Evaluating sex-specific prediction models for colorectal cancer risk using a genome-wide polygenic risk score and lifestyle factors in a Japanese population

Abstract

Background

The predictive performance of a colorectal cancer (CRC) risk prediction model incorporating genome-wide polygenic risk scores (PRSs) and lifestyle factors remains unclear in Asian populations. This study aimed to develop and evaluate the Asian-specific models using a Japanese population-based prospective study.

Methods

We derived 31 genome-wide PRSs using a genome-wide association study of CRC from the Biobank Japan and selected the best-performing PRS with the highest C-index in development case-cohort, including 200 incident cases. In evaluation case-cohort, including 693 incident cases, we assessed the discrimination accuracy (C-index, integrated discrimination improvement (IDI), and net reclassification improvement (NRI)) of lifestyle, PRS, and combined models using 5-fold cross-validation methods and estimated 10-year absolute risk.

Results

Of the 31 derived PRSs, the PRS aggregating 104,677 variant risks performed best in the development case-cohort. The men and women in the highest quintiles of the PRS had an approximately three-fold and two-fold higher risk of CRC, respectively, than those in the lowest in the evaluation case-cohort. Meanwhile, the association of lifestyle factors with CRC risk was observed only in men. Incorporating the PRS into a lifestyle model improved the C-index from 0.64 to 0.66 for men and from 0.61 to 0.63 for women. The IDI and NRI values supported this improvement. The 10-year absolute risk was 3.3 % and 1.6 % for high-risk men and women, respectively, and 0.5 % for both low-risk men and women.

Conclusions

This study suggests that the CRC risk prediction model utilizing genome-wide PRS for Asians is valuable; however, further improvement is needed before clinical implementation.