The frequency patterns of HLA class I and II were first evaluated in the healthy control and melanoma populations. Eleven HLA alleles were found to be significantly different between the melanoma patients and the healthy control population. For the remainder, no significant differences between the distribution of HLA genotype were found in the melanoma population compared with the healthy control population. The alleles that differed among melanoma patients were HLA-A*03, HLA-B*037, *53, *54, *58, *59, *78, HLA-DRB1*1102, HLA-DQB1*0201, and *0302 with corresponding P values .042, .033, .050, .046, .001, .021, .004, .021, .006 and .033, respectively. Most of these alleles are represented in <10% in both populations.
Patient demographics and baseline characteristics have been described elsewhere.6,21
With a median follow-up of 70.67 months (only among patients alive [censored values], range, 7.1-138.7 months), there were 156 recurrences and 104 deaths. The association of HLA genotype with absence of recurrence is presented in Table . Correlations between nonrecurrence (alive/well vs recurrent/dead) and the presence of MHC class I allele HLA-Cw*06 and MHC class II allele HLA-DRB1*1501 were noted with the corresponding P
values .035 and .045, although these are not corrected for multiple testing. When we examined the gene frequency, no further information was acquired on the predictive value of HLA. This was the case also in comparing gene frequency between healthy controls and melanoma patients.
Association Between HLA Genotype and Recurrence in High-Risk Melanoma Patients Receiving Adjuvant Interferon (Median Follow-up, 71 Months)
Comparison of RFS and OS for previously defined alleles of interest also showed no statistically significant difference among the other alleles investigated (A*02, B*35, Cw*04, Cw*07, DQB1*0301, DRB1*1101, DRB1*03, DRB1*04, DQB1*05 + *06) with the exception of Cw*06. The median RFS of the Cw*06-positive cohort was 100.2 months (range, 2.7-100.2; 95% confidence interval [CI] not evaluable) versus 37.3 months in the Cw*06-negative cohort (range, 1.1-115.1; 95% CI 21.1-53.5; P = .013). The median OS of the Cw*06-positive cohort has not yet been reached (observed minimum and maximum death times 9.8 and 84.2 months; 95% CI not evaluable) versus 78.9 months in the Cw*06-negative cohort (range, 2.3-86.1; 95% CI not evaluable), with a P value of .025. Apart from HLA-Cw*06, stage was found to be an independent predictor for RFS (P = .029) and OS (P = .013) in univariate analysis. The association of HLA-Cw*06 with RFS and OS was explored through a multivariate Cox model including lymph node involvement, ulceration, and stage. In the presence of HLA-Cw*06 in the Cox model, only stage was found to be an independent predictor for RFS (P = .010) and OS (P = .0011). Controlling for disease stage, the P values for the association of Cw*06 with RFS and OS are .018 and .039, respectively. Treatment duration and interaction of treatment duration by HLA-Cw*06 were found nonsignificant, when included in the multivariate Cox models for RFS and OS (P = .20 and P = .46, respectively for RFS; P = .33 and P = .86, respectively for OS), along with HLA-Cw*06 and autoimmunity. The Kaplan-Meier curves for RFS and OS in the A*02 and Cw*06 cohorts are presented in Figures and . The median OS of the B*35-positive cohort was 69.2 months (range, 8.0-86.1; 95% CI, 49.3-89.1) and has not yet been reached for the B*35-negative cohort (range, 2.3-84.2; 95% CI not evaluable), with P = .040 and P = .1383 when controlling for disease stage. No differences were observed in RFS (P = .143 and P = .2141 when controlling for disease stage).
(Top) Relapse-free survival plot shows human leukocyte antigen (HLA)-A2 status (positive, n = 133; negative, n = 151). (Bottom) Overall survival plot shows HLA-A2 status (positive, n = 133; negative, n = 151).
(Top) Relapse-free survival plot shows Cw6 status (positive, n = 55; negative, n = 229). (Bottom) Overall survival plot shows Cw6 status (positive, n = 55; negative, n = 229). HLA indicates human leukocyte antigen.
In the cohort of patients included in the prospective autoimmunity study, allele frequency was investigated in 155 of 200 patients (47 in the autoimmunity group and 108 in the nonautoimmunity group). HLA-Cw*06 was expressed in 29.79% (14 of 47) of patients in the autoimmunity group and 15.38% (16 of 108) of patients in the nonautoimmunity group (P = .049). Statistical significant differences were seen in HLA-B*35 (21.28% vs 41.35%; P = .018) and HLA-DRB1*0701 (23.4% vs 7.69%; P = .015, respectively). No differences were seen in the expression of HLA-A*02 (31.91% vs 48.8%; P = .077).
In the landmark model for the association of Cw*06 with RFS, Cw*06 and autoimmunity were statistically significant (P = .020 and P < .0001, respectively). In the landmark model for the association of Cw*06 with OS, only autoimmunity was statistically significant (P < .0001). In the landmark models for the association of HLA-B*35, HLA-DRB1*0701, and HLA-A*02 with RFS or OS, only autoimmunity was found statistically significant (P < .0001). For all other cases, interaction between alleles and autoimmunity was also tested, but it was found to be insignificant.