Lack of association of the HLADRB1 shared epitope with rheumatoid nodulesAn individual patient data meta-analysis of 3272 Caucasian patients with rheumatoid arthritis.код для вставкиСкачать
ARTHRITIS & RHEUMATISM Vol. 50, No. 3, March 2004, pp 753–762 DOI 10.1002/art.20119 © 2004, American College of Rheumatology Lack of Association of the HLA–DRB1 Shared Epitope With Rheumatoid Nodules An Individual Patient Data Meta-Analysis of 3,272 Caucasian Patients With Rheumatoid Arthritis Jennifer D. Gorman,1 Eve David-Vaudey,2 Madhukar Pai,3 Raymond F. Lum,4 and Lindsey A. Criswell4 Objective. The objective of this individual patient data (IPD) meta-analysis was to examine the relationship of rheumatoid nodules to the HLA–DRB1 shared epitope (SE) and to individual SE genotypes. Methods. English-language studies that enrolled adult non-Hispanic Caucasian patients with rheumatoid arthritis (RA) were identified by searches of Medline and Embase, and by manual searches of medical journals. All authors were contacted for IPD. Metaanalysis was performed to assess the association of SE presence, dose, and genotype with rheumatoid nodules. Meta-analyses adjusted for disease duration and cumulative meta-analyses were also performed to assess the influence of RA duration and year of study publication on the results. Results. A total of 24 studies and 3,272 patients were available for analysis. IPD were obtained for 22 of the studies. There was a nonsignificant association between the presence of the SE (i.e., 1 or 2 alleles versus 0 alleles) and rheumatoid nodules (summary odds ratio [OR] 1.3, 95% confidence interval [95% CI] 0.97–1.6). Analysis by SE genotype, however, demonstrated a weak relationship with inheritance of a single DRB1*0401 SE allele (OR 1.4, 95% CI 1.1–1.8). No other genotypes achieved statistical significance in the adjusted or unadjusted analyses. Conclusion. The presence of the HLA–DRB1 SE does not appear to significantly increase the risk of rheumatoid nodules among Caucasian patients with RA. Analysis by DRB1 SE genotype was uninformative, suggesting only a potential (and at most modest) role of the DRB1*0401 SE allele. Results from this IPD metaanalysis implicate other genetic, stochastic, and/or environmental factors in the susceptibility to rheumatoid nodules. Marked heterogeneity exists in the clinical presentation of rheumatoid arthritis (RA) between individual patients. This is particularly true for development of extraarticular manifestations, typically defined as manifestations of disease distinct from the erosive and deforming joint process. One of the most commonly encountered extraarticular manifestations is rheumatoid nodules, which are estimated to affect ⬃20% of patients with RA (1). The presence of extraarticular manifestations, including nodules, has long been considered a sign of greater RA severity, as supported by results of recent studies demonstrating 2–4-fold increased mortality in patients with these manifestations (2,3). Genetic variation is believed to be important in determining both the susceptibility to and severity of Supported by the American College of Rheumatology, the NIH (grants AR-07304, AR-20684, and AR-02175, and the Fogarty AIDS International Training Program [grant 1-D43-TW00003]), the Arthritis Foundation, and the Rosalind Russell Medical Research Center for Arthritis. 1 Jennifer D. Gorman, MD, MPH: University of California, San Francisco, and School of Public Health, University of California, Berkeley; 2Eve David-Vaudey, MD: University of Paul Sabatier, Toulouse, France; 3Madhukar Pai, MD: School of Public Health, University of California, Berkeley; 4Raymond F. Lum, MPH, Lindsey A. Criswell, MD, MPH: University of California, San Francisco. Address correspondence and reprint requests to Lindsey A. Criswell, MD, MPH, Division of Rheumatology, University of California, San Francisco, 374 Parnassus Avenue, Box 0500, San Francisco, CA 94143-0500. E-mail: firstname.lastname@example.org. Submitted for publication August 8, 2003; accepted in revised form December 11, 2003. 753 754 GORMAN ET AL RA, and, to date, the most illuminating genetic theory has been that of the shared epitope (SE) (4). The SE hypothesis was proposed to explain susceptibility to RA, based on the observation that HLA–DR specificities associated with disease share a similar amino acid sequence in the third hypervariable region of the DRB1 molecule. Although preliminary research on the SE supported an association with RA susceptibility, results of subsequent investigations suggested that its primary role may be in disease severity. The most striking results were generated from one of the earliest analyses, demonstrating a marked association of nodules and other extraarticular manifestations with SE alleles in a highly selected group of patients with RA (5). Since then, numerous studies have pursued clarification of this relationship, although subsequent findings have been much less striking, with many investigations failing to demonstrate a significant association with outcome (6–8). The difficulty in interpreting the relationship of the SE with RA severity is likely attributable, at least in part, to differences in the populations studied, such as ethnicity, clinical characteristics, and the particular SE allele(s) inherited, as well as the relatively small size of most studies. The combination of the relative infrequency of nodules and the small size of individual studies motivated us to perform a meta-analysis. In order to investigate the relationship more thoroughly than is possible using traditional meta-analytic techniques (i.e., limited to summary data from published articles), we obtained individual patient data (IPD) from study authors. IPD meta-analysis has several advantages over the traditional approach, including the ability to incorporate unpublished