Hum Genet (2005) 118: 123–132 DOI 10.1007/s00439-005-0033-9

O RI GI N AL IN V ES T IG A T IO N

Shizhong Han Æ Yao Li Æ Yumin Mao Æ Yi Xie

Meta-analysis of the association of CTLA-4 exon-1 +49A/G polymorphism with rheumatoid arthritis

Received: 22 April 2005 / Accepted: 4 July 2005 / Published online: 17 August 2005
Springer-Verlag 2005

Abstract Rheumatoid arthritis (RA) is a common autoimmune disease. Cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) is a highly suspected candidate gene for RA susceptibility. However, association studies on the polymorphism of CTLA-4 exon-1 +49A/G in RA have shown conflicting results. Therefore, we performed a meta-analysis to better assess the purported association. In order to look for ethnic effect, we performed subgroup meta-analysis in populations of European descent and Asian descent. Meta-regression analysis was also performed to explore the possible heterogeneity between the two subgroups. Ten studies (11 comparisons) with the CTLA-4 exon-1 +49A/G genotyping on 2,315 patients with RA and 2,536 controls were selected for our meta-analysis. Overall, the fixed-effects odds ratio (OR) for the G versus A allele was 1.11 (P=0.02, 95% confidence interval (CI) 1.02– 1.21), with no between-study heterogeneity. Subgroup and meta-regression analysis according to the ethnicity (European or Asian) demonstrated different scenarios concerning the CTLA-4 exon-1 +49A/G polymorphism’s role in RA susceptibility for the two different subgroups. No effect of G on susceptibility was seen in European descent (five comparisons; OR=1.04, P=0.30, 95% CI 0.95–1.19; no significant betweenstudy heterogeneity). However, there is a significant association in Asian descent under both fixed [OR=1.21, 95% CI (1.06–1.39), P=0.005] and randomeffect models [OR=1.19, 95% CI (1.01–1.42), P=0.04]. Meta-regression analysis also supports the heterogeneity between the two subgroups (P=0.082). We also explored the role of this polymorphism on RA risk under Shizhong Han and Yao Li have contributed equally to this paper. S. Han Æ Y. Li Æ Y. Mao Æ Y. Xie (&) State Key Laboratory of Genetic Engineering, Institute of Genetics, School of Life Science, Fudan University, Shanghai, 200433 People’s Republic of China E-mail: [email protected] Tel.: +86-21-55520025 Fax: +86-21-65642502

other various interested genetic contrasts. These results further support that this polymorphism could not be a risk factor for Europeans. Interestingly, we find that in Asians the G allele has a greater tendency to cause RA in a recessive genetic model. However, sensitivity analysis showed that the combined result of Asian populations was unstable. In conclusion, our meta-analysis results suggest that CTLA-4 exon-1 +49G allele would not be a risk factor for RA in Europeans but might play a role in RA susceptibility for Asians.

Introduction Rheumatoid arthritis (RA) is a common autoimmune disease characterized by features of persistent inflammatory synovitis, potentially resulting in progressive destruction of joint structures. Multiple genetic and environmental factors are considered to be involved in the pathogenesis of rheumatic diseases (Singal et al. 1999). Association between human leukocyte antigen (HLA-DR4) and genetic susceptibility to RA has been widely documented for many populations. Recent genomewide linkage analyses in RA have also confirmed that the major histocompatibility complex is a major susceptibility locus for RA (Cornelis et al. 1998; Jawaheer et al. 2001, 2003; Osorio et al. 2004; Eyre et al. 2004). However, HLA has been estimated to account for approximately 40% of the genetic component of susceptibility to RA, so several to many other genes maybe involved in the pathogenesis of RA disease, each contributing a small effect to the total genetic component (Hasstedt et al. 1994). In order to have a better understanding about the etiology of RA, genes other than HLA must be explored. Evidence from animal models of inflammatory disease, the success of anti-T-cell therapy on RA, and the disease’s association with HLA suggest that RA is a T-cell-mediated autoimmune disease; therefore regulators of T cell activity are strong candidates for

124

influencing disease (Panayi et al. 1992). Cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) is a homologue of CD28. Both molecules are expressed on activated T cells and competitively bind to the B7 molecule on antigen-presenting cells. Whereas interaction of CD28 with B7 can increase and maintain a T cell response, interaction of CTLA-4 with the same ligands plays an inhibitory role in T cell activation and might contribute to peripheral tolerance (Tivol et al. 1996; Thompson et al. 1997). CTLA-4 has been of considerable interest as a candidate susceptibility gene for autoimmune diseases such as RA. Indeed, association with CTLA-4 polymorphisms and linkage with this locus have been demonstrated in several autoimmune disorders: Grave’s disease (GD), insulindependent diabetes mellitus (IDDM), and multiple sclerosis (MS) (Ueda et al. 2003). However, studies on the polymorphism of CTLA-4 exon-1 +49A/G in RA have shown conflicting results. Four studies (Seidl et al. 1998; Yanagawa et al. 2000; Lee et al. 2003; Lei et al. 2005) found an association between this polymorphism and RA independent of HLA-DRB1 alleles. Six studies (Barton et al. 2000; Matsushita et al. 2000; Hadj Kacem et al. 2001; Milicic et al. 2001; Vaidya et al. 2002; Lee et al. 2002; Miterski et al. 2004) found no such association. Three studies (Seidl et al. 1998; Yanagawa et al. 2000; Matsushita et al. 2000) further demonstrated that the association was more significant with respect to some specific HLADRB1 alleles. However, two studies (Barton et al. 2000; Milicic et al. 2001) found no evidence of such association. Considering the possible small effect size of this genetic polymorphism to RA and the relatively small sample size in each study, this discrepancy will become apparent because a single study may have been underpowered to detect a small but real association. Given the amount of accumulated data now available, it is important to perform a quantitative synthesis of the evidence using rigorous methods. The aim of this study was to assess the association of CTLA-4 exon-1 +49A/G polymorphism with risk of RA by conducting a meta-analysis of individual dataset from all eligible case-control studies published to date. Our results suggest that CTLA-4 exon-1 +49G allele would not be a risk factor for RA in Europeans but might play a role in RA susceptibility for Asians.

