The severity of joint destruction in rheumatoid arthritis (RA) is highly variable between patients. Recent twin and population studies indicated that the severity of joint destruction is influenced by genetic factors.1 ,2 Previously, we reported the association of rs10818488 (TRAF1-C5) and rs675520 (TNFAIP3-OLIG3) with progression of joint destruction.3 ,4 The genes near these loci encode for tumour necrosis factor receptor-associated factor-1 (TRAF1), complement component-5 (C5) and tumour necrosis factor α-induced protein-3 (TNFAIP3; a protein that inhibits NF-κ B activation). A basic principle in genetic association studies is to evaluate multiple cohorts to validate observed findings. We therefore studied both single-nucleotide polymorphisms in several RA cohorts with radiological follow-up data.

Six thousand two hundred and eighty-two x-rays of 2666 RA patients were studied: 147 patients from Lund (Sweden), 385 patients from Sheffield (UK), 285 patients from Iceland, 384 patients from the North American Rheumatoid Arthritis Consortium (NARAC), 756 patients from the National Databank of Rheumatic Diseases (NDB), 113 patients from Wichita and 596 patients from the Leiden Early Arthritis Clinic (Leiden-EAC) cohort (table 1). Detailed information on these datasets is provided elsewhere.2 ,5,–,10 Genotyping of Lund and Sheffield DNA was performed using Sequenom analysis, genotyping results of the Icelandic and NARAC patients were retrieved from genome-wide-association studies,2 ,7 and those of the Wichita and Leiden cohorts were done with the Illumina Immunochip.11 Radiographs were scored according to the Sharp-van der Heijde method (Iceland, Wichita, NDB, Leiden-EAC) or the Larsen method (Lund, Sheffield). For each dataset, the relative increase in the progression rate per year of follow-up in comparison with the reference genotype was estimated using multivariate normal regression (cohorts with repeated radiological measures—Leiden-EAC, Lund, Groningen and Wichita) or linear regression (datasets with single measurements per patient—Sheffield, NARAC, NDB and Iceland), after adjusting for age and gender. The RA patients studied were treated in an era when treatment strategies were not as intensive as they are today. In the Groningen and Leiden cohorts, adjustment for treatment strategies were applied12; in the other datasets, no treatment effects on the radiological progression rates were observed.

Considering the number of patients and x-rays studied, it was not expected that statistical significance would be obtained in the individual datasets. Therefore, the estimates of individual datasets, reflecting the relative increase in the rate of joint destruction per year, were summarised in a meta-analysis with inverse variance weighting. rs10818488 was previously associated with joint destruction in a dominant analysis performed on all RA patients and with rs675520 in a recessive analysis performed on anticyclic citrullinated protein (anti-CCP) positive RA patients. In line with this, analyses were primarily performed using a dominant model on all patients (rs10818488) and a recessive model on anti-CCP positive patients (rs675520). SPSS V.17.0 and Stata V.10.1 were used for statistical analysis.

Figure 1 summarises the results of the analyses. The effect estimates of rs10818488 were not consistent in their direction. Statistical significance was not obtained in the meta-analysis. Analysis of the subgroup of anti-CCP-positive patients also did not result in significant findings (data not shown). Similarly, rs675520 was not associated with progression of joint destruction in anti-CCP positive RA patients (figure 1) or in all patients (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Results of the analyses of the association of (A) TRAF1-C5 (rs10818488) and (B) -TNFAIP3-OLIG3 (rs675520) with progression of joint destruction in seven sets of rheumatoid arthritis patients. rs10818488 was tested in all RA patients (without selection on anti-cyclic citrullinated protein (anti-CCP) status) under a dominant model; patients homozygous for the minor allele together with heterozygous patients were compared with the reference group of patients homozygous for the common allele. rs675520 was tested in anti-CCP-positive patients under a recessive model; patients homozygous for the minor allele were compared with the reference group of heterozygous patients together with patients homozygous for the common allele. The effect sizes (ES) are the estimated relative progression rates per year for the tested genotype compared with the reference group. To illustrate the value of ES, an ES of 1.02 (rs10818488-Lund) per year equal a 1.15 higher rate of joint destruction over 7 years, which is similar to a 15% higher increase in the rate of joint destruction for patients with AA genotype versus AG and GG.

In the Leiden cohort, rs10818488 was previously found to be associated with joint destruction over the first 2 years in 278 RA patients and rs675520 with progression of joint destruction over 5 years in 181 anti-CCP positive patients. Extending the Leiden dataset to 596 RA patients and carrying out a yearly follow-up over 7 years resulted in a different finding for rs10818488 but not for rs675520 (figure 1).3 ,4 rs2900180 in TRAF1-C5 is observed to be associated with erosiveness in a UK study, but this variant shows low linkage disequilibrium (R²=0.67) with rs10818488.13

A drawback of the current study is that the studied datasets differed in designs; this heterogeneity may affect the findings. However, the present data do not support the initial findings that rs10818488 and rs675520 are associated with the severity of joint destruction in RA.