and/or updated data, evaluate and correct raw data, perform additional analyses of subgroups, or adjust for potential confounding factors (9). Therefore, our IPD meta-analysis allowed us the unique opportunity to examine the role of specific SE genotypes in the development of rheumatoid nodules and to investigate the influence of covariates, such as disease duration and sex. MATERIALS AND METHODS Inclusion criteria. Studies eligible for inclusion in this meta-analysis were those in which HLA–DRB1 alleles in RA patients were investigated, and in which information regarding extraarticular manifestations was reported. The HLA–DRB1 alleles defined as encoding the SE were as follows: HLA– DRB1*0101, *0102, and *0104 (DR1); HLA–DRB1*0401, *0404, *0405, *0408, *0409, *0410, *0413, *0416, *0419, and *0421 (DR4); HLA-DRB1*1001 (DR10); and HLA– DRB1*1402 and *1406 (DR14) (10). The frequency of HLA– DRB1 alleles varies according to ethnic and racial background, with some alleles being extremely rare; therefore, articles were not required to identify all 16 SE alleles to be eligible for inclusion. Studies selected were those that evaluated adult, nonHispanic, Caucasian patients and assessed for the presence of rheumatoid nodules. Non-Caucasian and Hispanic patients were excluded from this analysis in order to evaluate a more genetically homogenous population and due to the limited number of these studies available within the published literature. Only studies or letters written in English were included in our analysis. Search strategy. Three overlapping search strategies were used to identify appropriate articles published between January 1987 and June 1999. We searched Medline with the truncated keywords of rheumatoid arthritis and 1 or more of the following: human leukocyte antigen, gene, prognosis, or severity. Embase was searched with the exploded terms rheumatoid arthritis and 1 or more of the following: histocompatibility, HLA system, epitope, etiology, or genetics. The database search terms were designed to be exceedingly sensitive in order to identify all studies involving HLA–DRB1 genotyping in which there was any reference to RA severity. In addition, an updated PubMed search was performed to identify articles published between June 1999 and March 2003. Hand searches of 6 journals (Annals of the Rheumatic Diseases, Arthritis & Rheumatism, British Journal of Rheumatology, Journal of Rheumatology, Scandinavian Journal of Rheumatology, and Tissue Antigens) for articles (January 1987 through June 1999) were also performed. We selected the start of the time period for article identification based on the introduction of molecular DRB1 typing in 1987 (11). A single investigator (JDG) performed the literature searches and reviewed abstracts identified by the search. Training on database and hand searching was provided by formal instruction prior to the start of the study. Upon identification of articles appropriate for study inclusion, letters were mailed or electronically sent to the corresponding author to request missing information and/or IPD. If the same patients were described in more than 1 article (i.e., duplicate data) and IPD were not available, the publication that presented the most complete recent information was selected for inclusion in the analysis. One investigator (JDG) performed the data extraction from studies without IPD. Because the IPD were provided without personal identifiers, this study was certified as exempt from Human Subjects Review by the University of California, San Francisco Committee on Human Research. The influence of individual study quality on the results of meta-analyses has been well described, particularly for meta-analyses of clinical trials (12). Although assessment of study quality is considered to be an important component of systematic reviews and meta-analyses, it has been demonstrated that use of such scoring methods can be problematic (13,14) and may not accurately assess the quality measures of interest (15,16). Moreover, in practice, most systematic reviews of published studies assess the quality of data reporting, which does not always correspond precisely with the quality of the study conducted. Because we obtained IPD, we did not conduct a formal assessment of study reporting. However, we used Women only RA after vaccination Cross-sectional Cross-sectional Prospective cohort Case series Cross-sectional Case–control Multiplex§ 87 120 109 100 Case–control Case–control Case–control Cross-sectional 3,272 136.3 ⫾ 90.3 111 299 171 9 77 88 95 83 45 239 131 Cross-sectional Case–control Case–control Cross-sectional Cross-sectional 128 403 69 36 200 216 218 66 172 Case–control Prospective cohort Cross-sectional Cross-sectional Case–control Cross-sectional Cross-sectional Retrospective cohort Prospective cohort RF⫹, erosive disease Early RA RF⫹ Multiplex§ Multiplex§ Erosive disease Early RA Early RA Early RA Study design No. of patients† French French French French European American American Canadian European, Australian S. and N. French Turkish Greek Greek S. S. S. S. Swiss German British British N. French British N. French British British Swedish Swedish Ethnicity‡ 59.9 ⫾ 4.5 – 63 63 – – – – 57 – – 61 50 – 65 – – – 59 56 59 – 54 NA – Age, mean years 10.8 ⫾ 6.6 – 16 21 3 18 8 12 7 11 9 10 0.6 – 16 – 1 0.9 12 22 13 – 14 17 – Disease duration, mean years 73.6 ⫾ 10.4 73 56 100 56 66 85 80 82 87 80 74 79 70 74 – 66 67 80 83 72 – 84 77 64 % female 75.6 ⫾ 13.5 90 88 78 75 92 72 100 84 78 62 85 69 73 73 100 72 44 75 95 85 – 83 57 70 % RF⫹ * Individual patient data were not available for Hall et al (30) and Seidl et al (33). SE ⫽ shared epitope; RF ⫽ rheumatoid factor; NA ⫽ not available; early rheumatoid arthritis (RA) ⫽ disease duration ⬍2 years; N ⫽ northern; S ⫽ southern. † Available for at least 1 analysis. ‡ American, European, and