Threshold-Based Overlap of Breast Cancer High-Risk Classification Using Family History, Polygenic Risk Scores, and...

Area of Science:

• Genetics and Genomics
• Cancer Epidemiology
• Biostatistics

Background:

• Traditional breast cancer risk models like the Gail model and polygenic risk scores (PRS) provide independent information.
• The overlap and variation in performance of these models across different ancestries, age groups, and disease types remain unclear.

Purpose of the Study:

• To evaluate the performance and overlap of PRS and the Gail model in predicting breast cancer risk.
• To assess how model performance varies by ancestry, age, and disease type (invasive breast cancer, DCIS).

Main Methods:

• A retrospective case-control study of 180,398 women (European and Asian ancestry).
• Logistic regression models were used to estimate odds ratios (ORs) and area under the receiver operating characteristic curve (AUC).
• Analysis included interactions by age and ancestry, and calibration assessments.

Main Results:

• PRS showed stronger association in younger European women (AUC=0.622) than older (AUC=0.653), with lower performance in Asian women (AUC=0.551-0.600).
• Gail model performance was modest overall, and poor in younger Asian women (AUC=0.523-0.533).
• PRS identified more high-risk Asian individuals, while Gail identified more high-risk European individuals. Both models showed miscalibration.

Conclusions:

• Polygenic risk scores enhance breast cancer risk stratification beyond traditional models.
• PRS is particularly valuable for younger women and individuals of Asian ancestry.
• Further research is needed to address model miscalibration and improve risk prediction across diverse populations.