SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX A
  10. APPENDIX B

Objective

To define interactions between the HLA–DRB1 shared epitope (SE), PTPN22, and smoking in cyclic citrullinated peptide (CCP) antibody– and rheumatoid factor (RF)–positive and –negative rheumatoid arthritis (RA).

Methods

Data on ∼5,000 RA patients and ∼3,700 healthy controls recruited from 6 centers in the UK were analyzed; not all centers had both genotype data and smoking data available for study. The magnitude of association was assessed in autoantibody-positive and -negative subgroups. The effect of smoking on antibody status among cases was assessed following adjustment for year of birth and center, using Mantel-Haenszel analysis. Analyses of the combined effects of PTPN22, HLA–DRB1 SE, and smoking were performed using additive and multiplicative models of interaction within a logistic regression framework.

Results

The combined effects of PTPN22, HLA–DRB1 SE, and smoking were defined, with no evidence of departure from a multiplicative model. Within the case population, all 3 factors were independently associated with the generation of CCP antibodies (odds ratio [OR] 11.1, P < 0.0001), whereas only HLA–DRB1 SE and smoking were independently associated with RF production (OR 4.4, P < 0.0001). There was some evidence of increasing likelihood of antibody positivity with heavier smoking. Finally, we demonstrated that smoking was associated with the generation of both CCP and RF antibodies (OR 1.7, P = 0.0001).

Conclusion

PTPN22 appears to be primarily associated with anticitrulline autoimmunity, whereas HLA–DRB1 SE is independently associated with RF. This study has confirmed associations of specific gene–environment combinations with a substantially increased risk of developing RA. Further work is needed to determine how these data can be used to inform clinical practice.

Rheumatoid arthritis (RA [MIM no. 180300]) is a phenotypically heterogeneous, chronic destructive inflammatory disease of the synovial joints, with an estimated prevalence of 0.8% in the UK (1). A strong genetic component has been determined, with heritability estimates of 50–60% from twin studies, and up to an additional 50% contribution from environmental and/or physiologic risk factors (2). Approximately 40% of genetic susceptibility to RA is accounted for by HLA–DRB1, the major RA susceptibility locus (3), together with the PTPN22 gene, a second susceptibility gene confirmed in populations of Northern European ancestry, with an attributable risk fraction of ∼8% (4). A genome-wide association study of 1,860 RA cases and 2,938 healthy controls recently carried out by the Wellcome Trust Case Control Consortium (WTCCC) to identify variants underlying 7 common complex diseases, including RA, confirmed association at genome-wide significance levels (P < 10−7) with single-nucleotide polymorphisms in both the HLA–DRB1 region and PTPN22 (5). These included the known RA susceptibility variant that maps to the PTPN22 gene at 1q13. To date, cigarette smoking has been the best-characterized environmental trigger that appears to be associated with the development of RA (6, 7), perhaps more commonly in men (8).

RA is characterized by the presence of autoantibodies (rheumatoid factor [RF] and antibodies to cyclic citrullinated peptide [CCP]) in a significant majority of patients. Cigarette smoking has been recognized to increase the likelihood of RF positivity (6, 9–12) and to affect RA disease severity (11, 13). These effects appear to be dependent on the intensity and the duration of smoking (6, 7, 10, 14, 15).

A putative gene–environment interaction between smoking and the HLA–DRB1 alleles comprising the shared epitope (SE) in susceptibility to RF-positive RA was first reported in 2004 (16). More recent studies have demonstrated that the RA susceptibility factors HLA–DRB1 SE, PTPN22, and smoking have a stronger effect in the subgroup of RA patients who are positive for CCP autoantibodies (17–22). Indeed, there is now evidence to suggest that cigarette smoking contributes to citrullination of proteins in target tissues, such as the lungs, and to the triggering of anticitrulline autoimmunity (17), which appears to be restricted to specific HLA–DRB1 SE alleles (17, 20, 21). The proteins encoded by these HLA–DRB1 SE alleles have additionally been shown to bind citrullinated peptides with a higher affinity, resulting in an enhanced T helper cell response (23). Furthermore, PTPN22 encodes lymphoid tyrosine phosphatase, a protein with a potential function in the regulation of B and T cell activation thresholds (24–26). It has been proposed that carriage of the PTPN22 susceptibility variant leads to a failure to delete autoreactive T cells, thereby predisposing to autoimmunity in general (27). Variation in either gene could therefore predispose to RA, by creating a permissive environment for an up-regulated immune response to an endogenous or environmental autoantigen, such as the induction of anticitrulline immunity following exposure to cigarette smoke.

It remains unclear whether these putative gene–environment interactions also contribute to RA susceptibility in autoantibody (CCP and/or RF)–negative subjects, due to the high prevalence of these antibodies in RA. To date, it has also been difficult to confirm from published genetic studies whether the primary association is with RF or CCP antibodies. Previous attempts at stratification analysis, in subjects discordant for RF and CCP, supported the notion of a primary association between HLA–DRB1 SE alleles and CCP, rather than RF, autoantibodies (17, 20). Although the numbers studied have been relatively low, it appears that the interaction between HLA–DRB1 SE and smoking results primarily in anticitrulline autoimmunity (20). These data are supported by the results of 2 studies in which serum was collected prior to the development of RA; in those studies, circulating CCP autoantibodies were observed before RF production (28, 29). This provides biologic evidence to support the suggestion that RF arises secondary to the development of CCP antibodies, and perhaps secondary to immune complex deposition in tissues.

In the current study we have extended the work of Källberg et al (19) in defining the gene–environment interactions in autoantibody (CCP and RF)–positive and -negative RA in a large UK Caucasian population. This cohort is now large enough to provide sufficient power to investigate whether the primary association lies with RF or CCP autoantibody production.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX A
  10. APPENDIX B

The study sample consisted of a large UK Caucasian RA population from the WTCCC replication studies undertaken by the UK RA Genetics Consortium. We examined evidence for interaction between the previously reported RA susceptibility genes (HLA–DRB1 and PTPN22) and smoking with respect to RA susceptibility, particularly in subgroups classified according to autoantibody production.

Cases and controls.

Patients with RA who were age 18 years or older at disease onset (n = 5,020) were recruited from 6 centers across the UK (Manchester, Sheffield, Leeds, Aberdeen, Oxford, and London) (30). All cases were Caucasians of Northern European descent and fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) classification criteria for RA (31, 32). Healthy controls (n = 3,759) were recruited from 5 of the 6 centers (cases only recruited from London), and an additional 357 controls from the 1958 Birth Cohort (5) were included in the analyses. All participants were recruited after providing informed consent, and the study was approved by the North West Research Ethics Committee (MREC 99/8/84).

Smoking.

Smoking data were available on a subgroup of 2,703 RA cases (Leeds, London, Manchester, and Sheffield) and 1,396 controls (Leeds and Sheffield). Individuals were divided into “ever smokers” and “never smokers” according to whether they reported ever having smoked cigarettes. The control cohorts with smoking data (Leeds and Sheffield) differed in method of recruitment and in age. Those from Leeds were general population controls, free of known cardiovascular and autoimmune disease (33), with a median age of 76 years; those from Sheffield were all healthy, 90% were <65 years of age and employed full-time at the time of recruitment, and the median age was 44 years (34). A subset of 1,361 cases (750 smokers and 611 nonsmokers) from Sheffield and Leeds had reported lifetime smoking in terms of pack-years, and these cases were included in further analyses.

Immunoassays.

