast cancer risk, higher breast density, and lower comorbidity level experienced greater screening benefits than other groups. Annual screening of Black women from age 40 to 49 years with biennial screening thereafter reduced breast cancer mortality disparities while maintaining similar benefit-to-harm trade-offs as for all women. Conclusions: This modeling analysis suggests that biennial mammography screening starting at age 40 years reduces breast cancer mortality and increases life-years gained per mammogram. More intensive screening for women with greater risk of breast cancer diagnosis or death can maintain similar benefit-to-harm trade-offs and reduce mortality disparities.

Predicting five-year interval second breast cancer risk in women with prior breast cancer.

BACKGROUND: Annual surveillance mammography is recommended for women with a personal history of breast cancer. Risk prediction models that estimate mammography failures such as interval second breast cancers could help to tailor surveillance imaging regimens to women's individual risk profiles. METHODS: In a cohort of women with a history of breast cancer receiving surveillance mammography in the Breast Cancer Surveillance Consortium in 1996-2019, we used LASSO-penalized regression to estimate the probability of an interval second cancer (invasive cancer or ductal carcinoma in situ) in the one-year following a negative surveillance mammogram. Based on predicted risks from this one-year risk model, we generated cumulative risks of an interval second cancer for the five-year period following each mammogram. Model performance was evaluated using cross-validation in the overall cohort and within race and ethnicity strata. RESULTS: In 173,290 surveillance mammograms, we observed 496 interval cancers. One-year risk models were well-calibrated (expected/observed ratio = 1.00) with good accuracy (area under the receiver operating characteristic curve = 0.64). Model performance was similar across race and ethnicity groups. The median five-year cumulative risk was 1.20% (interquartile range 0.93-1.63%). Median five-year risks were highest in women who were under age 40 or pre- or peri-menopausal at diagnosis and those with estrogen receptor-negative primary breast cancers. CONCLUSIONS: Our risk model identified women at high risk of interval second breast cancers who may benefit from additional surveillance imaging modalities. Risk models should be evaluated to determine if risk-guided supplemental surveillance imaging improves early detection and decreases surveillance failures.

Artificial Intelligence-Driven Mammography-Based Future Breast Cancer Risk Prediction: A Systematic Review.

PURPOSE: To summarize the literature regarding the performance of mammography-image based artificial intelligence (AI) algorithms, with and without additional clinical data, for future breast cancer risk prediction. MATERIALS AND METHODS: A systematic literature review was performed using six databases (medRixiv, bioRxiv, Embase, Engineer Village, IEEE Xplore, and PubMed) from 2012 through September 30, 2022. Studies were included if they used real-world screening mammography examinations to validate AI algorithms for future risk prediction based on images alone or in combination with clinical risk factors. The quality of studies was assessed, and predictive accuracy was recorded as the area under the receiver operating characteristic curve (AUC). RESULTS: Sixteen studies met inclusion and exclusion criteria, of which 14 studies provided AUC values. The median AUC performance of AI image-only models was 0.72 (range 0.62-0.90) compared with 0.61 for breast density or clinical risk factor-based tools (range 0.54-0.69). Of the seven studies that compared AI image-only performance directly to combined image + clinical risk factor performance, six demonstrated no significant improvement, and one study demonstrated increased improvement. CONCLUSIONS: Early efforts for predicting future breast cancer risk based on mammography images alone demonstrate comparable or better accuracy to traditional risk tools with little or no improvement when adding clinical risk factor data. Transitioning from clinical risk factor-based to AI image-based risk models may lead to more accurate, personalized risk-based screening approaches.

Comparative Performance of Contrast-enhanced Mammography, Abbreviated Breast MRI, and Standard Breast MRI for Breast Cancer Screening.