Australian populations were composed entirely of Caucasians. § Family history of RA; only 1 family member was included in the meta-analysis. Total Mean ⫾ SD Scandinavian Turesson et al, 2000 (8) Eberhardt and Fex, 1998 (24) Northern European Vaidya et al, 2002 (25) Thomson et al, 1999 (26) Cortet et al, 1997 (27) McDonagh et al, 1997 (28) Perdriger et al, 1997 (29) Hall et al, 1996 (30) MacGregor et al, 1995 (31) Central European Bas et al, 2000 (32) Seidl et al, 1999 (33) Southern European Hayem et al, 1999 (34) Toussirot et al, 1999 (35) Cantagrel et al, 1999 (36) Benazet et al, 1995 (37) Mediterranean Saruhan-Direskeneli, 1998 (38) Stavropoulos et al, 1997 (39) Boki et al, 1993 (40) Other/mixed Caucasian Bali et al, 1999 (41) Meyer et al, 1999 (42) Criswell et al, 1998 (43) Pope et al, 1998 (44) Rowley et al, 1997 (45) Combe et al, 1995 (46) Patient selection requirements Characteristics of studies included in the meta-analysis of the association of the SE with rheumatoid nodules* Author, year (ref.) Table 1. NO RELATIONSHIP OF RHEUMATOID NODULES TO THE HLA–DRB1 SE 755 756 GORMAN ET AL Table 2. Clinical and genetic characteristics of RA patients with or without nodules* Characteristic Nodular RA (n ⫽ 1,092) RA without nodules (n ⫽ 2,180) P % male (n ⫽ 2,924) Age, mean ⫾ SD years (n ⫽ 1,125) Age at onset, mean ⫾ SD years (n ⫽ 2,898) Disease duration, mean ⫾ SD years (n ⫽ 2,536) % rheumatoid factor positive (n ⫽ 3,019) % with erosive disease (n ⫽ 2,162) 30.2 59.6 ⫾ 13.0 45.4 ⫾ 14.4 14.2 ⫾ 11.2 88.2 89.0 24.1 58.9 ⫾ 14.1 49.0 ⫾ 15.0 9.1 ⫾ 10.0 70.0 66.5 ⬍0.001 0.4 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 * P values were calculated by 2-tailed chi-square test or t-test. RA ⫽ rheumatoid arthritis; n ⫽ total number of patients available for analysis. study design as a quality variable (case–control/cross-sectional versus cohort designs) and performed subgroup analyses to evaluate its impact on effect measures. We also critically appraised the quality of included studies to evaluate factors such as bias in case selection or misclassification due to a shorter duration of RA. One reviewer (JDG) evaluated the impact of these quality variables. Statistical analysis. The odds ratio (OR) for the association between the SE and rheumatoid nodules was the effect measure of interest. In all cases, patients with no SE alleles served as the reference group for the computation of ORs. For the meta-analysis, summary ORs were computed using either fixed effects models (for data that did not demonstrate significant heterogeneity) or random effects models (for data demonstrating significant heterogeneity) (17). The Breslow-Day test for homogeneity was used to assess heterogeneity (18), and a significance level of P ⫽ 0.1 was established to test consistency of the OR across studies (19). Additional analyses were performed to account for differences in disease duration between individuals. This approach required the use of multivariate logistic regression analyses to calculate ORs for each of the individual studies along with 95% confidence intervals (95% CIs), with the presence of rheumatoid nodules as the outcome variable, the SE (i.e., presence, dose, or allelic genotype) as the explanatory variable, and disease duration as a covariate. IPD were required for this adjusted analysis because of the need for information regarding disease duration for each patient. The adjusted ORs after logistic regression were then pooled using fixed effects or random effects meta-analyses. In addition, a cumulative meta-analysis by publication year was performed to assess the pattern of results over time. Sensitivity analyses based on the definition of rheumatoid nodules by individual studies were also performed. Publication bias was assessed graphically by funnel plot (20) and was statistically evaluated by the regression asymmetry test described by Egger et al (21), which has been demonstrated to be a powerful test for assessment of publication bias (22). Specifically, the power to detect bias for a meta-analysis of 20 studies with a control group event rate of 20% is estimated to be ⬃50% for meta-analyses with moderate bias and exceeds 95% for those with severe bias (22). All analyses were performed using Stata statistical software, version 8.0 (23). RESULTS The database searches identified a total of 10,367 articles. Upon review, however, only 52 articles mentioned HLA–DRB1 and rheumatoid nodules in the title or abstract. Eighteen of these articles were excluded from our analysis: 12 because they evaluated nonCaucasian patients, and 6 that represented duplicate publications. Authors were contacted for additional data, and the authors of 22 of the 34 articles provided IPD, with information regarding genotype and rheumatoid nodules in Caucasian patients. Two additional articles from the original 34 did not include IPD, but enough information was available within the published articles regarding the SE genotype and presence of nodules to allow inclusion of these 2 studies. Therefore, our final meta-analysis consisted of 24 studies (22 with IPD) and 3,272 patients (8,24–46). Characteristics of the included studies are presented in Table 1. By definition, all studies evaluated non-Hispanic Caucasian patients, the majority of whom lived in European countries. The study designs ranged from prospective cohort studies to a case series, although one-half were of cross-sectional design and one-third were case–control studies. IPD were available for 92% of the studies. The mean duration of disease was 11 years, although 4 studies evaluated patients with early disease (i.e., disease duration of ⬍2 years). Seventy-four percent of the patients were female, and 76% of all patients were seropositive for rheumatoid factor (RF). A description of the criteria used for assessing rheumatoid nodules was available for 14 (58%) of the 24 studies. Of these, 13 studies required physician assessment of nodules by either direct clinical examination or chart review. A single study designated use of patient self-report to identify RA patients with nodular disease. NO RELATIONSHIP OF RHEUMATOID NODULES TO THE HLA–DRB1 SE 757 Figure 1. Summary meta-analysis of association of the shared epitope (SE) with rheumatoid nodules. Comparison of 2 or 1 SE alleles versus 0 SE alleles (referent). The odds ratios are graphed on a log scale. The relative weight of each study in the meta-analysis is indicated by the size of the associated box. The studies are listed according to ethnic subgroup (as in Table 1). 