Abstract 3594: Performance of common general-population breast cancer risk prediction models in 14 cohorts
Bibliographic record
Abstract
Abstract Background: Several breast cancer (BC) risk prediction models are used in clinical practice to identify women eligible for enhanced screening or prevention strategies. While these models have been independently validated in specific and often separate contexts, their performance has not been systematically evaluated and compared across a wide range of populations or age ranges. Methods: We collected individual-level baseline questionnaire data and incident cancer diagnoses from 14 cohorts participating in the Breast Cancer Risk Prediction Project (BCRPP), representing the United States (N=12), Canada (N=1) and Australia (N=1). After harmonizing data across cohorts, five-year absolute risk estimates for invasive breast cancer were derived using four models: BCRAT, iCARE-Lit, Tyrer-Cuzick (all estimating risk in women aged ≥20-75 years) and the Black Women’s Health Study (BWHS) calculator (estimating risk in Black women aged ≥30-70 years). Using the iCARE-calibrate function in R we estimated the area under the curve (AUC) and 95% confidence intervals (CI) within each cohort, using absolute risk designations to incorporate age. Expected to observed (E/O) absolute risks were estimated on average and within expected absolute risk deciles for each cohort. To align with its intended use in clinical practice, calibration of the BWHS calculator (and comparison to other models) was performed within the subset of Black women pooled from all cohorts. Results: A total of 1, 041, 708 women, enrolled in studies between 1976-2015, were included. Within five years of baseline cohort entry, 116, 113 invasive breast cancer cases were diagnosed. Mean age at baseline ranged from 34 to 62 years. Within cohorts, discrimination was similar across models but varied substantially across cohorts. Age-incorporated AUCs ranged from 0.55-0.77, with most cohort-specific AUCs under 0.65 and higher AUCs observed among younger cohorts. Calibration, measured by E/O ratios, differed substantially by both cohort and risk prediction model. For the majority of cohort-model combinations, risk was overestimated in the upper risk deciles. In general, E/O ratios were more similar for BCRAT and Tyrer-Cuzick models compared to iCARE-Lit, which tended to overestimate risk more on average. In the subset of Black women aged ≥30-70 years (N=84, 594, N=824 invasive cases in 5 years), age-incorporated AUCs ranged from 0.61 to 0.64. While risk was overestimated in the upper risk deciles for all models, overestimation was 13-118% lower for the BWHS calculator. Conclusion: The discrimination and calibration of existing risk prediction models varied across studies. Future model development including additional risk factors (e.g., genetic and mammographic information) should leverage diverse training data and flexible models to ensure risk estimates perform well across different regions, countries, and ethnicities. Citation Format: Kristen D. Brantley, Thomas U. Ahearn, Emily Norton, Julie Palmer, Gary Zirpoli, Matt Barnett, Marian L. Neuhouser, Lauren Teras, James Hodge, Thomas E. Rohan, Roger Milne, A. Heather Eliassen, Hongyan Huang, Yu Chen, Katie M. O'Brien, Cari Kitahara, Garnet Anderson, I-Min Lee, Nilanjan Chatterjee, Montserrat Garcia-Closas, Peter Kraft, Breast Cancer Risk Prediction Project. Performance of common general-population breast cancer risk prediction models in 14 cohorts [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3594.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".