The majority of RA cases were recruited from National Health Service rheumatology clinics throughout the UK, and IgM-RF was measured using standard nephelometric assays. The presence of IgG CCP antibodies was documented at a single time point in a proportion of the patients (n = 2,570) using the commercially available DIASTAT anti-CCP enzyme-linked immunosorbent assay (Axis-Shield Diagnostics, Cambs, UK), as previously reported (30, 34). Patients who had ever had an RF titer of ≥40 units/μl were considered positive for RF, and those who had ever had an anti-CCP titer of ≥5.5 units/μl were considered positive for CCP antibodies.

Genotyping.

HLA–DRB1*01–16 types were determined at each center, using commercially available semiautomated polymerase chain reaction–sequence-specific oligonucleotide probe (PCR-SSOP) typing techniques. In a subgroup of RA patients, HLA–DRB1 typing was undertaken using a research assay based on PCR-SSOP linear array technology, developed by Roche Molecular Systems (Pleasanton, CA). PCR products amplified with biotinylated primers were denatured and hybridized to an immobilized probe array. Labeled PCR products hybridized to specific probes were detected using streptavidin–horseradish peroxidase and a chromogenic substrate. For each sample, the probe binding patterns were scanned and the HLA–DRB1 genotype assigned using in-house software (StripScan). The HLA–DRB1 SE alleles were defined as HLA–DRB1*0101, *0102, *0401, *0404, *0405, *0408, and *1001. PTPN22 genotyping was performed as previously described (5, 30). When HLA–DRB1 genotyping data were not available at sufficient resolution to classify an allele as SE or not (e.g., HLA–DRB1*01), the most likely assignment was made when this could be done with at least 85% certainty based on population allele frequencies; otherwise, the data were treated as missing. A total of 282 alleles were incompletely resolved but likely to be HLA–DRB1 SE alleles with probabilities ranging from 0.85 to 0.93, and 6 alleles were coded as missing. We estimate that this will have led to an expected incorrect classification of 36 alleles, which is 0.35% of all alleles, and would therefore be unlikely to have had a significant effect on the analysis.

Statistical analysis.

Genotype distributions for PTPN22 and HLA–DRB1 SE alleles were compared between cases and controls using chi-square tests with 2 × 3 contingency tables, classifying subjects by the number of risk alleles carried. Estimated odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated using logistic regression, with those with no copies of the risk alleles used as the referent group. The analysis was repeated in subsets of cases classified based on the presence or absence of CCP and RF antibodies.

Analysis of the combined effect of PTPN22 and HLA–DRB1 SE alleles was undertaken. To aid in interpretation of these interaction analyses, a binary classification of genotypes, carriage versus noncarriage of the risk allele (i.e., a dominant model), was considered. Two different models of interaction, which were comparable with previous interaction analyses of these susceptibility factors in RA (19), were analyzed. First, departure from additivity was assessed (35). The relative excess risk due to interaction (RERI) was used to quantify the risk that was due to the interaction per se and beyond the independent effects of the 2 risk factors added together, with CIs calculated using variance estimates obtained from Taylor series expansion (36). In the absence of interaction, the RERI is 0. When significant evidence of interaction was found, the attributable proportion of risk due to interaction (i.e., the ratio of RERI to the overall relative risk from carrying both risk factors) was also calculated. Second, departure from a multiplicative combined effect on risk was tested using logistic regression, comparing models with and without an interaction term using the likelihood ratio test.

Because of differences between the control groups from the 2 centers at which smoking data were collected, the analysis of smoking was restricted to comparisons between the proportions of subjects who had ever smoked, or the level of smoking in pack-years, in different subsets of cases. The effect of smoking on antibody status among cases was assessed based on calculated ORs pooled across centers, adjusting for period of birth (essentially equivalent to age in the year 2000) in order to allow for secular trends in smoking prevalence, using Mantel-Haenszel analysis. The combined effect of smoking and each gene on antibody status was estimated from logistic regression analysis comparing, e.g., CCP antibody–positive versus CCP antibody–negative cases. The effect of pack-years of smoking was assessed by logistic regression of case type on a categorical variable with 4 levels: nonsmoker, light smoker (up to 10 pack-years), moderate smoker (>10–30 pack-years), and heavy smoker (>30 pack-years). All analyses were conducted using Stata, version 10 (StataCorp, College Station, TX).

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX A
  10. APPENDIX B

Characteristics of the RA patients and controls.

The RA population studied was comparable with other hospital-based series of RA subjects. Overall, 72% of the cases were female and, in those for whom data were available, 72% were seropositive for IgM-RF, 67% had CCP antibodies, 70% had erosive disease, and 34% had documented nodules. More detailed demographic and clinical data on the RA cases and demographic data on the controls are available online at http://limm.leeds.ac.uk/research_sections/musculoskeletal_disease/groups/morgan.htm. Center-specific data have been reported previously (30).

Contribution of PTPN22 and HLA–DRB1 SE alleles to RA.

The frequencies of the PTPN22 and the HLA–DRB1 SE alleles and the magnitude of association with RA, including stratification according to autoantibody status, are shown in Tables 1 and 2 (center-specific data available online at http://limm.leeds.ac.uk/research_sections/musculoskeletal_disease/groups/morgan.htm). The allele distributions of both genes conformed to Hardy-Weinberg equilibrium.

Table 1. Frequencies of HLA–DRB1 SE and PTPN22 genotypes in RA cases in the total cohort (n = 4,789), in CCP antibody–positive and –negative RA cases, and in RF-positive and -negative RA cases, relative to controls*
 Cases, no. (%)Controls, no. (%)OR (95% CI)P
  • *

    The control cohort consisted of 3,630 subjects. All 3,630 were included in the PTPN22 analyses and 1,357 in the HLA–DRB1 shared epitope (SE) analyses; n values for rheumatoid arthritis (RA) patients in each subanalysis are shown in parentheses (data for some analyses not available from all study subjects). CCP = cyclic citrullinated peptide; RF = rheumatoid factor; OR = odds ratio; 95% CI = 95% confidence interval.

  • From overall chi-square test for association, with 2 df.

Total RA cohort    
 PTPN22    
  GA1,208 (25.2)662 (18.2)1.5 (1.4, 1.7)<10−17
  AA116 (2.4)43 (1.2)2.3 (1.6, 3.2) 
 HLA–DRB1 SE (n = 3,652)    
  1 copy1,788 (49.0)530 (39.1)2.7 (2.4, 3.1)<10−91
  2 copies994 (27.2)121 (8.9)6.7 (5.4, 8.3) 
CCP+ RA    
 PTPN22 (n = 1,634)    
  GA445 (27.2)662 (18.2)1.7 (1.5, 2.0)<10−19
  AA54 (3.3)43 (1.2)3.2 (2.2, 4.9) 
 HLA–DRB1 SE (n = 1,473)    
  1 copy759 (51.5)530 (39.1)4.3 (3.6, 5.2)<10−106
  2 copies480 (32.6)121 (8.9)12.0 (9.3, 15.3) 
CCP− RA    
 PTPN22 (n = 810)    
  GA181 (22.4)662 (18.2)1.3 (1.1, 1.6)0.02
  AA7 (0.9)43 (1.2)0.8 (0.3, 1.7) 
 HLA–DRB1 SE (n = 724)    
  1 copy325 (44.9)530 (39.1)1.4 (1.2, 1.7)0.00006
  2 copies92 (12.7)121 (8.9)1.7 (1.3, 2.4) 
RF+ RA    
 PTPN22 (n = 3,223)    
  GA821 (25.5)662 (18.2)1.6 (1.4, 1.8)<10−15
  AA79 (2.5)43 (1.2)2.3 (1.6, 3.4) 
 HLA–DRB1 SE (n = 2,512)    
  1 copy1,263 (50.3)530 (39.1)3.5 (3.0, 4.0)<10−109
  2 copies762 (30.3)121 (8.9)9.1 (7.3, 11.4) 
RF− RA    
 PTPN22 (n = 1,249)    
  GA290 (23.2)662 (18.2)1.4 (1.2, 1.6)0.00002
  AA27 (2.2)43 (1.2)2.0 (1.2, 3.2) 
 HLA–DRB1 SE (n = 951)    
  1 copy444 (46.7)530 (39.1)1.8 (1.5, 2.2)<10−19
  2 copies182 (19.1)121 (8.9)3.3 (2.5, 4.3) 
Table 2. Frequencies of HLA–DRB1 SE and PTPN22 genotypes in subsets of RA cases categorized by CCP and RF antibody status, relative to controls*
 Cases, no. (%)OR (95% CI)P
  • *

    The control cohort consisted of 3,630 subjects. All 3,630 were included in the PTPN22 analyses and 1,357 in the HLA–DRB1 SE analyses; n values for RA patients in each subanalysis are shown in parentheses (data for some analyses not available from all study subjects). The PTPN22 GA and AA genotypes were present in 662 (18.2%) and 43 (1.2%) of controls, respectively. Five hundred thirty (39.1%) of the controls had 1 copy of the HLA–DRB1 SE, and 121 (8.9%) had 2 copies of the SE. See Table 1 for definitions.

  • From overall chi-square test for association, with 2 df.

PTPN22   
 CCP+/RF+ (n = 1,346)   
  GA366 (27.2)1.7 (1.5, 2.0)<10−16
  AA42 (3.1)3.0 (2.0, 4.7) 
 CCP+/RF− (n = 201)   
  GA51 (25.4)1.6 (1.1, 2.2)0.003
  AA6 (3.0)2.8 (1.2, 6.8) 
 CCP−/RF+ (n = 288)   
  GA52 (18.1)1.0 (0.7, 1.3)0.4
  AA1 (0.4)0.3 (0.0, 2.1) 
 CCP−/RF− (n = 448)   
  GA102 (22.8)1.3 (1.0, 1.7)0.07
  AA5 (1.1)1.0 (0.4, 2.5) 
HLA–DRB1 SE   
 CCP+/RF+ (n = 1,218)   
  1 copy635 (52.1)4.4 (3.6, 5.4)<10−95
  2 copies391 (32.1)11.9 (9.2, 15.4) 
 CCP+/RF− (n = 176)   
  1 copy87 (49.4)4.3 (2.7, 6.7)<10−30
  2 copies62 (35.2)13.4 (8.2, 21.9) 
 CCP−/RF+ (n = 272)   
  1 copy124 (45.6)1.6 (1.2, 2.2)<10−5
  2 copies46 (16.9)2.6 (1.8, 3.9) 
 CCP−/RF− (n = 391)   
  1 copy177 (45.3)1.3 (1.1, 1.7)0.05
  2 copies38 (9.7)1.3 (0.8, 1.9) 

PTPN22 was associated with RA (OR 1.5 for GA versus GG and 2.3 for AA versus GG, P < 10−17), and particularly with CCP-positive RA (OR 1.7 and 3.2, respectively, P < 10−19). Within this subgroup, the magnitude of risk did not vary substantially according to RF status (GA and AA ORs 1.7 and 3.0, respectively, for CCP-positive/RF-positive RA, versus 1.6 and 2.8 for CCP-positive/RF-negative RA) (Table 2). The association was much weaker in the CCP-negative subgroup (OR 1.3 and 0.8, respectively, P = 0.02). The association with PTPN22 was not as marked in relation to RF status (OR 1.6 and 2.3, respectively, for GA and AA in the RF-positive subgroup and 1.4 and 2.0, respectively, in the RF-negative subgroup) (Table 1). Interestingly, there was no evidence of an effect on risk of CCP-negative/RF-positive disease (OR 1.0 and 0.3, respectively, P = 0.4), suggesting that the primary effect is with CCP antibody generation. Although the frequency of homozygosity for the PTPN22 1858A allele was low both in cases and in controls, the magnitude of association in individuals was increased among those with 2 copies of the risk allele compared with those with 1 copy, in all subgroups except the CCP-negative group (Table 1). The difference between heterozygotes and homozygotes for the risk allele was statistically significant for all RA (P = 0.03), CCP-positive RA (P = 0.003), and RF-positive RA (P = 0.04), but not for RF-negative RA (P = 0.16).

In the study cohort, the HLA–DRB1 SE alleles were associated with RA (OR 2.7 for 1 SE allele and 6.7 for 2 SE alleles, P < 10−91). Again, the effect was much stronger for CCP-positive RA (OR 4.3 and 12.0 for 1 allele and 2 alleles, respectively, P < 10−106) compared with CCP-negative RA (OR 1.4 and 1.7, respectively, P = 0.00006) (Table 1). There was also a stronger effect on risk of RF-positive RA (OR 3.5 and 9.1, P < 10−109) than of RF-negative RA (OR 1.8 and 3.3, P < 10−19), but the difference was less marked than for CCP. Stratification by both antibodies (Table 2) revealed that SE genotype influences risk of all subtypes of RA, but with a strong effect only on CCP-positive RA, in which the magnitude of effect is similar irrespective of RF antibody status.

Combined analysis of PTPN22 and HLA–DRB1 SE alleles in susceptibility to RA.

Logistic regression analyses confirmed that under a dominant model, PTPN22 and the HLA–DRB1 SE alleles were independently associated with RA, including the CCP-positive, RF-positive, and RF-negative subgroups (data not shown). PTPN22 was not significantly associated with CCP-negative RA when adjusted for HLA–DRB1 SE alleles in a logistic regression framework (OR 1.20 [95% CI 0.96, 1.50] for presence versus absence of the PTPN22 A allele). The ORs for RA development with different combinations of PTPN22 and HLA–DRB1 SE alleles, including stratification by autoantibody status, are shown in Table 3. There was some evidence that the combined effect of presence of both types of risk allele was more than additive, for RA overall (RERI 2.0, P = 0.01) and within the CCP-positive RA subgroup (RERI 4.6, P = 0.02), the RF-positive RA subgroup (RERI 2.5, P = 0.03), and the RF-negative RA subgroup (RERI 1.2, P = 0.04); this was not evident for CCP-negative RA (RERI 0.5, P = 0.19). Where there was significant evidence of interaction, the estimated attributable proportion was similar in each group, ranging from 0.34 for RF-positive RA to 0.40 for CCP-positive RA. There was no evidence of departure from a multiplicative model for any RA subgroup, supporting the notion of a multiplicative combined effect of PTPN22 and HLA–DRB1 SE alleles on the risk of RA.

Table 3. ORs for developing RA according to the presence or absence of PTPN22 and HLA-DRB1 susceptibility alleles, by CCP and RF antibody status, relative to controls*
Antibody status, PTPN22/HLA–DRB1 SECases, no. (%)Controls, no. (%)OR (95% CI)RERI (95% CI)P
Deviation from additivityDeviation from multiplicity
  • *

    The control cohort for these analyses consisted of 1,334 subjects; n values for RA patients in each subanalysis are shown in parentheses. See Table 1 for other definitions.