95% CI ⫽ 95% confidence interval. † ⫽ P ⫽ 0.08 for heterogeneity. A comparison of patients with nodules and those without nodules (Table 2) revealed that patients with nodular RA were significantly more likely to be male, have a younger age at disease onset, and a longer disease duration compared with patients without nodules (P ⬍ 0.001 for each). Additionally, patients with nodular RA were also significantly more likely to be seropositive for RF and to exhibit erosive disease (P ⬍ 0.001 for each). Relationship of rheumatoid nodules to presence and dose of the SE. As shown in Figure 1, evaluation of the presence of the SE (i.e., 1 or 2 versus 0 alleles) did not demonstrate a significant association with disease (OR 1.3, 95% CI 0.97–1.6), although significant heterogeneity was present (P ⫽ 0.08). Because 2 studies were observed to have noticeably divergent results, these were excluded in a sensitivity analysis. After this exclusion the effect estimate remained stable (OR 1.3, 95% CI 1.1–1.6), although heterogeneity was reduced (P ⫽ 0.2). Of note, the divergent results did not appear to be explained by a particular study design, ethnic subgroup, or clinical characteristics such as sex, RF positivity, or erosive disease. Additional analyses of the association between the SE dose and the presence of nodules failed to provide evidence for a dose-dependent relationship, because the estimate of the association for either 1 or 2 SE alleles was comparable (Table 3). Additionally, there was no significant difference when inheritance of 2 SE alleles was directly compared with that of a single allele (OR 1.1, 95% CI 0.9–1.3). To assess the influence of study quality, we repeated the analysis excluding the cohort studies. This 758 GORMAN ET AL Table 3. Association of HLA–DRB1 SE genotype with rheumatoid nodules* Genotype No. of studies No. of patients with genotype Total no. of patients available for analysis OR (95% CI) P for heterogeneity† 2 SE alleles‡ 401/401 404/404 401/404 101/401 101/404 101/101 1 SE allele 401/X 404/X 405/X 408/X 101/X 1001/X 23 18 8 16 16 11 6 24 20 18 17 14 16 17 801 130 17 114 145 40 17 1,604 647 167 62 33 376 50 1,606 826 454 749 786 519 391 2,414 1,384 860 739 625 1,017 703 1.3 (0.9–1.8) 1.1 (0.7–1.6) 1.5 (0.6–3.5) 1.3 (0.8–2.0) 1.2 (0.8–1.8) 0.9 (0.5–1.8) 0.9 (0.3–2.6) 1.2 (1.0–1.5)§ 1.4 (1.1–1.8)¶ 1.3 (0.9–1.9) 1.1 (0.6–2.0) 1.9 (0.9–3.8) 1.0 (0.8–1.4) 1.7 (0.96–3.1) 0.06 0.69 0.31 0.74 0.13 0.90 0.89 0.39 0.36 0.10 0.80 0.98 0.28 0.80 * The referent group for all comparisons was 0 shared epitope (SE) alleles. OR ⫽ odds ratio; 95% CI ⫽ 95% confidence interval; X ⫽ non–SE allele. † Values ⬍0.1 represent significant heterogeneity. ‡ All applicable genotypes combined. § P ⬍ 0.04. ¶ P ⬍ 0.008. approach would be expected to result in an inflation of the effect estimate, because cross-sectional and case– control studies tend to demonstrate greater ORs compared with cohort studies; however, the estimate generated (OR 1.3, 95% CI 1.1–1.6) was strikingly similar to that for our primary analysis. Because of the prominent association of the SE with nodules that is observed in patients with familial RA (28,31,47,48), an analysis excluding 3 studies that enrolled exclusively multiplex RA cases was performed and revealed results similar to those from our main analysis (OR 1.2, 95% CI 1.1–1.6). Moreover, the results remained unchanged when stratified analyses for males (OR 1.3, 95% CI 0.8–1.9) and cases of seronegative RA (OR 1.0, 95% CI 0.7–1.6) were performed. Relationship of rheumatoid nodules to the DRB1 SE genotype. Because individual SE alleles and genotypes may differ in their association with specific outcomes (49), an analysis by individual genotypes was undertaken (Table 3). A total of 2,863 patients were available for analysis by SE genotype. Only one genotype demonstrated a statistically significant association with nodules, specifically that of a single DRB1*0401 SE allele (OR 1.4, 95% CI 1.1–1.8). The effect estimates for the other genotypes were relatively modest (i.e., OR ⬍ 2.0), and those with greater estimated ORs tended to have much broader confidence intervals (i.e., DRB1*0404 homozygotes, or a single DRB1*0408 or *1001 SE allele). Because of a prior analysis in which we demonstrated heterogeneity in the association of the SE with erosions among Greeks in comparison with other Caucasians (50), we performed a sensitivity analysis excluding individuals of Mediterranean descent. There was only one notable difference in this sensitivity analysis: specifically, the demonstration of an association of rheumatoid nodules with inheritance of a single DRB1*1001 allele (OR 2.4, 95% CI 1.3–4.7). In order to evaluate a more ethnically homogenous population, a separate analysis of HLA–DRB1 SE genotypes limited to northern European Caucasians, the largest ethnic subgroup (with 1,270 patients), was performed and yielded results similar to those of the analyses of all Caucasians combined. In order to assess whether the inclusion of studies comprising patients with early RA may have negatively influenced the association of rheumatoid nodules with genotypes, 2 separate analyses were