Obesity protects against radiographic joint damage in rheumatoid arthritis (RA) through poorly defined mechanisms. Adipocytokines are produced in adipose tissue and modulate inflammatory responses and radiographic joint damage in animal models. The purpose of this study was to examine the hypothesis that adipocytokines modulate inflammation and radiographic joint damage in patients with RA.
We compared serum concentrations of leptin, resistin, adiponectin, and visfatin in 167 RA patients and 91 control subjects. The independent association between adipocytokines and body mass index (BMI), measures of inflammation (C-reactive protein [CRP], interleukin-6 [IL-6], and tumor necrosis factor α [TNFα]), and radiographic joint damage (Larsen score; n = 93 patients) was examined in RA patients by multivariable regression analysis first controlling for age, race, and sex, and then for obesity (BMI) and inflammation (TNFα, IL-6, and CRP).
Concentrations of all adipocytokines were significantly higher in RA patients than in controls; for visfatin and adiponectin, this association remained significant after adjusting for BMI, inflammation, or both. Visfatin concentrations were associated with higher Larsen scores, and this association remained significant after adjustment for age, race, sex, disease duration, BMI, and inflammation (odds ratio [OR] 2.38 [95% confidence interval (95% CI) 1.32–4.29], P = 0.004). Leptin concentrations showed a positive association with the BMI (ρ = 0.58, P < 0.01) and showed a negative association with the Larsen score after adjustment for inflammation (OR 0.32 [95% CI 0.17–0.61], P < 0.001), but not after adjustment for BMI (OR 0.86 [95% CI 0.42–1.73], P = 0.67).
Concentrations of adipocytokines are increased in patients with RA and may modulate radiographic joint damage. Visfatin is associated with increased, and leptin with reduced, levels of radiographic joint damage.
Obesity is consistently associated with reduced levels of radiographic joint damage in patients with rheumatoid arthritis (RA) (1–3), but the underlying mechanisms are not known. Adipose tissue is a source of inflammatory cytokines, such as tumor necrosis factor α (TNFα) and interleukin-6 (IL-6) (4), and obesity is associated with a chronic inflammatory response. Therefore, the association between obesity and reduced radiographic joint damage appears to be paradoxical.
Adipose tissue is also a major source of several mediators, termed adipocytokines, that include leptin, resistin, adiponectin, and visfatin. They have profound effects not only on glucose homeostasis and appetite regulation, but also on inflammation (4). For example, leptin, resistin, and visfatin are proinflammatory (4), whereas adiponectin is antiinflammatory (5). Adipocytokines have diverse immunologic effects. Leptin, best known for its ability to regulate body weight by decreasing food intake and increasing energy expenditure, also has immunomodulatory effects. Leptin production is stimulated by inflammation and it is proinflammatory, inducing T cell activation and production of inflammatory mediators such as TNFα, IL-6, and nitric oxide (4, 6). Adiponectin transcription, on the other hand, is suppressed by inflammatory cytokines such as TNFα and IL-6 (4). Resistin, also known as FIZZ3 (7), induces TNFα and IL-6 (8). Visfatin (9), also known as pre–B cell colony-enhancing factor (10) or nicotinamide phosphoribosyltransferase (11), is produced by visceral adipose tissue as well as by other cells, such as macrophages and neutrophils. It is induced by inflammation and immune activation (12) and has immunomodulatory properties, including enhancement of B cell differentiation, induction of cytokines and matrix metalloproteinases (13), and inhibition of apoptosis (14).
Concentrations of adipocytokines have generally been reported to be higher in patients with RA than in control subjects (15–17). However, information regarding their relationship to inflammation and disease activity is conflicting (18, 19), and little is known about their role in radiographic joint damage. We examined the hypothesis that adipocytokines may affect inflammation and disease activity in RA and contribute to reduced radiographic joint damage in obesity.
PATIENTS AND METHODS
Patients and control subjects.
We studied 167 patients with RA and 91 control subjects who were frequency-matched for age, race, and sex. Study subjects were recruited through advertisements, referrals from local rheumatologists, or from a database of volunteers that is maintained by the General Clinical Research Center at Vanderbilt University. All study subjects were older than 18 years. Patients with RA fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) classification criteria for the diseases (20). The study sample was selected to be enriched with patients with relatively early RA or with established disease. Control subjects did not meet the classification criteria for RA and had no inflammatory disease. These subjects are part of an ongoing study, and enrollment procedures have previously been described in detail (21). This study was approved by the Vanderbilt University Institutional Review Board, and all subjects gave written informed consent.
Clinical and laboratory measurements.