Background Contrast-enhanced mammography (CEM) and abbreviated breast MRI (ABMRI) are emerging alternatives to standard MRI for supplemental breast cancer screening. Purpose To compare the diagnostic performance of CEM, ABMRI, and standard MRI. Materials and Methods This single-institution, prospective, blinded reader study included female participants referred for breast MRI from January 2018 to June 2021. CEM was performed within 14 days of standard MRI; ABMRI was produced from standard MRI images. Two readers independently interpreted each CEM and ABMRI after a washout period. Examination-level performance metrics calculated were recall rate, cancer detection, and false-positive biopsy recommendation rates per 1000 examinations and sensitivity, specificity, and positive predictive value of biopsy recommendation. Bootstrap and permutation tests were used to calculate 95% CIs and compare modalities. Results Evaluated were 492 paired CEM and ABMRI interpretations from 246 participants (median age, 51 years; IQR, 43-61 years). On 49 MRI scans with lesions recommended for biopsy, nine lesions showed malignant pathology. No differences in ABMRI and standard MRI performance were identified. Compared with standard MRI, CEM demonstrated significantly lower recall rate (14.0% vs 22.8%; difference, -8.7%; 95% CI: -14.0, -3.5), lower false-positive biopsy recommendation rate per 1000 examinations (65.0 vs 162.6; difference, -97.6; 95% CI: -146.3, -50.8), and higher specificity (87.8% vs 80.2%; difference, 7.6%; 95% CI: 2.3, 13.1). Compared with standard MRI, CEM had significantly lower cancer detection rate (22.4 vs 36.6; difference, -14.2; 95% CI: -28.5, -2.0) and sensitivity (61.1% vs 100%; difference, -38.9%; 95% CI: -66.7, -12.5). The performance differences between CEM and ABMRI were similar to those observed between CEM and standard MRI. Conclusion ABMRI had comparable performance to standard MRI and may support more efficient MRI screening. CEM had lower recall and higher specificity compared with standard MRI or ABMRI, offset by lower cancer detection rate and sensitivity compared with standard MRI. These trade-offs warrant further consideration of patient population characteristics before widespread screening with CEM. Clinical trial registration no. NCT03517813 © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Chang in this issue.

Impact of Surveillance Mammography Intervals Less Than One Year on Performance Measures in Women With a Personal History of Breast Cancer.

OBJECTIVE: When multiple surveillance mammograms are performed within an annual interval, the current guidance for one-year follow-up to determine breast cancer status results in shared follow-up periods in which a single breast cancer diagnosis can be attributed to multiple preceding examinations, posing a challenge for standardized performance assessment. We assessed the impact of using follow-up periods that eliminate the artifactual inflation of second breast cancer diagnoses. MATERIALS AND METHODS: We evaluated surveillance mammograms from 2007-2016 in women with treated breast cancer linked with tumor registry and pathology outcomes. Second breast cancers included ductal carcinoma in situ or invasive breast cancer diagnosed during one-year follow-up. The cancer detection rate, interval cancer rate, sensitivity, and specificity were compared using different follow-up periods: standard one-year follow-up per the American College of Radiology versus follow-up that was shortened at the next surveillance mammogram if less than one year (truncated follow-up). Performance measures were calculated overall and by indication (screening, evaluation for breast problem, and short interval follow-up). RESULTS: Of 117971 surveillance mammograms, 20% (n = 23533) were followed by another surveillance mammogram within one year. Standard follow-up identified 1597 mammograms that were associated with second breast cancers. With truncated follow-up, the breast cancer status of 179 mammograms (11.2%) was revised, resulting in 1418 mammograms associated with unique second breast cancers. The interval cancer rate decreased with truncated versus standard follow-up (3.6 versus 4.9 per 1000 mammograms, respectively), with a difference (95% confidence interval [CI]) of -1.3 (-1.6, -1.1). The overall sensitivity increased to 70.4% from 63.7%, for the truncated versus standard follow-up, with a difference (95% CI) of 6.6% (5.6%, 7.7%). The specificity remained stable at 98.1%. CONCLUSION: Truncated follow-up, if less than one year to the next surveillance mammogram, enabled second breast cancers to be associated with a single preceding mammogram and resulted in more accurate estimates of diagnostic performance for national benchmarks.

Breast cancer risk characteristics of women undergoing whole-breast ultrasound screening versus mammography alone.

