Statistics from Altmetric.com
Editor,—I read with great interest the paper by Kilmartin et al.1 The incidence of 25% (six patients out of 24) of HLA-B51 among the 24 patients with Behçet’s disease is interpreted as showing a “primary association” between this HLA antigen and Behçet’s disease in Ireland.
In my opinion, the far reaching conclusions of these authors is questionable as the control group consisting of “96 healthy voluntary blood donors to the Blood Transfusion Service Board” may not be an adequate sample for comparison.
The HLA frequencies in the 24 patients with Behçet’s disease as outlined in Table 1 of their paper show that of the 52 HLA antigens tested, 20 antigens were absent (0.00 frequency) in the group of Behçet’s patients and were present in the control group. Moreover, four other HLA antigens (of the 52 tested) were absent in the control group and detected only in the patient group. The absence of 46% of the tested HLA antigens in either group and their presence in the other group indicate unequivocally that the group of volunteers chosen as control for the frequency of the HLA-B51 is inadequate. This paper thus exemplifies once again the need for proper controls in order to circumvent the unavoidable pitfalls of possible “clustering” of antigens among a small group of patients (or controls).
When dealing with small groups of patients, as in this study, the proper control group for comparison should be collected from a proportionate number of volunteers living in the same region or the same village as the patients. Preferably, patients attending the same ophthalmology clinic serving the area should serve as controls. For example, if two patients with Behçet’s disease were diagnosed and referred from clinic X, then eight non-Behçet’s patients (to form the same ratio for the total number of patients to controls—one to four—as performed in this study) from the same ophthalmic clinic should be included within the control group. This is especially important in countries where heterogeneous genetic haplotypes may be found in different regions.
It is undeniable that the most important findings of this study are the observations that all six Behçet’s patients with HLA-B51 were male and the fact that 18 other Behçet’s patients were negative for this antigen. Therefore, crucial data to be analysed would be comparison of the various HLA antigens among the HLA-B51 positive and HLA-B51 negative Behçet’s patients. Furthermore, the study of HLA haplotypes of the family members of these two groups (HLA-B51 positive and negative Behçet’s patients) may shed more light on whether HLA-B51 is truly associated with Behçet’s disease in Ireland or is only an incidental finding as a result of clustering of antigens in various regions in this country.
Editor,—We would like to thank Professor BenEzra for the interest shown in our paper but wish to point out that the claim for a primary association between HLA-B51 and Behçet’s disease is strongly supported by the experimental data in our study. Although only 25% of our Behçet’s disease patients were B51 positive, when compared with the very low prevalence of HLA-B51 in the Irish population (2%, 372 of 16 682 healthy volunteer blood donors), and correspondingly in our control group (0%), a highly significant association is found suggesting a strong immunogenetic predisposition but also a multifactorial pathogenesis. BenEzra suggests that the absence of 42% (not 46% as quoted in his letter) of the tested HLA antigens in either the Behçet’s disease patient group or the control group, and their presence in the other group, “unequivocally” indicates that the group of volunteers chosen for the control group is inadequate. Clustering of HLA antigens may indeed occur in a small group of patients or controls, and as BenEzra admits, is an “unavoidable pitfall” associated with a small patient or control group. However, in our study this was due to the inevitably small Behçet’s disease patient group (and appropriate number of controls), owing to the very low prevalence of Behçet’s disease in north western Europe, and not the inadequate nature of our control group. The outstanding Behçet’s disease association in our study was with HLA-B51, with a relative risk of 6.3 and corrected exact p value of 0.002, comparable with previous reports from Japan and Turkey. Were clustering to have a significant effect, other spurious HLA haplotype associations would have been seen but were not, with none of the other HLA antigens showing a significantly increased frequency using the Bonferroni correction, even when an antigen was present in one group and absent in the other group.
We agree that in countries where heterogeneous genetic haplotypes are found in different regions, such as Israel, a proper control group should be from a proportionate number of volunteers living in the same region or village as the patients. In ethnic terms, however, the Republic of Ireland is a “village” and can be considered as one region with a high degree of ethnic homogeneity in this island population at the periphery of north western Europe. As pointed out in our paper, there is only one blood donor service in the Republic of Ireland, operating nationally with one central tissue typing laboratory, allowing healthy voluntary blood donors to be used as representative of the general population, and thus an appropriate control group. Moreover, the geographic distribution of our Behçet’s disease patient group very closely matches the current Irish demographic distribution from census data. We do not agree, however, that patients attending the same ophthalmic clinic serving the area should serve as controls, as de facto this group has ophthalmic disease and the control group should be healthy, as in our control group.
Further data analysis is given in Table 11 showing that clustering of various antigens was not found in the HLA-B51 positive Behçet’s disease patients compared with the HLA-B51 negative Behçet’s disease group. This would make it extremely unlikely that the significant prevalence of HLA-B51 in Irish patients with Behçet’s disease was due to clustering of HLA antigens.