Clinical information and laboratory data were obtained through a structured interview, self-reported questionnaires, physical examination, and blood tests. The degree of obesity was measured using the body mass index (BMI; in kg/m2). Disease activity was measured using the Disease Activity Score in 28 joints (DAS28) instrument (22). Functional capacity was measured using the modified Health Assessment Questionnaire (M-HAQ) (23), a standard 8-question instrument addressing activities of daily living.
Radiographs of the hands and feet were available for 93 patients. These had been obtained at a median of 1.9 years (interquartile range [IQR] 1.1–2.7) before the date of enrollment in this study. A single investigator (TS) who was blinded to the adipocytokine concentrations determined the degree of radiographic joint damage using the Larsen score, a standardized method for quantifying the amount of damage of the joints of the hands, wrists, and feet (24, 25). Twenty joints were evaluated: the wrists, the first to fifth metacarpophalangeal joints, and the second to fifth metatarsophalangeal joints. The total score ranged from 0 to 100.
Levels of C-reactive protein (CRP) and rheumatoid factor (RF; measured in 159 patients) were measured by the Vanderbilt University Medical Center Clinical Laboratory; CRP concentrations were measured by enzyme-linked immunosorbent assay (ELISA; Linco Research, St. Charles, MO) in all control subjects and in RA patients studied before 2003 with levels <3 mg/ml. Serum concentrations of TNFα, IL-6, leptin, resistin, and adiponectin were measured by multiplex ELISA (LincoPlex Multiplex Immunoassay kit; Linco Research). Visfatin was measured with a Visfatin C-terminal ELISA kit (Phoenix Pharmaceuticals, Burlingame, CA).
Descriptive statistics were calculated as the mean ± SD or as the median and IQR according to the distributions of the continuous variables. Concentrations of each adipocytokine were compared in RA patients and control subjects as well as in RF-positive and RF-negative RA patients using Wilcoxon's rank sum test. We assessed the relationship between adipocytokines and obesity (as indicated by the BMI), disease activity and damage (as indicated by the DAS28, M-HAQ, and Larsen scores), and inflammation (as indicated by the TNFα, IL-6, and CRP levels) using Spearman's rank correlation test.
The association between RA and each adipocytokine was assessed using linear regression models, with the concentration of each adipocytokine as the outcome variable and disease status (RA or control) as the predictor variable. Since adipocytokines are associated with both obesity and inflammation, we defined 4 models a priori in order to examine the contribution of each component separately and in combination. The base model was adjusted for age, sex, and race. In addition to the base model, we controlled for obesity (base plus BMI) and inflammation (base plus TNFα, IL-6, and CRP) separately, and in combination (full model, representing the base plus BMI plus inflammation). With logarithmic (log10) transformation of the outcome variable (adipocytokine concentration), the exponentiated regression coefficient (10b) for disease status can be interpreted as the average value of the serum adipocytokine concentration in the RA patients relative to that in the controls. For example, 10b = 1.2 indicates a 20% increase in the mean serum adipocytokine concentration among patients with RA as compared with controls.
The association between adipocytokine levels and the Larsen score was tested by applying a proportional odds logistic regression model because the Larsen score was heavily skewed. The approach to controlling for confounders was similar to that described above, using the same 4 models (base model, base plus BMI, base plus inflammation, and base plus BMI plus inflammation); however, since disease duration is a strong predictor of the Larsen score (26), it was included in the base model for these analyses (age, sex, race, and disease duration). For all models, each adipocytokine was logarithmically transformed to ensure linearity and to stabilize variance. The effect sizes are shown as the odds ratio (OR) per IQR difference with the 95% confidence interval (95% CI). Additional sensitivity analyses were performed to assess the effect of adipocytokines on the Larsen score among RA patients, with the inclusion of a further set of a priori–selected factors known to be associated with the Larsen score. Covariates included the time between the date of enrollment and the date of the radiograph, smoking status, DAS28, M-HAQ score, RF status, and duration of exposure to corticosteroids, methotrexate, leflunomide, hydroxychloroquine, and TNFα inhibitors. Data reduction, applying principal components of correlated variables for markers of inflammation, medication use, and disease severity, were used to minimize overfitting of the model. The distribution of variables in patients with and without available radiographs was compared using Wilcoxon's rank sum test for continuous variables and chi-square test for categorical variables.
Statistical significance was determined using a 2-sided 5% significance level (P < 0.05). Statistical analysis was performed using the R program, version 2.6.2 (http://www.r-project.org).
