Association between body mass index and anti–citrullinated protein antibody–positive and anti–citrullinated protein antibody–negative rheumatoid arthritis: Results from a population-based case–control study

Abstract

Objective

Being overweight or obese is associated with many chronic diseases, but previous studies of the association with rheumatoid arthritis (RA) have shown inconsistent results. The aim of this study was to investigate the association between body mass index (BMI) and the risk of developing the 2 main subtypes of RA.

Methods

At inclusion, cases and controls answered questions about their weight and height and donated blood samples. The presence of antibodies to citrullinated protein antigens (ACPAs) was analyzed among 2,748 cases and 3,444 controls (28% men). Odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated using conditional logistic regression.

Results

Compared to those with normal weight (BMI <25 kg/m²), the adjusted overall OR for developing ACPA-negative RA was 1.1 (95% CI 0.9–1.3) for overweight individuals (BMI ≥25 to <30 kg/m²) and 1.4 (95% CI 1.1–1.9) for obese individuals (BMI ≥30 kg/m²). When stratified by sex, the OR for ACPA-negative RA for obese women was 1.6 (95% CI 1.2–2.2), and there was no association between obesity and ACPA-negative RA in men (OR 1.1, 95% CI 0.6–1.8). In obese men compared to men with normal weight, the OR for ACPA-positive RA was 0.6 (95% CI 0.3–0.9), while there was no association between BMI and ACPA-positive RA among women (OR 1.0, 95% CI 0.8–1.2).

Conclusion

Our findings show that obesity is associated with developing ACPA-negative RA in women, and indicate an inverse association between BMI and ACPA-positive RA in men.

INTRODUCTION

Rheumatoid arthritis (RA) is the most common inflammatory joint disease, with a prevalence in Sweden of ∼0.7%, and a higher prevalence in women than in men (1). The number of overweight and obese people has dramatically increased worldwide in the last 2 decades (1). In Sweden, the prevalence of overweight or obese adults has doubled since the 1980s, and today every second man and every third woman is overweight or obese (1–3).

Being overweight or obese is associated with several chronic diseases, such as cardiovascular disease (CVD) (4, 5). Inflammation is involved in the pathogenesis of both RA and CVD, and some lifestyle factors that are associated with the risk of developing CVD, notably smoking, have also been shown to be associated with RA (6–8). Being overweight or obese has been shown to be associated with an overall increased level of inflammation (9). Taken together, this information indicates that the risk for RA may be associated with weight status.

To date, the published associations between being overweight or obese and RA have been contradictory, with some studies reporting no association (10–15), and others showing obesity to increase the risk of developing RA (16, 17). One relatively small case–control study (18) found an association of being overweight or obese to be confined to those with anti–citrullinated protein antibody (ACPA)–negative RA, a subgroup analysis that was not carried out in the earlier studies. Recent data on the major etiologic differences between ACPA-positive and ACPA-negative RA (19) have emphasized that analyses of potential risk factors for RA should always include a separate analysis of these disease subsets.

Because of the potentially important implications for understanding the disease-causing mechanisms in RA subgroups, we set out to investigate the association between body mass index (BMI) and RA specifically in the etiologically distinct groups of ACPA-positive and ACPA-negative RA using a large population-based study, the Epidemiological Investigation of Rheumatoid Arthritis (EIRA).

Significance & Innovations

• Obesity is associated with developing anti–citrullinated protein antibody (ACPA)–negative rheumatoid arthritis (RA) in women, and indicates an inverse association between body mass index (BMI) and ACPA-positive RA.

• The association between BMI and RA differs among women and men.

SUBJECTS AND METHODS

Study design and subjects.

The RA patients evaluated were recruited from EIRA, a population-based case–control study covering public rheumatology units and most private units in the middle and southern parts of Sweden. A detailed description of the design and methods of EIRA has been previously published (20, 21). Cases and controls included in these analyses were recruited between May 1996 and November 2009. A case was defined as an individual age 18–70 years in whom RA was newly diagnosed by a rheumatologist according to the American College of Rheumatology (ACR) 1987 revised criteria (22).

