Background: After treatment of a first primary cutaneous melanoma localised to the skin, individuals are at increased risk of both subsequent new primary melanoma and recurrence of the index melanomas. Current follow-up recommendations are largely based on AJCC stage and primarily reflect recurrence risk, despite subsequent new primary melanoma being a major concern after early-stage disease. Risk prediction models could support more personalised, risk-stratified surveillance, but their availability and scope are unclear. This systematic review aimed to identify and appraise validated risk prediction models for subsequent new primary melanoma in adults with a history of localised cutaneous melanoma, and to determine whether any models jointly predict both new primary melanoma and recurrence.
Methods: We searched PubMed, CENTRAL, and Embase from inception to 20 January 2026 for studies reporting risk prediction models in adults with a history of at least one primary cutaneous melanoma (invasive or in situ). Two reviewers independently screened titles and abstracts, assessed full-text articles, extracted data using the CHARMS checklist, and evaluated risk of bias with PROBAST.
Results: Of 3540 records identified, only one study met the inclusion criteria. This model, developed from an Australian population-based cohort linked to the New South Wales Cancer Registry, predicted risk of subsequent new primary melanoma using demographic, phenotypic, genetic, histopathological, and behavioural predictors. Model discrimination was moderate, with Harrell’s C-statistics of 0.73 for a second primary melanoma and 0.65 for third and fourth melanomas. Calibration suggested underestimation of second melanoma risk and overestimation of third melanoma risk. The included model had not undergone external validation and did not assess clinical utility. No study reported a model that jointly predicted both subsequent new primary melanoma and recurrence of the index melanoma.
Conclusions: There is a substantial evidence gap in risk-stratified melanoma surveillance. Only one eligible study identified in this review examined risk prediction for subsequent new primary melanoma, and no validated model was found that also incorporated recurrence risk. Development and external validation of models addressing both outcomes would support more personalised follow-up after localised melanoma.