Abstract
Objective. Obesity and overweight are increasing conditions. Adipose tissue with proinflammatory properties could be involved in rheumatoid arthritis (RA) activity and radiographic progression. This study aims to investigate the influence of overweight and obesity on RA activity and severity.
Methods. We conducted a systematic review and metaanalysis to assess the association of body mass index (BMI) categories with the Disease Activity Score in 28 joints (DAS28), functional disability [Health Assessment Questionnaire (HAQ)], and radiographic joint damage in patients with RA. We searched Medline through PubMed, EMBASE, and the Cochrane Database of Systematic Reviews for all studies assessing DAS28, HAQ, or/and radiographic damage according to predefined BMI groups.
Results. Among the 737 citations retrieved, 58 articles met the inclusion criteria and 7 were included in the metaanalysis. DAS28 was higher in obese (BMI > 30 kg/m2) than non-obese (BMI ≤ 30 kg/m2) patients (mean difference 0.14, 95% CI 0.01–0.27, p = 0.04, I2 = 0%). HAQ score was also higher among obese patients (mean difference 0.10, 95% CI 0.01–0.19, p = 0.03, I2 = 0%). Radiographic joint damage was negatively associated with obesity (standardized mean difference −0.15, 95% CI −0.29 to −0.02, p = 0.03, I2 = 38%).
Conclusion. Obesity in RA is associated with increased DAS28 and HAQ score and with lower radiographic joint damage. These associations mainly result from an increase of subjective components of the DAS28 (total joint count and global health assessment) in obese patients. Conflicting results were reported concerning inflammation markers (C-reactive protein and erythrocyte sedimentation rate).
About one-third of the world’s population is estimated to be overweight [body mass index (BMI) > 25 kg/m2)] or obese (BMI > 30 kg/m2)1. Among people with rheumatoid arthritis (RA), this proportion is higher by more than 50%2.
This problem of obesity and the resulting metabolic syndrome in patients with RА has attracted research attention3. Indeed, adipose tissue has immunomodulating and proinflammatory properties that could affect inflammatory diseases such as RA4. Among the cytokines produced by adipose tissue, tumor necrosis factor (TNF), interleukin 6 (IL-6), leptin, resistin, and visfatin are considered proinflammatory, whereas adiponectin has antiinflammatory and proinflammatory properties, depending on its molecular form5.
Serum IL-6, visfatin, and adiponectin levels in patients with RA were found to be positively associated with radiographic joint damage in cross-sectional and longitudinal studies6,7. Moreover, some studies have found that a high BMI was associated with poor RA disease outcome, whereas others have suggested that increased BMI may be associated with less radiographic joint damage8,9.
Thus, the effect of obesity among patients with RA has to be clarified.
We performed a systematic review and metaanalysis to assess the cross-sectional association between obesity and disease activity, functional disability, and radiographic joint damage in patients with RA in observational studies.
MATERIALS AND METHODS
Literature search
We searched Medline through PubMed, EMBASE, and the Cochrane Database of Systematic Reviews for articles published up to June 22, 2015, with the MeSH terms “body mass index” OR “obese” OR “obesity” AND “rheumatoid arthritis”, without any limits. We also performed a manual search, but no congress abstracts [European League Against Rheumatism (EULAR) and American College of Rheumatology (ACR)] were retrieved because we needed exhaustive data for our metaanalysis. Using our prespecified inclusion and exclusion criteria, 2 independent rheumatologist reviewers examined titles and abstracts for potential relevance and then chose full-text articles. Any discrepancies were resolved by discussion.
Study selection
Inclusion criteria were patients with a diagnosis of RA according to the 1987 ACR criteria (score ≥ 4/7 with symptoms persisting for more than 6 weeks) and/or the 2010 ACR/EULAR criteria (score ≥ 6/10) and predefined BMI groups, including a group of overweight/obese patients10,11. We searched for cohort, longitudinal, and cross-sectional studies. Studies with a group of patients with BMI ≥ 30 kg/m2 were included in the metaanalysis; studies with other cutoffs were included in the systematic review. Studies with results expressed in means were included in the metaanalysis; studies with results expressed in medians were included in the systematic review.
We excluded reviews and case reports. Articles with data expressed as medians were excluded from the metaanalysis, but were retained for the systematic review. Manual search was performed using references of the selected studies.
