Cancer
Comparison of Breast Cancer Risk Prediction in Models in a Population of Hispanic Women Molly Rogers* Molly Rogers Annie DiFrank Min Shi Clarice Weinberg Mary Diaz-Santana
Background: Among Hispanic women, breast cancer is the most frequently diagnosed malignancy and the leading cause of cancer death. Risk prediction models are used as a tool in precision prevention for risk stratification. However, widely used breast cancer risk prediction tools have been reported to overall either underestimate or overestimate breast cancer risk in Hispanic women. We used data on Hispanic women from the Sister Study to compare the performance of two frequently used breast cancer risk prediction models.
Methods: Using longitudinal data from Hispanic women in the Sister Study (N = 2,462) we compared the predictive performance of two breast cancer risk prediction models, the Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA) and the Breast Cancer Risk Assessment Tool (BCRAT). We compared model performance at 5 years follow-up in all Hispanic women, as well as stratified by country of birth Calibration was assessed by calculating the expected to observed ratios(E/O) and 95% confidence intervals. Model discrimination was examined using ROC curves and c-statistics.
Results: At 5 years follow-up 60 breast cancer cases had accrued. The BOADICEA model had better calibration than BCRAT overall (E/O 0.87(95%CI 0.68-1.12) vs. 0.76(95%CI 0.59-0.98)) and in US-born Hispanic women(E/O 0.74(95%CI 0.52-1.05) for BOADICEA, and 0.47(95%CI 0.33-0.67) for BCRAT). The two models were similarly calibrated in foreign-born Hispanic women(E/O 1.01(95%CI 0.70-1.46) for BOADICEA, and 1.06(95%CI 0.74-1.53) for BCRAT). The estimated c-statistic(AUC) for overall Hispanic women was .64(95%CI 0.57-0.71) for BOADICEA and .57(95%CI 0.50-0.64) for BCRAT.
Conclusions: Our results suggest that both models underestimate BC risk in Hispanic women. However, BOADICEA performs better than BCRAT, particularly within the US-born Hispanic population. Further research should be done to improve risk prediction models specific to Hispanic women.
