Melanoma Risk Prediction Models Show Moderate Accuracy in Large Australian Cohort
Key Takeaways
Six melanoma risk prediction models showed varying discriminatory performance in an Australian cohort of up to more than 43,000 adults.
Most models underestimated melanoma incidence, indicating limitations in predicting absolute 10-year risk, according to the new study in JAMA Dermatology.
The findings support the potential role of melanoma risk prediction models in risk stratification, the authors concluded.
Six published melanoma risk prediction models demonstrated varying accuracy in identifying individuals who developed melanoma, with most underestimating absolute risk, according to a cohort study published in JAMA Dermatology.
Melanoma Risk Prediction Accuracy and Study Findings
Researchers evaluated six models using data from the Sax Institute's 45 and Up Study, a population-based cohort of New South Wales residents aged 45 years or older. Participants were recruited between 2005 and 2009, with cancer registry data linked through December 2019.
The analysis included up to 43,235 participants who completed the 2020 follow-up survey. Median age was 57 years, and 55.9% were female. Investigators identified 946 incident invasive melanomas and 1,406 in situ melanomas.
The primary outcomes were discrimination, or the ability to distinguish individuals who developed melanoma from those who did not, and calibration, which measures agreement between predicted and observed 10-year melanoma risk.
Across the six models, area under the receiver operating characteristic curve (AUC) values ranged from 0.58 (95% CI, 0.56-0.60) to 0.70 (95% CI, 0.68-0.72).
Most models underestimated melanoma incidence. The two models with the best calibration had calibration-in-the-large values of −0.16 and 0.59, with corresponding calibration slopes of 0.74 and 0.72.
The investigators noted that recalibration may improve population-level absolute risk estimates. Study considerations include the Australian cohort, restriction to follow-up survey respondents, and potential differences in model performance across populations.
"In this cohort study, several prediction models demonstrated moderate discriminatory performance comparable to their original studies, supporting their potential use for risk stratification to guide prevention and early detection strategies," the authors wrote. "Some recalibration may be needed to improve estimation of absolute risks at the population level."
Source
Reyes-Marcelino G, et al. JAMA Dermatology. Published online October 21, 2026. Doi:10.1001/jamadermatol.2026.4097