For each case, a control subject (1 control per case from May 1996 to May 2006, and after a revision of the study protocol, 2 controls per case from June 2006 to November 2009) who was matched by age, sex, and location of residence (county or municipality area) was randomly selected using the Swedish national population registry (23). If a control declined to participate, was not traceable, or reported having RA, a new control was selected using the same procedure.

Data collection.

Consenting cases were given a self-administered questionnaire at the rheumatology unit, while the controls received an identical questionnaire by mail. The questionnaire contained a wide range of questions about lifestyle and environmental exposures. Of the 3,062 cases and 5,196 controls identified, 2,886 cases (94%) and 4,072 controls (78%) answered the questionnaire. The Regional Committee on Ethics at the Karolinska Institute approved the study.

Exposure and covariates.

Self-reported weight and height at inclusion were used to calculate BMI. We used the World Health Organization BMI classifications of underweight (BMI <18.5 kg/m²), normal weight (BMI ≥18.5 to <25 kg/m²), overweight (BMI ≥25 to <30 kg/m²), and obese (BMI ≥30 kg/m²). Less than 2% of the observations had a BMI <18.5 kg/m², so the categories for underweight and normal weight were combined (BMI <25 kg/m²). We performed the same categorization with information on weight and height at age 20 years. Since being overweight or obese is related to several other lifestyle factors that potentially can explain an association between BMI and RA, we adjusted for smoking (never/current/past/nonregular smoking), alcohol consumption (ever [consumed alcohol during the past 12 months]/never [no alcohol consumed in the past 12 months]), and educational level (no university degree/university degree, as a proxy for socioeconomic status) (21, 24, 25). We also considered the following additional potential confounding factors: menopause (using age 50 years as a proxy for menopause), civil status, born in Sweden (yes/no), physical activity (categorized both as ever [physically active at least sometimes per week]/never [very seldom or never] and sedentary/low/moderate/high), occupational class as a proxy for socioeconomic status (as an alternative to educational level) parity, and history of oral contraceptive use (yes/no) and fatty fish consumption (never/seldom, sometimes/month, sometimes/week).

Genetic and serologic analyses.

The presence of ACPA was determined using the Immunoscan RA Mark 2 enzyme-linked immunosorbent assay test (Euro-Diagnostica), as described previously (26). ACPA was considered present if the serum antibody level was >25 units/ml. We extracted DNA from the blood and genotyped HLA–DRB1 to establish the presence or absence of shared epitope (SE) alleles (DRB1*01, DRB1*04, and DRB1*10), as described in detail by Padyukov et al (27).

Statistical analysis.

Odds ratios (ORs) with 95% confidence interval (95% CIs) for developing RA were calculated using conditional logistic regression models. We investigated the effect of overweight (BMI ≥25 to <30 kg/m²) and obesity (BMI ≥30 kg/m²) using normal weight (BMI <25 kg/m²) as the reference. The final multivariate model was conditional on the matching variables (sex, age, and area of residence) and adjusted for smoking, alcohol consumption, and education. Due to missing information on exposure variables (present height or weight) or confounding variables (smoking, alcohol, and education), 138 cases (0.5%) and 628 controls (15%) and their matched case/control were excluded (Supplementary Table 1 and Supplementary Figure 1, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002(ISSN)2151-4658).

We aimed to stratify by APCA status and analyze women and men separately due to numerous prior indications of differing mechanisms causing disease in these groups (27). In secondary analyses, we stratified for the presence of any HLA–DRB1 SE alleles, menopause, parity, and oral contraceptive use and RA with BMI divided into 2 categories (BMI <25 kg/m² and BMI ≥25 kg/m²). All analyses were performed using SAS software, version 9.2.

RESULTS

We analyzed 2,748 cases (1,962 women and 786 men) and 3,444 controls (2,468 women and 976 men), where 696 cases had 2 controls each (Table 1). The median age among the participants was 54 years (range 18–70 years) for women and 56 years (range 19–70 years) for men. Sixty-three percent of the cases were ACPA positive.

