Materials and methods

Twenty four patients were recruited from the ophthalmology departments of university teaching hospitals and by referral from requests to ophthalmologists, rheumatologists, dermatologists, and genitourinary physicians throughout the Republic of Ireland. All were required to satisfy the ISG criteria for diagnosis of Behçet’s disease.23 Hospital charts were reviewed retrospectively and each patient was reassessed by one ophthalmologist (DK) and a blood sample for HLA typing was obtained after informed consent.

HLA typing was carried out by the National Tissue Typing Reference Laboratory of the Blood Transfusion Service Board. HLA-A and HLA-B typing was performed by the standard National Institutes of Health microlymphocytotoxicity method and HLA-DR typing was performed by DNA polymerase chain reaction for both patients and controls. The control group consisted of 96 healthy voluntary blood donors to the Blood Transfusion Service Board, of whom all had HLA-A and HLA-B typing and 84 had HLA-DR typing. Age and sex matched consecutively typed controls were chosen for each patient so that both patient and controls were typed at the same time using the same antisera. There is only one blood donor service in the Republic of Ireland which operates nationally with one central tissue typing laboratory allowing voluntary blood donors to be used as representative of the general population. Antigen frequencies were counted by direct counting. Statistical significance between the relative frequencies of antigens in patients and controls was determined by χ² and Fisher’s exact tests with Yates’s correction. Exact p values were corrected to avoid a type I error, as outlined by Meenken et al,24 for, in a sample of 53 informative comparisons, one may expect to observe, by chance alone, two to three ‘significant’ HLA associations. The correction is performed by multiplying the exact p value with the number of antigen frequencies tested (a factor of 53 in our study). All p values given are two tailed. Significance was evaluated by 0.05>p>0.01 is significant and p<0.01 is highly significant.

Any HLA-B51 association with certain clinical features, such as sex, age at first manifestation of disease, age at diagnosis, the presence of eye involvement, visual acuity worse than 6/60 in an affected eye, and requirement for systemic cyclosporin A therapy, was assessed. First manifestation of disease was defined as the first manifestation of one of the ISG diagnostic criteria (including recurrent mouth ulcers). Age at diagnosis was defined as the age at which the ISG criteria for diagnosis were fulfilled. Statistical association with age was assessed using the Mann-Whitney U test for non-parametric data.

Results

Of the 24 patients with Behçet’s disease, all had HLA-A and HLA-B typing and 21 underwent HLA-DR typing. There were 14 male and 10 female patients. All patients were white and of ethnic native Irish origin. HLA antigen frequencies of patients and controls are given in Table 1. A remarkably significant HLA association in patients with Behçet’s disease in the Republic of Ireland was found with HLA-B51 (exact p value = 0.00004, χ² = 20.3, relative risk = 6.3). When the exact p value was corrected for the number of HLA antigens tested (53), this HLA-B51 association remained highly significant (corrected exact p value = 0.002). The HLA-B51 antigen was found in six out of 24 patients (25%) compared with none of the 96 controls.

An increased frequency was observed for HLA-B7 (exact p value = 0.008) in the BD group compared with the control group. We also observed an increased frequency of two HLA class II antigens, HLA-DR6 (exact p value = 0.007, relative risk =5.7) and HLA-DR2 (exact p value = 0.041), in patients with Behçet’s disease compared with controls. However, after correction of the exact p values for the number of antigens tested (53), the observed differences for HLA-B7, HLA-DR6, and HLA-DR2 did not reach a level of significance.

Sex differences in HLA-B51 status are outlined in Table 2. Of patients with the HLA-B51 antigen, all six were male (exact p value = 0.024) with a male sex:HLA-B51 association achieving probable significance. All 10 female patients were negative for the HLA-B51 antigen. Age at first manifestation and at diagnosis of Behçet’s disease with respect to HLA-B51 status is given in Table 3. No difference was found within the BD group for age at first manifestation (exact p value = 0.160) between patients who were HLA-B51 antigen positive (mean age 25.5 years (SD 8.12)) and those who were HLA-B51 negative (mean age 19.2 years (7.28)). Mean age at diagnosis of BD was 28.5 years (8.26) in the HLA-B51 positive group compared with 30.0 years (5.73) in the HLA-B51 negative group with no significant difference (exact p value = 0.867).

Nineteen patients (79.2%) in the BD group had eye involvement with five possessing the HLA-B51 antigen while 14 patients did not. There was no association between HLA-B51 positivity and eye involvement (exact p value = 1.00). No association was seen between the HLA-B51 antigen and requirement for systemic cyclosporin A therapy (exact p value = 1.00) or visual acuity worse than 6/60 in patients with eye involvement (exact p value = 1.00, Table 4). Of six patients with a visual acuity worse than 6/60 in an affected eye, only two possessed the HLA-B51 antigen.

Discussion

The most significant finding in our study is a highly significant association between HLA-B51 and Behçet’s disease (corrected exact p value = 0.002) in the Republic of Ireland which has an ethnically homogeneous white population. Mizuki et al 5has suggested that the same HLA-B51 association exists worldwide and this has been confirmed in many different ethnic groups, but mostly from the Far East and Mediterranean countries.47-13This HLA-B51 association is unclear, however, in white patients from north western Europe and the USA.

