Predictive models of melanoma metastasis based on dermatoscopy in an international retrospective human reader study
Menée auprès de 30 dermatologues et à l'aide de 776 images dermatoscopiques de mélanomes cutanés de stade IB et supérieur, cette étude rétrospective internationale évalue la performance de trois modèles pour prédire le risque de métastase : le premier basé sur les caractéristiques des images dermatoscopiques, le deuxième sur les caractéristiques histopathologiques et le troisième sur une combinaison des deux
Current melanoma prognostic tools have limited clinical use at the bedside, highlighting the need for more effective biomarkers. Dermatoscopy correlates with established prognostic markers obtained through invasive procedures. However, its direct predictive value for metastasis remains unexplored. In this multinational study, 30 dermatologists evaluated 776 dermatoscopic images of melanomas (stage IB and above) for predefined criteria including structures, colors and vessels. Extensive dermatoscopic ulceration and blue-white veil are associated with increased risk of metastasis in the total cohort and reduced recurrence-free survival in early-stage melanomas, while extensive regression is associated with reduced metastasis risk and improved recurrence-free survival. Three predictive models of metastasis: (1) dermatoscopic features only, (2) histopathologic features only, and (3) a combination of both demonstrate comparable prognostic accuracy. Here, we show that dermatoscopy may offer valuable prognostic insights into melanoma’s biological behavior before excision and help guide therapeutic decisions. Prospective validation in future trials is essential.
Nature Communications , article en libre accès, 2025