Key result
Uncalibrated UKPDS OM2 underestimates 10-year mortality by up to ~30% in diabetic breast cancer survivors.
Why the study?
Older breast cancer survivors have high rates of T2DM and CVD mortality, but the applicability of the UKPDS OM2 microsimulation model to this population is unknown.
Does calibration of the UKPDS Outcomes model 2 improve 10-year mortality prediction in older breast cancer survivors with diabetes?
Observational (n=5,938)
Yes
Does calibration of the UKPDS Outcomes model 2 improve 10-year mortality prediction in older breast cancer survivors with diabetes?
The widely used UKPDS Outcomes model 2 underestimates mortality in older breast cancer survivors with diabetes, but calibration significantly improves its predictive accuracy.
Answers are grounded in this paper and cite their sources.
May support calibrating UKPDS OM2 for mortality prediction in this population; leaves open prospective validation before clinical use.
Introduction and Objective: Older breast cancer (BC) survivors have a high rates of type 2 diabetes (T2DM) and cardiovascular disease (CVD) mortality. The UKPDS Outcomes model 2 (OM2) is a widely used DM microsimulation model to predict health outcomes. Though validated in US populations, OM2 applicability to BC survivors, who have distinct mortality and CV risk, is unknown. As accurate risk prediction is critical for DM clinical decision making in BC survivors, we compared OM2-predicted with observed mortality in BC survivors and calibrated OM2 to improve mortality prediction. Methods: We identified women ≥66 years with Stage II-III BC and DM from 2008-2019 US Surveillance, Epidemiology, and End Results (SEER) data linked to Medicare. Model inputs were derived from claims; missing biometric variables were imputed with predictive means matching from NHANES. Using published OM2 equations, we predicted 10-year overall mortality and compared estimates to SEER-Medicare rates using Kaplan-Meier curves. Calibration iteratively scaled baseline hazards (time-dependent mortality risk) by BC stage to align predicted and observed survival curves, for which goodness of fit was assessed with observed expected (O/E) ratios. Results: Among 5938 survivors (77% Stage II), median age was 75 years (interquartile range 70-81), and 40% had preexisting CVD. Observed 10-year mortality was 64% (95% confidence interval [CI] 61%-66%) for Stage II and 76% (95% CI 72%-80%) for Stage III survivors. Uncalibrated OM2 underestimated 10-year mortality, predicting 60% [95% CI 58%-61%] for Stage II (O/E 1.1) and 60% [95% CI 57%-63%] for Stage III BC (O/E 1.3). After calibration, O/E ratios were 1.0 and predicted and observed mortality aligned across age groups. Conclusion: A widely used DM model substantially underestimated mortality in older BC survivors, with implications for DM clinical decision making. Calibration improved accuracy, highlighting the need to adapt T2DM models in cancer survivors. Disclosure A. Leiter: None. B.S. Ferket: None. T. Moin: None. J.J. Lin: None. C. Kong: None. J. Wisnivesky: Consultant; Current; Sanofi, PPD, Banook, AMA. Research Support; Current; Regeneron, Sanofi, Axella, BI.
No takes yet. Share an insight, caveat, or question.
LEITER et al. (2026) conducted an observational in Stage II-III breast cancer and type 2 diabetes (n=5,938). UKPDS Outcomes model 2 (OM2) vs. Observed SEER-Medicare mortality rates was evaluated on 10-year overall mortality (O/E ratio 1.1 for Stage II, 1.3 for Stage III). The uncalibrated UKPDS OM2 model underestimated 10-year mortality in older breast cancer survivors with diabetes (predicted 60% vs observed 64% for Stage II, and 60% vs 76% for Stage III).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: