Strong combined gene–environment effects in anti–cyclic citrullinated peptide–positive rheumatoid arthritis: A nationwide case–control study in Denmark

Authors


Abstract

Objective

To study the role of shared epitope (SE) susceptibility genes, alone and in combination with tobacco smoking and other environmental risk factors, for risk of subtypes of rheumatoid arthritis (RA) defined by the presence or absence of serum antibodies against cyclic citrullinated peptides (CCPs).

Methods

To address these issues, a nationwide case–control study was conducted in Denmark during 2002–2004, comprising incident cases of RA or patients with recently diagnosed RA (309 seropositive and 136 seronegative for IgG antibodies against CCP) and 533 sex- and age-matched population controls. Associations were evaluated by logistic regression analyses, in which odds ratios (ORs) served as measures of relative risk.

Results

Compared with individuals without SE susceptibility genes, SE homozygotes had an elevated risk of anti-CCP–positive RA (OR 17.8, 95% confidence interval [95% CI] 10.8–29.4) but not anti-CCP–negative RA (OR 1.07, 95% CI 0.53–2.18). Strong combined gene–environment effects were observed, with markedly increased risks of anti-CCP–positive RA in SE homozygotes who were heavy smokers (OR 52.6, 95% CI 18.0–154), heavy coffee drinkers (OR 53.3, 95% CI 15.5–183), or oral contraceptive users (OR 44.6, 95% CI 15.2–131) compared with SE noncarriers who were not exposed to these environmental risk factors.

Conclusion

Persons who are homozygous for SE susceptibility genes, notably those who are also exposed to environmental risk factors, have a markedly and selectively increased risk of anti-CCP–positive RA. A distinction between anti-CCP–positive RA and anti-CCP–negative RA seems warranted, because these RA subtypes most likely represent etiologically distinct disease entities.

Rheumatoid arthritis (RA), the most common of the autoimmune rheumatic joint diseases, is believed to occur as the result of actions of genetic and nongenetic factors that are currently poorly characterized. Recent studies support the view that RA may not be a single disease entity but rather a clinical syndrome consisting of at least 2 distinct diseases with different etiologies (1–3). Traditionally, RA has been divided into rheumatoid factor (RF)–positive and RF-negative RA (4). More recently, antibodies against cyclic citrullinated peptides (CCPs), which are highly specific for RA, have received attention as a possible basis for subdividing patients with RA into clinically and etiologically meaningful subgroups (2).

The HLA–DRB1 alleles DR4 and DR1, which for most subtypes encode for the conserved amino acid motif called the shared epitope (SE), are well-established genetic risk factors for RA (5, 6). These DRB1 alleles have been found to be more strongly associated with RF-positive RA than with RF-negative RA (7) and selectively with anti-CCP–positive RA (8). Recently, another DRB1 allele, DR3, was reported as being potentially associated with an increased risk of anti-CCP–negative RA (9).

Recently, the combination of SE carrier status and tobacco smoking, the most well-established environmental risk factor for RA, was found to be associated with a 15-fold increased risk of RF-positive RA (10), and the association may be even stronger for anti-CCP–positive RA (11, 12). Based on questionnaire-derived data in the present case–control study, we recently showed that smoking and a series of other nongenetic risk factors are selectively associated with either anti-CCP–positive or anti-CCP–negative RA (3), but whether these nongenetic risk factors operate independently of the genetic background is currently unknown. The aim of the present study was to examine the possible interaction between genetic and nongenetic risk factors in subtypes of RA defined by the presence or absence of anti-CCP antibodies.

PATIENTS AND METHODS

Patients with RA and controls.

The study was conducted as a frequency-matched case–control study. An appropriate size of the study was determined by power calculation, with a significance level of 0.05 and power of 0.80. Minimum statistically significant odds ratios (ORs) were calculated for different sample sizes, based on assumptions about frequencies of exposures in the healthy population and assuming a 3:1 ratio of female predominance among patients with RA. The sample size was chosen to be large enough to detect ORs of ∼1.5–2 as statistically significant for most exposures.