Materials and methods Identification and eligibility of relevant studies To identify all studies that examined the association of CTLA-4 exon-1 +49A/G polymorphism with RA, we conducted a computerized literature search of PubMed database (from January 1991 to April 2005) using the following keywords and subject terms: ‘‘Cytotoxic T lymphocyte associated antigen 4’’, ‘‘CTLA-4’’, ‘‘polymorphism’’, ‘‘rheumatoid arthritis’’, and ‘‘RA’’.

References of retrieved articles were also screened. When a study reported results on different subpopulations, we treated each subpopulation as a separate comparison in the meta-analysis. Studies included in the current meta-analysis had to meet all the following criteria: (1) patients must fulfill the 1987 revised criteria for RA (Arnett et al. 1988), (2) use an unrelated case-control design, (3) have available genotype frequency and (4) genotype distribution of control population must be in Hardy–Weinberg equilibrium (HWE). Data extraction Two investigators independently extracted the data and reached a consensus on all items. Data were collected on the CTLA-4 exon-1 +49A/G genotype, authors, journal, years of publication, country of origin, demographics, characteristics of RA cases, controls, and racial descent (categorized as Asian, European, or African descent). Statistical analysis Because case-control studies were involved, odds ratio (OR) was used to assess the strength of association between the CTLA-4 exon-1 +49A/G polymorphism and RA. For each study, the data were summarized in a 2·2 table with the allele or genotype of interest and RA defined as present or absent. OR was calculated according to the method of Woolf (1955). A chi-squarebased Q statistic test was performed to assess the between-study heterogeneity (Lau et al. 1997). Heterogeneity was considered significant for P<0.10. A fixedeffects model using the Mantel–Haenszel method and a random-effects model using the DerSimonian and Laird method were used to pool the results (Petitti 1994). In the absence of between-study heterogeneity, the two methods provide identical results. Random effects are more appropriate when heterogeneity is present. The significance of the pooled OR was determined by the Z test. For each genetic contrast, subgroup analysis was performed according to the racial descent of the population to estimate ethnic-specific OR. Since the case-control approach was used in all eligible studies in our meta-analysis, we also performed meta-regression analysis to explore the possible heterogeneity using ethnicity as a single covariate under various genetic contrasts. A random-effects weighted linear regression model was used, whereby the study-specific log(OR) was regressed on the ethnicity of each study (Thompson et al. 1999). The regression incorporated both the within-study variance as well as the betweenstudy variance, and the weights were estimated using restricted maximum likelihood. Statistical significance was defined as a P value less than 0.10 because of the relatively weak statistical power.

125

Publication bias was investigated by using a funnel plot, in which the standard error of log(OR) of each study was plotted against its OR. An asymmetric plot suggested possible publication bias. Funnel plot asymmetry was assessed by the method of Egger’s linear regression test, which used a linear regression approach to measure funnel plot asymmetry on the natural logarithm scale of the OR (Egger et al. 1997). Briefly, the standard normal deviation of the estimate effect (defined as the Ln(OR) divided by its standard error, termed ‘‘SND’’) was regressed against the estimate’s precision [defined as the inverse of the standard error of Ln(OR)] (SND = a + b Precision). The y-intercept (a) provides a measure of asymmetry (the larger its deviation from zero, the more pronounced the asymmetry). The significance of the intercept (a) was determined by the t test as suggested by Egger. Hardy–Weinberg equilibrium was tested by the chisquare test for goodness of fit. The power of each study was computed as the probability of detecting an association between the CTLA-4 exon-1 +49A/G polymorphism and RA at the 0.05 level of significance with the two-sided test, assuming a relative risk of either 1.5 or 1.2 under different genetic models. We calculated the power for meta-analysis by the method recommended by Hedges and Pigott (2001). The power of detecting some effect size of meta-analysis can be calculated from the two-tailed Z test, given a significant value 0.05. Analyses were performed using the software Stata version 7, ReviewManage 4.2 (Oxford, England) and QUANTO version 0.5 (http://hydra.usc.edu/gxe). All P values were two-sided.

Results Study inclusion and characteristics Through literature search and selection based on the inclusion criteria, 14 studies were found, but only 10 studies met our inclusion criteria, as listed in Table 1. For various reasons, four articles were not included. Specifically, in two studies (Liu et al. 2004; GonzalezEscribano et al. 1999), genotype distribution in control population significantly deviates from HWE. An article (Miterski et al. 2004) was not included because it did not contain genotype distribution information. Another article (Barton et al. 2004) was excluded from the meta-analysis because it used an affected sibling pair and unrelated control design instead of unrelated case-control design. In addition, the study by Vaidya et al. (2002) found an association between CTLA-4 exon-1 +49G/A polymorphism and RA, but this association was attributed to some RA patients coexisting with autoimmune endocrinopathies. When these patients were excluded, no association was detected. So data from this study should only include patients without coexisting autoimmune endocrinopathies in our meta-analysis. Notably, only two studies (Lee

