The NZ population-based Caucasian sample consists of 740 RA patients fulfilling the American College of Rheumatology (ACR) criteria for RA 22. Of the patients for whom data were available, 34.3% (234/683) were male, 82.9% (538/649) were rheumatoid factor (RF)-positive, 68.1% (275/404) were anti-cyclic citrullinated peptide (anti-CCP)-positive and 79.4% (576/725) carried the HLA-DRB1 shared epitope. Ethical approval for recruitment of cases was given by the New Zealand Multi-Region Ethics Committee, and recruitment of the controls was approved by the Lower South Ethics Committee. All patients provided written informed consent for the collection of samples and subsequent analysis. The control sample consisted of 553 NZ European Caucasians (226/552 male; 40.9%) with no history of autoimmune disease.

The 713 UK patients were recruited in Oxford, with informed consent from attendees at the rheumatology outpatient clinic at the Nuffield Orthopaedic Centre. All fulfilled the 1987 ACR criteria for RA; the average age of onset was 48 years, 28% were male, 77% were positive for RF and 77% carried the HLA-DRB1 shared epitope. Healthy ethnically matched controls (n = 515) were recruited from the same locale as that of blood donors. Approval for the study was given by the Oxford Research Ethics Committee (OxRec number C02.032).

DNA was extracted from peripheral blood samples of the RA patients and controls by means of guanidine isothiocyanate-chloroform extraction methods. NZ study participants were genotyped for the ITGAV single-nucleotide polymorphism (SNP) rs3738919 using a polymerase chain reaction-restriction fragment length polymorphic SNP genotyping assay as follows: forward primer CACTTTCTGTAAATTAGTGTTAGATCAAAAGG and reverse primer GCTTATAACTCACAATTCAGATTTTTGCC (primers from Sigma-Genosys, Sydney, Australia). The C allele (major allele) of the rs3738919 product (286 base pairs) was digested using the AluI restriction enzyme to form fragments of 223 and 63 base pairs. Oxford study participants were genotyped for rs3738919 using the TaqMan genotyping assay C___1278131_1, and both the NZ and Oxford samples were genotyped for the rs10174098 and rs3911238 ITGAV variants using the TaqMan genotyping assays C__30567648_10 and C___7617051_10, respectively, from Applied Biosystems (Scoresby, Australia).

Although the imputed rs3738919 data were available and were previously reported by Ahnert and Kirsten 21, we reimputed rs3738919 genotypes in order to provide information on imputation parameters. RA case and control genotypes were imputed from the WTCCC dataset using IMPUTE software 23. Genotypes were imputed from 89% of cases (n = 1,659) and 90% of controls (n = 2,639) using a 7-Mb region and a calling threshold of 0.7.

An a priori power calculation was made based on the combined French 20 and WTCCC 21 OR (OR_allelic = 1.16), using 800 cases and 600 controls with a major allele frequency of 65%. The power to detect association of rs3738919 to RA using either the NZ or Oxford sample set was estimated to be 47%, using α = 0.05. ITGAV SNPs were tested for deviation from Hardy-Weinberg equilibrium in both the control and RA samples using a chi-square goodness-of-fit test. The significance of differences in the minor allele frequency between RA patients and controls, and stratified patients, was assessed using the chi-square goodness-of-fit test.

Of the five ITGAV SNPs (rs3911238, rs2887827, rs10174098, rs13006571 and rs16828163) genotyped by the WTCCC and rs3738919, only three (rs3911238, rs10174098 and rs3738919) were required to tag all major haplotypes defined by the six variants. Haplotype association analysis was performed using the SHEsis software package 24, which uses a full-precise-iteration algorithm to construct haplotypes and estimate haplotype frequencies.

Meta-analysis combining the French 20, WTCCC 25, NZ and Oxford samples was performed using STATA version 8.0 (StataCorp LP, College Station, TX, USA). The Mantel-Haenszel test was used to estimate the average conditional common OR between the four independent sample sets and to test for any heterogeneity between the four groups using both fixed and random effects. R software 26 was used to perform the Cochran-Armitage trend test to determine recessive, additive and dominant trend values (Table 1).

We genotyped rs3738919 across the NZ and Oxford RA case control cohorts and found no evidence for an association between ITGAV and RA (P_allelic = 0.11; OR_allelic = 0.88 [95% CI 0.75 to 1.03] and P_allelic = 0.07; OR_allelic = 1.17 [95% CI 0.99 to 1.39], respectively) (Table 1). The direction of the NZ allele distribution is consistent with the previous French 20 and WTCCC 21 data, with the minor allele under-represented in the case groups in all three cohorts (Table 1). However, the allele distribution in the Oxford RA cohort differs in that the minor allele is over-represented in the patient group compared with the control group (Table 1).

Association analysis using our imputed genotypes for rs3738919 (P = 0.04) from the WTCCC data were consistent with the previous report of an association of this SNP with RA in the WTCCC 21. In our analysis, we obtained imputed genotypes from 89% of the WTCCC case control subjects. Presumably, this reflects the influence of the low linkage disequilibrium that exists between rs3738919 and the genotyped SNPs (0.01 <r² < 0.54 in CEU [Centre d'Etude du Polymorphisme Humain Utah] HapMap 27) on confidence calls of imputed genotypes. Haplotype analysis (Table 2) was performed using the actual genotypes from ITGAV SNPs rs3911238 and rs10174098 (Additional data file 1) and the imputed genotypes for rs3738919 (Table 1). One susceptibility haplotype (1-1-1), containing the major allele at each of the three SNPs, was significantly over-represented in cases compared with controls (OR = 1.15, 95% CI 1.02 to 1.30; P = 0.021).

Variants rs3911238 and rs10174098 were typed over the NZ and Oxford sample sets, and association of three-marker haplotypes was examined (Additional data file 1 and Table 2). There was no support for a positive association of the 1-1-1 haplotype with RA in either the NZ or Oxford sample set. There was a significant protective effect for this haplotype in the Oxford samples (OR = 0.78, 95% CI 0.62 to 0.99; P = 0.042).

Meta-analysis of all four sample sets, using a fixed effects model, revealed some evidence for an association of rs3738919 with RA (OR = 0.92, 95% CI 0.86 to 0.99; P = 0.021) (Figure 1a). Because of evidence for heterogeneity between the sample sets (Breslow-Day P = 0.011), a random effects model was also used for meta-analysis; this did not provide evidence for an association of rs3738919 with RA (OR = 0.92, 95% CI 0.80 to 1.07; P = 0.29) (Figure 1b). The Oxford sample set is significantly different from the other three sample sets at rs3738919; however, the Oxford patient sample set did not have any large differences in relation to gender (Oxford 28% male, NZ 34%, WTCCC 25%, France 13%), RF status (Oxford 77% positive, NZ 83%, WTCCC 84%, France 75%) or inheritance of the shared epitope (Oxford 77% positive, NZ 79%, WTCCC 79%, France 79%).

The NZ and Oxford samples were stratified according to gender, RF and shared epitope status (Table 3). The stratification results did not reveal any significant differences in the rs3738919 genotype distribution for RA patients. There was a difference in the rs3738919 genotype distribution of male and female patients in the NZ cases (P = 0.026) (Table 3). However, a significant difference was not evident in either the Oxford or WTCCC case sample set (P = 0.55 and 0.80, respectively) (Table 3).