Health Disparities
Development and Validation of a Breast Cancer Risk Prediction model for Puerto Rican women Mary V Diaz Santana* Mary Diaz Santana Molly Rogers Clarice Weinberg Dale Sandler
Background
Breast cancer (BC) is the leading cause of cancer death among Hispanic women. Country of birth has been reported to modify BC risk in this population of women. Compared to Hispanic women in the US, BC mortality rates are higher in Puerto Rico (PR). BC risk prediction models are needed to identify women at high risk who would benefit from early detection. Previously developed risk prediction tools for Hispanic women have been reported to underestimate or overestimate risk in Hispanic women. We seek to develop a BC risk prediction model, specific to Puerto Rican women, that includes socioeconomic determinants of health.
Methods
We used data on Puerto Rican women (315 cases; 348 controls) from the ATABEY case-control study to estimate BC relative risk and attributable risk for each woman. Puerto Rican-specific absolute risks were estimated by accounting for SEER age-specific and country-specific BC incidence and competing mortality from the PR Cancer Registry. We validated our model in prospective data from 1,108 Puerto Rican participants from the Sister Study. We assessed model performance by calibration (expected/observed (E/O) cases) and discrimination accuracy (AUC). The model was built using the Individualized Coherent Absolute Risk Estimator R package.
Results
The Puerto Rican-specific risk model included age at menarche, age at first full-term pregnancy, family history of breast cancer, biopsy for benign breast disease, and socioeconomic factors. In independent data from the Sister Study, the Puerto Rican-specific model was well calibrated (E/O) ratio 1.02 (95%CI, 0.71-1.47). The AUC was 0.71 (95%CI, 0.60-0.81).
Conclusions
This is the first absolute risk model that is specific to Puerto Rican women. Discriminatory accuracy for this model was higher than what has been reported using the Hispanic Risk Model. This suggests that the use of the Puerto Rican-specific risk prediction model can improve breast cancer risk stratification for Puerto Rican women.