et al. 2002; Lei et al. 2005) mentioned that the age and sex status were matched between case and control population, most of the other studies did not mention it or the age and sex status were not stated. Table 2 shows the power of individual studies to demonstrate an association between the CTLA-4 exon-1 +49A/G polymorphism and RA if the true relative risk was 1.2 or 1.5 under a dominant, recessive, or log additive genetic model. A statistical power greater than 80% was attained by only 1 (or 4) of the 11 comparisons if the genotypic risk was equal to or greater than 1.5 under a dominant genetic model (or log additive genetic model). However, if the genotypic risk was 1.2, none of the studies had power greater than 80% to demonstrate an association no matter which genetic model was used. We calculated the power of the meta-analysis to detect association if the overall effect size (allele G versus allele A) is 1.2 or 1.5 in the two separate populations. The results illustrated that in European studies the power is 0.867004 and 0.999999 and in Asian studies the power is 0.743329 and 0.999942 at the effect sizes of 1.2 and 1.5, respectively. Meta-analysis database Ten population-based association studies met our criteria for the meta-analysis. One of the eligible studies (Barton et al. 2000) contained subjects of two different racial descents, thus a total of 11 separate comparisons were considered. Table 1 shows these studies in detail. A total of 2,315 patients and 2,536 controls were investigated. Four studies were from Europe, five from Asia, and one from Africa. Table 3 shows both genotype and allele frequencies of RA patients and controls in the selected studies. The G allele was more highly represented among controls of Asian descent [64.4% (95% confidence interval (CI) 57.1–71.7)] and African descent [61%] than in controls of European descent [38.4% (95% CI 32.0–44.9)]. Overall, the prevalence of A/A homozygosity was 13.1, 37.5, and 13.3% in control subjects of Asian, European, and African descents, respectively. The respective prevalence rates of A/G heterozygosity were 50.0, 48.0, and 41.3%, and the respective rates for G/G homozygosity were 41.9, 14.4, and 45.3%. Interestingly, the allele and genotype distribution are very similar between Asian and African descent populations. Because there was only one African descent population study, subgroup analysis according to ethnicity was only considered for Asian or European descent population. Overall effects for alleles No significant heterogeneity existed between the 11 study comparisons (P=0.34). First, a fixed-effect model was used to pool the result (Fig. 1). There was evidence that the G allele resulted in increased susceptibility to

China (Asian)

China (Asian)

Korea (Asian)

UK (European)

Tunisia (African)

UK (European)

Japan (Asian)

UK (European)

Spain (European)

Japan (Asian)

German (European)

Lei (2005)

Lee (2003)

Lee (2002)

Vaidya (2002)

Hadj Kacem (2001)

Milicic (2001)

Yanagawa (2000)

Barton (2000)

Barton (2000)

Matsushita (2000)

Seidl (1998)

NA not available

Country (origin)

First author (year)

64 unrelated cases from the ARC National twin study and 128 RA probands from the ARC national respository; 83% female; mean age 40, range 16–78 136 patients with RA from the Galicia region of northern Spain; 71% female; mean age 49, range 16–78 461 unrelated Japanese patients with RA; % female NA; mean age NA 258 unrelated RA patients; % female NA; mean age 52, range 24–68

85 unrelated Japanese patients with RA; 84% female; mean age NA

421 Caucasian RA patients from Nuffield Orthopaedic Center, Oxford

86 patients from the Rheumatology Clinic of Korea University Hospital; 83.7% female; mean age 45, range 16–75 123 unrelated white probands with early RA from the north-east of England; 65% female; mean age 53.7, range 18–85; coexist AITD/IDDM 15% 60 RA patients; 80% female; mean age 45, range 18–80

186 unrelated adult patients living in northern Taiwan; 88.1% female; mean age NA

326 unrelated patients living in China; 72.1% female; mean age 48.4 (SD=±8.13)

Case

Table 1 Characteristics of studies included in the meta-analysis

150 unrelated healthy Japanese; % female NA; mean age NA 456 unrelated healthy controls without family history of any autoimmune disease; % female NA; mean age NA

144 controls from the same population; 59% female; mean age NA, range 16–90

250 unrelated patients living in China; 70.8% female; mean age 47.8 (SD=±11.05) 203 randomly selected normal adults who had no history of autoimmune disease living in northern Taiwan; mean age NA 86 sex and age-matched healthy Korean controls from the Rheumatology Clinic of Korea University Hospital 349 local white subjects without clinical evidence or family history of any autoimmune disease; % female NA; mean age NA 150 unrelated healthy controls without clinical evidence or family history of RA or inflammatory joint disease; 53% female; mean age NA 452 ethnically matched healthy controls from Oxford Regional Transfusion Center 200 unrelated healthy Japanese without clinical evidence or family history of any autoimmune disease; 52% female; mean age NA 96 normal individuals were recruited from general practice who had no history of inflammatory joint disease; 57% female; mean age 48, range 19–71

Control

1.14

1.17

0.92–1.43

0.89–1.53

0.66–1.34

0.60–1.20

0.84

0.94

0.77–1.61

0.89–1.30

0.52–1.24

0.88–1.66

0.47–1.20

1.02–1.89

1.07–1.73

95% CI

1.11

1.07

0.80

1.20

0.75

1.39

1.36

OR (G versus A)

126

127 Table 2 Power of studies included in the meta-analysis First author (year)

Country (origin)

Genetic models Dominant

Lei (2005) Lee (2003) Lee (2002) Vaidya (2002) Hadj Kacem (2001) Milicic (2001) Yanagawa (2000) Barton (2000) Barton (2000) Matsushita (2000) Seidl (1998)