Table 1. Characteristics stratified by ACPA*

	ACPA-positive RA			ACPA-negative RA
	BMI <25 kg/m^2	BMI ≥25 to <30 kg/m^2	BMI ≥30 kg/m^2	BMI <25 kg/m^2	BMI ≥25 to <30 kg/m^2	BMI ≥30 kg/m^2
* Values are the number (percentage) of cases/controls unless otherwise indicated. ACPA = anti–citrullinated protein antibody; RA = rheumatoid arthritis; BMI = body mass index.
Disease duration, median days	205	213	218	212	183	177
Age at onset, median years	52	56	55	53	57	56
Sex
Women	379 (59)/1,455 (59)	355 (28)/708 (29)	159 (13)/305 (12)	379 (54)/1,455 (59)	208 (30)/708 (29)	117 (17)/305 (12)
Men	229 (46)/360 (37)	217 (43)/475 (49)	57 (11)/141 (14)	364 (96)/360 (37)	143 (51)/475 (49)	44 (16)/141 (14)
Cigarette smoking
Never	299 (61)/787 (53)	145 (30)/476 (32)	45 (9)/216 (15)	192 (51)/787 (53)	127 (34)/476 (32)	55 (15)/216 (15)
Current	318 (59)/383 (55)	167 (31)/232 (33)	55 (10)/78 (11)	124 (54)/383 (55)	73 (32)/232 (33)	31 (14)/78 (11)
Past	268 (46)/450 (47)	215 (37)/376 (40)	100 (17)/126 (13)	102 (36)/450 (47)	120 (42)/376 (40)	64 (22)/126 (13)
Nonregular	88 (59)/195 (61)	45 (30)/99 (31)	16 (11)/26 (8)	57 (58)/195 (61)	31 (31)/99 (31)	11 (11)/26 (8)
Alcohol consumption
None	96 (45)/138 (46)	82 (38)/105 (35)	36 (17)/59 (20)	57 (46)/138 (46)	34 (28)/105 (35)	32 (26)/59 (20)
Ever	872 (57)/1,671 (53)	490 (32)/1,077 (34)	180 (12)/382 (12)	417 (48)/1,671 (53)	316 (37)/1,077 (34)	127 (15)/382 (12)
University degree
No	710 (52)/1,169 (48)	478 (35)/896 (37)	182 (13)/356 (15)	342 (45)/1,169 (48)	280 (37)/896 (37)	140 (18)/356 (15)
Yes	263 (67)/646 (63)	94 (24)/287 (28)	34 (9)/90 (9)	133 (59)/646 (63)	71 (32)/287 (28)	21 (9)/90 (9)

In this study sample, 53% of the cases and controls were normal weight (BMI <25 kg/m²), 34% of the cases and controls were overweight (BMI ≥25 to <30 kg/m²), and 14% of the cases and 13% of the controls were obese (BMI ≥30 kg/m²). The cases were more likely to smoke, whereas the controls were slightly more likely to drink alcohol, reflecting known associations (21, 25). A higher proportion of controls (30%) than cases (22%) had a university degree, also reflecting a known association (24).

Association with RA overall and by ACPA and sex.

There was no association between BMI and RA overall (BMI ≥25 to <30 kg/m²: OR 0.9, 95% CI 0.8–1.1; BMI ≥30 kg/m²: OR 1.0, 95% CI 0.9–1.2). For ACPA-negative RA, a BMI ≥30 kg/m² was associated with a significantly increased risk (OR 1.4, 95% CI 1.1–1.9) in women and men combined. Stratifying by sex, the increased risk for ACPA-negative RA in obese individuals was found to be restricted to women (OR 1.6, 95% CI 1.0–3.3); the OR for obese men was 1.1 (95% CI 0.6–1.8) (Table 2).