BACKGROUND: There are no consensus guidelines for supplemental breast cancer screening with whole-breast ultrasound. However, criteria for women at high risk of mammography screening failures (interval invasive cancer or advanced cancer) have been identified. Mammography screening failure risk was evaluated among women undergoing supplemental ultrasound screening in clinical practice compared with women undergoing mammography alone. METHODS: A total of 38,166 screening ultrasounds and 825,360 screening mammograms without supplemental screening were identified during 2014-2020 within three Breast Cancer Surveillance Consortium (BCSC) registries. Risk of interval invasive cancer and advanced cancer were determined using BCSC prediction models. High interval invasive breast cancer risk was defined as heterogeneously dense breasts and BCSC 5-year breast cancer risk ≥2.5% or extremely dense breasts and BCSC 5-year breast cancer risk ≥1.67%. Intermediate/high advanced cancer risk was defined as BCSC 6-year advanced breast cancer risk ≥0.38%. RESULTS: A total of 95.3% of 38,166 ultrasounds were among women with heterogeneously or extremely dense breasts, compared with 41.8% of 825,360 screening mammograms without supplemental screening (p < .0001). Among women with dense breasts, high interval invasive breast cancer risk was prevalent in 23.7% of screening ultrasounds compared with 18.5% of screening mammograms without supplemental imaging (adjusted odds ratio, 1.35; 95% CI, 1.30-1.39); intermediate/high advanced cancer risk was prevalent in 32.0% of screening ultrasounds versus 30.5% of screening mammograms without supplemental screening (adjusted odds ratio, 0.91; 95% CI, 0.89-0.94). CONCLUSIONS: Ultrasound screening was highly targeted to women with dense breasts, but only a modest proportion were at high mammography screening failure risk. A clinically significant proportion of women undergoing mammography screening alone were at high mammography screening failure risk.

Digital Breast Tomosynthesis versus Digital Mammography Screening Performance on Successive Screening Rounds from the Breast Cancer Surveillance Consortium.

Background Prior cross-sectional studies have observed that breast cancer screening with digital breast tomosynthesis (DBT) has a lower recall rate and higher cancer detection rate compared with digital mammography (DM). Purpose To evaluate breast cancer screening outcomes with DBT versus DM on successive screening rounds. Materials and Methods In this retrospective cohort study, data from 58 breast imaging facilities in the Breast Cancer Surveillance Consortium were collected. Analysis included women aged 40-79 years undergoing DBT or DM screening from 2011 to 2020. Absolute differences in screening outcomes by modality and screening round were estimated during the study period by using generalized estimating equations with marginal standardization to adjust for differences in women's risk characteristics across modality and round. Results A total of 523 485 DBT examinations (mean age of women, 58.7 years ± 9.7 [SD]) and 1 008 123 DM examinations (mean age, 58.4 years ± 9.8) among 504 863 women were evaluated. DBT and DM recall rates decreased with successive screening round, but absolute recall rates in each round were significantly lower with DBT versus DM (round 1 difference, -3.3% [95% CI: -4.6, -2.1] [P < .001]; round 2 difference, -1.8% [95% CI: -2.9, -0.7] [P = .003]; round 3 or above difference, -1.2% [95% CI: -2.4, -0.1] [P = .03]). DBT had significantly higher cancer detection (difference, 0.6 per 1000 examinations [95% CI: 0.2, 1.1]; P = .009) compared with DM only for round 3 and above. There were no significant differences in interval cancer rate (round 1 difference, 0.00 per 1000 examinations [95% CI: -0.24, 0.30] [P = .96]; round 2 or above difference, 0.04 [95% CI: -0.19, 0.31] [P = .76]) or total advanced cancer rate (round 1 difference, 0.00 per 1000 examinations [95% CI: -0.15, 0.19] [P = .94]; round 2 or above difference, -0.06 [95% CI: -0.18, 0.11] [P = .43]). Conclusion DBT had lower recall rates and could help detect more cancers than DM across three screening rounds, with no difference in interval or advanced cancer rates. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Skaane in this issue.

Variation in second breast cancer risk after primary invasive cancer by time since primary cancer diagnosis and estrogen receptor status.

BACKGROUND: In women with previously treated breast cancer, occurrence and timing of second breast cancers have implications for surveillance. The authors examined the timing of second breast cancers by primary cancer estrogen receptor (ER) status in the Breast Cancer Surveillance Consortium. METHODS: Women who were diagnosed with American Joint Commission on Cancer stage I-III breast cancer were identified within six Breast Cancer Surveillance Consortium registries from 2000 to 2017. Characteristics collected at primary breast cancer diagnosis included demographics, ER status, and treatment. Second breast cancer events included subsequent ipsilateral or contralateral breast cancers diagnosed >6 months after primary diagnosis. The authors examined cumulative incidence and second breast cancer rates by primary cancer ER status during 1-5 versus 6-10 years after diagnosis. RESULTS: At 10 years, the cumulative second breast cancer incidence was 11.8% (95% confidence interval [CI], 10.7%-13.1%) for women with ER-negative disease and 7.5% (95% CI, 7.0%-8.0%) for those with ER-positive disease. Women with ER-negative cancer had higher second breast cancer rates than those with ER-positive cancer during the first 5 years of follow-up (16.0 per 1000 person-years [PY]; 95% CI, 14.2-17.9 per 1000 PY; vs. 7.8 per 1000 PY; 95% CI, 7.3-8.4 per 1000 PY, respectively). After 5 years, second breast cancer rates were similar for women with ER-negative versus ER-positive breast cancer (12.1 per 1000 PY; 95% CI, 9.9-14.7; vs. 9.3 per 1000 PY; 95% CI, 8.4-10.3 per 1000 PY, respectively). CONCLUSIONS: ER-negative primary breast cancers are associated with a higher risk of second breast cancers than ER-positive cancers during the first 5 years after diagnosis. Further study is needed to examine the potential benefit of more intensive surveillance targeting these women in the early postdiagnosis period.