China (Asian) China (Asian) Korea (Asian) UK (European) Tunisia (African) UK (European) Japan (Asian) UK (European) Spain (European) Japan (Asian) German (European)

Recessive

Log additive

RR=1.2

RR=1.5

RR=1.2

RR=1.5

RR=1.2

RR=1.5

0.1090 0.0900 0.0675 0.1329 0.0699 0.2591 0.0742 0.1082 0.1136 0.0974 0.2029

0.3279 0.2388 0.1322 0.4449 0.1427 0.8202 0.1626 0.3378 0.3613 0.2772 0.6920

0.1887 0.1442 0.0909 0.0969 0.0898 0.1538 0.1073 0.0822 0.0855 0.1601 0.1357

0.6680 0.5097 0.2618 0.3087 0.2578 0.5729 0.3438 0.2251 0.2447 0.5712 0.4983

0.3066 0.2253 0.1261 0.2259 0.1285 0.4543 0.1561 0.1721 0.1840 0.2506 0.3673

0.8864 0.7447 0.4144 0.7687 0.4252 0.9849 0.5337 0.6142 0.6539 0.8127 0.9571

RA on a worldwide population. The summary OR was 1.11 by fixed effects (P=0.02) and 1.10 by random effects (P=0.07). Second, in order to look for ethnic effect we performed the subgroup meta-analysis in populations of European descent and Asian decent, respectively. No effect of G on susceptibility was seen in subgroups of European descent (five comparisons; OR=1.06, P=0.31, 95% CI 0.95–1.19; no significant between-study heterogeneity). However, there was a significant association in Asian descent under both fixed [OR=1.21, 95% CI (1.06–1.39), P=0.005] and randomeffect models [OR=1.19, 95% CI (1.01–1.42), P=0.04]. Meta-regression analysis also supported the heterogeneity between the two subgroups (P=0.082; Table 4).

European and Asian subgroups, respectively. The pooled results consistently encompassed 1.0 in the European subgroup under fixed or random-effect model indicating that the significance of pooled ORs was not excessively influenced by any single study in the European subgroup (data not shown). However, when Lei’s study was excluded from the Asian subgroup, the pooled result was not significant for G versus A allele under both fixed [OR=1.15, 95% CI (0.98–1.36), P=0.10] and random-effect models [OR=1.13, 95% CI (0.92–1.40), P=0.25]. Since only four studies were included, this result must be explained with caution. More Asian studies should be recruited to clarify this possible association. Publication bias

Other contrasts We found no evidence of an association of the G/G genotype with the risk of RA relative to the A/A genotype on a worldwide population. However, subgroup analysis supported a significant association under both fixed and random-effect models for Asians but not for Europeans. No evidence for an association with RA was discerned as we compared G/A genotype with A/A genotype for Asians or Europeans, but significant between-study heterogeneity did exist in this contrast. Also, no evidence for a significant association with RA was documented when recessive or dominant models were examined for the effect of G. Between-study heterogeneity existed in the Asian subgroup under a recessive model. In addition, it can be found that the mean effect size of G allele in Asians was always higher than in Europeans except under the G/A versus A/A contrast. Meta-regression analysis also supported our subgroup analysis. More details are shown in Table 4. Sensitivity analysis Sensitivity analysis was performed by sequential omission of individual studies under various contrasts in

Figure 2 shows the funnel plots for comparisons of G versus A in the OR analysis. Furthermore, an Egger’s test was performed to provide statistical evidence for funnel plot symmetry. These data provided no significant evidence for publication bias.

Discussion This meta-analysis demonstrated a significant association between the CTLA-4 exon-1 +49A/G polymorphism and RA on a worldwide population under an overall effect size of allele G versus A. Subgroup analysis according to the origin of the population investigated suggested this polymorphism was a risk factor for susceptibility to RA for Asians but not for Europeans. Meta-analysis under other various interested genetic contrasts suggested that this polymorphism more likely has a trend to affect RA in a recessive genetic model in Asians. Still we cannot detect any significant association in Europeans regardless of which genetic contrast was performed. Association with CTLA-4 polymorphisms and linkage with this locus have been demonstrated in several autoimmune disorders (Nistico et al. 1996; Marron

297 (59.4) 270 (66.5) 127 (73.8) 198 (66.0) 244 (61.0) 184 (61.3) 248 (35.5 359 (39.7) 89 (46.4) 94 (32.6) 346 (37.9) 434 (66.6) 273 (73.4) 117 (68.0) 73 (60.8) 108 (63.5) 599 (65.0) 83 (40.0) 349 (41.4) 162 (42.2) 85 (31.3) 212 (40.9) 203 (40.6) 136 (33.5) 45 (26.2) 102 (34.0) 156 (39.0) 116 (38.7) 450 (64.5) 545 (60.3) 103 (53.6) 194 (67.4) 568 (62.1) 218 (33.4) 99 (26.6) 55 (32.0) 47 (39.2) 62 (36.5) 323 (35.0) 125 (60.1) 493 (58.6) 222 (57.8) 187 (68.8) 304 (59.1) (34.4) (41.9) (57.0) (45.3) (39.0) (37.3) (12.9) (16.2) (19.8) (8.3) (14.9) 86 85 49 68 78 56 45 73 19 12 68 148 (45.4) 103 (55.4) 41 (47.7) 23 (38.3) 29 (34.1) 200 (43.3) 14 (13.5) 63 (15.0) 38 (19.8) 14 (10.3) 37 (14.0) 125 (50.0) 100 (49.2) 29 (33.7) 62 (41.3) 88 (44.0) 72 (48.0) 158 (45.2) 213 (47.1) 51 (53.1) 70 (48.6) 210 (46.1) 138 (42.3) 67 (36.0) 35 (40.7 27 (45.0) 50 (58.8) 199 (43.2) 55 (52.9) 223 (53.0) 86 (44.8) 57 (41.9) 138 (53.9) 39 (15.6) 18 (8.9) 8 (9.3) 20 (13.3) 34 (17.0) 22 (14.7) 146 (41.8) 166 (36.7 26 (27.1 62 (43.1) 179 (39.0) 40 (12.3) 16 (8.6) 10 (11.6) 10 (16.7) 6 (7.1) 62 (13.4) 35 (33.7) 135 (32.1) 68 (35.4) 65 (47.8) 83 (32.2) Asian Asian Asian African Asian Asian European European European European European Lei (2005) Lee (2003) Lee (2002) Hadj Kacem (2001) Yanagawa (2000) Matsushita (2000) Vaidya (2002) Milicic (2001) Barton (2000), UK Barton (2000), Spain Seidl (1998)