  • † Relative excess risk due to interaction (RERI) = relative risk (AE) − relative risk (ĀE) − relative risk (AĒ) + 1, where A and E denote the presence and Ā and Ē the absence of 2 risk factors. This was used to quantify the risk that was due to the interaction per se and beyond the independent effects of the 2 risk factors added together. Departure from a multiplicative joint effect on risk was tested using logistic regression, comparing models with and without the interaction term using the likelihood ratio test.

Total RA (n = 3,548)      
 No PTPN22/No HLA–DRB1 SE619 (17.4)547 (41.0)1.0 (referent)2.0 (0.4, 3.5)0.010.15
 No PTPN22/Any HLA–DRB1 SE1,952 (55.0)522 (39.1)3.3 (2.8, 3.8)   
 Any PTPN22/No HLA–DRB1 SE224 (6.3)147 (11.0)1.3 (1.1, 1.7)   
 Any PTPN22/Any HLA–DRB1 SE753 (21.2)118 (8.9)5.6 (4.5, 7.1)   
CCP+RA (n = 1,430)      
 No PTPN22/No HLA–DRB1 SE149 (10.4)547 (41.0)1.0 (referent)4.6 (0.6, 8.5)0.020.94
 No PTPN22/Any HLA–DRB1 SE839 (58.7)522 (39.1)5.9 (4.8, 7.3)   
 Any PTPN22/No HLA–DRB1 SE76 (5.3)147 (11.0)1.9 (1.4, 2.6)   
 Any PTPN22/Any HLA–DRB1 SE366 (25.6)118 (8.9)11.4 (8.6, 15.0)   
CCP− RA (n = 697)      
 No PTPN22/No HLA–DRB1 SE233 (33.4)547 (41.0)1.0 (referent)0.5 (−0.3, 1.3)0.190.16
 No PTPN22/Any HLA–DRB1 SE305 (43.8)522 (39.1)1.4 (1.1, 1.7)   
 Any PTPN22/No HLA–DRB1 SE63 (9.0)147 (11.0)1.0 (0.7, 1.4)   
 Any PTPN22/Any HLA–DRB1 SE96 (13.8)118 (8.9)1.9 (1.4, 2.6)   
RF+ RA (n = 2,440)      
 No PTPN22/No HLA–DRB1 SE341 (14.0)547 (41.0)1.0 (referent)2.5 (0.3, 4.7)0.030.45
 No PTPN22/Any HLA–DRB1 SE1,426 (58.4)522 (39.1)4.4 (3.7, 5.2)   
 Any PTPN22/No HLA–DRB1 SE134 (5.5)147 (11.0)1.5 (1.1, 1.9)   
 Any PTPN22/Any HLA–DRB1 SE539 (22.1)118 (8.9)7.3 (5.8, 9.3)   
RF− RA (n = 921)      
 No PTPN22/No HLA–DRB1 SE241 (26.2)547 (41.0)1.0 (referent)1.2 (0.0, 2.3)0.040.05
 No PTPN22/Any HLA–DRB1 SE445 (48.3)522 (39.1)1.9 (1.6, 2.4)   
 Any PTPN22/No HLA–DRB1 SE70 (7.6)147 (11.0)1.1 (0.8, 1.5)   
 Any PTPN22/Any HLA–DRB1 SE165 (17.9)118 (8.9)3.2 (2.4, 4.2)   

Contribution of smoking to RA susceptibility.

There was a statistically significant difference between centers in the incidence of smoking among the RA cases (Manchester 62%, Leeds 50%, Sheffield 59%, London 67%; P < 0.001) and controls (Leeds 48%, Sheffield 41%; P = 0.04). Smoking status differed significantly by sex (women 49%, men 65%; P < 0.001) and age in the year 2000 (period of birth: age <40 years 40%, 40–49 years 49%, 50–89 years 60%, >90 years 36%; P < 0.001). Overall, after adjustment for age and sex, smoking was associated with increased risk of RA (OR 1.9 [95% CI 1.6, 2.2], P < 0.001), particularly CCP-positive RA (OR 2.1 [95% CI 1.8, 2.5], P < 0.001) and RF-positive RA (OR 2.2 [95% CI 1.9, 2.6], P < 0.001), although association with CCP-negative RA and RF-negative RA was also demonstrated (OR 1.34 [95% CI 1.1, 1.6], P = 0.004 and OR 1.5 [95% CI 1.2, 1.8], P < 0.001). It was not possible to achieve homogeneity of ORs, after adjustment for age and sex, between the 2 centers from which both cases and controls were analyzed with regard to smoking status after adjustment (P = 0.008), probably reflecting differences in the method of recruitment of controls at each of these centers. These risk estimates should thus be interpreted with caution.

Subsequent analyses, therefore, focused on a case–case analysis of smoking, which was considered to be more robust since smoking data were collected similarly between different subgroups of cases within a center. After adjustment, it was found that sex was not associated with autoantibody status in cases, removing this as a potential confounder. Homogeneity of ORs was observed following adjustment for age (at year 2000) and center (P for homogeneity = 0.41 for RF-positive versus RF-negative RA; P = 0.17 for CCP-positive versus CCP-negative RA). This enabled us to include in the analyses data from all 4 centers from which smoking data had been obtained. Within the RA population, smoking was strongly predictive of the presence of CCP antibodies (Mantel-Haenszel–estimated OR 1.7 [95% CI 1.4, 2.1], P < 0.001) and of RF (Mantel-Haenszel–estimated OR 1.5 [95% CI 1.2, 1.8], P < 0.001) (Table 4); center-specific data available online at http://limm.leeds.ac.uk/research_sections/musculoskeletal_disease/groups/morgan.htm. In an analysis of pack-years of smoking, there was some evidence of increasing likelihood of antibody positivity among patients with heavier smoking. The ORs for CCP antibody positivity for light, moderate, and heavy smokers versus nonsmokers were 1.8, 2.0, and 2.7, respectively, and for RF positivity the corresponding ORs were 1.6, 1.8, and 2.1.

Table 4. ORs for developing CCP-positive compared with CCP-negative RA and RF-positive compared with RF-negative RA, according to the presence or absence of PTPN22 alleles, HLA–DRB1 SE alleles, and history of smoking*
Antibody and allele analyzed in relation to smokingAntibody positive, no. (%)Antibody negative, no. (%)OR (95% CI)RERI (95% CI)
  • *

    See Table 1 for other definitions.

  • OR from Mantel-Haenszel analysis stratified by center and adjusted for period of birth; in test for homogeneity of ORs, P values for CCP and RF were not significant.

  • Relative excess risk due to interaction (RERI) = relative risk (AE) − relative risk (ĀE) − relative risk (AĒ) + 1, where A and E denote the presence and Ā and Ē the absence of 2 risk factors. This was used to quantify the risk that was due to the interaction per se and beyond the independent effects of the 2 risk factors added together. Departure from a multiplicative joint effect on risk was tested using logistic regression, comparing models with and without the interaction term using the likelihood ratio test. The P value for the deviation from additivity or multiplicity was not significant in any of these analyses.

