OBJECTIVE:Reconciling cutoff thresholds for short-term (5-year) and long-term (lifetime) breast cancer risk could support tailored and evidence-based approaches to supplemental screening and risk management most relevant to short-term clinical actions. This study aims to consistently classify women at increased risk and provide 5-year risk cutoff that corresponds to a 20% lifetime risk. METHODS:Using U.S. Surveillance, Epidemiology and End Results (SEER) program population incidence data for women 40 to 74 years of age, this study reports both lifetime and 5-year population-based risk estimates controlling for competing risk and age varying breast cancer incidence. A cut point for 5-year risk equivalent to lifetime risk of 20% which triggers increased screening is generated. This computation is a weighted average incorporating age, remaining life expectancy, and population risk distribution. The primary outcome is breast cancer incidence (in situ and invasive). RESULTS:A lifetime risk threshold of 20% corresponded to markedly age-dependent 5-year risk cut points, increasing from ∼1.3% at ages 40-44 to ∼10.9% at ages 70-74. For women 40-74, 20% lifetime risk corresponds to a 5-year risk cut-off of 3.16%. CONCLUSIONS:Aligning lifetime risk of ≥20% and the 5-year breast cancer risk cutoff enhances consistency of classification of women at increased risk and clinical decision-making. Women with a ≥3.16% 5-year risk of breast cancer have risk equivalent to a lifetime risk of ≥20% on average. This can facilitate rational and evidence-based approaches to short-term and long-term risk assessment results for both risk reduction and tailored screening.

Supplementary Figure S1 shows distribution of density measures in external cohort

Abstract Purpose: Existing breast cancer risk prediction models are ineffective for women with benign breast disease (BBD) and do not stratify risk by key histologic subtypes, especially in racially and ethnically diverse populations. Although BBD and breast cancer share several risk factors, women with BBD have a higher underlying risk due to accumulated changes in benign breast tissue compared to the general population. The magnitude of risk varies by morphologic subtype—modestly increased for proliferative disease without atypia (PDWA), and highest for atypical hyperplasia (AH). We aimed to refine a prediction model for breast cancer through risk stratification by PDWA and AH subtypes, using data from a contemporary and racially diverse cohort of women. Methods: 8,870 women diagnosed with histologically confirmed benign lesions were identified at the Siteman Cancer Center in St. Louis from 2010-2023 (followed until June 10, 2025). Risk modeling was performed with Cox regression and internally validated through a bootstrap resampling method (i.e., corrected for overfitting). Model discrimination was estimated with Harrell's C-index. Model calibration was assessed by the 5-year observed-to-expected (O/E) ratio. Positive and negative predictive values (PPV/NPV) were estimated for a 3% 5-year risk threshold per national guidelines to define increased risk. Results: Among the 8,870 women with BBD, there were 362 subsequent breast cancers at least 6 months following a benign biopsy. 11.2% were triple negative tumors (ER-PR-HER2-); 73.8% were invasive breast cancers. The cohort is 64.3% White, 32.5% Black, and 3.2% Asian. On average, women were 50 (SD=13) years of age at biopsy. Average age at menarche was 13 (SD=1.8) and average BMI 29.6 kg/m2 (SD=7.7). Harrell’s C-index was 0.68 (95% CI: 0.65-0.71) in the original sample and 0.67 (95% CI: 0.65-0.70) in the bootstrap sample. At 5 years, the overall model calibration was 0.99 (95% CI: 0.86-1.13). Women predicted to be in the high-risk group (≥3%, 5y) included those with atypia and half of those with proliferative disease without atypia. Of these women, 7.0% developed breast cancer, and in the average-risk group, 97.5% did not in the 5 years following a benign biopsy. Conclusion: Our model demonstrates sound discriminatory ability in the internal validation within a racially diverse cohort of women followed after benign biopsy. With a 5-year PPV of 7%, the model effectively flags women at high-risk (≥3%, 5y), providing a meaningful basis for intensified surveillance or preventive measures for early detection after benign biopsy showing proliferative disease with or without atypia. Citation Format: Alzina Koric, Yikyung Park, Shu Jiang, Fouad Boulos, Debbie L. Bennett, Graham A. Colditz. Breast cancer risk prediction model for racially diverse women with benign breast disease [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 937.