AG

Case n (%) Control n (%) Case n (%) Control n (%) Case n (%) Control n (%) Case n (%) Control n (%) Case n (%) Control n (%)

A AA

GG

Allele Genotype Origin First author (year)

Table 3 Distribution of CTLA-4 exon-1 +49A/G genotype and allele among RA cases and controls included in the meta-analysis

G

128

et al. 1997; Yanagawa et al. 1995; Kristiansen et al. 2000) such as GD, IDDM, and MS. However, using regression analysis, the exon-1 +49A/G polymorphism has been excluded as a disease causal variant for GD and IDDM, and CT60 was considered the most likely causal polymorphism (Ueda et al. 2003). Reported associations about the exon-1 +49A/G polymorphism with autoimmune disease might attribute to the linkage disequilibrium (LD) between this polymorphism and the real causal variant. We must mention that the exon-1 +49G polymorphism that causes an unsynonymous amino acid change in the signal peptide of CTLA-4 protein can change the cell surface expression of CTLA4, thus leading to decreased inhibitory function of CTLA-4 (Anjos et al. 2002; Ueda et al. 2003; Kouki et al.2000). So it seems reasonable to take this polymorphism as the possible causal variant for autoimmune diseases, but because this allele is so widely distributed among the worldwide populations (Asian, European, African), selection pressure might exist to maintain its high frequency, such as a more effective defense against infectious bacterial diseases. Arguably, single marker locus analysis provides limited power to detect an association if a real causal variant stands by the marker. Haplotype analysis is therefore recommended to improve the power, due to the fact that if a disease-related allele exists, it should lie on one or more associated haplotypes. A recent study (Barton et al. 2004) used five haplotype-tagging SNP (htSNP) including the exon-1 +49A/G polymorphism that can capture at least 80% of the common haplotype in the European population to test the association of the CTLA-4 gene with RA. An affected sibling pair–control group comprising 96 affected sibling pairs and 173 unrelated controls and 122 RA simplex families were investigated but neither single-locus analysis nor Haplotype analysis revealed any evidence of association. Furthermore, a large case-control design of 759 unrelated RA patients and 755 controls was also performed in this study to investigate another two highly suspected polymorphism, CT60 and rs1863800, concerning their role in the susceptibility of RA. Results were still negative: no association was detected. Another large casecontrol study (Orozco et al. 2004) also discarded the possible role of CT60 as a causal variant. These large studies demonstrated that CTLA-4 could not be a major risk factor for RA in Europeans. Moreover, one whole genome screen for susceptibility loci in RA revealed only nominal evidence of linkage with the region 2q33 (Cornelis et al. 1998). Another large linkage analysis (Jawaheer et al. 2001, 2003; Osorio et al. 2004; Eyre et al. 2004) did not find any evidence of nominal linkage. These linkage and association studies performed on European populations support our meta-analysis that CTLA-4 polymorphisms could not be a major risk factor for Europeans. However, due to low resolution of linkage studies and limited association studies on other polymorphisms within CTLA-4 gene, we cannot exclude the possibility that CTLA-4 gene might be a modifying

129

Fig. 1 Overall meta-analysis for CTLA-4 exon-1 +49A/G gene polymorphisms and RA: G versus A allele. The study is shown by a point estimate of the OR and the accompanying 95% CIs using a fixed-effect model. n indicates the total number of G alleles; N indicates the total number of G alleles plus A alleles

factor rather than a major risk factor in European populations. In our subgroup meta-analysis, we found a significant association of CTLA-4 exon-1 +49A/G polymorphism to RA susceptibility in Asian populations, although sensitivity analysis showed that the combined result was not stable and one study influenced the Asian group result greatly. Interestingly, from Table 4 we can find that this polymorphism had a tendency to affect RA in a recessive genetic model in Asians. This kind of genetic model indicates that one G allele might not be sufficient to exhibit a risk effect but that both G alleles are needed. In addition, the mean effect size of G allele in Asians was

always higher than in Europeans except under G/A versus A/A contrasts, but this inconsistency may be partially explained by the study heterogeneity within each subgroup. Meta-regression analysis also supported our subgroup analysis. The reasons that the same polymorphism plays a different role in different ethnic populations or across different studies may arise from many aspects. First, RA is a complex disease and genetic heterogeneity exists in different populations. Whole genome linkage studies on RA have confirmed this genetic heterogeneity (Cornelis et al. 1998; Jawaheer et al. 2001, 2003; Osorio et al. 2004; Eyre et al. 2004). Secondly, clinical heterogeneity may also explain the discrepancy. Potential contribution of differences in patient populations (e.g., age and years from onset, female proportion, disease severity, and so on) might cause different results. Sometimes association can only be found in stratification analysis according to