Risk of bias assessment
We used the Quality in Prognostic Studies tool to evaluate the quality of studies included in the metaanalysis because they were prospective studies12.
Data collected
We collected data on the 28-joint count Disease Activity Score (DAS28) and/or its components [tender joint count (TJC), swollen joint count (SJC), global health assessment (GHA; visual analog scale from 0–100), erythrocyte sedimentation rate (ESR), C-reactive protein level (CRP)], Health Assessment Questionnaire (HAQ) score, and patient’s global pain assessment (GPA; visual analog scale from 0–100). We collected radiographic joint damage scored by Ratingen, Simple Erosion Narrowing Score (SENS), or modified Sharp/van der Heijde (SvdH) scales13,14.All these assessments had to be analyzed by BMI categories. For data collected in clinical trials or cohort studies, only baseline information was considered.
Statistical analysis
Results of each selected study were presented as mean differences with their 95% CI. Metaanalyses were carried out using the inverse variance method, pooling estimates of each study using fixed or random-effects model according to the level and significance of heterogeneity. Heterogeneity was tested with the Cochran Q-test and evaluated by the I2 statistic with the following classification: 0–30% indicating negligible heterogeneity, 30–50% moderate heterogeneity, 50–75% substantial heterogeneity, and considerable heterogeneity for values larger than 75%. For the Q-test, a p value < 0.10 was considered significant and a random-effects model was used. Standardized means were used for the analysis of radiographic joint damage because of the varied radiographic scores used among studies. Funnel plots and the Egger test assessed publication biases. Analysis involved the use of RevMan 5.1.6 (The Nordic Cochrane Centre, The Cochrane Collaboration, 2011) for the metaanalysis calculations, and STATA 13.1 software (StataCorp) was used to perform publication bias assessment. P values < 0.05 were considered statistically significant.
RESULTS
The literature search yielded 58 articles for the systematic review; 7 articles (3787 patients) were included in the metaanalysis (Figure 1).
Flowchart of articles selected for the metaanalysis. BMI: body mass index.
Description of studies retrieved
Studies were mainly prospective, conducted in various centers and countries, except for 3 retrospective studies15,16,17. The diagnosis of RA was according to the revised 1987 ACR criteria, except for 1 study, in which 63% of patients fulfilled the 2010 ACR/EULAR criteria and 42% fulfilled the revised 1987 ACR criteria18. In most of the studies, patients were divided by BMI in accordance with the World Health Organization definition or with a lower boundary of 20 kg/m2 for normal BMI, except in 5 studies19. One study defined obesity as BMI > 25 kg/m2 because the population was Asian and 1 divided BMI into 3 categories (< 20, 20–30, and > 30 kg/m2) to have large groups20,21,22. One study defined normal BMI as 18.5–22.9 kg/m2 and obesity as > 28 kg/m2 because patients with RA may have decreased fat-free mass because of inflammation and increased fat mass secondary to inactivity, thereby suggesting that BMI cutoffs in the RA population may be more appropriate if they were reduced by about 2 kg/m223,24. In 1 study, obese patients were compared with only non-obese patients (BMI < 30 kg/m2) and in others, underweight and normal weight patients were included in the same group (BMI < 25 kg/m2)15,16,17,25,26.
DAS28 assessment was detailed in 6 studies, with TJC and SJC assessed by experienced rheumatologists or nurses18,20,27,28,29,30. Radiographic evaluation was detailed in 3 studies. One study used the mean scores of 2 blinded centralized readers31. In another study, physicians who scored the radiographs were unaware of the clinical data, and the BeSt study (Behandel Strategieën, i.e., Treatment Strategies Study) used the mean score from the 2 readers29. Another study used the radiograph scores from 1 investigator30.
Risk of bias assessment
Articles were rated low bias considering the population studied, except for 1 in which patients were not randomized32. The evaluation of study attrition was not applicable for 3 studies because they were not designed to evaluate RA activity and severity by BMI31,32,33. The remaining 4 studies had low to moderate bias25,29,30,34. All studies were rated low bias for prognostic factor measurement and study confounding. The outcome measurement was also rated low bias, except for 2 studies rated as moderate bias because the DAS28 assessment method was not well described25,33. Studies were rated as low bias because of the statistical analysis and results reported, except for 1 study rated moderate bias because the statistical analysis was not well explained25.