Conventional prediction models incorporating genetic and clinical factors including breast density underperform in non- European populations. We investigated an artificial intelligence-derived mammogram risk score (MRS), a summary of texture features that captures intrinsic breast- tissue characteristics-the substrate for cancer devel-opment. This study leveraged data from two North American screening cohorts totaling >226,000 women, includ-ing non- Hispanic white, non- Hispanic Black, East Asian, South Asian, and Indigenous women. MRS distributions showed nonsignificant shifts (and similar SDs) between cohorts and across race and ethnic subgroups. MRS in-creased with age and was significantly associated with breast cancer risk, with hazard ratios per SD ranging from 2.24 [95% confidence interval (CI), 2.03 to 2.46] to 2.32 (95% CI, 2.25 to 2.39) after age adjustment. Associations remained significant within all subgroups. Calibration was excellent across the racial and ethnic groups and across full- field digital mammograms and tomosynthesis. These findings establish MRS as a strong predictor that is inde-pendent of race or ethnicity, demonstrating its potential for broader clinical utility.

Supplementary Figure S3 shows density estimation by radiologists vs deep learning model for digital breast tomosynthesis

Multi-omics analysis offers unparalleled insights into the interlinked molecular interactions that govern the underlying biological processes. In the era of big data, driven by the emergence of high-throughput technologies, it is possible to gain a more comprehensive and detailed understanding of complex systems. Nevertheless, the challenges lie in developing methods to effectively integrate and analyze this wealth of data. This challenge is even more apparent when the type of-omics data (e.g., pathomics) lacks pixel-to-pixel or region-to-region correspondence across the population. A novel sample-specific cooperative learning framework is introduced, designed to adaptively manage diverse multi-omics data types, even when there is no direct correspondence between regions. The proposed framework is defined for both continuous and categorical outcomes, with theoretical guarantees based on finite samples. Model performance is demonstrated and compared with existing methods using real-world datasets involving proteomics and metabolomics, and radiomics and pathomics.

BACKGROUND:Adult survivors of childhood cancer are at higher risk of premature aging compared to their cancer-free peers due to the cancer and its treatments. However, little is known about the effect of adherence to healthy dietary patterns on aging in childhood cancer survivors. METHODS:A cross-sectional analysis was conducted of 3322 participants (mean age, 30.5 years; standard deviation [SD], 8.4) from the St. Jude Lifetime Cohort Study. Diet was measured by a food frequency questionnaire and used to assess the Healthy Eating Index (HEI)-2015 and alternate Mediterranean diet (aMED) scores. Premature aging was assessed by the deficit accumulation index and categorized into low, medium, and high risk. Multinomial logistic regressions adjusting for confounders were used to estimate odds ratios (ORs) with 95% confidence intervlas (CIs). RESULTS:The mean (SD) HEI-2015 score was 60.0 (10.9) of 100, and the aMED score was 4.2 (2.0) of 9. Twenty percent and 8% of survivors were in the medium and high deficit accumulation index categories, respectively. Higher adherence to HEI-2015 (ORhigh vs. low = 0.80; 95% CI, 0.69-0.93 per 10-point increment) and aMED (ORhigh vs. low = 0.91; 95% CI, 0.84-0.98 per 1-point increment) were associated with a lower risk of premature aging. The associations remained consistent among survivors who received radiation or chemotherapy. CONCLUSION:Adherence to a healthy diet may contribute to reducing the premature aging risk in adult survivors of childhood cancer. Interventions that support healthy eating in this population could potentially have benefits for long-term health outcomes.

Abstract Integrating longitudinal data with survival models is a prevalent strategy for dynamic survival risk prediction while accounting for subjects' longitudinally observed variables. However, existing methods primarily focus on scalar longitudinal data and seldom tackle the complexities associated with high‐dimensional longitudinal imaging data. This article introduces a new approach that effectively incorporates longitudinal medical images as features for dynamic risk prediction. This approach ensures interpretability, enhances computational efficiency, and performs robustly even with small datasets. Our method achieves high prediction accuracy, as validated through extensive simulation studies and a real‐world application to Alzheimer's disease data. To the best of our knowledge, this is the first attempt to use longitudinal medical images for predicting dynamic survival risk.

