HLA class II alleles predict recurrence and pattern of failure in early-stage melanoma patients

8503 Background: Prior reports have suggested a link between specific HLA class II alleles, including HLA-DRB1*1101, and melanoma recurrence. We therefore investigated the relationship between HLA class II alleles and clinical outcome in a large population of patients who presented with localized melanoma. Methods: HLA class II alleles were determined by typing of genomic DNA from patients who presented with localized melanoma (AJCC stage IA-IIC). Patients found to have occult regional disease by sentinel node biopsy were excluded. Log-rank and Cox analysis of outcome versus standard prognostic factors and common HLA class II alleles (=5% of the population) was performed. Results: 1,241 patients underwent HLA typing. The median primary tumor thickness was 1.1 mm, 16% were ulcerated, and the median follow-up was 56 months. Tumor thickness, ulceration, and Clark level were independent predictors of recurrence. The HLA class II allele HLA-DRB1*1101 (11% of the population) independently predicted a higher risk of recurrence [hazard ratio (HR) 2.04, P=0.004] while HLA-DRB1*0401 (20% of the population) predicted a lower risk of recurrence (HR 0.42, P=0.004). HLA- DRB1*1101 was a particularly strong predictor of recurrence in stage I patients (HR 2.73, P<0.0001). Among patients who recurred, HLA- DRB1*1101 was an independent predictor of local-regional vs. distant recurrence (HR 2.7, P=0.007), and was more common in those whose first recurrence was isolated to the regional nodal basin, compared to those whose first recurrence was elsewhere or in multiple locations (24% HLA-DRB1*1101-positive vs. 14%, P<0.009). Conclusions: These results confirm that HLA-DRB1*1101 is a genetic marker of increased risk of melanoma recurrence, while HLA-DRB1*0401 is a marker of decreased risk of recurrence. HLA-DRB1*1101 is a particularly strong marker among the group of patients predicted clinically to be at the lowest overall risk for recurrence, and independently predicts pattern of recurrence (regional nodal failure). This information is of value in the development of surveillance strategies for melanoma patients, as well as the selection, stratification and randomization of early-stage melanoma patients for clinical trials. No significant financial relationships to disclose.