Prediction models for breast cancer prognosis among Asian women

Background Robust and reliable prognosis prediction models have not been developed and validated for Asian patients with breast cancer, a rapidly growing yet understudied population in the United States. Methods We used longitudinal data from the Shanghai Breast Cancer Survival Study, a population-based prospective cohort study (n = 5042), to develop prediction models for 5- and 10-year disease-free survival (DFS) and overall survival (OS). The initial models considered age at diagnosis, tumor grade, tumor size, number of positive nodes, TNM stage, chemotherapy, tamoxifen therapy, and estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) status. We then evaluated whether the addition of modifiable lifestyle factors (physical activity, soy isoflavones intake, and postdiagnostic weight change) improved the models. All final models have been validated internally and externally in the National Cancer Database when applicable. Results Our final models included age at diagnosis, tumor grade, tumor size, number of positive nodes, TNM stage, chemotherapy, tamoxifen therapy, ER status, PR status, 6-month postdiagnostic weight change, interaction between ER status and tamoxifen therapy, and interaction between age and TNM stage. The internal validation yielded C-statistics of 0.76, 0.74, 0.78, and 0.75 for 5-year DFS, 10-year DFS, 5-year OS, and 10-year OS, respectively. The external validation yielded C-statistics of 5- and 10-year OS both at 0.78 for Chinese ethnicity, 0.79 for East Asian ethnicity, and 0.75 and 0.76 for all ethnic groups combined. Conclusion We developed prediction models for breast cancer prognosis from a large prospective study. Our prognostic models performed very well in women from the United States-particularly in Asian American women-and demonstrated high prediction accuracy and generalizability.

Automated summary

Robust and reliable prognosis prediction models have been developed for Asian patients with breast cancer, incorporating age, tumor characteristics, treatment information, and lifestyle factors. The models showed high prediction accuracy and generalizability, particularly in Asian American women, after internal and external validation.