Supplementary Figure S2 shows density estimation by radiologists vs deep learning model

Purpose: Personalized risk communication (PRCom) can help individuals make informed decisions about electing to undergo cancer screening based on their specific risk. In this systematic review, we examined existing risk communication strategies used in the presentation of PRCom and how they may impact screening-related cognitive and behavioral outcomes. Methods: We searched Embase.com, Ovid-Medline All, Scopus, CINAHL Complete, and CENTRAL Trials (via Cochrane Library) for randomized controlled trials addressing interventions with an element of “personalized” cancer risk communication based on the individual’s risk factors. Results: Eighteen unique studies were included. From these 18 studies, we extracted data on PRCom presentation methods and outcome measures such as knowledge and screening completion. Text was used in the majority of PRCom (56

Supplementary Figure S3: Simple directed acyclic graph showing a new marker partially mediated through an existing predictor

Importance:For breast cancer risk prediction to be clinically useful, it must be accurate and applicable to diverse groups of women across multiple settings. Objective:To examine whether a dynamic risk prediction model incorporating prior mammograms, previously validated in Black and White women, could predict future risk of breast cancer across a racially and ethnically diverse population in a population-based screening program. Design, Setting, and Participants:This prognostic study included women aged 40 to 74 years with 1 or more screening mammograms drawn from the British Columbia Breast Screening Program from January 1, 2013, to December 31, 2019, with follow-up via linkage to the British Columbia Cancer Registry through June 2023. This provincial, organized screening program offers screening mammography with full field digital mammography (FFDM) every 2 years. Data were analyzed from May to August 2024. Exposure:FFDM-based, artificial intelligence-generated mammogram risk score (MRS), including up to 4 years of prior mammograms. Main Outcomes and Measures:The primary outcomes were 5-year risk of breast cancer (measured with the area under the receiver operating characteristic curve [AUROC]) and absolute risk of breast cancer calibrated to the US Surveillance, Epidemiology, and End Results incidence rates. Results:Among 206 929 women (mean [SD] age, 56.1 [9.7] years; of 118 093 with data on race, there were 34 266 East Asian; 1946 Indigenous; 6116 South Asian; and 66 742 White women), there were 4168 pathology-confirmed incident breast cancers diagnosed through June 2023. Mean (SD) follow-up time was 5.3 (3.0) years. Using up to 4 years of prior mammogram images in addition to the most current mammogram, a 5-year AUROC of 0.78 (95% CI, 0.77-0.80) was obtained based on analysis of images alone. Performance was consistent across subgroups defined by race and ethnicity in East Asian (AUROC, 0.77; 95% CI, 0.75-0.79), Indigenous (AUROC, 0.77; 95% CI 0.71-0.83), and South Asian (AUROC, 0.75; 95% CI 0.71-0.79) women. Stratification by age gave a 5-year AUROC of 0.76 (95% CI, 0.74-0.78) for women aged 50 years or younger and 0.80 (95% CI, 0.78-0.82) for women older than 50 years. There were 18 839 participants (9.0%) with a 5-year risk greater than 3%, and the positive predictive value was 4.9% with an incidence of 11.8 per 1000 person-years. Conclusions and Relevance:A dynamic MRS generated from both current and prior mammograms showed robust performance across diverse racial and ethnic populations in a province-wide screening program starting from age 40 years, reflecting improved accuracy for racially and ethnically diverse populations.

A functional multistate model is presented, which accommodates Markov processes governing disease transition in a finite set of states. Importantly, we consider a setting where the set of predictors contains a high-dimensional image with the goal of quantifying the association between the image and the transition of disease states. In the motivating application of breast cancer, women start from normal breast tissue, go through benign lesions, and then to the onset of DCIS/invasive cancer. As in the real data application, we consider the setting in which the individuals are observed intermittently and the transition times are interval censored. A score test is developed to test the nullity of the coefficient function for the image predictor at different transitions between states. The asymptotic distribution of the score statistic is provided. An application involving progression to the development of breast cancer with mammogram image data provides illustration. Our results demonstrate an important association between the mammogram image and the probability of transition in breast cancer.