Patients with RA were identified in rheumatology and internal medicine departments throughout Denmark, which has a predominantly white population comprising ∼5.2 million persons. Selected departments were those responding to a general invitation to major departments of rheumatology and internal medicine in Denmark that were expected to be able to recruit >20 patients with RA to the study. Of 22 invited departments, 21 departments participated in the study. To be included, patients had to be age 18–65 years at the time of diagnosis of RA fulfilling the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) 1987 classification criteria (4), between August 1998 and July 2003. Information about the date of diagnosis, defined as the date when the RA diagnosis was clinically confirmed by a rheumatologist, and cumulative fulfillment of the ACR 1987 classification criteria for RA was obtained from medical records by a rheumatologist at each department or by the project coordinator (MP) and a rheumatologist (MK) from the study team.

Control subjects, who were frequency-matched by sex and birth year, were randomly selected from the Danish population by means of the Civil Registration System, a national database that keeps track of all demographic changes in Denmark (13). Using identical invitation letters to patients and control subjects, we aimed at a 1:1 case:control ratio for women and a 1:2 case:control ratio for men; however, all invited individuals who agreed to participate were included.

The study was approved by the Scientific Ethical Committees for Copenhagen and Frederiksberg (KF 01-039/01) and the Danish Data Protection Agency (2001–41-0658).

Interview data.

The procedure for the blinded data collection was tested in a pilot experiment from April to August 2002, to which 100 individuals (50 patients and 50 control subjects) were invited; data for these individuals are not included in this report. Three trained female medical students carried out all interviews between September 2002 and February 2004. Bimonthly meetings were held to ensure that all interviews were performed in a uniform manner. Interviews were conducted as computer-assisted telephone interviews, and answers were entered directly into a database. Logical tests were built into the program to keep data entry errors at a minimum. Each telephone interview took approximately half an hour and covered a broad spectrum of factors potentially associated with RA risk, including marital status, menstrual and reproductive factors, sexual behavior, tobacco smoking, coffee and alcohol consumption 10 years prior to the interview (to avoid problems that might arise if patients with RA modify these lifestyle habits after disease onset), variables measuring socioeconomic status and physical activity, and prior diseases in respondents and their first-degree relatives (3, 14).

Anti-CCP antibody status.

Blood samples were collected at rheumatology departments (patients) or by general practitioners (patients and controls), and serum was stored at −20°C. Anti-CCP IgG antibodies were determined by a second-generation enzyme-linked immunosorbent assay using the Immunoscan RA kit (Euro-Diagnostica, Malmö, Sweden). We used a cutoff value of 25 arbitrary units (AUs) per milliliter to distinguish between anti-CCP–positive (>25 AU/ml) and anti-CCP–negative (≤25 AU/ml) cases of RA, in accordance with routine clinical guidelines to determine anti-CCP serostatus at the Department of Autoimmunity, Statens Serum Institut, Copenhagen, Denmark.

HLA–DRB1 genotyping.

Genomic DNA was isolated from EDTA-preserved blood cells, using a QIAamp Maxi Kit (Qiagen, Chatsworth, CA) in accordance with the manufacturer's instructions, and stored at −20°C. Low-resolution HLA–DRB1 tissue typing was performed by polymerase chain reaction–based sequence-specific oligonucleotide probing, as described elsewhere (15). Here, we define the SE as the presence of HLA–DRB1*04 and/or HLA–DRB1*01. Low-resolution HLA class II typing allows distinction between heterozygous and homozygous DR genotypes but does not distinguish between various subtypes of DRB1*04. This type includes the following subtypes containing the SE: HLA–DRB1*0401, 0404, 0405, and 0408 (16), which together comprise ∼80% of the DRB1*04 allele frequency (17). DRB1*01 has 2 subtypes containing the SE: DRB1*0101 and 0102, which comprise ∼97% of the DRB1*01 allele frequency (17). Accordingly, our definition of the SE is slightly conservative in terms of demonstrating associations with RA.

Statistical analysis.

In order to make exposure information comparable for patients and controls, a pseudo-year of diagnosis was attributed to control subjects, according to the frequency distribution of the year of RA diagnosis for patients of the same sex. Throughout, we disregarded information about exposures after the year (patients) or pseudo-year (controls) of diagnosis.

In a series of logistic regression analyses, we calculated ORs with 95% confidence intervals (95% CIs) for the associations of genotypes and nongenetic risk factors with the risk of RA, with adjustment for sex, birth year, and year (or pseudo-year) of diagnosis. To examine possible sex-specific risk factor differences, we introduced terms for statistical interaction with sex for all studied genotypes and nongenetic risk factors in our statistical models, but none of these interaction terms was statistically significant (for all likelihood ratio tests, P > 0.05). Therefore, all reported analyses are based on the combined data set for women and men. To study possible differences in genotypes associated with anti-CCP–positive and anti-CCP–negative RA, we used polytomous logistic regression to calculate ORs associated with carrier status (noncarrier, heterozygous carrier, homozygous carrier) for HLA–DRB1 alleles DR1, DR4, the SE, and DR3, with adjustment for sex, birth year, and year (or pseudo-year) of diagnosis.

Elsewhere, we reported statistically significant associations between a series of nongenetic exposures and the risk of anti-CCP–positive RA, including tobacco smoking, alcohol and coffee consumption, use of oral contraceptives, marital status, employment status, and having a first-degree relative with schizophrenia (3). Here, using logistic regression, we examined the combined effects of these risk factors and SE carrier status, adjusting for potential confounding by sex, birth year, year (or pseudo-year) of diagnosis, place of residence (5 categories), education status (4 categories), pack-years of smoking (3 categories), alcohol consumption (3 categories), coffee consumption (3 categories), and use of oral contraceptives (never versus ever). To evaluate whether SE carrier status and nongenetic risk factors for anti-CCP–positive RA operate independently of each other, we introduced terms for statistical interaction between SE carrier status and each of the examined nongenetic risk factors in our logistic regression models. All logistic regression analyses were carried out using SAS software version 9.1 (PROC LOGISTIC and PROC GENMOD procedures) (SAS Institute, Cary, NC). Throughout, 2-sided P values less than 0.05 were considered significant.

The population attributable risk percentage (PAR%) due to smoking among SE carriers was estimated using the approach described by Bruzzi et al (18). Assuming a causal relationship, the estimated PAR% can be interpreted as the proportion of anti-CCP–positive RA cases in the population that theoretically can be prevented if SE carriers abstain from tobacco smoking.

RESULTS

The overall study population consisted of 515 patients with RA (participation rate 83%) and 769 population controls (participation rate 64%), for whom selected demographic characteristics are shown in Table 1. The mean disease duration at the time of interview was 2.3 years (range 0–5 years). Four hundred fifty-six patients with RA (89% of those interviewed) provided blood samples and were successfully HLA–DRB1 genotyped. Of these, 309 (69%) were positive and 136 (31%) were negative for anti-CCP antibodies (samples from 11 patients were not examined for anti-CCP antibodies). Among controls, 533 (69% of those interviewed) provided blood samples and were HLA–DRB1 genotyped.

Table 1. Demographic characteristics of Danish patients with rheumatoid arthritis (RA) and population controls, 1998–2003*
CharacteristicParticipants who completed interviewParticipants who completed interview and genetic testing
RA patients (n = 515)Population controls (n = 769)RA patients (n = 456)Population controls (n = 533)
  • *

    Except where indicated otherwise, values are the number (%) of subjects.

  • The age of population controls was calculated as the pseudo-year of diagnosis minus the birth year.

  • Information missing for 1 population control; current students were categorized as “no post-school education.”

Female sex366 (71.1)478 (62.2)322 (70.6)327 (61.4)
Male sex149 (28.9)291 (37.8)134 (29.4)206 (38.7)
Age at diagnosis, mean (range) years49 (18–65)48 (16–68)49 (18–65)50 (18–67)
Birth year    
 <194071 (13.8)98 (12.7)64 (14.0)76 (14.3)
 1940–1949171 (33.2)250 (32.5)152 (33.3)190 (35.6)
 1950–1959145 (28.2)223 (29.0)131 (28.7)156 (29.3)
 1960–196977 (15.0)129 (16.8)67 (14.7)80 (15.0)
 ≥197051 (9.9)69 (9.0)42 (9.2)31 (5.8)
Place of residence    
 Copenhagen86 (16.7)102 (13.3)72 (15.8)63 (11.8)
 Suburb of Copenhagen90 (17.5)137 (17.8)79 (17.3)98 (18.4)
 Other town with ≥100,000 inhabitants78 (15.1)77 (10.0)68 (14.9)46 (8.6)
 Other town with 10,000–99,999 inhabitants93 (18.1)197 (25.6)80 (17.5)141 (26.5)
 Rural area/town with <10,000 inhabitants168 (32.6)256 (33.3)157 (34.4)185 (34.7)
Education    
 No post-school education123 (23.9)120 (15.6)105 (23.0)77 (14.5)
 Semi-skilled worker, advanced studies for  <1 year, or apprentice204 (39.6)293 (38.2)180 (39.5)205 (38.5)
 Advanced studies for 1–4 years156 (30.3)265 (34.5)139 (30.5)192 (36.1)
 Advanced studies for >4 years32 (6.2)90 (11.7)32 (7.0)58 (10.9)

Genotypes constituting the SE and DRB1*03.

Individuals with HLA–DRB1 genotypes constituting the SE were at statistically significantly increased risk of anti-CCP–positive RA but not anti-CCP–negative RA (Table 2). Specifically, DR4 homozygotes (OR 17.5, 95% CI 9.78–31.3) and SE homozygotes (OR 17.8, 95% CI 10.8–29.4) were at markedly elevated risk compared with their respective reference groups of noncarriers of DR4.

Table 2. HLA–DRB1 genotypes and risk of rheumatoid arthritis (RA), Denmark, 1998–2003*
 No. of anti-CCP–positive RA patientsNo. of anti-CCP–negative RA patientsNo. of population controlsOdds ratio (95% confidence interval)P for homogeneity
RA overallAnti-CCP–positive RAAnti-CCP–negative RA
  • *

    Odds ratios were adjusted for sex, birth year, and year (or pseudo-year) of diagnosis. Shared epitope heterozygotes comprise individuals with DRB1*04/X and DRB1*01/X genotypes, where X indicates all DR alleles other than 04 and 01. Shared epitope homozygotes comprise individuals with DRB1*04/04, DRB1*01/01, and DRB1*01/04 genotypes.

  • Anti–cyclic citrullinated peptide (anti-CCP)–positive versus anti-CCP–negative RA.

  • Tests for trend: RA overall, P < 0.001; anti-CCP–positive RA, P < 0.001; anti-CCP–negative RA, P = 0.92.

  • §

    Tests for homogeneity: RA overall, P = 0.10; anti-CCP–positive RA, P = 0.03; anti-CCP–negative RA, P = 0.88.

  • Tests for trend: RA overall, P < 0.001; anti-CCP–positive RA, P < 0.001; anti-CCP–negative RA, P = 0.97.

DR4       
 Noncarriers82863401 reference1 reference1 reference 
 Heterozygotes156441732.32 (1.74–3.09)3.86 (2.75–5.41)0.93 (0.60–1.43) 
 Homozygotes716208.33 (4.85–14.3)17.5 (9.78–31.3)1.13 (0.43–2.99)<0.001
DR1§       
 Noncarriers2311084311 reference1 reference1 reference 
 Heterozygotes or homozygotes78281021.31 (0.95–1.81)1.48 (1.04–2.11)1.04 (0.63–1.70)0.18
Shared epitope       
 Noncarriers47642571 reference1 reference1 reference 
 Heterozygotes146592332.08 (1.54–2.81)3.72 (2.52–5.49)0.96 (0.63–1.46) 
 Homozygotes11613437.19 (4.69–11.0)17.8 (10.8–29.4)1.07 (0.53–2.18)<0.001

Compared with the risk in DR3 noncarriers, a reduced risk of anti-CCP–positive RA was observed in the combined group of homozygous and heterozygous DR3 carriers (OR 0.64, 95% CI 0.45–0.91), but there was no significant association of DR3 carrier status with anti-CCP–negative RA (OR 1.21, 95% CI 0.78–1.86). Because DR3 noncarriers include high-risk SE carriers, we subsequently restricted the analyses to SE noncarriers. With this restriction, risks associated with DR3 were nonsignificantly increased for both anti-CCP–positive RA (OR 1.93, 95% CI 0.98–3.79) and anti-CCP–negative RA (OR 1.45, 95% CI 0.81–2.61).

Combined effects of SE carrier status and nongenetic risk factors in anti-CCP–positive RA.

Having documented a strong positive association between SE carrier status and the risk of anti-CCP–positive RA, we subsequently examined possible interactions between SE carrier status and a series of recently identified nongenetic risk factors for this serologic subtype of RA. Upon introduction of terms for statistical interaction in our multiplicative logistic regression models, we found no evidence of such interaction for any combination of genetic predisposition risk factors (noncarrier, heterozygous carrier, homozygous carrier) and the presently studied environmental risk factors (all P > 0.10), suggesting that genetic and nongenetic risk factors operated independently.

Consequently, markedly elevated ORs were observed among SE carriers who were also exposed to the studied environmental risk factors (Table 3). Specifically, SE homozygotes with a cumulative tobacco consumption of >20 pack-years had markedly increased odds of anti-CCP–positive RA compared with SE noncarriers who had never smoked (OR 52.6, 95% CI 18.0–154). Drinking >5 cups of coffee per day (OR 53.3, 95% CI 15.5–183) and ever use of oral contraceptives (OR 44.6, 95% CI 15.2–131) were also associated with high risks of anti-CCP–positive RA among SE homozygotes compared with noncarriers who did not drink coffee or use oral contraceptives, respectively. Abstinence from alcohol was associated with a high risk of anti-CCP–positive RA in genetically predisposed individuals (OR 50.1, 95% CI 8.26–304, for nondrinking SE homozygotes versus noncarriers with moderate alcohol consumption) (Table 3).

Table 3. Combined effects of shared epitope carrier status and nongenetic risk factors in anti-CCP–positive rheumatoid arthritis, Denmark, 1998–2003*
 Shared epitope noncarriersShared epitope heterozygotesShared epitope homozygotes
No. of patients/controlsOR (95% CI)No. of patients/controlsOR (95% CI)No. of patients/controlsOR (95% CI)
  • *

    Questionnaire information about nongenetic risk factors was missing for some individuals. Shared epitope heterozygotes comprise individuals with DRB1*04/X and DRB1*01/X genotypes, where X indicates all DR alleles other than 04 and 01. Shared epitope homozygotes comprise individuals with DRB*04/04, DRB1*01/01, and DRB1*01/04 genotypes. Odds ratios (ORs) were adjusted for sex, birth year, year (or pseudo-year) of diagnosis, education status, place of residence, tobacco smoking, alcohol consumption, coffee consumption, and use of oral contraceptives (OCs). Alcohol and coffee consumption refer to consumption 10 years prior to the interview. Anti-CCP = anti–cyclic citrullinated peptide; 95% CI = 95% confidence interval.

  • One pack-year = 7,300 cigarettes (20 cigarettes/day for 1 year).

  • One drink = 1 bottle of normal beer (⅔ bottle of strong beer) or 1 glass of wine, dessert wine, or spirits.

Tobacco smoking      
 Never smokers16/1031 reference33/873.31 (1.58–6.96)34/2017.4 (7.22–41.8)
 ≤20 pack-years18/811.74 (0.76–3.98)57/915.07 (2.48–10.4)47/1357.4 (22.0–150)
 >20 pack-years12/671.22 (0.48–3.08)53/519.66 (4.38–21.3)32/852.6 (18.0–154)
Alcohol consumption      
 0 drinks/week11/311.73 (0.71–4.21)33/198.72 (4.00–19.0)14/250.1 (8.26–304)
 1–10 drinks/week30/1551 reference82/1463.25 (1.92–5.51)75/2427.0 (13.6–53.7)
 >10 drinks/week5/650.30 (0.10–0.88)28/642.18 (1.11–4.30)24/1510.5 (4.41–25.2)
Coffee consumption      
 0 cups/day7/371 reference21/373.99 (1.34–11.9)15/813.0 (3.33–50.8)
 1–5 cups/day18/1081.27 (0.43–3.75)40/1023.03 (1.09–8.47)50/2327.4 (9.04–83.3)
 >5 cups/day21/1061.13 (0.38–3.38)82/906.84 (2.45–19.1)48/1053.3 (15.5–183)
OC use, women only      
 Never8/561 reference32/396.47 (2.41–17.4)24/1232.3 (9.93–105)
 Ever26/962.27 (0.85–6.04)61/1015.38 (2.11–13.7)54/1744.6 (15.2–131)

Moreover, after adjustment for sex, birth year, year (or pseudo-year) of diagnosis, education level, place of residence, pack-years smoked, intake of coffee and alcohol, and use of oral contraceptives, rather extreme ORs for anti-CCP–positive RA were seen among SE homozygotes who were unmarried (OR 177, 95% CI 18.3–1,715) and those without a job (OR 243, 95% CI 23.7–2,501) compared with SE noncarriers who were married or cohabiting and those who had a job, respectively. In a similarly adjusted analysis, having a first-degree relative with schizophrenia was associated with a high risk of anti-CCP–positive RA (OR 33.2, 95% CI 7.45–148) in homozygous and heterozygous SE carriers compared with noncarriers reporting no schizophrenia among their first-degree relatives.

Based on the ORs in Table 3, we estimate that smoking among heterozygous and homozygous SE carriers accounts for 36% of all cases of anti-CCP–positive RA in the population.

DISCUSSION

The present study provides strong support for the recent proposal that RA may be composed of more than a single disease entity (1–3). We also confirm strong links between HLA–DRB1 alleles encoding for the SE and the risk of anti-CCP–positive RA (8). The major novel finding, however, is that the combined effects of SE carrier status and several nongenetic risk factors produce highly increased risks of anti-CCP–positive RA, an observation that has previously been described only for smoking (10, 11).

In a recent Swedish case–control study, SE homozygotes who had ever smoked had a 21-fold increased risk of anti-CCP–positive RA compared with never-smoking noncarriers of the SE (11). In the present study, heavy smoking in an SE homozygote was associated with an OR for anti-CCP–positive RA of ∼50 compared with never-smoking noncarriers of the SE. Assuming causality, the PAR% for smoking among SE carriers was estimated at 36%, provided that anti-CCP–positive RA risk levels among tobacco-smoking heterozygous and homozygous SE carriers can be reduced to corresponding levels for nonsmokers of the same genotype. Theoretically, therefore, in Denmark and countries with similar levels of tobacco smoking, we estimate that ∼1 of 3 new anti-CCP–positive RA cases in the population can be avoided if all persons with an SE predisposition abstain from tobacco smoking. However, people generally do not know their HLA type, so major public health implications of this finding are unlikely. Nevertheless, smoking should be strongly discouraged among persons with first-degree relatives affected by anti-CCP–positive RA. Carriers of the SE will remain at markedly increased risk compared with noncarriers, but in terms of absolute risk reduction, such efforts at targeted smoking prevention may markedly reduce the risk at the individual level.

Heavy coffee drinking has, in several studies, been shown to be associated with RA (19, 20). In concordance with these studies, we observed that heavy coffee drinking was associated with an increased risk of anti-CCP–positive RA. We were not able to examine caffeinated and decaffeinated coffee separately. However, the rate of decaffeinated coffee intake in Denmark is negligible, and our results therefore represent associations between caffeinated coffee and risk of anti-CCP–positive RA.

Conflicting results have been obtained for the association between use of oral contraceptives and the risk of RA (21). However, none of the previous studies assessed subtypes of RA according to anti-CCP status, which might explain why the association between use of oral contraceptives and anti-CCP–positive RA has not been reported until now.

In the present study, we observed that tobacco smoking was not the only modifiable risk factor to produce markedly increased risks of anti-CCP–positive RA when combined with susceptibility genes for RA. Even after adjustment for the effect of tobacco smoking and other nongenetic risk factors, rather extreme ORs for anti-CCP–positive RA were also seen for SE homozygotes who were heavy coffee drinkers, oral contraceptive users, unmarried, or without a job, suggesting that these lifestyle factors or their correlates might be part of an etiologic mechanism involving both genetic predisposition and environmental factors. Recently, it was shown that conversion of arginine to citrulline, a process catalyzed by the enzyme peptidylarginine deiminase, results in peptides with significantly increased DR4 affinity and increased T cell activation in mice (22). It is conceivable that tobacco smoking and other environmental risk factors that we and other investigators have shown to be associated with anti-CCP–positive RA might somehow operate in this citrullination process. Although currently substantiated only for tobacco smoking (11), such a mechanism might explain the observed strong associations between environmental risk factors, SE carrier status, and risk of anti-CCP–positive, but not anti-CCP–negative, RA.

Elsewhere, we have shown that alcohol is significantly inversely associated with the risk of anti-CCP–positive RA, suggesting a protective effect (3). Here, we observed that SE homozygotes reporting no weekly alcohol consumption 10 years before the interview had a markedly higher risk (OR 50.1) of anti-CCP–positive RA than noncarriers who consumed 1–10 alcoholic drinks per week. Other observations support an inverse association between alcohol intake and RA risk. In one study, alcohol intake at the time of the first visit to a rheumatology department was lower among women with RA than among women with soft tissue rheumatism or osteoarthritis (23), and other researchers have suggested a protective effect of alcohol that may be more pronounced in RF-positive than in RF-negative RA (24). If immune mechanisms involved in the pathogenesis of anti-CCP–positive RA are affected by alcohol, the nonspecific immune suppression thus exerted (25) might be beneficial in relation to RA risk.

Having a first-degree relative with schizophrenia was associated with a highly and significantly increased risk of anti-CCP–positive RA in SE carriers. Because RA and schizophrenia are inversely associated in studies of intraindividual disease correlations (26), this was not an a priori expectation, but anti-CCP–positive RA is not the first autoimmune disease to be linked with schizophrenia. An increased risk of insulin-dependent diabetes mellitus has been reported among first-degree relatives of patients with schizophrenia (27). The magnitude and statistical significance of the association observed in the present study and its specificity to the anti-CCP–positive subtype of RA render it unlikely to be a result of chance or spurious recall problems. The finding is potentially interesting, because previous observations point to possible common etiologic features between some autoimmune diseases and schizophrenia, although conflicting results have been obtained for the association between the HLA–DR4 genotype and the risk of schizophrenia (28, 29). Further examination of possible links between schizophrenia and anti-CCP–positive RA is warranted.

We observed no evidence of statistical interaction for any combination of genetic predisposition and the presently studied environmental risk factors, suggesting that genetic and nongenetic risk factors operate independently. Therefore, because we observed a highly increased risk of anti-CCP–positive RA associated with being an SE carrier and a moderately increased risk associated with exposure to the studied environmental risk factors, ORs for combinations of these genetic and environmental risk factors were highly increased. The lack of significant statistical interactions found in this study might, however, be attributable to inadequate statistical power to detect such interactions, notably because of the small number of SE noncarriers among patients with anti-CCP–positive RA.

There are no well-established genetic and nongenetic risk factors for anti-CCP–negative RA. However, the DRB1 allele DR3 was recently shown to be associated with an increased risk of anti-CCP–negative RA in a Dutch study (9), and DR3 carriers may have lower levels of anti-CCP antibodies compared with DR3 noncarriers (30). Our findings do not support an association between DR3 and the risk of anti-CCP–negative RA. Among SE noncarriers, we found no statistically significant association of DR3 with either anti-CCP–positive or anti-CCP–negative RA. Further studies are needed to assess genetic and nongenetic risk factors for anti-CCP–negative RA. Possibly, the lack of convincing etiologic candidates may reflect that this RA subtype, characterized merely by the absence of anti-CCP antibodies, comprises a heterogeneous group of etiologically distinct rheumatic diseases.