CCP (n = 1,447 positive, 670 negative)    
 Ever smoked894 (61.8)348 (51.9)1.7 (1.3, 2.1) 
RF (n = 1,777 positive, 712 negative)    
 Ever smoked1,111 (62.5)388 (54.5)1.5 (1.2, 1.8) 
CCP (n = 1,382 positive, 634 negative) and PTPN22    
 No PTPN22/Never smoked355 (25.7)231 (36.4)1.0 (referent)−0.1 (−1.0, 0.8)
 No PTPN22/Ever smoked599 (43.3)251 (39.6)1.6 (1.2, 1.9) 
 Any PTPN22/Never smoked168 (12.2)69 (10.9)1.6 (1.1, 2.2) 
 Any PTPN22/Ever smoked260 (18.8)83 (13.1)2.0 (1.5, 2.7) 
RF (n = 1,694 positive, 669 negative) and PTPN22    
 No PTPN22/Never smoked448 (26.5)226 (33.8)1.0 (referent)−0.2 (−0.9, 0.5)
 No PTPN22/Ever smoked761 (44.9)264 (39.5)1.5 (1.2, 1.8) 
 Any PTPN22/Never smoked185 (10.9)77 (11.5)1.2 (0.9, 1.7) 
 Any PTPN22/Ever smoked300 (17.7)102 (15.3)1.5 (1.1, 2.0) 
CCP (n = 1,347 positive, 625 negative) and HLA–DRB1 SE    
 No HLA–DRB1 SE/Never smoked80 (5.9)122 (19.5)1.0 (referent)1.7 (−0.3, 3.7)
 No HLA–DRB1 SE/Ever smoked133 (9.9)147 (23.5)1.4 (1.0, 2.0) 
 Any HLA–DRB1 SE/Never smoked417 (31.0)169 (27.0)3.8 (2.7, 5.3) 
 Any HLA–DRB1 SE/Ever smoked717 (53.2)187 (29.9)5.8 (4.2, 8.1) 
RF (n = 1,662 positive, 657 negative) and HLA–DRB1 SE    
 No HLA–DRB1 SE/Never smoked112 (6.7)100 (15.2)1.0 (referent)0.7 (−0.6, 2.0)
 No HLA–DRB1 SE/Ever smoked205 (12.3)136 (20.7)1.3 (1.0, 1.9) 
 Any HLA–DRB1 SE/Never smoked488 (29.4)191 (29.1)2.3 (1.7, 3.1) 
 Any HLA–DRB1 SE/Ever smoked857 (51.6)230 (35.0)3.3 (2.4, 4.5) 

Combined analysis of PTPN22, HLA–DRB1 SE alleles, and smoking in susceptibility to RA.

Logistic regression analyses within the RA case population confirmed that under a dominant model, PTPN22, HLA–DRB1 SE alleles, and smoking were independently associated with the development of CCP autoantibodies, but that only HLA–DRB1 SE alleles and smoking were independently associated with RF (data not shown). The Mantel-Haenszel OR estimates (dominant model, stratified by center) for CCP-positive RA were as follows: for PTPN22 (adjusted for HLA–DRB1 SE alleles and smoking status) OR 1.5 (95% CI 1.2, 1.8), P = 0.002; for HLA–DRB1 SE (adjusted for PTPN22 and smoking status) OR 3.9 (95% CI 3.1, 4.9), P < 0.0001; and for smoking (adjusted for PTPN22 and HLA–DRB1 SE alleles) OR 1.5 (95% CI 1.2, 1.9), P = 0.0001. The corresponding values for RF-positive RA were as follows: for PTPN22 OR 1.1 (95% CI 0.9, 1.3), P = 0.51; for HLA–DRB1 SE alleles OR 2.3 (95% CI 1.9, 2.9), P < 0.0001; and for smoking OR 1.5 (95% CI 1.2, 1.8), P = 0.0001.

The ORs for developing RA with different combinations of PTPN22 or HLA–DRB1 SE alleles and smoking status are shown in Table 4. The combined effect of PTPN22 or HLA–DRB1 SE alleles with smoking, within the population of cases, was consistent with both multiplicative and additive models for the development of either CCP or RF autoantibodies. There was some suggestion of a greater than additive effect of smoking and HLA–DRB1 SE alleles on CCP positivity status (RERI 1.7 [95% CI −0.3, 3.7]), but this was not statistically significant.

Table 5 shows estimates of the effect of various combinations of the PTPN22, HLA–DRB1 SE allele, and smoking risk factors on the likelihood of antibody positivity among cases. Smokers with 2 HLA–DRB1 SE alleles were more likely to be positive than negative for CCP, irrespective of their PTPN22 status (OR 13.3 with no PTPN22 risk allele and 11.1 with at least 1 PTPN22 risk allele, P < 0.0001 for both). Similarly this combination was strongly predictive (although less so) of RF positivity irrespective of PTPN22 genotype (OR 4.4 and 4.2, respectively, P < 0.0001 for both).

Table 5. ORs for developing CCP-positive compared with CCP-negative RA and RF-positive compared with RF-negative RA, according to combinations of all 3 risk factors studied*
Presence or absence of allele(s) and smoking history, by antibody statusAntibody positive/negative, no.OR (95% CI)
  • *

    See Table 1 for definitions.

  • OR from Mantel-Haenszel analysis stratified by center and adjusted for period of birth.

CCP-positive vs. CCP-negative  
 No PTPN22/No HLA–DRB1 SE/Never smoked54/931.0 (referent)
 Any PTPN22/No HLA–DRB1 SE/Never smoked22/221.7 (0.9, 3.4)
 No PTPN22/Single HLA–DRB1 SE/Never smoked166/893.2 (2.1, 4.9)
 Any PTPN22/Single HLA–DRB1 SE/Never smoked76/314.2 (2.5, 7.2)
 No PTPN22/Double HLA–DRB1 SE/Never smoked105/315.8 (3.5, 9.8)
 Any PTPN22/Double HLA–DRB1 SE/Never smoked60/119.4 (4.5, 19.4)
 No PTPN22/No HLA–DRB1 SE/Ever smoked83/1081.3 (0.9, 2.1)
 Any PTPN22/No HLA–DRB1 SE/Ever smoked47/352.3 (1.3, 4.0)
 No PTPN22/Single HLA–DRB1 SE/Ever smoked305/1134.6 (3.1, 6.9)
 Any PTPN22/Single HLA–DRB1 SE/Ever smoked128/326.9 (4.1, 11.5)
 No PTPN22/Double HLA–DRB1 SE/Ever smoked185/2413.3 (7.7, 22.8)
 Any PTPN22/Double HLA–DRB1 SE/Ever smoked77/1211.1 (5.5, 22.1)
RF-positive vs. RF-negative  
 No PTPN22/No HLA–DRB1 SE/Never smoked83/731.0 (referent)
 Any PTPN22/No HLA–DRB1 SE/Never smoked26/181.3 (0.6, 2.5)
 No PTPN22/Single HLA–DRB1 SE/Never smoked209/971.9 (1.3, 2.8)
 Any PTPN22/Single HLA–DRB1 SE/Never smoked86/342.2 (1.3, 3.7)
 No PTPN22/Double HLA–DRB1 SE/Never smoked120/382.8 (1.7, 4.5)
 Any PTPN22/Double HLA–DRB1 SE/Never smoked56/192.6 (1.4, 4.8)
 No PTPN22/No HLA–DRB1 SE/Ever smoked140/961.3 (0.9, 1.9)
 Any PTPN22/No HLA–DRB1 SE/Ever smoked60/351.5 (0.9, 2.5)
 No PTPN22/Single HLA–DRB1 SE/Ever smoked379/1142.9 (2.0, 4.3)
 Any PTPN22/Single HLA–DRB1 SE/Ever smoked146/442.9 (1.8, 4.6)
 No PTPN22/Double HLA–DRB1 SE/Ever smoked213/434.4 (2.8, 6.9)
 Any PTPN22/Double HLA–DRB1 SE/Ever smoked85/184.2 (2.3, 7.6)

Exploration of whether the primary association with smoking is with the generation of CCP or RF autoantibodies.

