43P - Performance evaluation of a combined risk model for breast cancer risk prediction in Indonesian population (TRIP Study)

Background

Breast cancer remains a significant concern worldwide. Risk-based screening, which tailors screening recommendations to individual risk levels, has been shown to enhance patient stratification. However, most research on model development focuses on Western populations, leaving the predictive accuracy of these models for Southeast Asian populations largely uncharacterized.

Methods

We conducted an observational case-control study comprising 305 Indonesian women to assess the applicability of published risk models to the local population. We employed a combined risk model to classify patients as either elevated or average risk. Our combined model evaluates two separate risk factors: genetic risk, assessed using ancestry-adjusted PRS scores based on the Mavaddat model, and clinical risk, evaluated using the Gail model. The performance of each individual model and their combined effectiveness were analyzed using the Area Under the Curve (AUC) and Odds Ratio (OR).

Results

Individual risk models retained their predictive efficacy in the Indonesian context. Specifically, the AUC achieved for genetic risk is AUC of 0.674 (p = 1.28 x 10-3; Risk group: OR = 3.16; p = 2.5 x 10-1). For clinical risk, the AUC stands at 0.674 (p = 5.16 x 10-4; Risk group: OR = 7.636; p = 6.1 x 10-3). Remarkably, when combined, the AUC increased to 0.701 (Risk group: OR= 3.897; p = 4.28 x 10-2), signifying the benefits of a multi-factor model. Based on a subset of the samples taken from this study, the Nala Breast Cancer Risk genetic risk algorithm generated higher AUC when compared to a leading third-party software that uses the same PRS model for breast cancer (0.63 vs 0.55). This improvement is primarily due to our method of translating PRS scores into categorical outcomes, which integrates localized disease incidence and mortality rates.

Conclusions

Our findings demonstrate for the first time the applicability of the Polygenic Risk Score using Mavaddat model and clinical score using Gail model to Indonesian populations. In addition, our study shows that, within this demographic, combined risk models provide a superior predictive framework compared to single-factor approaches.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

NalaGenetics.

Disclosure

The author has declared no conflicts of interest.