Previous older studies from Britain1819 and the USA1620 in the 1970s do not show any HLA-B5 association. O’Duffy et al found no HLA association in 26 US patients16 and more recently stated in an editorial in theNew England Journal of Medicine17 that no HLA association existed in either US or British patients. In the largest American series of BD patients to date (35 patients), Foster et al could not find a HLA-B51 association.22 Previous North American studies in whites have been, however, from ethnically diverse groups.162223 Lehner et al 19 typed 52 white British patients and found no HLA-B5 association, verifying the results of a previous British study,18 though there was an association (uncorrected p = 0.002) with the ‘ocular type’ using an unusual classification system. The only previous Irish study on Behçet’s disease found no HLA-B5 association in a small group of six typed patients.25 The results of these older studies are difficult to interpret as they were performed before the HLA-B5 complex was differentiated into its subantigens, HLA-B51, B52, and B53. In addition, different and older diagnostic criteria for diagnosis were used before the advent of the ISG criteria, therefore making comparison different. A more recent study in 14 ethnically homogeneous Scottish patients using the ISG criteria found no HLA-B5 proband association.26 However, the HLA-B5 antigen was not subtyped and while only 13% Scottish BD patients were HLA-B5 positive the authors did not report the HLA-B5 frequency in normal Scottish controls precluding statistical comparison.

In Germany, however, two studies reported an increased frequency of HLA-B5 in BD patients of 23% (compared with 14% of controls)27 and 67%28 using the ISG criteria. Comparison was made of German BD patients with Mediterranean patients and not with normal controls in the latter study.28 Zouboulis et al 27reported a statistically significant HLA-B5 association (p<0.05) with a low relative risk of 2.6 but did not correct p values to avoid a type I error24 which would not produce a significant difference between groups.

Both these studies had inappropriate statistical analyses,2728 comparing a small number of patients with a very much greater number of controls.29 Associations between HLA and diseases should be investigated with an equal number of patients and controls for a near optimal estimation of relative risk.30 However, the optimal sampling ratio in our group for a precise estimate of relative risk is a control group four times the number of patients,30 and this increases the power of the test to an optimal level beyond which diminishing power gains will be achieved.31

Support for an immunogenetic HLA-B51 predisposition came from Villaneuva et al 32 in a Mediterranean family from the south of Spain. They described three sisters with BD (ISG criteria) out of six siblings, with all affected siblings possessing identical HLA haplotypes and all siblings possessing the HLA-B51 antigen.

Behçet’s disease is relatively rare in north western Europe and the significance of studies on smaller series of European patients is particularly difficult to interpret.3 There are no prevalence data for BD in Ireland and our study is the largest Irish series to date. Assuming a roughly similar prevalence to Yorkshire18 of 0.6/100 000 population, one would expect there to be 21 or 22 patients with BD in the Republic of Ireland from current census data.33 Our study on 24 patients has probably recruited the vast majority of BD patients in the Republic of Ireland. Despite HLA-B51 being present in only 25% of Irish BD patients, when compared with B51 positivity in none of the controls, this produces a relative risk of 6.3. While the proportion of HLA-B51 in BD patients is much higher in Japan45 and Turkey,3435 57% and 82% with relative risks of 7.9 and 14 respectively, the high relative risk in our study is similar. While our study reports a highly significant HLA-B51 association and strong immunogenetic predisposition for BD in those who possess the HLA-B51 antigen, 75% of Irish BD patients were HLA-B51 negative and indicates a multifactorial pathogenesis for BD. This is supported by the Spanish familial study,32 and another Japanese family with BD,36 where all affected and unaffected siblings possessed the HLA-B51 antigen. The B51 antigen has been recently subdivided into three alleles, HLA-B*5101, HLA-B*5102, and HLA-B*5103, with all B51 positive patients in a Japanese study37 possessing HLA-B*5101. This allele differs from the other two by a single amino acid substitution in each case, and this would merit further study in our B51 positive Irish patients.

In support of previous studies of BD,1927 we have found a significant association of HLA-B51 positivity with male sex (exact p value = 0.024). However, with respect to other clinical associations, such as age at first presentation/diagnosis, eye involvement, and severity of ocular disease/poor visual acuity, we found no HLA-B51 association, in common with many other studies from the Far East and Europe.2738-41 Wechsler et al have questioned whether the HLA-B51 antigen is relevant for the severity of Behçet’s disease.42 We would suggest that it is not, and we attribute particular importance to the distinct lack of HLA-B51 association with severity of ocular disease found in this study. Two recent studies from Japan39 and Korea41 also could not find a HLA-B51 association with poor visual prognosis.

In conclusion, our study found a highly significant HLA-B51 association with BD in Ireland, with a high relative risk. A positive HLA-B51:male sex association and no association with visual prognosis in Behçet’s disease was confirmed. This study supports a HLA-B51 immunogenetic predisposition in an ethnically homogeneous white population in north western Europe, far from the Silk Route.