Performance of Gene Expression Profiling Versus Nomograms in Predicting Sentinel Lymph Node Positivity and Melanoma Recurrence, a Single Institution Retrospective Cohort Study
Recommended Citation
El Jbeily R, Zhao R, Young A, Meltan S, Tisack A, Friedman BJ, Matthews NH. Performance of Gene Expression Profiling Versus Nomograms in Predicting Sentinel Lymph Node Positivity and Melanoma Recurrence, a Single Institution Retrospective Cohort Study. J Am Acad Dermatol 2026; 95(1):AB337.
Document Type
Conference Proceeding
Publication Date
7-1-2026
Publication Title
J Am Acad Dermatol
Keywords
Dermatology
Abstract
Background: Gene Expression Profiling (GEP), like the integrated 31-GEP by Castle Biosciences, and clinicopathologic nomograms, like the Melanoma Institute Australia (MIA) and Memorial Sloan Kettering Cancer Center (MSKCC), were developed to assess risk of sentinel lymph node biopsy (SLNB) positivity in patients with thin-melanomas. Direct head-to-head comparisons of their performance remain limited. Methods: We retrospectively reviewed charts of all melanoma patients who underwent i31-GEP testing at a single institution and calculated MIA and MSKCC nomogram scores. Sensitivity, specificity, AUC, and Net Reclassification Index (NRI) were calculated for SLN positivity prediction at 5% and 10% risk thresholds. Recurrence discrimination was evaluated using AUROC for MIA prognostic score and GEP. Results: Among 20 patients (18 SLN–, 2 SLN+) with results from all tools, MIA and MSKCC achieved perfect sensitivity for SLN positivity at 5% threshold, but low specificity (0.17 and 0.22). i31-GEP showed lower sensitivity (0.50) and moderate specificity (0.44). Among 31 patients (7 recurrences) with MIA and i31-GEP results (MSKCC does not calculate recurrence), the MIA prognostic score (AUROC 0.87, 95% CI 0.74–1.00) demonstrated stronger discrimination for recurrence than i31-GEP (AUROC 0.82, 95% CI 0.66–0.99). i31-GEP positivity was significantly associated with recurrence in non-T1a tumors (Fisher’s exact p = 0.0238; OR = ∞ [ 1.06–∞]) but showed no significant link in T1a tumors (p = 0.111; OR = ∞ [ 0.21–∞]). Conclusions: Clinicopathologic nomograms outperform i31-GEP at predicting SLN sensitivity and disease recurrence. Larger prospective studies are needed to validate these findings.
Volume
95
Issue
1
First Page
AB337