Introduction:The mammogram risk score (MRS), an AI-driven texture feature derived from digital mammograms, strongly predicts breast cancer risk independently of breast density, though underlying mechanisms remain unclear. This study investigated relationships between established breast cancer risk factors, covering anthropometrics, reproductive factors, family history, and mammographic density metrics, and MRS. Methods:Using data from the Nurses' Health Study II (292 cases, 561 controls), we validated MRS's association with breast cancer using logistic regression and evaluated its relationships with risk factors through: linear regressions of MRS on observed risk factors and polygenic scores associated with risk factors, and Mendelian randomization (MR) analysis via two-stage least squares regression. We conducted two-sample MR of MRS using summary statistics from genome-wide association studies of risk factors. Results:MRS was significantly associated with breast cancer risk before adjustment for BI-RADS density (OR=1.92 per SD increase in MRS; 95%CI:1.57-2.33; AUC=0.69) and after (OR=1.85; 95%CI:1.49-2.30). Early life body size and adult body mass index (BMI) were inversely associated with MRS, while history of benign breast disease and BI-RADS density showed positive associations; after adjusting for BI-RADS density, associations between MRS and the other three risk factors attenuated. Higher polygenic score for dense area was associated with increased MRS (β=0.16 SD increase in MRS per SD increase in polygenic score; 95%CI: 0.06-0.25), as was percent density (β=0.14; 95%CI:0.05-0.23). Two-sample MR identified associations between genetically predicted dense area (β=0.83 SD increase in MRS per SD increase in dense area; 95%CI:0.39-1.27) and percent density (β=1.14; 95%CI:0.55-1.74) with MRS. After adjusting for BI-RADS density and BMI, higher waist-to-hip ratio was significantly associated with increased MRS in polygenic score and two-sample MR analyses. No significant associations were observed with other risk factors. Conclusion:We validated MRS's association with breast cancer risk in cases diagnosed 0.5-10.1 years (median 2.6) after mammogram acquisition. Our findings reveal robust associations between breast density measures and MRS and suggest a potential impact of central obesity on MRS. Future larger-scale studies are crucial to validate these results and explore their potential to enhance our understanding of breast cancer etiology and refine risk prediction models.

Screening digital breast tomosynthesis (DBT) aims to identify breast cancer early when treatment is most effective, leading to reduced mortality. In addition to early detection, the information contained within DBT images may also inform subsequent risk stratification and guide risk-reducing management. Using transfer learning, we refined a model in the Joanne Knight Breast Health Cohort at Washington University, a cohort of 5,066 women with DBT screening (mean age, 54.6), among whom 105 were diagnosed with breast cancer (26 ductal carcinoma in situ). We applied the model to external data from the Emory Breast Imaging Dataset, a cohort of 7,017 women free from cancer (mean age, 55.4), among whom 111 pathology-confirmed breast cancer cases were diagnosed more than 6 months after initial DBT (17 ductal carcinoma in situ). We obtained a 5-year AUC of 0.75 [95% confidence interval (CI), 0.73-0.78] in the internal validation. The model validated in external data gave an AUC of 0.72 (95% CI, 0.69-0.75). The AUC was unchanged when age and Breast Imaging-Reporting and Data System density were added to the model with synthetic DBT images. The model significantly outperforms the Tyrer-Cuzick model, with a 5-year AUC of 0.56 (95% CI, 0.54-0.58; P < 0.01). Our model extends risk prediction applications to synthetic DBT, provides 5-year risk estimates, and is readily calibrated to national risk strata for clinical translation and guideline-driven risk management. The model could be implemented within any digital mammography program. Prevention Relevance: We develop and externally validate a 5-year risk prediction model for breast cancer using synthetic DBT and demonstrate clinical utility by calibrating to the national risk strata as defined in breast cancer risk management guidelines.

Abstract Mammographic density is a strong risk factor for breast cancer and is reported clinically as part of Breast Imaging Reporting and Data System (BI-RADS) results issued by radiologists. Automated assessment of density is needed that can be used for both full-field digital mammography (FFDM) and digital breast tomosynthesis (DBT) as both types of exams are acquired in standard clinical practice. We trained a deep learning model to automate the estimation of BI-RADS density from a prospective Washington University clinic-based cohort of 9,714 women, entering into the cohort in 2013 with follow-up through October 31, 2020. The cohort included 27% non-Hispanic Black women. The trained algorithm was assessed in an external validation cohort that included 18,360 women screened at Emory from January 1, 2013, and followed up through December 31, 2020, that included 42% non-Hispanic Black women. Our model-estimated BI-RADS density demonstrated substantial agreement with the density as assessed by radiologists. In the external validation, the agreement with radiologists for category B 81% and C 77% for FFDM and B 83% and C 74% for DBT shows important distinction for separation of women with dense breast. We obtained a Cohen’s κ of 0.72 (95% confidence interval, 0.71–0.73) in FFDM and 0.71 (95% confidence interval, 0.69–0.73) in DBT. We provided a consistent and fully automated BI-RADS estimation for both FFDM and DBT using a deep learning model. The software can be easily implemented anywhere for clinical use and risk prediction. Prevention Relevance: The proposed model can reduce interobserver variability in BI-RADS density assessment, thereby providing more standard and consistent density assessment for use in decisions about supplemental screening and risk assessment.

Supplementary Table S6: Hazards ratio of association among premenopausal women for baseline risk factors and breast cancer risk utilizing FULL-VPD

Supplementary Figure S2: Distribution of estimated risk scores at cohort entry