Performance of Statistical and Machine Learning Risk Prediction Models for Surveillance Benefits and Failures in Breast Cancer Survivors.

BACKGROUND: Machine learning (ML) approaches facilitate risk prediction model development using high-dimensional predictors and higher-order interactions at the cost of model interpretability and transparency. We compared the relative predictive performance of statistical and ML models to guide modeling strategy selection for surveillance mammography outcomes in women with a personal history of breast cancer (PHBC). METHODS: We cross-validated seven risk prediction models for two surveillance outcomes, failure (breast cancer within 12 months of a negative surveillance mammogram) and benefit (surveillance-detected breast cancer). We included 9,447 mammograms (495 failures, 1,414 benefits, and 7,538 nonevents) from years 1996 to 2017 using a 1:4 matched case-control samples of women with PHBC in the Breast Cancer Surveillance Consortium. We assessed model performance of conventional regression, regularized regressions (LASSO and elastic-net), and ML methods (random forests and gradient boosting machines) by evaluating their calibration and, among well-calibrated models, comparing the area under the receiver operating characteristic curve (AUC) and 95% confidence intervals (CI). RESULTS: LASSO and elastic-net consistently provided well-calibrated predicted risks for surveillance failure and benefit. The AUCs of LASSO and elastic-net were both 0.63 (95% CI, 0.60-0.66) for surveillance failure and 0.66 (95% CI, 0.64-0.68) for surveillance benefit, the highest among well-calibrated models. CONCLUSIONS: For predicting breast cancer surveillance mammography outcomes, regularized regression outperformed other modeling approaches and balanced the trade-off between model flexibility and interpretability. IMPACT: Regularized regression may be preferred for developing risk prediction models in other contexts with rare outcomes, similar training sample sizes, and low-dimensional features.

Radiologist Preferences for Artificial Intelligence-Based Decision Support During Screening Mammography Interpretation.

BACKGROUND: Artificial intelligence (AI) may improve cancer detection and risk prediction during mammography screening, but radiologists' preferences regarding its characteristics and implementation are unknown. PURPOSE: To quantify how different attributes of AI-based cancer detection and risk prediction tools affect radiologists' intentions to use AI during screening mammography interpretation. MATERIALS AND METHODS: Through qualitative interviews with radiologists, we identified five primary attributes for AI-based breast cancer detection and four for breast cancer risk prediction. We developed a discrete choice experiment based on these attributes and invited 150 US-based radiologists to participate. Each respondent made eight choices for each tool between three alternatives: two hypothetical AI-based tools versus screening without AI. We analyzed samplewide preferences using random parameters logit models and identified subgroups with latent class models. RESULTS: Respondents (n = 66; 44% response rate) were from six diverse practice settings across eight states. Radiologists were more interested in AI for cancer detection when sensitivity and specificity were balanced (94% sensitivity with <25% of examinations marked) and AI markup appeared at the end of the hanging protocol after radiologists complete their independent review. For AI-based risk prediction, radiologists preferred AI models using both mammography images and clinical data. Overall, 46% to 60% intended to adopt any of the AI tools presented in the study; 26% to 33% approached AI enthusiastically but were deterred if the features did not align with their preferences. CONCLUSION: Although most radiologists want to use AI-based decision support, short-term uptake may be maximized by implementing tools that meet the preferences of dissuadable users.

Risk of Lobular Neoplasia Upgrade with Synchronous Carcinoma.