Finally, we used the large number of cases in this cohort to evaluate whether smoking is primarily associated with the generation of CCP or RF autoantibodies. In total 1,968 cases were available for this analysis, of whom 1,190 (60.5%) were positive for both antibodies, 173 (8.8%) were positive for only CCP antibodies, 237 (12.0%) were positive for RF only, and 368 (18.7%) were negative for both CCP and RF (Table 6). Smoking was associated with CCP-positive/RF-positive RA (OR 1.7 [95% CI 1.3, 2.2], P = 0.0001 versus CCP-negative/RF-negative RA), but was not associated with positivity for CCP only or RF only.

Table 6. ORs for developing CCP-negative/RF-positive RA, CCP-positive/RF-negative RA, and CCP-positive/RF-positive RA compared with CCP-negative/RF-negative RA, according to smoking status*
 No. of smokers/ total (%)OR (95% CI)P
  • *

    Of the total of 2,703 patients included in the analysis, 1,607 (59.5%) were ever smokers. See Table 1 for definitions.

  • OR from Mantel-Haenszel analysis stratified by center and adjusted for age (year 2000) or period of birth; in test for homogeneity of ORs, P values were not significant.

CCP-negative/RF-negative195/368 (53.0)1.0 (referent)
CCP-negative/RF-positive128/237 (54.0)1.0 (0.7, 1.5)1.0
CCP-positive/RF-negative92/173 (53.2)1.2 (0.8, 1.9)0.4
CCP-positive/RF-positive756/1,190 (63.5)1.7 (1.3, 2.2)0.0001

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX A
  10. APPENDIX B

The present investigation is the best-powered study reported to date on putative interactions between the 2 major RA susceptibility variants and smoking. The results showed that together, HLA–DRB1 SE alleles, PTPN22, and smoking account for ∼10% of the variance in RA susceptibility, increasing to 17% of variance in CCP-positive RA susceptibility (pseudo-R2 from logistic regression analysis of our data). RA is a clinically and phenotypically heterogeneous disease, and there is now an accumulating body of evidence indicating that there are at least 2 genetically distinct subgroups, characterized by the presence or absence of CCP autoantibodies (37), which may in part be determined by exposure to cigarette smoke (17, 38). With the large sample size, robust estimates of the effect sizes could be inferred, although further larger and longitudinal studies, and potentially meta-analyses, will be needed to derive similarly robust data on the CCP-negative/RF-negative cohorts.

The study population originated from 6 different centers in the UK, which raises the possibility that the results may have been affected by differences in recruitment, population substructure, and heterogeneity. There was no discernible heterogeneity in the PTPN22 and HLA–DRB1 SE associations when a stratified analysis by center was undertaken (HLA–DRB1 SE typing in controls was available from only 3 of 5 centers), and there was no evidence of a difference in allele frequencies among controls for the 2 loci considered (P = 0.63 for HLA–DRB1 SE alleles, P = 0.71 for PTPN22) (detailed data available online at http://limm.leeds.ac.uk/research_sections/musculoskeletal_disease/groups/morgan.htm).

However, some heterogeneity between centers with respect to the smoking data was observed. None of these studies were designed to primarily address environmental exposures, although similar definitions were used and similar data were collected. Differences were most marked for the control cohorts, and the analyses of smoking presented herein are therefore restricted to comparisons between cases. This highlights the importance of careful study design for assessing gene–environment interactions. In the cases, homogeneity of ORs was observed following adjustment for age and stratification by center. When interpreting the data on smoking, it is important to recognize that the ORs represent the increase in risk associated with the presence of CCP or RF antibodies compared with RA cases who are negative for the respective antibodies. However, the strength of the current study lies in the large number of cases and controls, providing increased statistical power for interaction analyses.

In this study, the magnitude of association of carriage of PTPN22 risk alleles (OR 1.5 for 1 allele and 2.3 for 2) and HLA–DRB1 SE (OR 2.7 for 1 SE allele and 6.7 for 2) with RA was comparable with findings previously described in the literature (19). Carriage of both risk alleles was significantly associated with autoantibody (CCP and RF)–positive and -negative RA, albeit at a lower magnitude for subjects without antibodies (Table 3). The presence of 2 copies of the disease-associated allele was consistently associated with greater risk of RA, overall and in subsequent subgroup analyses. The only exception was for PTPN22 and CCP-negative RA, but this may reflect the rarity of the PTPN22 AA genotype combined with the relatively small number of CCP-negative cases (Table 1). A combined analysis of PTPN22 and HLA–DRB1 SE alleles revealed that, with the exception of CCP-negative RA, both genes were independently associated with RA and the different autoantibody subgroups.

There are different definitions of interaction, and, using a method similar to that described by Källberg et al (19), we considered 2 distinct definitions: departure from an additive effect and departure from a multiplicative effect. We found evidence of statistically significant departure from the additive model, but no evidence of departure from a multiplicative model for any RA subgroup. The observed combined effect of HLA–DRB1 SE alleles and PTPN22 was greater than would be expected under an additive model in all RA cohorts studied, with the exception of the CCP-negative subgroup. The magnitude of this increase in risk was quantified as the RERI (Table 3). Biologic interaction between 2 factors can be defined as participation in the same sufficient cause (i.e., pathway to disease) or in terms of counterfactuals (i.e., the effect of one factor depends on the value of the other factor). The relationship between biologic interaction and statistical interaction is complex, but it has been argued that departure from additivity implies the presence of some type of biologic interaction, and that a multiplicative model (homogeneity of risk ratios) is consistent with biologic interaction (38). These data thus confirm the previously reported presence of interaction between HLA–DRB1 SE alleles and PTPN22 in CCP-positive RA (19), although we were unable to substantiate the previously reported departure from a multiplicative model (19) in this study.

Smoking was independently associated with the presence of both CCP and RF antibodies in cases, and to a comparable magnitude (Table 4). We demonstrated that within the case population, HLA–DRB1 SE alleles, PTPN22, and smoking were independently associated with the development of anticitrulline autoimmunity. Results of interaction analyses were consistent with the notion that smoking acts either additively or multiplicatively with both PTPN22 and HLA–DRB1 SE alleles in the generation of CCP antibodies (Table 4). Although this does not preclude biologic interaction, it provides no direct evidence of it. Previous interaction analyses compared cases and controls and yielded comparable results for PTPN22 and smoking (19). However, the findings of previous case–control analyses were consistent with the notion of a greater than additive effect for HLA–DRB1 SE alleles and smoking (17), which was not observed in the within-case analysis in the current study.

In contrast, only HLA–DRB1 SE and smoking were independently associated with RF production, suggesting that PTPN22 may be associated primarily with the development of CCP, rather than RF, antibodies. Smokers with 2 HLA–DRB1 SE alleles were much more likely to be CCP-positive than CCP-negative, and, to a lesser extent, were more likely to be RF-positive than RF-negative, irrespective of PTPN22 status (Table 5). In subsequent analyses, smoking appeared to be associated with a significantly increased risk of the presence of both antibodies, with some evidence of increasing likelihood of antibody positivity with heavier smoking, as previously described (7, 17, 21). The within-case study design precluded evaluation of the risk of smoking in the subgroup of cases who were negative for both autoantibodies, which was the referent group for these analyses.

