Evaluating BRCA mutation risk predictive models in a Chinese cohort in Taiwan

Area of Science:

• Genetics
• Oncology
• Medical Informatics

Background:

• Accurate estimation of cancer susceptibility gene mutation carrier probabilities is crucial for genetic counseling.
• Existing predictive models for BRCA mutations have uncertain applicability in Asian populations.

Purpose of the Study:

• To evaluate the performance of five BRCA mutation risk predictive models in a Chinese cohort.
• To assess model applicability across different patient subgroups, including those with personal and/or family history of breast or ovarian cancer.

Main Methods:

• A cohort of 647 Chinese women underwent germline DNA sequencing for cancer susceptibility genes.
• Performance of five models (BOADICEA, BRCAPRO, Penn II, Myriad, Tyrer-Cuzick) was assessed using areas under the curve (AUCs) on receiver operating characteristic (ROC) curves.
• Subgroup analyses were conducted based on personal and family history, and tumor characteristics (ER/PR/HER2 status).

Main Results:

• Models showed comparable performance to Western cohorts overall (AUCs 0.68-0.75).
• Excellent performance was observed for unaffected women with family history (AUCs 0.92-0.93).
• Models performed poorly for women with personal history but no family history; BOADICEA underestimated and BRCAPRO overestimated risks.
• Models performed better for triple-negative breast cancer (AUC 0.74-0.80) than other subtypes, but incorporating ER/PR/HER2 status did not improve BOADICEA's accuracy.

Conclusions:

• BRCA mutation predictive models exhibit variable utility in Chinese women, with performance dependent on clinical presentation.
• Models are most reliable for unaffected individuals with a family history of breast or ovarian cancer.
• Further refinement of predictive models is needed for the Asian population, particularly for individuals with personal history only.