Rxivist logo

Prediction of Colorectal Cancer Risk Based on Profiling with Common Genetic Variants

By Xue Li, Maria Timofeeva, Athina Spiliopoulou, Paul McKeigue, Yazhou He, Xiaomeng Zhang, Victoria Svinti, Harry Campbell, Richard Houlston, Ian PM Tomlinson, Susan M Farrington, Malcolm G Dunlop, Evropi Theodoratou

Posted 29 Oct 2019
medRxiv DOI: 10.1101/19010116

BackgroundStratifying the risk of colorectal cancer (CRC) based on polygenic risk scores (PRSs) within populations has the potential to optimize screening and develop targeted prevention strategies. MethodsA meta-analysis of eleven genome-wide association studies (GWAS), comprising 16 871 cases and 26 328 controls, was performed to capture CRC susceptibility variants. Genetic models with several candidate PRSs were generated from Scottish CRC case-control studies (6478 cases and 11 043 controls) for prediction of overall and site-specific CRC. Model performance was validated in UK Biobank (4800 cases and 20 287 controls). The 10-year absolute risk of CRC was estimated by modelling PRS with age and sex using the CRC incidence and mortality rates in the UK population. FindingsA weighted PRS including 116 CRC SNPs (wPRS116) showed the strongest performance. Deconstructing the PRS into multiple genetic risk regional scores or inclusion of additional SNPs that did not reach genome-wide significance did not provide any further improvement on predictive performance. The odds ratio (OR) for CRC risk per SD of wPRS116 in Scottish dataset was 1{middle dot}46 (95%CI: 1{middle dot}41-1{middle dot}50, c-statistics: 0{middle dot}603). Consistent estimates were observed in UK Biobank (OR=1{middle dot}49, 95%CI: 1{middle dot}44-1{middle dot}54, c-statistics: 0{middle dot}610) and showed no substantial heterogeneity among tumor sites. Compared to the middle quintile, those in the highest 1% of PRSs had 3{middle dot}25-fold higher risk and those in the lowest 1% had 0{middle dot}32-fold lower risk of developing CRC. Modelling PRS with age and sex in the general UK population allows the identification of a high-risk group with 10-year absolute risk [≥]5%. InterpretationBy optimizing wPRS116, we show that genetic factors increase predictive performance but this increment is equivalent to the extraction of only one-tenth of the genetic susceptibility. When employing genetic risk profiling in population settings it provides a degree of risk discrimination that could, in principle, be integrated into population-based screening programs.

Download data

  • Downloaded 212 times
  • Download rankings, all-time:
    • Site-wide: 99,252
    • In epidemiology: 4,044
  • Year to date:
    • Site-wide: 106,396
  • Since beginning of last month:
    • Site-wide: 90,781

Altmetric data

Downloads over time

Distribution of downloads per paper, site-wide


Sign up for the Rxivist weekly newsletter! (Click here for more details.)