Study characteristics
Table 1 shows the characteristics of the 7 studies included in the metaanalysis. Mean age of patients ranged from 47.6 ± 13.6 to 63.5 ± 12.3 years. Patients were mostly women, representing 67% to 100% of the patients. Mean duration of RA ranged from 25 weeks (range 15–45) to 15.5 ± 12.3 years. Smoking was assessed in 3 studies and no clear difference was found between obese and normal BMI patients29,31,33. Obese patients more often had seronegative RA [rheumatoid factor (RF) and anticyclic citrullinated peptide antibodies] than patients with normal BMI29,30,31. There were no differences for disease-modifying anti-rheumatic drug (DMARD) and corticosteroid use. Patients were classified in different groups by BMI: low BMI (< 18.5 or < 20 kg/m2), normal BMI (18.5–25 or 20–25 kg/m2), overweight (25–30 kg/m2), and obese (> 30 kg/m2).
Characteristics of studies included in the metaanalysis. Values are % unless otherwise specified.
DAS28
DAS28 according to BMI was reported in 15 studies; 4 studies involving 1402 patients were included in the metaanalysis because they showed similar BMI groups and comparable expression of DAS28 (means and not medians)31,32,33,34. DAS28 was higher for obese than non-obese patients (mean difference 0.14, 95% CI 0.01–0.27, p = 0.04; Figure 2), with no heterogeneity among studies (I2 = 0%). Among the 11 remaining studies, 9 supported this result for a trend of higher DAS28 in obese patients than those with normal BMI15,17,18,20,22,23,28,30,35 (Table 2)15,16,17,18,20,22,23,27,28,29,30,31,32,33,35,36. Two studies showed conflicting results, but in 1 study, obese and overweight patients were in the same BMI group (> 25 kg/m2), and the second study was a retrospective study16,36.
Forest plot of the effect of BMI (obesity ≥ 30 kg/m2 vs normal weight 18.5–25 or 20–25 kg/m2) on DAS28. BMI: body mass index; DAS28: Disease Activity Score in 28 joints; IV: independent variable.
Systematic review of the literature for DAS28, ESR, CRP, TJC, SJC, and GPA, and GHA. GPA was evaluated by VAS from 0 (no pain) to 100 (worst imaginable pain). GHA was evaluated by VAS from 0 (best imaginable health state) to 100 (worst imaginable health state). Values are mean (SD) or median (IQR).
TJC and SJC
Metaanalysis was not feasible for the analysis of TJC or SJC because of different BMI groups and different measurement tools (means or medians). TJC was assessed in 6 studies, which showed a trend of higher TJC for obese patients than those with normal BMI (Table 2)17,20,27,28,29,31. SJC was assessed in 5 studies showing conflicting results (Table 2)17,20,27,28,31.
GPA and GHA
Metaanalysis was not feasible for GPA or GHA because of different BMI groups and different measurement tools (means or medians). Six and 3 studies reported GPA and GHA, respectively, by BMI groups, revealing a trend of increased GPA and GHA for obese patients over those with normal BMI (Table 2)16,20,28,29,30,31,33.
ESR and CRP level
Metaanalysis was not feasible for ESR or CRP because of lack of comparable BMI groups and different measurement tools (means or medians). The association of ESR and BMI was assessed in 12 studies showing conflicting results (Table 2)15,16,17,20,22,23,27,28,31,32,33,36. The association of CRP level and BMI was assessed in 12 studies showing conflicting results, but a trend of lower CRP in obese patients than those with normal BMI (Table 2)15,16,17,20,22,23,29,30,31,33,35,36.