The patients in our study, who were identified at hospital departments of rheumatology and internal medicine throughout Denmark over a 5-year diagnostic period, are likely to be representative of hospital outpatients with RA. Our findings may not necessarily apply to milder cases of RA managed in outpatient settings, although we have no reason to believe that the strong combined gene–environment effects we observed for anti-CCP–positive RA would be different in other RA populations. Although the participation rate was high among patients with RA, invited population controls were slightly more reluctant to provide both interview data and blood samples. Theoretically, such a difference might lead to biased risk associations for some of the nongenetic factors studied, if these risk factors were also associated with the decision to accept or decline our invitation to participate. Examination of the influence of self-selection is limited by the fact that information about exposures is not available for nonresponders. However, assuming that similar selection takes place when participants accept or decline to provide a blood sample, a rough estimate of the influence of self-selection can be made. No differences were observed in pack-years smoked among controls who did and those who did not provide a blood sample, indicating that self-selection alone does not explain the observed associations between smoking and risk of anti-CCP–positive RA. Additionally, because tobacco and alcohol consumption are positively correlated behaviors, the observed strong inverse association between alcohol intake and risk of anti-CCP–positive RA cannot plausibly be explained by the lower participation rate among controls.

In our study, nongenetic risk factors were assessed retrospectively by means of telephone interviews; therefore, the possible influence of recall bias among study participants needs consideration. Because the nongenetic exposures studied are not broadly recognized as RA risk factors, we believe that misclassification arising from recall problems would most likely have been nondifferential and tended to produce conservative ORs. Accordingly, the systematic and strong combined effects of genetic and nongenetic risk factors observed for anti-CCP–positive RA, but not for anti-CCP–negative RA, cannot plausibly be the result of spurious recall problems among our study participants.

In conclusion, our study provides strong support for the notion that RA consists of at least 2 etiologically distinct disease entities identified by the presence or absence of anti-CCP antibodies. For anti-CCP–positive RA, we document markedly increased risks in SE homozygotes, notably those who are also exposed to environmental risk factors. Physicians should encourage patients with a family history of anti-CCP–positive RA to abstain from tobacco smoking, and associations of patients with rheumatic diseases should seriously consider campaigns to discourage tobacco smoking among persons with a family predisposition to anti-CCP–positive RA.

AUTHOR CONTRIBUTIONS

Merete Pedersen 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 design. M. Pedersen, Jacobsen, Wohlfahrt, Frisch.

Acquisition of data. M. Pedersen, Jacobsen, Garred, Madsen, Klarlund, Svejgaard.

Analysis and interpretation of data. M. Pedersen, Jacobsen, Garred, Madsen, Klarlund, B. Pedersen, Wohlfahrt, Frisch.

Manuscript preparation. M. Pedersen, Jacobsen, Garred, Madsen, Klarlund, B. Pedersen, Wohlfahrt, Frisch.

Statistical analysis. M. Pedersen, Garred, B. Pedersen, Wohlfahrt.

Acknowledgements

We thank the following individuals for providing patients for the study: Drs. J. K. Pedersen (Kong Christian X's Gigthospital, Gråsten), S. Freiesleben-Sørensen (Bispebjerg Hospital), A. Rødgaard (Roskilde County Hospital Køge), A. Hansen and M. Sejer Hansen (Copenhagen County Hospital Glostrup), L. Juul (Frederiksberg Hospital), H. H. Mogensen (Hørsholm Hospital), B. Unger (Holstebro Hospital), P. Mosborg Petersen (Randers Hospital), J. Christensen (Næstved Hospital), R. Pelck (Roskilde County Hospital Roskilde), M. Graakjær Nielsen (Aarhus University Hospital), N. Gregersen (Bornholm Hospital), and J. Sylvest (Amager Hospital). We also thank the rheumatology departments at the following hospitals for permission to identify patients for the study: Hvidovre Hospital, University Hospital of Copenhagen, Herlev University Hospital, Esbjerg Hospital, Gentofte Hospital, Odense University Hospital, Slagelse Hospital, and Århus University Hospital.

Ancillary