Quantifying the additional prognostic value of sentinel lymph node status over established clinicopathologic primary tumor features for predicting survival outcomes in melanoma patients.

Abstract

e22105 Background: Sentinel node (SN) staging not only provides important prognostic information for patients with primary cutaneous melanoma but it also assists in assessing eligibility for adjuvant systemic therapy and clinical trials. This study aimed to quantify the incremental value of SN status over established clinicopathologic prognostic factors in predicting survival outcomes. Methods: The training cohort (n = 9272) consisted of all Dutch melanoma patients who underwent SN biopsy (SNB) between 2004 and 2014. Melanoma patients having SNB during the same period were identified from the Melanoma Institute Australia (MIA) database as the validation cohort (n = 5644). Outcomes investigated included overall survival (OS), disease-free survival (DFS) and melanoma-specific survival (MSS, available for the MIA cohort only). Multivariable Cox proportional hazards analyses were performed with and without SN status and compared for each of the outcomes. The effect of adding SN status on the predictive performance of the models was quantified through discrimination index (Harrell’s C-statistic), calibration (plots) and reclassification (Net Reclassification Index (NRI)) at fixed follow-up time points. Results: In the Dutch cohort adding SN status to a model with Breslow thickness, ulceration, mitotic rate, age, gender, melanoma subtype, anatomic site and regression, improved the C-statistic from 0.72 (95% CI 0.70-0.73) to 0.76 (95% CI 0.75-0.77) for DFS and from 0.75 (95% CI 0.73-0.76) to 0.78 (95% CI 0.77-0.79) for OS. These findings were confirmed in the validation cohort, with the C-statistic increasing from 0.69 (95% CI 0.66-0.72) to 0.73 (95% CI 0.70-0.76) for DFS, 0.69 (95% CI 0.66-0.73) to 0.73 (95% CI 0.70-0.76) for OS and 0.71 (95% CI 0.69-0.73) to 0.76 (95% CI 0.74-0.78) for MSS. When setting 10% as the cut-off risk in NRI analysis, 54 Dutch and 26 MIA patients who experienced disease progression within the first 3 years after melanoma diagnosis, but were classified as low-risk patients, would have been reclassified as high-risk with addition of SN status. At the same time, 209 Dutch and 230 MIA patients who were disease-free within the first 3 years, but classified as high-risk would have been reclassified as low-risk with addition of SN status. Conclusions: Adding SN status significantly improved the predictive accuracy for OS, DFS and MSS models in melanoma patients.