SEARCH

SEARCH BY CITATION

Rheumatoid arthritis (RA) is a systemic autoimmune disease with an estimated prevalence of ∼0.5% in the Spanish population (1). As in most complex diseases, the genetic component for host susceptibility is still unknown. Genome-wide linkage studies have proven unable to pinpoint with enough confidence genomic regions other than HLA, thus leaving the main portion of the genetic influence unidentified (2–5). For this reason, the “positional candidate” strategy has been regarded as the dominant approach to investigate genetic associations with RA.

Understanding of the interactions between neuroendocrine and immune-mediated inflammatory reactions has improved enormously in recent years. In particular, there is mounting evidence that corticotropin-releasing hormone (CRH), through its role as central regulator of the hypothalamic–pituitary–adrenal axis, could be crucial in the onset of RA (6, 7). Single-nucleotide polymorphisms in the promoter region have been studied in relation to RA, but the results have not been conclusive (8–10).

Two recent studies from the same group have shown nominal linkage and positive association of the locus with the disease, by means of family-based association studies and the discovery of the highly polymorphic markers CRHRA1 and CRHRA2 (25 kb and 20 kb downstream of the CRH coding sequences, respectively) (11, 12). We performed the first study designed to replicate these results in a different population (121 simplex Spanish families), using the transmission disequilibrium test (TDT), but no significant association was found (13). Family-based association studies are less powerful than traditional case–control studies, but their robustness to population stratification has made them very popular in the last decade (14). However, case–control analysis methods and strategies that allow population stratification have recently been developed (15–17). The aim of the present investigation was to clarify the effect of the CRH locus in RA by performing the first population association study of the markers CRHRA1 and CRHRA2 in a study with a relatively high number of RA cases and the use of the “hypernormal control group” strategy proposed by Morton and Collins (15).

We recruited a group of 257 patients (the 121 probands from the original data set plus 136 new patients) who fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 criteria for RA (18). There were 198 women (77%) and 59 men. The mean ± SD age of the patients was 69 ± 16.6 years, and the mean ± SD disease duration was 17 ± 9.3 years. Sixty-eight percent of the patients were positive for rheumatoid factor, 76% had erosive lesions, and 22% had nodules. From a well-characterized control group of 409 individuals referred from the HOMFOL group (Hospital Universitari de San Joan, Reus Spain), we selected only those who were >39 years old (age at which risk for RA is increased) at the time of the clinical interview. From these, we subsequently applied the next selection criteria: same geographic area as the cases (Spain), same ethnicity (Mediterranean European Caucasian), and same origin (Spanish grandparents). The presence of RA or other autoimmune diseases in the controls or their first-order relatives (parents, siblings, or descendants) was treated as a potential confounder and was a criterion for exclusion from the study. After application of the inclusion and exclusion criteria, the final number of “hypernormal” controls (i.e., individuals who were demographically similar to the patients, but without RA or any other autoimmune disease in the individual or a first-order relative), was 178.

Genotyping of samples was performed as previously described (13). Allele frequencies for CRHRA1 and CRHRA2 were calculated (Table 1), and Hardy-Weinberg equilibrium and linkage disequilibrium between the two loci were evaluated using the ARLEQUIN program (19). For each marker, statistically significant differences between cases and controls were analyzed by the CLUMP program (20), which uses a Monte Carlo approach to assess the significance of the difference in allele frequencies between cases and controls when loci with multiple alleles are studied. Haplotype case–control studies were performed using the FASTEHPLUS program (21), which this is based on the EH program and performs model-free analysis and permutation tests of allelic association (22). Power estimates were performed with the Genetic Power Calculator (http://www.statgen.iop.kcl.ac.uk), using the following parameters: disease prevalence 0.005, D′ between disease and marker alleles 0.8, 2-fold increased risk of disease in heterozygotes under a multiplicative model, α = 0.05. We estimate that with a sample of the size used in this study, there is 94% power to detect association when disease and marker allele frequency (A) = 0.25, 90% power when A = 0.15, and 78% power when A = 0.1.