Table 2. ORs of developing RA when comparing overweight and obese to normal weight according to the World Health Organization classification, stratified by ACPA status and sex*

	No. of exposed cases/controls	OR (95% CI)†	OR (95% CI)‡
* OR = odds ratio; RA = rheumatoid arthritis; ACPA = anti–citrullinated protein antibody; 95% CI = 95% confidence interval; BMI = body mass index. † Logistic regression model conditional on applicable sex, age, and area of residence. ‡ Logistic regression model conditional on applicable sex, age, and area of residence and adjusted for smoking (ever/never), alcohol consumption (ever/never during the past 12 months), and education (university education; yes/no).
All
ACPA positive
BMI <25 kg/m^2	973/1,815	1.0 (–)	1.0 (–)
BMI ≥25 to <30 kg/m^2	572/1,183	0.9 (0.8–1.1)	0.9 (0.7–1.0)
BMI ≥30 kg/m^2	216/446	0.9 (0.7–1.1)	0.8 (0.7–1.0)
ACPA negative
BMI <25 kg/m^2	475/1,815	1.0 (–)	1.0 (–)
BMI ≥25 to <30 kg/m^2	351/1,183	1.1 (0.9–1.3)	1.1 (0.9–1.3)
BMI ≥30 kg/m^2	161/446	1.5 (1.1–1.9)	1.4 (1.1–1.9)
Women
ACPA-positive RA
BMI <25 kg/m^2	744/1,455	1.0 (–)	1.0 (–)
BMI ≥25 to <30 kg/m^2	355/708	1.0 (0.9–1.2)	1.0 (0.8–1.2)
BMI ≥30 kg/m^2	159/305	1.0 (0.8–1.3)	1.0 (0.8–1.2)
ACPA-negative RA
BMI <25 kg/m^2	379/1,455	1.0 (–)	1.0 (–)
BMI ≥25 to <30 kg/m^2	208/708	1.1 (0.9–1.4)	1.1 (0.7–1.6)
BMI ≥30 kg/m^2	117/305	1.6 (1.2–2.2)	1.6 (1.2–2.2)
Men
ACPA-positive RA
BMI <25 kg/m^2	229/360	1.0
BMI ≥25 to <30 kg/m^2	217/475	0.8 (0.6–1.0)	1.1 (0.7–1.8)
BMI ≥30 kg/m^2	57/141	0.7 (0.5–1.0)	0.6 (0.3–0.9)
ACPA-negative RA
BMI <25 kg/m^2	96/360	1.0 (–)	1.0 (–)
BMI ≥25 to <30 kg/m^2	143/475	1.0 (0.7–1.5)	1.0 (0.7–1.4)
BMI ≥30 kg/m^2	44/141	1.2 (0.7–1.9)	1.1 (0.6–1.8)

Analysis of ACPA-positive RA among men revealed an inverse association with BMI. The OR for BMI ≥30 kg/m² compared to normal weight was 0.6 (95% CI 0.3–0.9) (Table 2), whereas there was no association between BMI and ACPA-positive RA in women (OR 1.0, 95% CI 0.8–1.2).

Having a university degree or not and alcohol consumption were the only factors that substantially (>10%) altered the OR, but since smoking is associated with both RA and BMI, it was included in the model (28, 29). Additionally adjusting the analysis for civil status, being born in Sweden, physical activity, occupational class (as an alternative to educational level), parity, or history of oral contraceptive use and fatty fish consumption did not change the results. Stratification for having reached menopause age and the presence of any HLA–DRB1 SE alleles did not result in meaningful changes to the results (data not shown).

Symptom duration.

We stratified the analyses by duration from symptom onset to diagnosis in 3 categories (≤180 days, >180 to ≤365 days, and >365 days). The strength of the association increased slightly in ACPA-positive men (OR 0.8, 95% CI 0.5–1.3 for symptom duration of 180 days to 365 days compared to no stratification for symptom duration), and remained unchanged for ACPA-negative women when stratifying for symptom duration (OR 1.6, 95% CI 0.8–3.2).

DISCUSSION

In this large population-based study of incident RA, obesity was significantly associated with developing ACPA-negative RA in women. In contrast, we found a significant inverse association among men between BMI and ACPA-positive RA that was not seen among women.

This study was relatively large and had extensive information about various known and potential risk factors for RA that could be tested as confounders. All of the cases had received a first-time diagnosis of RA and fulfilled the ACR criteria as assessed by a rheumatologist. Available information on ACPA status allowed separate analyses of the 2 distinct subgroups of RA.