The characteristics of patients with RA and control subjects and the concentrations of adipocytokines in the 2 groups are shown in Table 1. Concentrations of leptin, resistin, adiponectin, and visfatin were significantly higher in patients with RA than in control subjects. In RA patients, leptin concentrations were lower in smokers than in nonsmokers (P = 0.009), but the concentrations of the other adipocytokines and the DAS28 and Larsen scores did not differ significantly between smokers and nonsmokers (P > 0.05 for all comparisons). Visfatin concentrations were significantly higher in patients with seropositive RA (n = 114; median 6.4 ng/ml [IQR 5.2–9.3]) than in patients with seronegative RA (n = 45; median 5.3 ng/ml [IQR 3.8–7.0]) (P < 0.001). Concentrations of leptin, resistin, and adiponectin did not differ significantly among patients with seropositive and seronegative RA.
Table 1. Clinical characteristics and adipocytokine concentrations in the control and RA groups*
Controls (n = 91)
RA patients (n = 167)
P values were determined by Wilcoxon's rank sum test or chi-square test. RA = rheumatoid arthritis; BMI = body mass index; IQR = interquartile range; DAS28 = Disease Activity Score 28-joint assessment; M-HAQ = modified Health Assessment Questionnaire; CRP = C-reactive protein; TNFα = tumor necrosis factor α; IL-6 = interleukin-6.
Rheumatoid factor (RF) was measured in 159 patients.
When only the patients with available radiographs were considered, the median disease duration was 5.0 years (IQR 2.0–20.0) at the time blood was drawn and 2.7 years (IQR 0.58–17.9) at the time the radiograph was taken. The clinical characteristics in the 93 patients for whom radiographs were available are shown in Table 2 and are categorized according to evidence of radiographic joint damage (Larsen score >0) or no evidence of radiographic joint damage (Larsen score 0). Patients with radiographic joint damage were older, had longer disease duration, and had higher concentrations of IL-6 and visfatin.
Table 2. Clinical variables and adipocytokine concentrations in the 93 RA patients with available radiographs, according to the presence or absence of radiographic joint damage*
No radiographic joint damage (n = 44)
Radiographic joint damage (n = 49)
Radiographic joint damage was defined as a Larsen score >0. P values were determined by Wilcoxon's rank sum test or chi-square test. RA = rheumatoid arthritis; BMI = body mass index; IQR = interquartile range; RF = rheumatoid factor; DAS28 = Disease Activity Score 28-joint assessment; M-HAQ = modified Health Assessment Questionnaire; CRP = C-reactive protein; TNFα = tumor necrosis factor α; IL-6 = interleukin-6.
Age, mean ± SD years
53.3 ± 10.5
58.5 ± 11.6
Sex, % male
Race, % white
BMI, mean ± SD kg/m2
28.3 ± 5.8
26.8 ± 5.0
Current smokers, %
Disease duration, median (IQR) years
RF, % positive
DAS28, median (IQR)
M-HAQ score, median (IQR)
CRP, median (IQR) mg/liter
TNFα, median (IQR) pg/ml
IL-6, median (IQR) pg/ml
Current medications, %
Adipocytokines, median (IQR)
The Spearman correlations between adipocytokines, BMI, and measures of inflammation and disease activity are shown in Table 3. The BMI correlated positively with leptin and negatively with adiponectin, but was not correlated with resistin or visfatin. Consistent with the univariate correlations with BMI as a continuous variable, when patients with RA were categorized according to BMI subgroups as normal weight (BMI <25 kg/m2; n = 54), overweight (25 ≤ BMI < 30 kg/m2; n = 42), and obese (BMI ≥30 kg/m2; n = 71), the leptin concentration was significantly associated with higher BMI and the adiponectin concentration with lower BMI (P < 0.001 for both comparisons, by Kruskal-Wallis test). Resistin and visfatin concentrations did not differ significantly across BMI categories.
Table 3. Correlation (ρ) between adipocytokines and clinical variables in patients with rheumatoid arthritis*
Adiponectin and visfatin concentrations correlated significantly with disease duration. With regard to mediators of inflammation and indices of RA (Table 3), the most striking associations were observed with visfatin, which showed a positive correlation with the TNFα, IL-6, and CRP concentrations, the neutrophil count, and the M-HAQ and Larsen scores. Leptin was positively correlated with the DAS28, the IL-6 level, and the CRP concentration. As expected, adiponectin and leptin were significantly inversely correlated (ρ = –0.20, P = 0.01), but there were no statistically significant correlations between the other adipocytokines (P > 0.05).
Analyses performed to examine whether higher concentrations of individual adipocytokines associated with RA were accounted for by BMI or inflammation are shown in Figure 1. Visfatin and adiponectin remained significantly higher in RA patients after adjustment for age, race, and sex (base model), as well as after additional adjustment for BMI, inflammation, or both. The higher leptin concentrations in RA patients remained statistically significant after adjustment for BMI, but not after adjustment for inflammation or for both BMI and inflammation. Resistin was significantly higher in RA in the base model and after additional adjustment for BMI, but not after adjustment for inflammation or for both BMI and inflammation.