Table 4 Summary of ORs for various contrasts and meta-regression results Contrast

Comparisons

Random-effects OR (95% CI)

Fixed-effects OR (95% CI)

P value of meta-regression coefficient

G versus A alleles European Asian G/G versus A/A European Asian G/A versus A/A European Asian (G/G +G/A) versus A/A European Asian G/G versus (G/A+A/A) European Asian

11 5 5 11 5 5 11 5 5 11 5 5 11 5 5

1.10 1.06 1.19 1.16 1.07 1.41 1.11 1.11 1.13 1.14 1.11 1.24 1.14 0.97 1.21

1.11 1.06 1.21 1.17 1.07 1.41 1.15 1.18 1.13 1.16 1.15 1.26 1.17 0.97 1.30

0.082a

a

(0.99–1.22) (0.95–1.19) (1.01–1.42) (0.96–1.41) (0.83–1.38) (1.04–1.91) (0.90–1.36)b (0.84–1.48)b (0.76–1.68)b (0.96–1.36) (0.87–1.42) (0.92–1.66) (0.95–1.37) (0.77–1.22) (0.88–1.66)b

Indicates a significant difference of the effect size between the two nations Indicates a significant heterogeneity (0.01
b

(1.02–1.21) (0.95–1.19) (1.06–1.39) (0.97–1.41) (0.83–1.38) (1.04–1.91) (0.99–1.34) (0.99–1.40) (0.84–1.52) (1.01–1.34) (0.97–1.36) (0.95–1.67) (1.02–1.34) (0.77–1.22) (1.07–1.56)

0.061a 0.612 0.432 0.154

130 Fig. 2 Funnel plot for comparison of G allele versus A allele to determine publication bias. Formal statistical criteria by Egger’s test was also performed to investigate the symmetry of the funnel plot (t = 1.81, P=0.1045 for G allele versus A allele)

the clinical character. Thirdly, population structure difference may also contribute to the discrepancy. Different populations often have different LD patterns. The same polymorphism plays a different role in disease susceptibility in different ethnic populations, implicating that this polymorphism might not be a causal variant. The fact is that this polymorphism may be in LD with a nearby causal variant in one ethnic population but not in another. Fourthly, the difference might arise from chance, such as type I error, or because of multiple testing which inflates the type I error. Last but not the least, phenotype characters would influence study results greatly. Prospective studies have shown that RA patients often coexist with autoimmune thyroid disease (AITD) or IDDM. One study (Vaidya et al. 2002) stated that the autoimmune endocrinopathies, which the RA patients were coexisting with, largely explained the association between the CTLA-4 exon-1 +49G allele and RA. It will also be helpful to understand the differences between study results, considering the different distributions of other risk factors, known or unknown, which may interact with genetic factors in the development of RA. Interestingly our meta-analysis result is in accordance with another meta-analysis study, which investigated CTLA-4 exon-1 +49A/G polymorphism’s role in another autoimmune disease: systemic lupus erythematosus (Lee et al. 2005). This study also supports exon-1 +49G allele as a significant risk factor for Asians but not for Europeans. We cannot be sure if this is only a coincidence or if this is an indicator of a relationship between this polymorphism and autoimmune diseases in different ethnicities. In addition, Lei et al. (2005) also performed a meta-analysis, in their study, concerning the same polymorphism with RA. There are differences between our results and their results: they support that the G allele might be a risk factor for Europeans. This discrepancy

can be attributed to the difference of study inclusion criteria after careful comparison of the two studies. However, we think that our study inclusion criteria for meta-analysis should be more reasonable. In two studies (Liu et al. 2004; Gonzalez-Escribano et al. 1999), genotype distribution in the control population significantly deviates from HWE, and thus were excluded from our meta-analysis but included in their study. An article (Barton et al. 2004) was excluded from our meta-analysis because it used an affected sibling pair and unrelated control design instead of unrelated case-control design. We used data from Vaidya et al.’s (2002) study but only included patients without coexisting autoimmune endocrinopathies; Lei et al. used all the patient data. In conclusion, our meta-analysis, along with other studies, suggests that the CTLA-4 exon-1 +49G allele might be a risk factor for RA in Asians but not in Europeans. More studies or large case-control studies should be performed to clarify its possible role in Asian populations. Acknowledgements We are grateful to two anonymous reviewers for their thoughtful criticisms, comments, and suggestions on early versions of the manuscript. We thank editors for proofreading the manuscript. We also thank Mr. Alonso Fuentes for revision of the manuscript. This research is supported by a grant 2002AA2Z2002 from the National High Technology Research and Development Program of China (863 Program) and this research is also a part of projects 60473104 and 30371422 supported by the National Natural Science Foundation of China.