performed. First, meta-analyses with adjustment for disease duration were performed for studies that provided IPD. These analyses demonstrated results similar to those of the main analysis and resulted in more heterogeneity and less precise results. In addition, a meta-analysis that excluded the 4 studies of patients with early RA was performed, with results that corresponded to those of the main analysis. Cumulative meta-analysis. To investigate the strength of the evidence for an association of the SE with nodules as it was collected over time, a cumulative meta-analysis by year of publication was performed for the genotype containing a single DRB1*0401 SE allele (Figure 2). Results using this meta-analytic technique NO RELATIONSHIP OF RHEUMATOID NODULES TO THE HLA–DRB1 SE 759 Figure 2. Cumulative meta-analysis of association of a single HLA–DRB1*0401 shared epitope allele with rheumatoid nodules. The referent group was 0 shared epitope alleles. Studies are listed in order of the publication date (from oldest to most recent). The odds ratios are graphed on a log scale. revealed the most striking association with nodules initially, with regression to the eventual modest estimate (i.e., OR 1.4) upon the addition of relatively few studies. Sensitivity analysis by definition of rheumatoid nodules. In order to determine whether different definitions for rheumatoid nodules used by individual studies influenced our results, we performed a sensitivity analysis excluding studies that either did not describe how rheumatoid nodules were defined or that evaluated the presence of nodules by patient report. There was no considerable change in the results, although 3 SE genotypes with borderline significance in the unadjusted analyses did achieve statistical significance in this sensitivity analysis. These genotypes included the presence of the SE (OR 1.4, 95% CI 1.1–1.8), the presence of 2 SE alleles (OR 1.5, 95% CI 1.1–2.1), and the presence of DRB1*1001/X (OR 2.4, 95% CI 1.1–5.3). Assessment of publication bias. A funnel plot of the association of SE positivity with rheumatoid nodules appeared relatively symmetric (data not shown). Moreover, statistical assessment by the regression asymmetry test described by Egger et al (21) did not reveal significant evidence for publication bias (P ⫽ 0.8). DISCUSSION Results from this meta-analysis suggest that, at least among Caucasians patients, no substantial associ- ation exists between the SE and rheumatoid nodules. Of all the genotypes examined, only those with a single DRB1*0401 SE allele demonstrated a significant association with nodules, and this relationship was extremely modest. Moreover, there was no evidence to suggest a significant association of nodules with a double dose of the SE or a contribution of DRB1*0401 homozygotes or *0401/*0404 compound heterozygotes above that of other genotypes, as has been suggested by the results of smaller individual studies (28). The results from our meta-analysis are intriguing, because it has been widely regarded that the most striking associations of the SE with RA severity exist with rheumatoid nodules and other extraarticular manifestations (51). However, although several studies of Caucasian patients support such a relationship (5,31,52), many investigations have failed to replicate these findings (53–55). Moreover, it should not be surprising that the genetic relationship with rheumatoid nodules may be more complex than one that involves only DRB1 alleles. For example, genetic interaction has recently been demonstrated between a tumor necrosis factor microsatellite allele and DRB1*0401 homozygosity in nodular RA (56). Additionally, linkage disequilibrium, the nonrandom association of alleles at different loci, is particularly striking across the HLA region and must be considered when interpreting genetic associations with disease. 760 Furthermore, the importance of environmental factors and gene–environmental interactions in the development of rheumatoid nodules should not be ignored and is beginning to be delineated. Specifically, Mattey et al (57) recently described the relationship of a drugmetabolizing enzyme and cigarette smoke to RA severity. In that study, the presence of the null GSTM1 genotype was significantly associated with RF positivity and titer and more severe erosive disease in RA patients who smoked cigarettes. This finding is of particular interest given the well-described association of cigarette smoking with the development of rheumatoid nodules (56,58,59). Another environmental, or specifically xenobiotic, risk factor known to be associated with the development of rheumatoid nodules is methotrexate. It is estimated that drug-induced nodules will develop in ⬃10% of methotrexate-treated patients with RA (60); such nodules are histopathologically similar to spontaneously occurring nodules. Interestingly, methotrexateinduced nodules have been associated with the DRB1*0401 allele, although in the same study a genetic association of this allele with spontaneous rheumatoid nodules was not observed (61). There are several potential reasons why this meta-analysis did not confirm an association of nodules with the SE or particular DRB1 alleles. A common problem in all clinical research relates to false-positive results (Type I errors). In fact, this point has been elegantly illustrated by a cumulative meta-analysis of genetic association studies, in which initial studies tended to demonstrate more striking results than did subsequent investigations, particularly when the earliest studies included fewer than 150 individuals and more than 15 subsequent publications followed (62). Results from the extensive cumulative meta-analyses performed by Ioannidis et al (62), supported by our analyses performed for DRB1*0401 and nodules, reinforce the notion that caution should be exercised