Table 1. Analysis of allele frequency differences between rheumatoid arthritis cases (n = 257) and “hypernormal” controls (n = 178)*
AlleleCRHRA1CRHRA2
CasesControlsCasesControls
  • *

    Values are the number (%).

  • Data were analyzed with the CLUMP program (version 2.2). P values are based on the T4 statistic after 100,000 simulations (random number seed = 2,000) (20).

40 (0.00)2 (0.56)19 (3.70)20 (5.62)
51 (0.19)2 (0.56)1 (0.19)0 (0.00)
60 (0.00)0 (0.00)1 (0.19)2 (0.56)
70 (0.00)1 (0.28)2 (0.39)1 (0.28)
815 (2.92)10 (2.81)1 (0.19)1 (0.28)
952 (10.12)30 (8.43)1 (0.19)2 (0.56)
10129 (25.10)96 (26.97)9 (1.75)5 (1.40)
11264 (51.36)173 (48.60)51 (9.92)30 (8.43)
1235 (6.81)30 (8.43)128 (24.90)87 (24.44)
139 (1.75)8 (2.25)80 (15.56)50 (14.04)
147 (1.36)3 (0.84)87 (16.93)61 (17.13)
151 (0.19)1 (0.28)65 (12.65)49 (13.76)
160 (0.00)0 (0.00)46 (8.95)33 (9.27)
170 (0.00)0 (0.00)14 (2.72)10 (2.81)
180 (0.00)0 (0.00)7 (1.36)3 (0.84)
191 (0.19)0 (0.00)2 (0.39)2 (0.56)
P0.86130.9302

Control and RA samples were in Hardy-Weinberg equilibrium for both of the markers (P > 0.05; 10,000 dememorization steps), and there was significant linkage disequilibrium between them (P < 0.0001). CLUMP analysis revealed that neither CRHRA1 nor CRHRA2 showed significant allele frequency differences between the cases and the controls (Table 1). Extended analysis of pairs of allelic combinations from the two sites near the CRH gene by the FASTEHPLUS program, did not improve the significance (data not shown).

Results of the present study reinforce our previous findings of a lack of association between the CRH locus and RA. To our knowledge, this is the first published report of a population-based association study with the microsatellites CRHRA1 and CRHRA2. Over the last decade, family-based association studies have become popular in the study of complex diseases because they are robust to population subdivision and admixture (14). However, very few reports have described confounding by population stratification, and recent work with experimental data suggests that it is not a substantial source of bias (23). Furthermore, Ardlie et al point out that with proper study design, population stratification could be a negligible factor in investigations of US and European populations (24).

For the above reasons, the shift toward the less powerful but more robust family-based association studies in the study of complex diseases (such as RA) in large populations (e.g., English and Spanish populations) does not seem justified. In addition, a number of strategies that maximize population-based association studies, either by circumventing the stratification problem or by increasing analytic efficiency, have recently been proposed (15–17). In particular, the “extreme discordance” strategy proposed by Morton and Collins (15) should give our study at least a 6-fold efficiency against TDT trios based on the null hypothesis. We therefore started by selecting an age-of-RA-risk group of Caucasian controls who were of Spanish origin, going back at least 2-generations. Discordance between case and control samples was then increased by excluding from the control group all individuals with RA or another autoimmune disease. We further increased the “hypernormality” of the control group by also excluding those individuals who had a first-order relative with RA or another autoimmune disease. Through the use of this selective strategy, our study should have the highest statistical power attainable in this type of investigation.

Based on our data, we conclude that the CRH locus is not associated with RA in the Spanish population, although the possibility of a very weak effect cannot be definitively excluded. This is the first population-based association study of the CRH locus and RA using the markers CRHRA1 and CRHRA2. By using controls with the lowest liability for development of RA we maximized the power obtained from standard estimators. Most association studies are performed with small-to-moderate–sized single samples, thus making replication in different populations necessary (25). If the absence of association we report here were to be replicated in other populations, there would be good reason to dismiss the initially reported association (26). Therefore, we encourage groups working with different populations to perform similar studies in order to help clarify the conflicting results, or even to conduct meta-analyses of the existing data.

Acknowledgements

  1. Top of page
  2. Acknowledgements

Supported by the Societat Catalana de Reumatologia.