Recommendations for Primary Prevention of Skin Melanoma

Abstract

Melanoma (MEL) is an aggressive form of skin cancer, causing over 60,000 deaths every year and it is considered one of the fastest-growing cancer forms. Genome-wide association studies have identified numerous genetic variants (SNPs) independently associated with MEL. The effects of such SNPs can be combined into a single polygenic risk score (PRS). Stratification of individuals according to PRS could be introduced to the primary prevention of melanoma. Our aim was to combine PRS with health behavior recommendations to develop a personalized recommendation for primary prevention of melanoma. Previously published PRS models for predicting the risk of melanoma were collected from the literature. Models were validated on the UK Biobank dataset consisting of a total of 487,410 quality-controlled genotypes with 3791 prevalent and 2345 incident cases. The best performing sex-specific models were selected based on the AUC in prevalent data and independently validated on an independent UKBB incident dataset for females and males separately. The best performing model included 28 SNPs. The C-index of the best performing model in the dataset was 0.59 (0.009) and hazard ratio (HR) per unit of PRS was 1.38 (standard error of log (HR) = 0.03) for both males and females. We performed absolute risk simulations on the Estonian population and developed individual risk-based clinical follow-up recommendations. Both models were able to identify individuals with more than a 2-fold risk increase. The observed 10-year risks of developing melanoma for individuals in the 99th percentile exceeded the risk of individuals in the 1st percentile more than 4.5-fold. We have developed a PRS-based recommendations pipeline for individual health behavior suggestions to support melanoma prevention.