Statistics from Altmetric.com
We read with great interest the paper by Laivoranta-Nyman et al.1 They studied Finnish patients with rheumatoid arthritis (RA) and controls, and suggested that there exist susceptibility, neutral and protective HLA-DR-DQ haplotypes that do not match the predictions of current hypotheses for the mechanism of association of the HLA class II region and RA.
The analysis performed, however, does not take into account the effect of marked overrepresentation of some HLA-DRB1 alleles among the cases on the relative frequencies of other alleles in the cases and controls. We believe that this is the reason for the unexpected findings. If a relative predispositional effects model is used, which does take this bias into account, it can be seen that the results are quite different (table 1).
We have called alleles “significant” if they achieved a Bonferroni corrected p value of <0.05, correcting for 21 alleles (equivalent to an uncorrected p value of <0.0024). This analysis shows that five haplotypes are overrepresented: (DRB1*0401-DQA1*03-DQB1*0301; DRB1*0408-DQA1*03-DQB1*0301; DRB1*0401-DQA1*03-DQB1*0302; DRB1*0404-DQA1*03-DQB1*0302; DRB1*(01/10)-DQA1*01-DQB1*0501), and that all other haplotypes are not significantly different from neutral (pcorr >0.2). All these haplotypes have previously been shown to be associated with RA, and carry the HLA-DRB1 shared epitope.
These findings do not support either the Zanelli/rheumatoid arthritis protection model,2 or the model proposed by de Vries et al,3 as no protective haplotypes were observed. The lack of support for these models may represent type II error, and the question of the protective effect of DRB1*0103 has not been dealt with because the DRB1*1 subtype was not further subdivided. None the less, in general, the results are most consistent with the shared epitope hypothesis.4
Our analysis agrees with the finding that different DRB1*0401-DQ haplotypes have different strengths of disease association. This is consistent with previous studies, which demonstrate differential association of major histocompatibility complex (MHC) class III-DRB1*0401 haplotypes with RA (reviewed by Newton et al5). What the true disease associated gene is on these haplotypes is as yet unknown, but the findings of Laivorante-Nyman et al1 strongly support the existence of further MHC genes influencing RA.
We thank Dr Harney and collaborators for their interest and comments on our recent paper.1 We agree that the use of the relative predispositional effects (RPE) in the comparison of the frequencies of the HLA haplotypes between patients and controls is a valuable method in clarifying the primary and secondary associations.
When dealing with less common haplotypes, and especially when applied to genotype analysis, the statistical power of the method is, however, limited. One possibility would of course have been to classify haplotypes based on the presence of shared epitope and DERAA motif, which might have given a significant protective effect also using the RPE method. We decided primarily to group haplotypes simply based on their increase or decrease among patients without any preassumptions of mechanisms, although we called the haplotype groups susceptibility and protective. This grouping was further used to search for genotype effects and its correlation with the presented models of susceptibility and protection was compared.
Results obtained are far from conclusive but demonstrate lines for further studies still much needed in the field, and we agree with the conclusion that the existence of further major histocompatibility genes influencing RA susceptibility is probable.