Development and validation of time-to-event models to predict metastatic recurrence of localized cutaneous melanoma.

BACKGROUND: The recent expansion of immunotherapy for stage IIB/IIC melanoma highlights a growing clinical need to identify patients at high risk of metastatic recurrence and, therefore, most likely to benefit from this therapeutic modality. OBJECTIVE: To develop time-to-event risk prediction models for melanoma metastatic recurrence. METHODS: Patients diagnosed with stage I/II primary cutaneous melanoma between 2000 and 2020 at Mass General Brigham and Dana-Farber Cancer Institute were included. Melanoma recurrence date and type were determined by chart review. Thirty clinicopathologic factors were extracted from electronic health records. Three types of time-to-event machine-learning models were evaluated internally and externally in the distant versus locoregional/nonrecurrence prediction. RESULTS: This study included 954 melanomas (155 distant, 163 locoregional, and 636 1:2 matched nonrecurrences). Distant recurrences were associated with worse survival compared to locoregional/nonrecurrences (HR: 6.21, P < .001) and to locoregional recurrences only (HR: 5.79, P < .001). The Gradient Boosting Survival model achieved the best performance (concordance index: 0.816; time-dependent AUC: 0.842; Brier score: 0.103) in the external validation. LIMITATIONS: Retrospective nature and cohort from one geography. CONCLUSIONS: These results suggest that time-to-event machine-learning models can reliably predict the metastatic recurrence from localized melanoma and help identify high-risk patients who are most likely to benefit from immunotherapy.

Spatial habitat heterogeneity influences host-pathogen dynamics in a patchy population of Ranchman's tiger moth.

Host-pathogen dynamics are influenced by many factors that vary locally, but models of disease rarely consider dynamics across spatially heterogeneous environments. In addition, theory predicts that dispersal will influence host-pathogen dynamics of populations that are linked, although this has not been examined empirically in natural systems. We examined the spatial dynamics of a patchy population of tiger moths and its baculovirus pathogen, in which habitat type and weather influence dynamics. Theoretical models of host-baculovirus dynamics predict that such variation in dynamics between habitat types could be driven by a range of factors, of which we predict two are likely to be operating in this system: (1) differences in the environmental persistence of pathogens or (2) differences in host intrinsic rates of increase. We used time series models and monitored infection rates of hosts to characterize population and disease dynamics and distinguish between these possibilities. We also examined the role of host dispersal (connectivity) and weather as important contributors to dynamics, using time series models and experiments. We found that the population growth rate was higher, delayed density dependence was weaker, and long-period oscillations had lower amplitudes in high-quality habitat patches. The infection rate was higher on average in high-quality habitat, and this was likely to have been driven by higher mean population densities and no differences in pathogen persistence in different habitats (delayed density dependence). Time series modeling and experiments also showed an interactive effect of temperature and precipitation on moth population growth rates (likely caused by variation in host plant quality and quantity), and an effect of connectivity. Our results showed that spatial heterogeneity, connectivity, climate, and their interactions were important in driving host-baculovirus dynamics. In particular, our study found that connected patches and spatial heterogeneity generated differences in dynamics that only partially aligned with theoretical predictions.