A case–control study, where information of exposures is collected retrospectively, has a common potential disadvantage due to the risk of information bias when cases and controls recall information on exposures differently. However, although recall of current height and weight may not be perfect, there is no reason to believe that it would be systematically different between RA cases and controls.

Weight loss might appear as an early symptom of RA, which could affect the results of our study (30, 31). If individuals with RA lose weight due to their disease before they are diagnosed, it may appear as if higher weight was protective. If such a bias would exist, it could contribute to a spurious finding of a diminished risk for ACPA-positive RA among men, whereas it would weaken the main result of our study, i.e., the higher risk for ACPA-negative RA in obese women. In an attempt to identify this type of bias, we stratified the analyses by duration from symptom onset to diagnosis. Although the strength of the association between BMI and ACPA-positive RA among men seemed to increase slightly with duration, the CIs were completely overlapping and therefore not lending substantive support for this type of bias.

The association between BMI and RA has previously been studied. The majority of previous studies did not find any significant association between BMI and RA overall (11–15). The most vital difference between these studies and the current study is that we looked at RA in the 2 major subsets and not only in RA overall.

In the Nurses' Health Study (14) and a Norwegian study (13), nonsignificant tendencies toward an association between RA and obesity were found. A study from the UK reported an increased risk of developing RA among premenopausal women with a BMI of >25.83 kg/m² (17).

Although the sample sizes of these studies were relatively small, it would be of interest to reanalyze them for effects of obesity/overweight on ACPA-positive and ACPA-negative RA. The only study that did investigate BMI in the 2 subgroups identified by ACPA status indeed found a similar association as in our study for ACPA-negative RA (18).

In order to exclude the possibility that the observed association we see between BMI and RA is partly explained by confounding from lifestyle, we adjusted the analysis for civil status, being born in Sweden, physical activity, occupational class, and fatty fish consumption, but none of these factors turned out to be a confounder.

Under the hypothesis that BMI-related RA pathogenesis may be mediated by hormones, we stratified by postmenopausal age, use of oral contraceptives, and parity, but found no modification of the association between BMI and RA.

The differences we see in men and women could be associated with fat distribution/composition and the biochemical consequences thereof, since fat distribution is known to differ between women and men (32). Unfortunately, we did not collect measures of waist circumference or waist to hip ratio and cannot explore this hypothesis further.

There is no known or obvious biologic explanation for the association between certain subsets of RA and BMI. There is a general association between excess body fat (white adipose tissue) and low-grade inflammation as visualized from increased C-reactive protein levels and the fact that adipose tissue in humans expresses proinflammatory cytokines (33). Baseline levels of adipokines have been shown to predict radiographic progression in early RA (34), but it is not known whether adipokines also contribute to the initiation of arthritis. Specifically, leptin is an adipokine that is increased in obese individuals and that is also involved in inflammatory processes. This fact may be of specific relevance in relation to our results in women, since leptin levels are higher in women as compared to men in certain contexts, and may therefore be involved in mechanisms related to inflammation that display sex differences and are related to obesity. Therefore, the relationship between leptin levels and RA needs further investigation.

In conclusion, our study shows that obesity is associated with developing ACPA-negative RA in women, and indicates an inverse association between BMI and ACPA-positive RA in men. Based on current knowledge of the pathogenesis of RA, we cannot further explain these findings. We know that ACPA status defines 2 etiologically different RA subsets that need to be investigated separately, but an RA risk factor having opposing effects in women and men was unexpected and needs corroboration.

AUTHOR CONTRIBUTIONS

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. Ms Wesley 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. Wesley, Klareskog, Alfredsson, Wedrén.

Acquisition of data. Wesley, Bengtsson, Klareskog, Alfredsson, Wedrén.

Analysis and interpretation of data. Wesley, Bengtsson, Elkan, Alfredsson, Wedrén.

Acknowledgements

The authors thank Marie-Louise Serra and Lena Nise for their assistance in the data collection.