BACKGROUND: Atypical lobular hyperplasia (ALH) and classic lobular carcinoma in situ encompass a spectrum of proliferative lesions known as lobular neoplasia (LN). When imaging-concordant and found in isolation on core needle biopsy (CNB), LN infrequently upgrades to carcinoma on surgical excision, and routine excision is not indicated. Upgrade rates in the setting of synchronous carcinoma are not well studied. PATIENTS AND METHODS: Patients with radiology-pathology concordant synchronous LN and separately biopsied ipsilateral (n = 35) or contralateral (n = 15) carcinoma who underwent excision between 2010 and 2021 were retrospectively identified. Frequency of upgrade, to either invasive or in situ carcinoma, was quantified, and factors associated with upgrade were assessed using Fisher's exact test. RESULTS: The median age was 55 (range 33-74) years. The upgrade rate of LN was 6% and not significantly different between ipsilateral (2.9%) and contralateral (13.3%) carcinoma (p = 0.15). All upgraded LN lesions were ALH on CNB and detected as non-mass enhancement on magnetic resonance imaging (MRI). No additional disease was demonstrated after excision at the site of the original LN CNB in 22.9% (8 out of 35) of ipsilateral and 13.3% (2 out of 15) of contralateral patients. Upgrade was not associated with family history, menopausal status, imaging modality used to detect LN, or extent of LN on CNB (p > 0.05). CONCLUSIONS: Our results demonstrate a low upgrade rate (6%) in our study cohort of LN with synchronous ipsilateral or contralateral carcinoma, which suggests that not all LN mandates excision with synchronous carcinoma. Larger, multi-institution studies are needed to validate these findings.

Trends in Annual Surveillance Mammography Participation Among Breast Cancer Survivors From 2004 to 2016.

BACKGROUND: Annual mammography is recommended for breast cancer survivors; however, population-level temporal trends in surveillance mammography participation have not been described. Our objective was to characterize trends in annual surveillance mammography participation among women with a personal history of breast cancer over a 13-year period. METHODS: We examined annual surveillance mammography participation from 2004 to 2016 in a nationwide sample of commercially insured women with prior breast cancer. Rates were stratified by age group (40-49 vs 50-64 years), visit with a surgical/oncology specialist or primary care provider within the prior year, and sociodemographic characteristics. Joinpoint models were used to estimate annual percentage changes (APCs) in participation during the study period. RESULTS: Among 141,672 women, mammography rates declined from 74.1% in 2004 to 67.1% in 2016. Rates were stable from 2004 to 2009 (APC, 0.1%; 95% CI, -0.5% to 0.8%) but declined 1.5% annually from 2009 to 2016 (95% CI, -1.9% to -1.1%). For women aged 40 to 49 years, rates declined 2.8% annually (95% CI, -3.4% to -2.1%) after 2009 versus 1.4% annually in women aged 50 to 64 years (95% CI, -1.9% to -1.0%). Similar trends were observed in women who had seen a surgeon/oncologist (APC, -1.7%; 95% CI, -2.1% to -1.4%) or a primary care provider (APC, -1.6%; 95% CI, -2.1% to -1.2%) in the prior year. CONCLUSIONS: Surveillance mammography participation among breast cancer survivors declined from 2009 to 2016, most notably among women aged 40 to 49 years. These findings highlight a need for focused efforts to improve adherence to surveillance and prevent delays in detection of breast cancer recurrence and second cancers.

Error Reduction and Diagnostic Concordance in Breast Pathology.

Errors in anatomic pathology can result in patients receiving inappropriate treatment and poor patient outcomes. Policies and procedures are necessary to decrease error and improve diagnostic concordance. Breast pathology may be more prone to diagnostic errors than other surgical pathology subspecialties due to inherit borderline diagnostic categories such as atypical ductal hyperplasia and low-grade ductal carcinoma in situ. Mandatory secondary review of internal and outside referral cases before treatment is effective in reducing diagnostic errors and improving concordance. Assessment of error through amendment/addendum tracking, implementing an incident reporting system, and multidisciplinary tumor boards can establish procedures to prevent future error.

Breast Cancer Screening Strategies for Women With ATM, CHEK2, and PALB2 Pathogenic Variants: A Comparative Modeling Analysis.

