Web-based predictive tool for vaginal and vulvar melanomas: a machine learning study.

Background: One in every 41 women develops malignant melanoma in their lifetime, with noncutaneous melanomas arising in areas such as the genitourinary (GU) system being particularly rare and aggressive. We used machine learning (ML) to build prognostic models for vaginal (VaM) and vulvar (VuM) melanomas and developed the first predictive web-based tool for survival in these cancers.

Methods: We leveraged the SEER database (2000-2020) to assemble our cohort and extract relevant clinical and demographic variables. Prognostic factors were screened using univariate and multivariate Cox proportional hazards regression analyses. Subsequently, we developed five machine-learning classifiers to predict 5-year survival. The discrimination of each model was assessed using the area under the receiver operating characteristic curve (AUC-ROC), and calibration was examined to ensure reliability. Kaplan-Meier analyses were performed to visualize survival distributions across key subgroups.

Results: This study included 1575 patients, of whom 372 and 1203 had VaM and VuM, respectively. The median patient age was 67 years, and the median tumor size was 2.4 cm. The 5-year survival rate of patients with VuM (45.4%) was significantly higher than that of patients with VaM (15.2%) (P < 0.001).

Conclusions: This study highlights the aggressive nature of rare GU melanomas and the importance of surgical intervention and caution in the use of chemotherapy and radiotherapy.