Evaluation of a polygenic risk score for breast cancer in Chinese women: A three-stage case-control study of 13,715 subjects.

10553 Background: Polygenic risk score (PRS) is a valuable tool for predicting the risk of breast cancer (BC). However, limited studies have been conducted in Chinese women. This study aimed to develop and validate a PRS which could be used to identify individuals with high risk of breast cancer. The associations between the PRS and patients' clinicopathological characteristics or survival outcomes were also evaluated. Methods: The PRS was developed based on findings from genome-wide association studies (GWAS) and validated in four independent cohorts with a three-stage design. A total of 7,056 patients and 6,659 controls were enrolled from Fujian Medical University Union Hospital (FJMUUH) and Shanghai Breast Cancer Genetics Study (SBCGS). Five approaches were utilized to calculate the PRS, including repeated logistic regression (RLR), logistic ridge regression (LRR), artificial neural network (ANN), random forest (RF) and support vector machine (SVM). Logistic regression analyses were performed to assess the association between established PRS and clinicopathological characteristics. The correlation between PRS and patients' survival outcomes was evaluated by cox regression models. Results: The LRR-based PRS was indicated to have the best predictive accuracy among five approaches (AUC = 0.601, OR per 1 SD increase = 1.39, Table 1). Women in the top 5% and 80-95% percentiles of PRS had a 1.43-fold and a 1.34-fold elevated risk of developing breast cancer compared with those at average risk (PRS in 40-60th percentiles). The predictive performance of PRS for patients with HER-2 positive tumors was demonstrated to be higher than that of HER-2 negative tumors (AUC = 0.612 vs 0.585, OR = 1.47 vs 1.35). It was also identified that the PRS was not correlated with age at diagnosis nor tumor characteristics. In survival analyses, an increase in PRS was associated with unfavorable disease-free survival (DFS) for ER negative patients (HR = 1.48, 95% CI = 1.08-2.03, p = 0.016). However, this association diminished after adjusting for clinicopathological characteristics. Regarding overall survival (OS), we observed significant correlations between increased PRS and overall survival (adjusted HR = 1.35, 95% CI = 1.05-1.75, p = 0.021), especially among ER negative patients (adjusted HR = 1.57, 95% CI = 1.07-2.30, p = 0.021) in the multivariate model. Conclusions: The PRS could provide additional information for Chinese women at high risk of breast cancer and holds significant value for BC screening. An increase in PRS also indicates an unfavorable prognosis and could play a crucial role in the clinical management of breast cancer patients at the time of diagnosis.