Critical Assessment of Clinical Prognostic Tools in Melanoma.

The 7th edition American Joint Committee on Cancer (AJCC) melanoma staging system classifies patients according to prognosis. Significant within-stage heterogeneity remains and the inclusion of additional clinicopathologic and other host- and tumor-based prognostic factors have been proposed. Clinical prognostic tools have been developed for use in clinical practice to refine survival estimates. Little is known about the comparative features of tools in melanoma. We performed a systematic search of the scientific published literature for clinical prognostic tools in melanoma and web-based resources. A priori criteria were used to evaluate their quality and clinical relevance, and included intended clinical use, model development approaches, validation strategies, and performance metrics. We identified 17 clinical prognostic tools for primary cutaneous melanoma. Patients with stages I-III and T1 or thin melanoma were the most frequently considered populations. Seventy-five percent of tools were developed using data collected from patients diagnosed in 2006 or earlier, and the well-established factors of tumor thickness, ulceration, and age were included in 70 % of tools. Internal validity using cross-validation or bootstrapping techniques was performed for two tools only. Fewer than half were evaluated for external validity; however, when done, the appropriate statistical methodology was applied and results indicated good generalizability. Several clinical prognostic tools have the potential to refine survival estimates for individual melanoma patients; however, there is a great opportunity to improve these tools and to foster the development of new, validated tools by the inclusion of contemporary clinicopathological covariates and by using improved statistical and methodological approaches.

Duke Scholars

Author Susan Halabi Biostatistics & Bioinformatics, Division of Biostatistics