Background/Purpose: Several human leukocyte antigen (HLA) Class I loci encoding the major histocompatibility complex (MHC) have been implicated both in psoriatic arthritis (PsA) susceptibility and in overall severity of disease.  However, associations of these HLA alleles with specific manifestations of PsA have received little study.  In particular, the contribution of HLA Class I alleles to peripheral arthritis mutilans (PAM), a potentially devastating manifestation of PsA, is unclear.

Methods: Patients from two prospective PsA cohorts were genotyped for HLA-B and C alleles.  PAM was defined radiographically as osteolysis affecting≥50% of the articular surface on both sides of a joint.   Associations of alleles with PAM were explored in multivariable logistic regression models adjusting for potentially confounding characteristics.  An allele scores was constructed as the sum of an individual’s positively associated alleles minus the sum of their inversely associated alleles.

Results: A total of 501 PsA patients were studied [mean age=56 years, 53% female, median PsA duration=18 years].  PAM was observed in 52 (10%).  B*27:05:02, B*35:01:01, and C*06:02:01 were significantly associated with PAM in multivariable analyses (Fig 1A). There were no significant inversely associated alleles.  One PAM-associated allele was present in 208 (42%) and 35 (7%) had two alleles. Compared with those with no PAM-associated alleles, the odds of PAM was nearly 3.5-fold higher for those with one allele (OR=3.59; p<0.001), and more than 4.5-fold higher for those with two alleles (OR=4.51; p=0.004) after adjusting for PsA duration, which was the only meaningful confounder of the PAM/allele association detected (Fig 1B).  The strongest associations were observed for B*27:05:02 in the context of its B*27:05:02-C*02:02:02 haplotype (OR=4.70 compared with those with no B*27:05:02; p=0.001), B*35:01:01 in the context of its B*35:01:01-C*04:01:01 haplotype (OR=3.46 compared with those with no B*35:01:01; p=0.014), and C*06:02:01 in the context of its two common haplotypes (B*13:02:01-C*06:02:01 and B*57:01:01-C*06:02:01: OR=1.93 compared with those with no C*06:02:01; p=0.041).  The area under the receiver operator curve (AUC) for detecting PAM knowing only the PAM-associated HLA-B and C alleles in their haplotype contexts was 0.682 (95% CI 0.610, 0.766).  Adding PsA duration to the model increased the AUC to 0.738 (95% CI 0.672, 0.805).

Conclusion: The associations of specific HLA-B and C alleles observed with PAM supports the concept that the particular PsA sub-phenotype observed in an individual is genetically determined.  Additional study is warranted to define how penetrance is determined among those with susceptibility alleles