Functional disability: HAQ score
HAQ score was assessed by BMI in 5 studies, 2 involving 1264 patients who were included in the metaanalysis because the studies involved similar BMI groups and comparable expression of HAQ (means and not medians)31,33. HAQ score was higher for obese than non-obese patients (mean difference 0.10, 95% CI 0.01–0.19, p = 0.03; Figure 3), with no heterogeneity among studies (I2 = 0%). Among the remaining 3 studies, 1 found a median HAQ score of 0.75 [interquartile range (IQR) 0.250–1.375], 0.875 (0.312–1.562), and 1.375 (0.625–1.870) for patients with BMI < 25 kg/m2, 30–35 kg/m2, and > 35 kg/m2, respectively18. Another study found a median HAQ score of 1.46 (IQR 0.39–2) and 2.25 (1.71–2.86) for patients with BMI 19.5–22.9 kg/m2 (considered normal BMI for patients with RA in this study) and > 28 kg/m2 (considered obese), respectively23. The last study reported a median HAQ score of 1.5 (1.4), 1.5 (1.5), and 1.75 (1.5) for patients with BMI < 25 kg/m2, 25–30 kg/m2, and > 30 kg/m2, respectively15. These findings suggested higher HAQ score for obese patients than for those with normal BMI.
Forest plot of the effect of BMI (obesity ≥ 30 kg/m2 vs normal weight 18.5–25 or 20–25 kg/m2) on HAQ score. BMI: body mass index; HAQ: Health Assessment Questionnaire; IV: independent variable.
Radiographic joint damage
Radiographic joint damage at baseline was assessed by BMI in 8 studies, 3 involving 1465 patients were included in the metaanalysis because the studies involved similar BMI groups and comparable expression of radiographic joint damage (means and not medians)25,29,30. Radiographic joint damage was lower among obese than non-obese patients (standardized mean difference −0.15, 95% CI −0.29 to −0.02, p = 0.03; Figure 4). Heterogeneity was acceptable with I2 = 38%. Among the remaining 5 studies, 1 found a median modified SvdH score of 4.5 (IQR 1.5–11), 5.5 (2.5–20), and 7 (2–27.5) for BMI ≥ 30 kg/m2 (obesity), 25–29.9 kg/m2 (overweight), and 20–24.9 kg/m2, respectively31. Another study reported a median modified SvdH score of 5.5 (1.5–18), 7 (2.5–25), and 12 (2.5–46.5) for patients with BMI ≥ 30 kg/m2, 25–30 kg/m2, and < 25 kg/m2, respectively26. The third study showed less erosive RA among obese patients than patients with BMI 20–30 kg/m2 (53% vs 85%)22. The fourth study, supporting the finding of lower radiographic joint damage in obese patients than non-obese patients, found a mean SENS of 11.27 ± 7.40 and 16.05 ± 8.34 for patients with BMI > 25 kg/m2 and 18.5–24.9 kg/m2, respectively37. The last study had conflicting results33.
Forest plot of the effect of BMI (obesity ≥ 30 kg/m2 vs non-obese < 30 kg/m2) on radiographic joint damage (standardized means). BMI: body mass index; IV: independent variable.
Analysis with the overweight group
Analyses were made comparing the normal BMI group with the overweight and the obese groups (BMI > 25 kg/m2), but no significant results were found, although they showed a trend toward higher DAS28 and HAQ and lower radiographic joint damage among the overweight and obese groups. The forest plots are in the Supplementary Data (available online at jrheum.org).
Publication bias
Publication bias was not assessed for HAQ because only 2 studies were included. We found no publication bias for DAS28 (Egger test p = 0.294) or radiographic joint damage (Egger test p = 0.716). The funnel plots are in the Supplementary Data (available online at jrheum.org).
DISCUSSION
Our systematic review with metaanalysis found significantly higher RA disease activity in obese than non-obese patients. As well, obesity in RA was associated with increased HAQ score. Conversely, radiographic joint damage was lower for obese than non-obese patients.
This increase of 0.14 in DAS28 for obese patients with RA may seem low and with a small clinical relevance, but this result was statistically significant with no heterogeneity among selected studies. If all articles could have been included in the metaanalysis, we probably would have found a higher difference and a stronger association. Indeed, studies included in our systematic review showed increased DAS28 among obese patients, between 0.1 and 1.6.
To better understand why DAS28 was increased in obese patients with RA, we evaluated each of its components. Increased DAS28 could result from an overestimation of SJC among obese patients because of evaluation difficulties. However, our systematic review did not reveal an association of SJC and obesity. Moreover, a study found that SJC predicted RA disease activity and severity in obese subjects at least as well as in subjects with lower BMI38. Another cause for the increased DAS28 could be the increased levels of inflammation markers among obese patients independent of RA because of the production of IL-6 by adipose tissue, which stimulates hepatic secretion of CRP39. However, our systematic review revealed lower CRP levels in obese patients than others. The increased DAS28 may mainly result from increased scores for subjective measures such as TJC, GPA, and GHA. Indeed, our systematic review revealed higher TJC, GPA, and GHA in obese than non-obese patients. Moreover, obesity is known to be associated with pain and reduced quality of life40. Therefore, the increase in subjective measure scores may be explained by comorbidities of obese patients, independent of RA features. Mechanisms of this association remain to be explored in prospective studies to evaluate the respective roles of subjective measures and adipokines in increased DAS28.