These results are from the largest published cohort to date, but further refinements may be possible with further increases in cohort size or meta-analyses, although the latter have the disadvantage of introducing geographic variation in allele and smoking frequencies. The data support the multistep model of RA pathogenesis proposed by Klareskog et al (17) and the clinical observations that CCP antibodies predate RF (28, 29). PTPN22, in particular, appears to be primarily associated with anticitrulline immunity, whereas the HLA–DRB1 SE remains independently associated with RF production in the absence of CCP antibodies. One important factor may be the sensitivity of currently available CCP assays to detect all autoantibodies directed at citrullinated epitopes. The current CCP-2 assays use a mixture of citrullinated peptides; however, it is possible that autoantibodies to alternative citrullinated epitopes are present but not detected with these assays. In addition, many of the RA patients included in this study have had RF measurements at several time points, whereas the CCP antibodies were measured at only a single time point, on stored serum samples.

It remains to be determined how many CCP-negative/RF-positive subjects might become positive for CCP on serial testing, thus further reducing the apparent association between HLA–DRB1 SE alleles and RF production. The stability of CCP antibodies has been investigated in some small cohorts, and the results suggest that seroconversion over time does occur in a small proportion of individuals (39, 40). Further large longitudinal studies are therefore needed to determine whether the apparent association between HLA–DRB1 SE alleles and RF is due to the misclassification of CCP-negative patients due to variability in either the level of antibodies or the assay used.

Importantly, this study has confirmed the association of specific gene–environment combinations with a substantially increased risk of developing RA. The challenge is to determine how these data can be used to inform clinical practice. There is now an exciting opportunity to begin investigating screening programs within the community, and further studies will determine the potential of these variants as prognostic and predictive biomarkers that may guide treatment decisions.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX A
  10. APPENDIX B

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Morgan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Morgan, Steer, Barrett.

Acquisition of data. Morgan, Martin, Carter, Erlich, Barton, Reid, Harrison, Wordsworth, Steer, Worthington, Emery, Wilson.

Analysis and interpretation of data. Morgan, Thomson, Hocking, Barrett.

Acknowledgements

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX A
  10. APPENDIX B

Genotyping data from The Wellcome Trust Case Control Consortium were used in this study.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX A
  10. APPENDIX B
  • 1
    Symmons D, Turner G, Webb R, Asten P, Barrett E, Lunt M, et al. The prevalence of rheumatoid arthritis in the United Kingdom: new estimates for a new century. Rheumatology (Oxford) 2002; 41: 793800.
  • 2
    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: 307.
  • 3
    MacKay K, Eyre S, Myerscough A, Milicic A, Barton A, Laval S, et al. Whole-genome linkage analysis of rheumatoid arthritis susceptibility loci in 252 affected sibling pairs in the United Kingdom. Arthritis Rheum 2002; 46: 6329.
  • 4
    Lee YH, Rho YH, Choi SJ, Ji JD, Song GG, Nath SK, et al. The PTPN22 C1858T functional polymorphism and autoimmune diseases—a meta-analysis. Rheumatology (Oxford) 2007; 46: 4956.
  • 5
    The Wellcome Trust Case Control Consortium. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature 2007; 447: 66178.
  • 6
    Stolt P, Bengtsson C, Nordmark B, Lindblad S, Lundberg I, Klareskog L, et al, and the other members of the EIRA Study Group. Quantification of the influence of cigarette smoking on rheumatoid arthritis: results from a population based case-control study, using incident cases. Ann Rheum Dis 2003; 62: 83541.
  • 7
    Costenbader KH, Feskanich D, Mandl LA, Karlson EW. Smoking intensity, duration, and cessation, and the risk of rheumatoid arthritis in women. Am J Med 2006; 119: 503.e19.
  • 8
    Krishnan E, Sokka T, Hannonen P. Smoking–gender interaction and risk of rheumatoid arthritis. Arthritis Res Ther 2003; 5: R15862.
  • 9
    Tuomi T, Heliovaara M, Palosuo T, Aho K. Smoking, lung function, and rheumatoid factors. Ann Rheum Dis 1990; 49: 7536.
  • 10
    Karlson EW, Lee IM, Cook NR, Manson JE, Buring JE, Hennekens CH. A retrospective cohort study of cigarette smoking and risk of rheumatoid arthritis in female health professionals. Arthritis Rheum 1999; 42: 9107.
  • 11
    Masdottir B, Jonsson T, Manfredsdottir V, Vikingsson A, Brekkan A, Valdimarsson H, et al. Smoking, rheumatoid factor isotypes and severity of rheumatoid arthritis. Rheumatology (Oxford) 2000; 39: 12025.
  • 12
    Harrison BJ, Silman AJ. Does smoking influence disease outcome in patients with rheumatoid arthritis? J Rheumatol 2000; 27: 56970.
  • 13
    Saag KG, Cerhan JR, Kolluri S, Ohashi K, Hunninghake GW, Schwartz DA. Cigarette smoking and rheumatoid arthritis severity. Ann Rheum Dis 1997; 56: 4639.
  • 14
    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: 23138.
  • 15
    Hutchinson D, Shepstone L, Moots R, Lear JT, Lynch MP. Heavy cigarette smoking is strongly associated with rheumatoid arthritis (RA), particularly in patients without a family history of RA. Ann Rheum Dis 2001; 60: 2237.
  • 16
    Padyukov L, Silva C, Stolt P, Alfredsson L, Klareskog L, for the Epidemiological Investigation of Rheumatoid Arthritis Study Group. A gene–environment interaction between smoking and shared epitope genes in HLA–DR provides a high risk of seropositive rheumatoid arthritis. Arthritis Rheum 2004; 50: 308592.
  • 17
    Klareskog L, Stolt P, Lundberg K, Kallberg H, Bengtsson C, Grunewald J, et al, and the Epidemiological Investigation of Rheumatoid Arthritis Study Group. A new model for an etiology of rheumatoid arthritis: smoking may trigger HLA–DR (shared epitope)–restricted immune reactions to autoantigens modified by citrullination. Arthritis Rheum 2006; 54: 3846.
  • 18
    Pedersen M, Jacobsen S, Klarlund M, Pedersen BV, Wiik A, Wohlfahrt J, et al. Environmental risk factors differ between rheumatoid arthritis with and without auto-antibodies against cyclic citrullinated peptides. Arthritis Res Ther 2006; 8: R133.
  • 19
    Kallberg H, Padyukov L, Plenge RM, Ronnelid J, Gregersen PK, van der Helm-van Mil AH, et al. Gene-gene and gene-environment interactions involving HLA-DRB1, PTPN22, and smoking in two subsets of rheumatoid arthritis. Am J Hum Genet 2007; 80: 86775.
  • 20
    Linn-Rasker SP, van der Helm-van Mil AH, van Gaalen FA, Kloppenburg M, de Vries RR, le Cessie S, et al. Smoking is a risk factor for anti-CCP antibodies only in rheumatoid arthritis patients who carry HLA-DRB1 shared epitope alleles. Ann Rheum Dis 2006; 65: 36671.
  • 21
    Michou L, Teixeira VH, Pierlot C, Lasbleiz S, Bardin T, Dieude P, et al. Associations between genetic factors, tobacco smoking and autoantibodies in familial and sporadic rheumatoid arthritis. Ann Rheum Dis 2008; 67: 46670.
  • 22
    Wesoly J, van der Helm-van Mil AH, Toes RE, Chokkalingam AP, Carlton VE, Begovich AB, et al. Association of the PTPN22 C1858T single-nucleotide polymorphism with rheumatoid arthritis phenotypes in an inception cohort. Arthritis Rheum 2005; 52: 294850.
  • 23
    Hill JA, Southwood S, Sette A, Jevnikar AM, Bell DA, Cairns E. Cutting edge: the conversion of arginine to citrulline allows for a high-affinity peptide interaction with the rheumatoid arthritis-associated HLA-DRB1*0401 MHC class II molecule. J Immunol 2003; 171: 53841.
  • 24
    Lee AT, Li W, Liew A, Bombardier C, Weisman M, Massarotti EM, et al. The PTPN22 R620W polymorphism associates with RF positive rheumatoid arthritis in a dose-dependent manner, but not with HLA-SE status. Genes Immun 2005; 6: 12933.
  • 25
    Rieck M, Arechiga A, Onengut-Gumuscu S, Greenbaum C, Concannon P, Buckner JH. Genetic variation in PTPN22 corresponds to altered function of T and B lymphocytes. J Immunol 2007; 179: 470410.
  • 26
    Bottini N, Vang T, Cucca F, Mustelin T. Role of PTPN22 in type 1 diabetes and other autoimmune diseases. Semin Immunol 2006; 18: 20713.
  • 27
    Vang T, Congia M, Macis MD, Musumeci L, Orru V, Zavattari P, et al. Autoimmune-associated lymphoid tyrosine phosphatase is a gain-of-function variant. Nat Genet 2005; 37: 13179.
  • 28
    Rantapaa-Dahlqvist S, de Jong BA, Berglin E, Hallmans G, Wadell G, Stenlund H, et al. Antibodies against cyclic citrullinated peptide and IgA rheumatoid factor predict the development of rheumatoid arthritis. Arthritis Rheum 2003; 48: 27419.
  • 29
    Nielen MM, van Schaardenburg D, Reesink HW, van de Stadt RJ, van der Horst-Bruinsma IE, de Koning MH, et al. Specific autoantibodies precede the symptoms of rheumatoid arthritis: a study of serial measurements in blood donors. Arthritis Rheum 2004; 50: 3806.
  • 30
    Thomson W, Barton A, Ke X, Eyre S, Hinks A, Bowes J, et al. Rheumatoid arthritis association at 6q23. Nat Genet 2007; 39: 14313.
  • 31
    Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS, et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 1988; 31: 31524.
  • 32
    MacGregor AJ, Bamder S, Silman AJ. A comparison of the performance of different methods of disease classification for rheumatoid arthritis: results from an analysis from a nationwide twin study. J Rheumatol 1994; 21: 14206.
  • 33
    Catto AJ, Carter AM, Stickland MH, Bamford JM, Davies JA, Grant PJ. Plasminogen activator inhibitor-1 (PAI-1) 4G/5G promoter polymorphism and levels in subjects with cerebrovascular disease. Thromb Haemost 1997; 77: 7304.
  • 34
    Mewar D, Coote A, Moore D, Marinou I, Keyworth J, Dickson M, et al. Independent associations of anti-cyclic citrullinated peptide antibodies and rheumatoid factor with radiographic severity of rheumatoid arthritis. Arthritis Res Ther 2006; 8: R128.
  • 35
    Rothman KJ. Epidemiology: an introduction. New York: Oxford University Press; 2002.
  • 36
    Hosmer DW, Lemeshow S. Confidence interval estimation of interaction. Epidemiology 1992; 3: 4526.
  • 37
    Van der Helm-van Mil AH, Huizinga TW. Advances in the genetics of rheumatoid arthritis point to subclassification into distinct disease subsets. Arthritis Res Ther 2008; 10: 205.
  • 38
    Huizinga TW, Amos CI, van der Helm-van Mil AH, Chen W, van Gaalen FA, Jawaheer D, et al. Refining the complex rheumatoid arthritis phenotype based on specificity of the HLA–DRB1 shared epitope for antibodies to citrullinated proteins. Arthritis Rheum 2005; 52: 34338.
  • 39
    Meyer O, Nicaise-Roland P, Santos MD, Labarre C, Dougados M, Goupille P, et al. Serial determination of cyclic citrullinated peptide autoantibodies predicted five-year radiological outcomes in a prospective cohort of patients with early rheumatoid arthritis. Arthritis Res Ther 2006; 8: R40.
  • 40
    Kastbom A, Strandberg G, Lindroos A, Skogh T. Anti-CCP antibody test predicts the disease course during 3 years in early rheumatoid arthritis (the Swedish TIRA project). Ann Rheum Dis 2004; 63: 10859.

