PURPOSE: With the development of PARP inhibitors for treatment of cancer patients with an altered BRCA1 or BRCA2 gene, there is an urgent need to ensure that there are appropriate strategies for identifying mutation carriers whilst balancing the increased demand for and cost of cancer genetics services. To date, the majority of mutation prediction tools have been developed in women of European descent where the age and cancer-subtype distributions are different from that in Asian women. METHODS: In this study, we built a new model (ARiCa: Asian Risk Calculator) for estimating the likelihood of carrying a pathogenic variant in BRCA1 or BRCA2 gene, using germline BRCA genetic testing results in a cross-sectional population-based study of 8,162 Asian breast cancer patients. We compared the model performance to existing mutation prediction models. The models were evaluated for discrimination and calibration. RESULTS: ARiCa included age of diagnosis, ethnicity, bilateral breast cancer, tumour biomarkers, and family history of breast cancer or ovarian cancer as predictors. The inclusion of tumour grade improved significantly the model performance. The full model was calibrated (Hosmer-Lemeshow p-value=0.614) and discriminated well between BRCA and non-BRCA pathogenic variant carriers (Area Under Receiver Operating Curve 0.80, 95% Confidence Interval=0.75-0.84). Addition of grade to the existing clinical genetic testing criteria targeting breast cancer patients below 45 years reduced the proportion of patients referred for genetic counselling and testing from 37% to 33% (p-value=0.003), thereby improving the overall efficacy. CONCLUSION: Population-specific customisation of mutation prediction models and clinical genetic testing criteria improved the accuracy of BRCA mutation prediction in Asian patients.