when interpreting results of individual genetic association studies, and that meta-analysis can serve as a useful tool in the systematic evaluation of any purported relationship. An additional possibility is that we had insufficient power to detect a significant association of the SE genotype with rheumatoid nodules. In fact, the only DRB1 SE genotype to demonstrate statistical significance was also the largest (647 patients with a single *0401 allele [1,384 patients total]). Unfortunately, in contrast to methods available for determining the required sample size in other study designs, no widely accepted approach for estimating the power of a metaanalysis exists, and, in fact, the primary concern regarding meta-analyses is overrepresentation of significant GORMAN ET AL results due to publication bias (63). Of interest, from our findings we would estimate that an individual study would require in excess of 400 cases and 400 controls to demonstrate with 80% power the same effect size for the association with a single DRB1*0401 allele. Other potential reasons for our negative results include variation within the population studied, such as ethnic heterogeneity, differences in patient selection, or imprecision in the definition of nodules. For example, although our meta-analysis was performed among Caucasians, the most striking results were mainly limited to a more defined subgroup, that of northern European Caucasians (28,31). Moreover, studies examining individuals with early RA could lead to misclassification of patients, because in cases categorized as non-nodular, insufficient time may have elapsed in order for nodular disease to develop. In addition, studies relying on patient self-report of nodules may not be as sensitive and/or specific compared with those incorporating physical examination and/or medical record review. It is also possible that systematic differences existed between studies demonstrating positive versus negative results. Because of these concerns, we performed several separate analyses to examine these individual possibilities. This included subgroup analyses limited to northern Europeans and those that used more precise definitions of rheumatoid nodules, adjusted analyses incorporating information regarding disease duration, and analyses that excluded RA patients with familial disease. In all cases, because there were no considerable differences between these results and those from the primary analysis, we do not believe that our negative findings were greatly influenced by these issues. In summary, this IPD meta-analysis represents a powerful and systematic approach to investigating the association of the SE with rheumatoid nodules. Because of the generous contribution of IPD regarding 2,925 RA patients from 10 different countries by 22 researchers, we were able to perform more sophisticated and detailed analyses than is possible using traditional meta-analytic methods based on published summary data. This approach allowed us to fully investigate associations with the SE and the SE allelic genotype, and our results do not support a significant association of rheumatoid nodules with either. A more complete understanding of the pathogenetic mechanisms involved in the development of nodules will likely require consideration of other genetic and nongenetic influences. However, due to the relative infrequency of rheumatoid nodules, individual genotypes and particular exposures, collaborative and/or prospective pooled-analytic methods will be essential in further characterizing these associations. NO RELATIONSHIP OF RHEUMATOID NODULES TO THE HLA–DRB1 SE ACKNOWLEDGMENTS We thank Kirsten A. Pfeiffer, BA, Glenys Thomson, PhD, John J. Chen, PhD, Maria E. Suarez-Almazor, MD, PhD, Catherine Mallon, RN, John M. Colford, MD, MPH, PhD, and Mark Segal, PhD, for their insight and assistance with this project. We are indebted to numerous colleagues who generously contributed data from their independent research for inclusion in this meta-analysis. These contributors include Graciela S. Alarcón, MD, MPH, Birmingham, AL; Sylvette Bas, PhD, Geneva, Switzerland; Alain Cantagrel, MD, Toulouse, France; Bernard Combe, MD, PhD, Montpellier, France; Bernard Cortet, MD, PhD, Lille, France; Anne Crilly, PhD, Glasgow, Scotland; Anne Davidson, MB, BS, FRACP, Bronx, NY; Kerstin Eberhardt, MD, PhD, Lund, Sweden; Jean-François Eliaou, MD, PhD, Montpellier, France; Markku Hakala, MD, Heinola, Finland; Ho-Youn Kim, MD, Seoul, Korea; Alex J. MacGregor, MD, London, UK; Loreto Massardo, MD, Santiago, Chile; Janet E. McDonagh, MD, MRCP, Birmingham, UK; Lachy McLean, MD, PhD, Loughborough, UK; Fiona M. McQueen, MBChB, MD, FRACP, Auckland, New Zealand; Olivier Meyer, MD, Paris, France; Timo Möttönen, MD, Paimio, Finland; Haralampos M. Moutsopoulos, MD, FACP, FRCP (Edin), Athens, Greece; George Moxley, MD, Richmond, VA; Antonio Núñez-Roldan, MD, PhD, Seville, Spain; James R. O’Dell, MD, Omaha, NB; William E. R. Ollier, PhD, Manchester, UK; Aleth Perdriger, MD, PhD, Rennes, France; Janet E. Pope, MD, MPH, FRCPC, London, Ontario, Canada; Jean Roudier, MD, PhD, Marseilles, France; Merrill Rowley, PhD, Victoria, Australia; Güher Saruhan-Direskeneli, MD, Istanbul, Turkey; Michael F. Seldin, MD, PhD, Davis, CA; Alan J. Silman, MD, Manchester, UK; Dharam P. Singal, PhD, Hamilton, Ontario, Canada; Fujio Takeuchi, MD, Tokyo, Japan; Wendy Thomson, PhD, Manchester, UK; Pierre Tiberghien, MD, PhD, Besançon, France; Yoshitaka Toda, MD, Osaka, Japan; Éric Toussirot, MD, Besançon, France; Carl G. Turesson, MD, PhD, Malmö, Sweden; Cor L. Verweij, PhD, Amsterdam, The Netherlands; Daniel Wendling, MD, PhD, Besançon, France; Jehsye-Hsien Yen, MD, Kaohsiung City, Taiwan; Steven A. Young-Min, BA, BMBCh, MRCP, Newcastle-upon-Tyne, UK. REFERENCES 1. Vlak T. Incidence of rheumatoid nodule in Dalmatia: similarities and differences among populations. Arch Med Res 2003;34:56–9. 