Photo Credit: Stefanamer
The following is a summary of “Development and validation of time-to-event models to predict metastatic recurrence of localized cutaneous melanoma,” published in the FEBRUARY 2024 issue of Dermatology by Wan, et al.
The expanding use of immunotherapy for stage IIB/IIC melanoma underscores the importance of identifying patients at high risk of metastatic recurrence to optimize therapeutic outcomes. For a study, researchers sought to develop predictive models for the time-to-event risk of melanoma metastatic recurrence and to create time-to-event risk prediction models for melanoma metastatic recurrence.
The study included patients diagnosed with stage I/II primary cutaneous melanoma between 2000 and 2020 at Mass General Brigham and Dana-Farber Cancer Institute. Melanoma recurrence dates and types were determined through chart reviews, and thirty clinicopathologic factors were extracted from electronic health records. Three types of time-to-event machine-learning models were internally and externally evaluated for predicting distant versus locoregional/nonrecurrence.
A total of 954 melanomas were included (155 distant, 163 locoregional, and 636 1:2 matched nonrecurrences). Distant recurrences were associated with worse survival compared to locoregional/non-recurrences (hazard ratio [HR]: 6.21, P < .001) and locoregional recurrences only (HR: 5.79, P < .001). The Gradient Boosting Survival model demonstrated the best performance (concordance index: 0.816; time-dependent area under the curve: 0.842; Brier score: 0.103) in external validation.
Time-to-event machine-learning models exhibited reliable predictive abilities for metastatic recurrence in localized melanoma. These models can aid in identifying high-risk patients who are more likely to benefit from immunotherapy.