APPENDIX A

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX A
  10. APPENDIX B

MEMBERS OF THE YEAR CONSORTIUM

Members of the Yorkshire Early Arthritis Register Consortium are as follows:

Management team: Paul Emery, Philip Conaghan, Ann W. Morgan, Anne-Maree Keenan, Elizabeth Hensor, Julie Kitcheman (University of Leeds, Leeds, UK); Mark Quinn (York District Hospital, York, UK).

Consultants: Andrew Gough (Harrogate District Hospital, Harrogate, UK); Michael Green (York District Hospital, York, UK and Harrogate District Hospital, Harrogate, UK); Richard Reece (Huddersfield Royal Infirmary, Huddersfield, UK); Lesley Hordon (Dewsbury District and General Hospital, Dewsbury, UK); Philip Helliwell, Richard Melsom (St Luke's Hospital, Bradford, UK); Sheelagh Doherty (Hull Royal Infirmary, Hull, UK); Ade Adebajo (Barnsley District General Hospital, Barnsley, UK); Andrew Harvey, Steve Jarrett, Zunaid Karim (Pinderfields General Hospital, Wakefield, UK); Gareth Huson, Mike Martin, Colin Pease, Sally Cox (University of Leeds, Leeds, UK); Amanda Isdale (York District Hospital, York, UK); Dennis McGonagle (Calderdale Royal Hospital, Halifax, UK).

Specialist registrars: Victoria Bejarano, Jackie Nam (University of Leeds, Leeds, UK).

Nurses: Claire Brown, Christine Thomas, David Pickles, Alison Hammond, Belinda Rhys-Evans, Barbara Padwell, Sally Smith, Heather King (University of Leeds, Leeds, UK); Beverley Neville (Harrogate District Hospital, Harrogate, UK); Alan Fairclough, Caroline Nunns (Huddersfield Royal Infirmary, Huddersfield, UK); Anne Gill, Julie Green (York District Hospital, York, UK), Julie Madden, Lynda Taylor (Calderdale Royal Hospital, Halifax, UK); Jill Firth, Linda Sigsworth (St Luke's Hospital, Bradford, UK); Jayne Heard (Hull Royal Infirmary, Hull, UK).

Laboratory staff: Diane Corscadden, Karen Henshaw, Lubna-Haroon Rashid, Stephen G. Martin, James I. Robinson (University of Leeds, Leeds, UK).

APPENDIX B

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES
  9. APPENDIX A
  10. APPENDIX B

MEMBERS OF THE UK RHEUMATOID ARTHRITIS GENETICS CONSORTIUM

Members of the UKRAG Consortium are as follows: Stephen Eyre, Anne Hinks, Laura J. Gibbons, John Bowes, Edward Flynn, Paul Martin, Wendy Thomson, Anne Barton, Jane Worthington (University of Manchester, Manchester, UK); Stephen G. Martin, James I. Robinson, Ann W. Morgan, Paul Emery (University of Leeds, Leeds, UK); Anthony G. Wilson (University of Sheffield, Sheffield, UK); Sophia Steer (Kings College Hospital National Health Service Foundation Trust, London, UK); Lynne Hocking, David M. Reid (University of Aberdeen, Aberdeen, UK); Pille Harrison, Paul Wordsworth (University of Oxford, Oxford, UK)