IMPORTANCE: Screening mammography and magnetic resonance imaging (MRI) are recommended for women with ATM, CHEK2, and PALB2 pathogenic variants. However, there are few data to guide screening regimens for these women. OBJECTIVE: To estimate the benefits and harms of breast cancer screening strategies using mammography and MRI at various start ages for women with ATM, CHEK2, and PALB2 pathogenic variants. DESIGN, SETTING, AND PARTICIPANTS: This comparative modeling analysis used 2 established breast cancer microsimulation models from the Cancer Intervention and Surveillance Modeling Network (CISNET) to evaluate different screening strategies. Age-specific breast cancer risks were estimated using aggregated data from the Cancer Risk Estimates Related to Susceptibility (CARRIERS) Consortium for 32 247 cases and 32 544 controls in 12 population-based studies. Data on screening performance for mammography and MRI were estimated from published literature. The models simulated US women with ATM, CHEK2, or PALB2 pathogenic variants born in 1985. INTERVENTIONS: Screening strategies with combinations of annual mammography alone and with MRI starting at age 25, 30, 35, or 40 years until age 74 years. MAIN OUTCOMES AND MEASURES: Estimated lifetime breast cancer mortality reduction, life-years gained, breast cancer deaths averted, total screening examinations, false-positive screenings, and benign biopsies per 1000 women screened. Results are reported as model mean values and ranges. RESULTS: The mean model-estimated lifetime breast cancer risk was 20.9% (18.1%-23.7%) for women with ATM pathogenic variants, 27.6% (23.4%-31.7%) for women with CHEK2 pathogenic variants, and 39.5% (35.6%-43.3%) for women with PALB2 pathogenic variants. Across pathogenic variants, annual mammography alone from 40 to 74 years was estimated to reduce breast cancer mortality by 36.4% (34.6%-38.2%) to 38.5% (37.8%-39.2%) compared with no screening. Screening with annual MRI starting at 35 years followed by annual mammography and MRI at 40 years was estimated to reduce breast cancer mortality by 54.4% (54.2%-54.7%) to 57.6% (57.2%-58.0%), with 4661 (4635-4688) to 5001 (4979-5023) false-positive screenings and 1280 (1272-1287) to 1368 (1362-1374) benign biopsies per 1000 women. Annual MRI starting at 30 years followed by mammography and MRI at 40 years was estimated to reduce mortality by 55.4% (55.3%-55.4%) to 59.5% (58.5%-60.4%), with 5075 (5057-5093) to 5415 (5393-5437) false-positive screenings and 1439 (1429-1449) to 1528 (1517-1538) benign biopsies per 1000 women. When starting MRI at 30 years, initiating annual mammography starting at 30 vs 40 years did not meaningfully reduce mean mortality rates (0.1% [0.1%-0.2%] to 0.3% [0.2%-0.3%]) but was estimated to add 649 (602-695) to 650 (603-696) false-positive screenings and 58 (41-76) to 59 (41-76) benign biopsies per 1000 women. CONCLUSIONS AND RELEVANCE: This analysis suggests that annual MRI screening starting at 30 to 35 years followed by annual MRI and mammography at 40 years may reduce breast cancer mortality by more than 50% for women with ATM, CHEK2, and PALB2 pathogenic variants. In the setting of MRI screening, mammography prior to 40 years may offer little additional benefit.

Independent External Validation of Artificial Intelligence Algorithms for Automated Interpretation of Screening Mammography: A Systematic Review.

PURPOSE: The aim of this study was to describe the current state of science regarding independent external validation of artificial intelligence (AI) technologies for screening mammography. METHODS: A systematic review was performed across five databases (Embase, PubMed, IEEE Explore, Engineer Village, and arXiv) through December 10, 2020. Studies that used screening examinations from real-world settings to externally validate AI algorithms for mammographic cancer detection were included. The main outcome was diagnostic accuracy, defined by area under the receiver operating characteristic curve (AUC). Performance was also compared between radiologists and either stand-alone AI or combined radiologist and AI interpretation. Study quality was assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 tool. RESULTS: After data extraction, 13 studies met the inclusion criteria (148,361 total patients). Most studies (77% [n = 10]) evaluated commercially available AI algorithms. Studies included retrospective reader studies (46% [n = 6]), retrospective simulation studies (38% [n = 5]), or both (15% [n = 2]). Across 5 studies comparing stand-alone AI with radiologists, 60% (n = 3) demonstrated improved accuracy with AI (AUC improvement range, 0.02-0.13). All 5 studies comparing combined radiologist and AI interpretation with radiologists alone demonstrated improved accuracy with AI (AUC improvement range, 0.028-0.115). Most studies had risk for bias or applicability concerns for patient selection (69% [n = 9]) and the reference standard (69% [n = 9]). Only two studies obtained ground-truth cancer outcomes through regional cancer registry linkage. CONCLUSIONS: To date, external validation efforts for AI screening mammographic technologies suggest small potential diagnostic accuracy improvements but have been retrospective in nature and suffer from risk for bias and applicability concerns.