ARTHRITIS & RHEUMATISM Vol. 48, No. 6, June 2003, pp 1762–1763 © 2003, American College of Rheumatology CONCISE COMMUNICATION DOI 10.1002/art.11022 Female sex increases risk for rheumatoid arthritis only in individuals encoding low-risk HLA–DRB1 alleles The first association of a genetic marker with rheuma- toid arthritis (RA) was observed in 1800, when Landre ´- Beauvais described “la goutte asthenique primitive.” In a limited sample of 9 patients, all were female. The association of this genetic marker (sex) with RA was subsequently con- firmed in 1967 by Charcot. In 1976, Stastny (1) reported the association of a second genetic marker with RA, a lymphocyte- defined determinant called R, later designated HLA–Dw4. Subsequent studies led to the formulation of the shared epitope (SE) hypothesis, in which a small amino acid sequence in the third hypervariable region of the HLA–DRB1 gene determines susceptibility to RA (for review, see refs. 2 and 3). Defining the interaction between both genetic markers may help to develop a pathogenetic model of RA. Further- more, better estimates of individual relative disease risks increase the power of linkage studies screening the human genome for other genetic susceptibility genes. To this end, we reanalyzed data from our inception cohort of Caucasian pa- tients with recent-onset RA according to the 1987 revised criteria of the American College of Rheumatology (formerly, the American Rheumatism Association) (3,4). Data on HLA– DRB1 typing and sex were available for 167 unrelated Dutch Caucasian patients (65% women) and 160 unrelated healthy Dutch Caucasian blood donors. The HLA–DRB1 alleles en- coding the amino acid sequences QRRAA (DRB1*0101, *0102, *0404, *0405, *0408), QKRAA (DRB1*0401, *0409), and RRRAA (DRB1*1001) were considered to encode the shared epitope. Controls and patients were assigned the geno- types SE/SE, SE/non-SE, and non-SE/non-SE. Here, non-SE represents any HLA–DRB1 allele not encoding the shared epitope. The resulting genotype distribution differed signifi- cantly between men and women (P  0.05 by chi-square analysis) (see Table 1). Compared with female patients, male patients showed a significantly higher frequency of SE alleles (odds ratio [OR] 1.77, 95% confidence interval [95% CI] 1.13–2.80, P = 0.016 by Fisher’s 2-tailed exact test), SE homozygosity (OR 2.17, 95% CI 1.04–4.51, P = 0.054), DRB1*0401 (OR 1.72, 95% CI 1.01–2.91, P = 0.053), and DRB1*0404 (OR 2.43, 95% CI 0.98–6.05, P = 0.056), but not DRB1*0101 (OR 1.00, 95% CI 0.52–1.93, P = 1.000). Thus, a significant interaction between HLA–DRB1 and sex was ob- served. Subsequently, we estimated the relative risk associated with each combination of sex and HLA genotype. As expected under Hardy-Weinberg equilibrium and random mating, the genotype distribution (non-SE/non-SE, SE/non-SE, SE/SE) in male and female controls did not differ. Therefore, the gene frequency of the shared epitope in the pooled controls was estimated using maximum-likelihood techniques (P = 0.253). The genotype frequencies in controls were adjusted to Hardy- Weinberg equilibrium, which results in smaller variance of the estimates and increased power (5). Data from 1988 on the ratio of men to women at ages 55 years and 65 years were obtained from the Dutch Central Bureau of Statistics (CBS). Given that at age 55 years (the median age of patients in our study) 49.8% of the Dutch population is male, the relative risks of RA for the genotypes non-SE/non-SE, SE/non-SE, and SE/SE, respec- tively, compared with the male non-SE/non-SE genotype were 1:3.76 (95% CI 1.74–8.11):15.0 (95% CI 5.62–40.3) in men and 3.25 (95% CI 1.55–6.78):7.05 (95% CI 3.41–14.6):15.7 (5.90– 41.8) in women. Stratifying for the genotypes non-SE/non-SE, SE/non-SE, and SE/SE, the relative risks of RA in women were 3.25 (95% CI 1.55–6.78), 1.88 (95% CI 1.05–3.35), and 1.04 (95% CI 0.36–3.04), respectively. These data show that the predisposing effect of female sex is maximal in individuals not encoding the shared epitope, decreases with increasing numbers of SE alleles, and is prac- tically absent in individuals homozygous for the shared epitope. This information is new, although several authors have noted differences in the frequencies of HLA–DRB1 alleles between men and women and suggested that sex might influence the association between HLA and RA (6–10). Sex-specific SE genotype data are described in the studies by MacGregor et al (6) and Meyer et al (7). However, both studies are biased in terms of the ratio of male and female patients. If we assume that 46.5% of the controls are male (CBS data for age 63 years), that 31% of RA patients in the UK are male (the frequency in a large, unbiased UK RA data set [8]), and that in the study by Meyer et al 54% of the patients are male (after exclusion of the 41 male veterans), we can estimate sex-specific genotype risks in both studies. Com- pared with the male non-SE/non-SE genotype, the relative risks associated with the non-SE/non-SE, SE/non-SE, and SE/SE genotypes are 1:2.97 (95% CI 0.99–8.92):19.60 (95% CI 6.01–63.91) in men and 2.74 (95% CI 1.02–7.39):7.53 (2.86– 19.84):24.24 (95% CI 7.99–73.49) in women in the study by MacGregor et al, and 1:3.58 (95% CI 2.17–5.91):12.94 (95% CI 6.46–25.92) in men and 2.03 (95% CI 1.21–3.38):5.88 (95% CI 3.62–9.57):16.17 (95% CI 8.07–32.39) in women in the study by Meyer et al. Again, the risks encoded by the male and female SE/SE genotype were almost identical. The mechanism by which HLA and sex interact is not Table 1. Cross tabulation of shared epitope genotypes by sex* Men Women OR 95% CI P† Controls (n = 160) SE/SE 8 (8) 4 (6) 1.00 – – SE/non-SE 34 (35) 23 (37) 1.35 0.36–5.02 0.753 Non-SE/non-SE 55 (57) 36 (57) 1.31 0.37–4.67 0.762 RA patients (n = 167)‡ SE/SE 19 (33) 20 (18) 1.00 – – SE/non-SE 28 (48) 53 (49) 1.80 0.83–3.91 0.164 Non-SE/non-SE 11 (19) 36 (33) 3.11 1.24–7.82 0.022 * Except where indicated otherwise, values are the number (percent- age of individuals of the given sex who encode the indicated genotype). SE represents any HLA–DRB1 allele encoding the shared epitope; non-SE represents any HLA–DRB1 allele not encoding the shared epitope. The odds ratio (OR) compares each genotype with the SE/SE genotype, and actually gives a crude estimate for the relative risk that the patient encoding this genotype is female. 95% CI = 95% confi- dence interval. † By Fisher’s 2-tailed exact test. ‡ P  0.05 by chi-square analysis, overall genotype distribution in male RA patients vs. female RA patients. 1762