2. Turesson C, O’Fallon WM, Crowson CS, Gabriel SE, Matteson EL. Occurrence of extraarticular disease manifestations is associated with excess mortality in a community based cohort of patients with rheumatoid arthritis. J Rheumatol 2002;29:62–7. 3. Gabriel SE, Crowson CS, Kremers HM, Doran MF, Turesson C, O’Fallon WM, et al. Survival in rheumatoid arthritis: a populationbased analysis of trends over 40 years. Arthritis Rheum 2003;48: 54–8. 4. Gregersen PK, Silver J, Winchester RJ. The shared epitope hypothesis: an approach to understanding the molecular genetics of susceptibility to rheumatoid arthritis. Arthritis Rheum 1987;30: 1205–13. 5. Weyand CM, Hicok KC, Conn DL, Goronzy JJ. The influence of HLA-DRB1 genes on disease severity in rheumatoid arthritis. Ann Intern Med 1992;117:801–6. 6. Hakala M, Silvennoinen-Kassinen S, Ikaheimo I, Isosomppi J, 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 761 Tiilikainen A. HLA markers in a community-based rheumatoid arthritis series. Ann Med 1997;29:291–6. Salvarani C, Macchioni PL, Mantovani W, Bragliani M, Collina E, Cremonesi T, et al. HLA-DRB1 alleles associated with rheumatoid arthritis in Northern Italy: correlation with disease severity. Br J Rheumatol 1998;37:165–9. Turesson C, Jacobsson L, Bergstrom U, Truedsson L, Sturfelt G. Predictors of extra-articular manifestations in rheumatoid arthritis. Scand J Rheumatol 2000;29:358–64. Egger M, Smith GD, Altman DG. Systematic reviews in health care: meta-analysis in context. 2nd ed. London: BMJ; 2001. p. 313–35. Reveille JD. The genetic contribution to the pathogenesis of rheumatoid arthritis. Curr Opin Rheumatol 1998;10:187–200. Apple RJ, Erlich H. HLA class II genes: structure and diversity. In: Browning MJ, McMichael AJ, editors. HLA and MHC: genes, molecules and function. Oxford (UK): BIOS Scientific Publishers, Ltd; 1996. p. 100. Egger M, Juni P, Bartlett C, Holenstein F, Sterne J. How important are comprehensive literature searches and the assessment of trial quality in systematic reviews? Empirical study. Health Technol Assess 2003;7:1–76. Juni P, Witschi A, Bloch R, Egger M. The hazards of scoring the quality of clinical trials for meta-analysis. JAMA 1999;282: 1054–60. Conn VS, Rantz MJ. Research methods: managing primary study quality in meta-analyses. Res Nurs Health 2003;26:322–33. Hill CL, LaValley MP, Felson DT. Discrepancy between published report and actual conduct of randomized clinical trials. J Clin Epidemiol 2002;55:783–6. Huwiler-Muntener K, Juni P, Junker C, Egger M. Quality of reporting of randomized trials as a measure of methodologic quality. JAMA 2002;287:2801–4. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959;22: 719–48. Breslow NE, Day NE. Statistical methods in cancer research. Volume I: the analysis of case-control studies. IARC Sci Publ 1980;32:5–338. Rothman KJ, Greenland S. Approaches to statistical analysis. In: Rothman KJ, Greenland S, editors. Modern epidemiology. 2nd ed. Philadelphia: Lippincott-Raven; 1998. p. 190–1. Sterne JA, Egger M. Funnel plots for detecting bias in metaanalysis: guidelines on choice of axis. J Clin Epidemiol 2001;54: 1046–55. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315: 629–34. Sterne JA, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. J Clin Epidemiol 2000;53:1119–29. Stata statistical software: release 8.0. College Station (TX): Stata Corporation, 2003. Eberhardt K, Fex E. Clinical course and remission rate in patients with early rheumatoid arthritis: relationship to outcome after 5 years. Br J Rheumatol 1998;37:1324–9. Vaidya B, Pearce SH, Charlton S, Marshall N, Rowan AD, Griffiths ID, et al. An association between the CTLA4 exon 1 polymorphism and early rheumatoid arthritis with autoimmune endocrinopathies. Rheumatology (Oxford) 2002;41:180–3. Thomson W, Harrison B, Ollier B, Wiles N, Payton T, Barrett J, et al. Quantifying the exact role of HLA–DRB1 alleles in susceptibility to inflammatory polyarthritis: results from a large, population-based study. Arthritis Rheum 1999;42:757–62. Cortet B, Perez T, Roux N, Flipo RM, Duquesnoy B, Delcambre B, et al. Pulmonary function tests and high resolution computed tomography of the lungs in patients with rheumatoid arthritis. Ann Rheum Dis 1997;56:596–600. McDonagh JE, Dunn A, Ollier WE, Walker DJ. Compound 762 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. GORMAN ET AL heterozygosity for the shared epitope and the risk and severity of rheumatoid arthritis in extended pedigrees. Br J Rheumatol 1997;36:322–7. Perdriger A, Chales G, Semana G, Guggenbuhl P, Meyer O, Quillivic F, et al. Role of HLA-DR-DR and DR-DQ associations in the expression of extraarticular manifestations and rheumatoid factor in rheumatoid arthritis. J Rheumatol 1997;24:1272–6. Hall FC, Weeks DE, Camilleri JP, Williams LA, Amos N, Darke C, et al. Influence of the HLA-DRB1 locus on susceptibility and severity in rheumatoid arthritis. QJM 1996;89:821–9. MacGregor A, Ollier W, Thomson W, Jawaheer D, Silman A. HLA-DRB1*0401/0404 genotype and rheumatoid arthritis: increased association in men, young age at onset, and disease severity. J Rheumatol 1995;22:1032–6. Bas S, Perneger TV, Mikhnevitch E, Seitz M, Tiercy JM, RouxLombard P, et al. Association of rheumatoid factors and antifilaggrin antibodies with severity of erosions in rheumatoid arthritis. Rheumatology (Oxford) 2000;39:1082–8. Seidl C, Kasser UR, Fischer B, Koch U, Meier L, Fischer P, et al. HLA-DR/DQ interaction in patients with erosive rheumatoid arthritis presenting articular and extraarticular disease manifestations. Eur J Immunogenet 1999;26:19–27. Hayem G, Chazerain P, Combe B, Elias A, Haim T, Nicaise P, et al. Anti-Sa antibody is an accurate diagnostic and prognostic marker in adult rheumatoid arthritis [published erratum appears in J