References Anjos S, Nguyen A, Ounissi-Benkalha H, Tessier MC, Polychronakos C (2002) A common autoimmunity predisposing signal peptide variant of the cytotoxic T-lymphocyte antigen 4 results in inefficient glycosylation of the susceptibility allele. J Biol Chem 277(48):46478–46486

131 Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS, Healey LA, Kaplan SR, Liang MH, Luthra HS (1988) The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 31:315–324 Barton A, Myerscough A, John S, Gonzalez-Gay M, Ollier W, Worthington J (2000) A single nucleotide polymorphism in exon 1 of cytotoxic T-lymphocyte-associated-4 (CTLA-4) is not associated with rheumatoid arthritis. Rheumatology (Oxford) 39(1):63–66 Barton A, Jury F, Eyre S, Bowes J, Hinks A, Ward D, Worthington J (2004) Haplotype analysis in simplex families and novel analytic approaches in a case-control cohort reveal no evidence of association of the CTLA-4 gene with rheumatoid arthritis. Arthritis Rheum 50(3):748–752 Cornelis F, Faure S, Martinez M, Prud’homme JF, Fritz P, Dib C, Alves H, Barrera P, de Vries N, Balsa A, Pascual-Salcedo D, Maenaut K, Westhovens R, Migliorini P, Tran TH, Delaye A, Prince N, Lefevre C, Thomas G, Poirier M, Soubigou S, Alibert O, Lasbleiz S, Fouix S, Weissenbach J (1998) New susceptibility locus for rheumatoid arthritis suggested by a genome-wide linkage study. Proc Natl Acad Sci USA 95:10746–10750 Egger M, Davey Smith G, Schneider M, Minder C (1997) Bias in meta-analysis detected by a simples, graphical test. Br Med J 315:629–634 Eyre S, Barton A, Shephard N, Hinks A, Brintnell W, MacKay K, Silman A, Ollier W, Wordsworth P, John S, Worthington J (2004) Investigation of susceptibility loci identified in the UK rheumatoid arthritis whole-genome scan in a further series of 217 UK affected sibling pairs. Arthritis Rheum 50(3):729–735 Gonzalez-Escribano MF, Rodriguez R, Valenzuela A, Garcia A, Garcia-Lozano JR, Nunez-Roldan A (1999) CTLA4 polymorphisms in Spanish patients with rheumatoid arthritis. Tissue Antigens 53(3):296–300 Hadj Kacem H, Kaddour N, Adyel FZ, Bahloul Z, Ayadi H (2001) HLA-DQB1 CAR1/CAR2, TNFa IR2/IR4 and CTLA-4 polymorphisms in Tunisian patients with rheumatoid arthritis and Sjogren’s syndrome. Rheumatology (Oxford) 40(12):1370–1374 Hasstedt SJ, Clegg DO, Ingles L, Ward RH (1994) HLA-linked rheumatoid arthritis. Am J Hum Genet 55:738–746 Hedges LV, Pigott TD (2001) The power of statistical tests in metaanalysis. Psychol Methods 6:203–217 Jawaheer D, Seldin MF, Amos CI, Chen WV, Shigeta R, Monteiro J, Kern M, Criswell LA, Albani S, Nelson JL, Clegg DO, Pope R, Schroeder HW Jr, Bridges SL Jr, Pisetsky DS, Ward R, Kastner DL, Wilder RL, Pincus T, Callahan LF, Flemming D, Wener MH, Gregersen PK (2001) A genomewide screen in multiplex rheumatoid arthritis families suggests genetic overlap with other autoimmune diseases. Am J Hum Genet 68:927–936 Jawaheer D, Seldin MF, Amos CI, Chen WV, Shigeta R, Etzel C, Damle A, Xiao X, Chen D, Lum RF, Monteiro J, Kern M, Criswell LA, Albani S, Nelson JL, Clegg DO, Pope R, Schroeder HW Jr, Bridges SL Jr, Pisetsky DS, Ward R, Kastner DL, Wilder RL, Pincus T, Callahan LF, Flemming D, Wener MH, Gregersen PK (2003) North American rheumatoid arthritis consortium. Screening the genome for rheumatoid arthritis susceptibility genes: a replication study and combined analysis of 512 multicase families. Arthritis Rheum 48:906–916 Kouki T, Sawai Y, Gardine CA, Fisfalen ME, Alegre ML, DeGroot LJ (2000) CTLA-4 gene polymorphism at position 49 in exon 1 reduces the inhibitory function of CTLA-4 and contributes to the pathogenesis of Graves’ disease. J Immunol 165:606–611 Kristiansen OP, Larsen ZM, Pociot F (2000) CTLA-4 in autoimmune diseases a general susceptibility gene to autoimmunity? Genes Immun 1:170–184 Lau J, Ioannidis JP, Schmid CH (1997) Quantitative synthesis in systematic reviews. Ann Intern Med 127:820–826 Lee YH, Choi SJ, Ji JD, Song GG (2002) No association of polymorphisms of the CTLA-4 exon 1(+49) and promoter ( 318) genes with rheumatoid arthritis in the Korean population. Scand J Rheumatol 31(5):266–270

