Evaluation and Enhancement of the Prognostic Ability of the Eighth Edition of TNM Staging in Cutaneous Malignant Melanoma: A Population-Based Study of 111,817 Cases Using Machine Learning

Comprehensive Summary

This population-based study evaluated the prognostic performance of the 8th edition TNM staging system (TNM-8) for overall survival (OS) and melanoma-specific survival (MSS) in patients with cutaneous malignant melanoma (CMM), using data from the SEER database (2018-2022). A total of 111,871 adult patients were included. Kaplan-Meier survival analysis demonstrated that most TNM-8 stage categories showed clear survival discrimination; however, overlap was observed between N1 and N2 stages at 12 months and between TNM stages II and III at multiple time points (12,36,48, and 59 months). Find Gray competing-risk and accelerated failure time (AFT) models identified several independent prognostic factors influencing OS and MSS, including age, sex, race/ethnicity, tumor site, histologic subtype, ulceration status, Breslow thickness, mitotic rate, number of positive lymph nodes, and M stage. Machine learning (ML) models were then developed to compare prognostic performance using TNM staging alone versus TNM combined with clinically relevant variables.

Outcomes and Implications

While TNM-8 staging retained meaningful prognostic value, ML-based models demonstrated superior discrimination when incorporating additional clinicopathologic variables. TNM alone showed moderate predictive ability (testing C-index range: 0.67-0.72 for OS, 0.82-0.87 for MSS), whereas integrated ML models achieved substantially higher discrimination (testing C-index range: 0.84-0.85 for OS, 0.89-0.92 for MSS). Notably, TNM staging performed better in predicting MSS than OS, highlighting limitations in capturing non-melanoma-related mortality risks. These findings suggest that AI-enhanced prognostic modeling may complement traditional staging systems, enabling more individualized risk stratification. The study supports the development of interactive clinical decision-support tools that combine TNM staging with patient-specific variables, advancing precision oncology in melanoma care.