Systematic evaluation of narrow‐sense validity of polygenic risk score for prostate cancer in a Chinese prostate biopsy cohort

• Published in: Clinical genetics Vol. 103; no. 6; pp. 636 - 643
• Main Authors: Wu, Yishuo, Ruan, Xiaohao, Gao, Peng, Da, Huang, Fang, Zujun, Xu, Danfeng, Jiang, Haowen, Ding, Qiang, Lin, Xiaoling, Lu, Daru, Na, Rong
• Format: Journal Article
• Language: English
• Published: Oxford, UK Blackwell Publishing Ltd 01.06.2023
• Subjects: benchmark; Biopsy; Calibration; East Asian People; Genetic diversity; Genetic Predisposition to Disease; Genome-Wide Association Study; Humans; Male; polygenic risk score; Population studies; Prospective Studies; Prostate - pathology; Prostate cancer; Prostatic Neoplasms - genetics; Prostatic Neoplasms - pathology; Risk assessment; Risk Assessment - methods; Risk Factors; Validity; biopsy; prostate cancer; validity; polygenic risk score; benchmark
• ISSN: 0009-9163; 1399-0004; 1399-0004
• DOI: 10.1111/cge.14315

The aim of this study was to assess the narrow‐sense validity of polygenic risk score (PRS) for prostate cancer (PCa) in a Chinese prostate biopsy cohort. We performed an observational prospective study with 2640 men who underwent prostate biopsy. Germline DNA samples were genotyped and PRS was calculated for each subject using 17 PCa risk‐associated genetic variants. Additional GWAS data of the ChinaPCa dataset was also used to compliment the evaluation process. The mean PRS was 1.02 in patients with negative biopsy results, which met the baseline benchmark. The mean PRS was significantly higher in the PCa cases (1.32 vs. 1.02, p = 5.56 × 10−17). Significant dose–response associations between PRS values and odds ratios for PCa were observed. However, the raw calibration slope was 0.524 and the average bias score between the observed risk and uncorrected PRS value was 0.307 in the entire biopsy cohort. After applying a correction factor derived from a training set, the corrected calibration slope improved to 1.002 in a testing set. Similar and satisfied results were also seen in the ChinaPCa dataset and two datasets combined, while the calibration results were inaccurate when the calibration process were performed mutually between two different study populations. In conclusion, assessing the narrow‐sense validity of PRS is necessary prior to its clinical implementation for accurate individual risk assessment. We performed an observational prospective study with 2640 Chinese men who underwent prostate biopsy to assess the narrow‐sense validity of polygenic risk score (PRS) for prostate cancer (PCa). Polygenic risk score (PRS) for prostate cancer (PCa) was calculated for each subject using 17 PCa risk‐associated genetic variants. Additional GWAS data of the ChinaPCa dataset was also used to compliment the evaluation process. Both baseline and calibration benchmarks were evaluated in two datasets. Calibration process by Platt scaling utilizing training and testing sets in both biopsy and ChinaPCa datasets showed improved calibration slope and reduced bias score, indicating the scaled PRS would be more reflective of the relative risk of PCa in specific study populations. Therefore, we propose that assessing the narrow‐sense validity of PRS is necessary prior to its clinical implementation for accurate individual risk assessment.