The association of obesity and increased HAQ score in RA revealed by our metaanalysis may seem irrelevant because the difference was 0.1 and only 2 studies were included. However, we may have found a stronger association if all articles had been included in the metaanalysis. Indeed, the other 3 studies in the systematic review found an increased HAQ score of 0.25 to 0.79 among obese patients. This association could reveal greater severity of RA in obese than non-obese patients, but as demonstrated before, those patients had higher subjective measure scores. Therefore, this association may result from increased joint pain and comorbidities in obese patients. Indeed, several studies revealed an association of increased BMI and impaired health-related quality of life41. The association of HAQ score and obesity among patients with RA may be associated with obesity, independent of RA course. Obese patients also seem to be less responsive to treatment. Indeed, 1 study showed lower response to DMARD among obese than non-obese patients42. Others found lower remission rate with anti-TNF-α therapy, particularly infliximab (IFX)16,22,43. However, 1 recent retrospective study reported conflicting results with tocilizumab (TCZ)17. This finding could also be explained by increased DAS28, independent of RA course. Therefore, DAS28 may not be the best way to evaluate therapeutic response in obese patients, and objective measures such as ultrasonography could be considered in some cases.
The lower radiographic joint damage among obese compared with non-obese patients with RA could be considered as a paradoxical effect in considering the results for DAS28 and HAQ score. As seen before, the association of obesity with increased DAS28 and impaired HAQ score results from subjective measures and may not reflect the exact disease severity. Obese patients may have lower radiographic joint damage and a slower radiographic joint damage progression as reported in 1 study, for early RA after 3 years of followup, but other longitudinal studies assessing radiographic progression among the obese are needed30. This favorable effect on radiographic joint damage may have different explanations. First of all, obese patients are more likely to develop seronegative RA, which is known to be associated with a better structural prognosis. However, in 3 studies, radiographic damage and progression remained significantly lower in obese patients after adjustment for anticitrullinated protein antibodies (ACPA) status and for ACPA and RF status25,30,31. Then, adiponectin is decreased in obese patients and is associated with radiographic damage44,45. This may be explained by a proinflammatory effect of adiponectin on fibroblast-like synoviocytes with an increase of vascular endothelial growth factor, matrix metalloproteinase (MMP) 1, and MMP-13 production46. Thus, obese patients with RA may have a lower level of circulating adiponectin that may protect against radiographic joint damage. Indeed, several studies found an association of radiographic disease progression and increased adiponectin level7,47. Another explanation could be a more intensive therapy because of high DAS28 levels and/or high plasmatic drug concentrations for treatment with dosage adapted to the weight, such as IFX, abatacept, and TCZ. Finally, it could also result from an inverse correlation between BMI and osteoporosis because patients with osteoporosis present with more radiographic joint damage attributable to common pathophysiological mechanisms48,49. This might explain in part the increased risk of seronegative RA, the lower radiographic joint damage, and the poorer therapeutic response9,16,22,31,43,50.
The main limitation of our metaanalysis is the lack of comparable studies, which led to the inclusion of only 7 articles, whereas more studies were available in the literature. Indeed, data for BMI groups and measurements (means or medians) were often different. Another limitation is the difference of obesity and overweight BMI intervals between studies included in our metaanalysis. However, almost all studies included in our systematic review found the same trends.
Obesity is associated with increased disease activity and HAQ score, but reduced radiographic joint damage in RA. Further studies are needed to reveal the involved mechanisms and to evaluate the effect of obesity on RA course.
ONLINE SUPPLEMENT
Supplementary data for this article are available online at jrheum.org.
Acknowledgment
We thank the AbbVie company for providing us an educational program on reviews and metaanalysis.
- Accepted for publication August 13, 2015.
REFERENCES
In this issue
