Background: Most patients with early-stage invasive melanoma have excellent outcomes, yet a minority experience disease progression and melanoma-specific death.Identifying these high-risk patients at diagnosis remains a significant clinical challenge.We previously validated the MEL12 microRNA signature as a prognostic biomarker in plasma samples.Here, we evaluate whether this signature retains its prognostic value when measured in formalin-fixed paraffin-embedded (FFPE) tumour tissue and assess its performance across two microRNA expression profiling platforms. Methods:We conducted a retrospective analysis of 277 stage I-II melanoma patients with up to 15-year survival data, stratified into training (n=179) and independent validation (n=98) cohorts.MicroRNA expression profiling was performed using both NanoString and RNA-sequencing platforms.A principal component-based Cox regression model was developed using leave-one-out cross-validation in the training cohort, with frozen model parameters applied to the validation cohort.The primary endpoint was melanoma-specific survival (MSS), with overall survival (OS) as secondary endpoint.Results: In the training cohort, the MEL12 score showed strong prognostic performance for MSS (C-index=0.82,HR=5.32 per SD, p<0.001) and OS (C-index=0.73,HR=2.90 per SD, p<0.001).These findings were confirmed in independent validation for both MSS (C-index=0.77,HR=3.34, p=0.002) and OS (C-index=0.63,HR=1.28, p=0.17).High-risk patients experienced significantly worse 15-year MSS in both cohorts (training HR=9.61, p<0.001; validation HR=8.40, p<0.001).In multivariable analysis adjusting for clinicopathologic variables, the MEL12 score remained independently associated with MSS and platform analysis revealed no significant interaction between MEL12 score and measurement platform. Conclusions:The MEL12 microRNA signature represents a potentially valuable prognostic biomarker for identifying early-stage melanoma patients at high risk of poor outcomes, maintaining consistent performance when measured in FFPE tissue across different expression profiling platforms.These results support further investigation into integrating genomic risk stratification with traditional histopathologic assessment.
Laar et al. (Sun,) studied this question.