Lee CS, Lee YJ, Liu HF, Su CH, Chang SC, Wang BR, Chen TL, Liu TL (2003) Association of CTLA4 gene A-G polymorphism with rheumatoid arthritis in Chinese. Clin Rheumatol 22(3):221–224 Lee YH, Harley JB, Nath SK (2005) CTLA-4 polymorphisms and systemic lupus erythematosus (SLE): a meta-analysis. Hum Genet 116(5):361–367 Lei C, Dongqing Z, Yeqing S, Oaks MK, Lishan C, Jianzhong J, Jie Q, Fang D, Ningli L, Xinghai H, Daming R (2005) Association of the CTLA-4 gene with rheumatoid arthritis in Chinese Han population. Eur J Hum Genet 2005 Apr 20 [Epub ahead of print] Liu MF, Wang CR, Chen PC, Lin TL (2004) CTLA-4 gene polymorphism in promoter and exon-1 regions is not associated with Chinese patients with rheumatoid arthritis. Clin Rheumatol 23(2):180–181 Marron MP, Raffel LJ, Garchon HJ, Jacob CO, Serrano-Rios M, Martinez Larrad MT, Teng WP, Park Y, Zhang ZX, Goldstein DR, Tao YW, Beaurain G, Bach JF, Huang HS, Luo DF, Zeidler A, Rotter JI, Yang MC, Modilevsky T, Maclaren NK, She JX (1997) Insulin-dependent diabetes mellitus (IDDM) is associated CTLA-4 polymorphisms in multiple ethnic groups. Hum Mol Genet 6:1275–1282 Matsushita M, Tsuchiya N, Oka T, Yamane A, Tokunaga K (2000) New polymorphisms of human CD80 and CD86: lack of association with rheumatoid arthritis and systemic lupus erythematosus. Genes Immun 1(7):428–434 Milicic A, Brown MA, Wordsworth BP (2001) Polymorphism in codon 17 of the CTLA-4 gene (+49 A/G) is not associated with susceptibility to rheumatoid arthritis in British Caucasians. Tissue Antigens 58(1):50–54 Miterski B, Drynda S, Boschow G, Klein W, Oppermann J, Kekow J, Epplen JT (2004) Complex genetic predisposition in adult and juvenile rheumatoid arthritis. BMC Genet 45(1):2 Nistico L, Buzzetti R, Pritchard LE, Van der Auwera B, Giovannini C, Bosi E, Larrad MT, Rios MS, Chow CC, Cockram CS, Jacobs K, Mijovic C, Bain SC, Barnett AH, Vandewalle CL, Schuit F, Gorus FK, Tosi R, Pozzilli P, Todd JA (1996) The CTLA-4 gene region of chromosome 2q33 is linked to and associated with type 1 diabetes. Hum Mol Genet 5:1075–1080 Orozco G, Torres B, Nunez-Roldan A, Gonzalez-Escribano MF, Martin J (2004) Cytotoxic T-lymphocyte antigen-4-CT60 polymorphism in rheumatoid arthritis. Tissue Antigens 64(6):667–670 Osorio Y, Fortea J, Bukulmez H, Petit-Teixeira E, Michou L, Pierlot C, Cailleau-Moindrault S, Lemaire I, Lasbleiz S, Alibert O, Quillet P, Bardin T, Prum B, Olson JM, Cornelis F (2004) Dense genome-wide linkage analysis of rheumatoid arthritis, including covariates. Arthritis Rheum 50(9):2757–2765 Panayi GS, Lanchbury JS, Kingsley GH (1992) The importance of the T cell in initiating and maintaining the chronic synovitis of rheumatoid arthritis. Arthritis Rheum 35:729–735 Petitti DB (1994) Meta-analysis, decision analysis, and cost-effectiveness analysis. Oxford University Press, New York Seidl C, Donner H, Fischer B, Usadel KH, Seifried E, Kaltwasser JP, Badenhoop K (1998) CTLA4 codon 17 dimorphism in patients with rheumatoid arthritis. Tissue Antigens 51:62–68 Singal DP, Li J, Zhu Y (1999) Genetic basis for rheumatoid arthritis. Arch Immunol Ther Exp 47:307–311 Thompson CB, Allison JP (1997) The emerging role of CTLA-4 as an immune attenuator. Immunity 7:445–450 Thompson SG, Sharp SJ (1999) Explaining heterogeneity in metaanalysis: a comparison of methods. Stat Med 18:2693–2708 Tivol EA, Schweitzer AN, Sharpe AH (1996) Costimulation and autoimmunity. Curr Opin Immunol 8:822–830 Ueda H, Howson JM, Esposito L, Heward J, Snook H, Chamberlain G, Rainbow DB, Hunter KM, Smith AN, Di Genova G, Herr MH, Dahlman I, Payne F, Smyth D, Lowe C, Twells RC, Howlett S, Healy B, Nutland S, Rance HE, Everett V, Smink LJ, Lam AC, Cordell HJ, Walker NM, Bordin C, Hulme J, Motzo C, Cucca F, Hess JF, Metzker ML, Rogers J, Gregory S, Allahabadia A, Nithiyananthan R, Tuomilehto-Wolf E,

132 Tuomilehto J, Bingley P, Gillespie KM, Undlien DE, Ronningen KS, Guja C, Ionescu-Tirgoviste C, SavageDA, Maxwell AP, Carson DJ, Patterson CC, Franklyn JA, Clayton DG, Peterson LB, Wicker LS, Todd JA, Gough SC (2003) Association of the T-cell regulatory gene CTLA4 with susceptibility to autoimmune disease. Nature 423:506–511 Vaidya B, Pearce SH, Charlton S, Marshall N, Rowan AD, Griffiths ID, Kendall-Taylor P, Cawston TE, Young-Min S (2002) An association between the CTLA4 exon 1 polymorphism and early rheumatoid arthritis with autoimmune endocrinopathies. Rheumatology (Oxford) 41(2):180–183

Woolf B (1955) On estimating the relationship between blood group and disease. Ann Human Genet 19:251–253 Yanagawa T, Hidaka Y, Guimaraes V, Soliman M, DeGroot LJ (1995) CTLA-4 gene polymorphism associated with Graves’ disease in a Caucasian population. J Clin Endocrinol Metab 80:41–58 Yanagawa T, Gomi K, Nakao EI, Inada S (2000) CTLA-4 gene polymorphism in Japanese patients with rheumatoid arthritis. J Rheumatol 27(12):2740–2742