Purpose
Molecular classification using gene expression profiling (GEP) is currently the most accurate single predictor of mortality from uveal melanoma. The purpose of this study was to determine whether clinical or cytopathologic features may provide independent prognostic value in assessing mortality risk.
Methods
336 consecutive cases of uveal melanoma were analyzed using Kaplan-Meier survival analysis and Cox regression. The primary outcome measure was 5-year all cause mortality. Dependent variables included GEP molecular classification (class 1 vs. 2), patient age and sex, tumor thickness and basal diameter, ciliary body involvement and cell type.
Results
After controlling for GEP molecular classification, no clinicopathologic features provided independent prognostic information except basal tumor diameter (p=0.002, Cox regression). For both class 1 and class 2 tumors, the best dichotomous cutoff for basal tumor diameter was <12mm (small) vs. ≥12mm (large). The risk of mortality relative to small class 1 tumors was 8.6 for large class 1 tumors, 9.1 for small class 2 tumors, and 78.1 for large class 2 tumors (all p<0.05). Similar results were demonstrated for melanoma-specific mortality. The model-based predicted 5-year survival was 98% for small class 1 tumors, 85-87% for large class 1 and small class 2 tumors, and 16% for large class 2 tumors.
Conclusions
For both class 1 and class 2 uveal melanomas, basal tumor diameter was the only clinicopathologic factor that provided additional prognostic information that enhanced the accuracy of GEP molecular classification. Basal tumor diameter, in combination with GEP, may be a useful metric for risk stratification, especially for tumors less than 12mm in diameter.