A Risk Prediction Tool for Invasive Melanoma.

Importance: Increasingly, strategies to systematically detect melanomas invoke targeted approaches, whereby those at highest risk are prioritized for skin screening. Many tools exist to predict future melanoma risk, but most have limited accuracy and are potentially biased.

Objectives: To develop an improved melanoma risk prediction tool for invasive melanoma.

Design, setting, and participants: This population-based prospective cohort study (the QSkin Study) in Queensland, Australia, involved 10 years of follow-up from the baseline survey in 2011 and included individuals aged between 40 to 69 years who were melanoma-free at baseline and completed a comprehensive risk factor survey at recruitment. The data analysis was conducted from October 2024 to April 2025.

Exposures: Thirty-one candidate variables collected at baseline were identified a priori as potential predictors of future risk of invasive melanoma.

Main outcomes and measures: Histologically confirmed invasive melanomas newly diagnosed from baseline through to December 31, 2021, captured by data linkage to the Queensland Cancer Register. Follow-up was censored on diagnosis of melanoma in situ or death. Cox proportional hazards models with forward and backward selection approaches were used to identify the best-fitting model.

Results: Of 41 919 eligible participants, 55% were female, and the mean (SD) age at baseline was 55.4 (8.2) years. A total of 706 new invasive melanomas were identified during 401 356 person-years of follow-up. The best-fitting model retained 14 predictors (age, sex, ancestry, nevus density, freckling density, hair color, tanning ability, adult sunburns, family history, other cancer prior to baseline, previous skin cancer excisions, previous actinic keratoses, smoking status, and height) and 2 statistical terms (age squared, age-by-sex interaction), yielding an apparent discriminatory accuracy of 0.74 (95% CI, 0.73-0.76). The Youden index was optimized at a screening threshold selecting the top 40% of predicted risk, which captured 74% of cases (number needed to screen = 32).

Conclusions and relevance: This cohort study has identified an improved tool that offers enhanced accuracy for predicting the future risk of invasive melanoma compared with existing tools.