Rheumatol 1999;26:2069). J Rheumatol 1999;26:7–13. Toussirot E, Auge B, Tiberghien P, Chabod J, Cedoz JP, Wendling D. HLA-DRB1 alleles and shared amino acid sequences in disease susceptibility and severity in patients from eastern France with rheumatoid arthritis. J Rheumatol 1999;26:1446–51. Cantagrel A, Navaux F, Loubet-Lescoulié P, Nourhashemi F, Enault G, Abbal M, et al. Interleukin-1␤, interleukin-1 receptor antagonist, interleukin-4, and interleukin-10 gene polymorphisms: relationship to occurrence and severity of rheumatoid arthritis. Arthritis Rheum 1999;42:1093–1100. Benazet JF, Reviron D, Mercier P, Roux H, Roudier J. HLADRB1 alleles associated with rheumatoid arthritis in southern France: absence of extraarticular disease despite expression of the shared epitope. J Rheumatol 1995;22:607–10. Saruhan-Direskeneli G, Istanbul Rheumatology Study Group. Shared epitope homozygosity is strongly associated with rheumatoid arthritis in Turkey. Br J Rheumatol 1998;37:1126–8. Stavropoulos C, Spyropoulou M, Koumantaki Y, Kappou I, Kaklamani V, Linos A, et al. HLA-DRB1 alleles in Greek rheumatoid arthritis patients and their association with clinical characteristics. Eur J Immunogenet 1997;24:265–74. Boki KA, Drosos AA, Tzioufas AG, Lanchbury JS, Panayi GS, Moutsopoulos HM. Examination of HLA-DR4 as a severity marker for rheumatoid arthritis in Greek patients. Ann Rheum Dis 1993;52:517–9. Bali D, Gourley S, Kostyu DD, Goel N, Bruce I, Bell A, et al. Genetic analysis of multiplex rheumatoid arthritis families. Genes Immun 1999;1:28–36. Meyer JM, Evans TI, Small RE, Redford TW, Han J, Singh R, et al. HLA-DRB1 genotype influences risk for and severity of rheumatoid arthritis. J Rheumatol 1999;26:1024–34. Criswell LA, Mu H, Such CL, King MC. Inheritance of the shared epitope and long-term outcomes of rheumatoid arthritis among community-based Caucasian females. Genet Epidemiol 1998;15: 61–72. Pope JE, Stevens A, Howson W, Bell DA. The development of rheumatoid arthritis after recombinant hepatitis B vaccination. J Rheumatol 1998;25:1687–93. Rowley MJ, Stockman A, Brand CA, Tait BD, Rowley GL, Sherritt MA, et al. The effect of HLA-DRB1 disease susceptibility markers on the expression of RA. Scand J Rheumatol 1997;26: 448–55. 46. Combe B, Eliaou JF, Daures JP, Meyer O, Clot J, Sany J. Prognostic factors in rheumatoid arthritis: comparative study of two subsets of patients according to severity of articular damage. Br J Rheumatol 1995;34:529–34. 47. MacGregor AJ, Snieder H, Rigby AS, Koskenvuo M, Kaprio J, Aho K, et al. Characterizing the quantitative genetic contribution to rheumatoid arthritis using data from twins. Arthritis Rheum 2000;43:30–7. 48. Jawaheer D, Lum RF, Amos CI, Gregersen PK, Criswell LA. Influence of sex and shared epitope status on the severity of familial RA [abstract]. Arthritis Rheum 2002;46 Suppl 9:S99. 49. Gorman JD, Criswell LA. The shared epitope and severity of rheumatoid arthritis. Rheum Dis Clin North Am 2002;28:59–78. 50. Gorman JD, Chen JJ, Thomson G, Suarez-Almazor ME, Criswell LA. Rheumatoid arthritis severity and dosage of the shared epitope: results of a meta-analysis [abstract]. Arthritis Rheum 1999;42 Suppl 9:S85. 51. Weyand CM, McCarthy TG, Goronzy JJ. Correlation between disease phenotype and genetic heterogeneity in rheumatoid arthritis. J Clin Invest 1995;95:2120–6. 52. Salvarani C, Macchioni P, Mantovani W, Rossi F, Veneziani M, Boiardi L, et al. Extraarticular manifestations of rheumatoid arthritis and HLA antigens in northern Italy. J Rheumatol 1992; 19:242–6. 53. Möttönen T, Paimela L, Leirisalo-Repo M, Kautiainen H, Ilonen J, Hannonen P. Only high disease activity and positive rheumatoid factor indicate poor prognosis in patients with early rheumatoid arthritis treated with “sawtooth” strategy. Ann Rheum Dis 1998; 57:533–9. 54. Suarez-Almazor ME, Tao S, Moustarah F, Russell AS, Maksymowych W. HLA-DR1, DR4, and DRB1 disease related subtypes in rheumatoid arthritis: association with susceptibility but not severity in a city wide community based study. J Rheumatol 1995;22:2027–33. 55. Ioannidis JP, Tarassi K, Papadopoulos IA, Voulgari PV, Boki KA, Papasteriades CA, et al. Shared epitopes and rheumatoid arthritis: disease associations in Greece and meta-analysis of Mediterranean European populations. Semin Arthritis Rheum 2002;31:361–70. 56. Mattey DL, Dawes PT, Fisher J, Brownfield A, Thomson W, Hajeer AH, et al. Nodular disease in rheumatoid arthritis: association with cigarette smoking and HLA-DRB1/TNF gene interaction. J Rheumatol 2002;29:2313–8. 57. Mattey DL, Hutchinson D, Dawes PT, Nixon NB, Clarke S, Fisher J, et al. Smoking and disease severity in rheumatoid arthritis: association with polymorphism at the glutathione S-transferase M1 locus. Arthritis Rheum 2002;46:640–6. 58. Wolfe F. The effect of smoking on clinical, laboratory, and radiographic status in rheumatoid arthritis. J Rheumatol 2000;27: 630–7. 59. Harrison BJ, Silman AJ, Wiles NJ, Scott DG, Symmons DP. The association of cigarette smoking with disease outcome in patients with early inflammatory polyarthritis. Arthritis Rheum 2001;44: 323–30. 60. Combe B, Didry C, Gutierrez M, Anaya JM, Sany J. Accelerated nodulosis and systemic manifestations during methotrexate therapy for rheumatoid arthritis. Eur J Med 1993;2:153–6. 61. Ahmed SS, Arnett FC, Smith CA, Ahn C, Reveille JD. The HLA-DRB1*0401 allele and the development of methotrexateinduced accelerated rheumatoid nodulosis: a follow-up study of 79 Caucasian patients with rheumatoid arthritis. Medicine (Baltimore) 2001;80:271–8. 62. Ioannidis JP, Ntzani EE, Trikalinos TA, Contopoulos-Ioannidis DG. Replication validity of genetic association studies. Nat Genet 2001;29:306–9. 63. Sutton AJ, Duval SJ, Tweedie RL, Abrams KR, Jones DR. Empirical assessment of effect of publication bias on meta-analyses. BMJ 2000;320:1574–7.