The associations between the Larsen score and the adipocytokine levels are shown in Figure 2. Visfatin was consistently associated with a higher Larsen score across all 4 models (base, base plus BMI, base plus inflammation, and base plus BMI plus inflammation), with an OR of 2.38 (95% CI 1.32–4.29) (P = 0.004) for the full model (i.e., base plus BMI plus inflammation). In contrast, leptin was marginally associated with a lower Larsen score (OR 0.58 [95% CI 0.33–1.01], P = 0.05 for the base model). This relationship was not significant after adjustment for the BMI (OR 0.86 [95% CI 0.42–1.73], P = 0.67), but was stronger after adjustment for inflammation (OR 0.32 [95% CI 0.17–0.61], P < 0.001). This negative relationship was attenuated when both BMI and inflammation were included in the model (OR 0.45 [95% CI 0.21–0.98], P = 0.04). Resistin was weakly associated with a higher Larsen score (OR 1.71 [95% CI 1.04–2.80], P = 0.03), and the association was maintained after adjustment for BMI (P = 0.02) and was attenuated by adjustment for inflammation (P = 0.07) and by adjustment for all variables in the fully adjusted model (P = 0.06). Adiponectin was marginally associated with a higher Larsen score (P = 0.05 for the base model), and this was attenuated by adjustments that included the BMI (P = 0.24 for the BMI model and P = 0.67 for the full model).
In a separate analysis, we examined the relationship between the BMI and the Larsen score. Consistent with the findings of previous studies, the BMI was significantly associated with a lower Larsen score (OR 0.39 [95% CI 0.18–0.84], P = 0.02, adjusted for age, race, sex, and disease duration) even after additional adjustment for inflammation (OR 0.27 [95% CI 0.12–0.61], P = 0.002). Patients in whom radiographs were (n = 93) or were not (n = 74) available for determination of the Larsen score did not differ significantly with regard to race, sex, current smoking status, disease duration, RF seropositivity, the M-HAQ score, or the DAS28 score (P > 0.05 for all comparisons). Patients with radiographs tended to have a lower median BMI (27.3 kg/m2 versus 29.3 kg/m2 in those without radiographs; P = 0.002) and to be older (median age 55.0 years versus 52.0 years in those without radiographs; P = 0.047).
A sensitivity analysis that included many factors in addition to age, race, sex, BMI, inflammation, and disease duration that could affect the Larsen score (i.e., time between the date of enrollment and the date of the radiograph, smoking, DAS28, M-HAQ score, RF status, and duration of corticosteroid, methotrexate, hydroxychloroquine, leflunomide, and TNFα inhibitor exposure) yielded almost identical findings. Visfatin was significantly associated with a higher Larsen score (P = 0.005), while leptin was marginally associated with a lower Larsen score (P = 0.045). This latter association was more significant when the BMI was not included in the model (P = 0.001), which is consistent with the findings of the primary analysis. Adiponectin was not associated with the Larsen score (P = 0.5), while resistin was marginally associated (P = 0.07), again consistent with the findings of the primary analysis.
There were several important findings of this study. First, visfatin concentrations were elevated in patients with RA and were associated with increased levels of radiographic joint damage, independently of the BMI and measures of inflammation. Second, leptin concentrations were higher in RA patients, independently of the BMI, and were associated with reduced levels of radiographic joint damage, particularly after adjustment for inflammation.
Although visfatin is an adipocytokine, its concentrations have generally not been found to be related to obesity (27, 28), but rather, have been associated with markers of inflammation in some studies of the general population (4, 29). Visfatin concentrations were reported to be increased in several small studies of patients with RA (12, 13, 16), and it has also been reported to be present in synovial fluid and tissue (12, 13). We found that visfatin concentrations were higher in patients with RA than in controls and that the concentrations correlated with those of mediators of inflammation. Notably, visfatin concentrations remained significantly higher in patients with RA after adjustment for BMI and measures of inflammation (TNFα, IL-6, and CRP), suggesting that visfatin may be independently associated with RA.
Visfatin expression has been shown to be increased in rheumatoid synovial fibroblasts, particularly at sites of invasion of cartilage (13). Our observation that visfatin was associated with radiographic joint damage provides further evidence that it may be a mediator of joint damage. Indeed, a recent study (30) identified visfatin as a key component of a newly described inflammatory pathway that leads to arthritis, showing that an inhibitor of visfatin markedly reduced inflammation, arthritis severity, and cartilage damage in a collagen-induced arthritis model. Visfatin mediates the rate-limiting step in the salvage pathway that forms nicotinamide-adenine dinucleotide (NAD) from nicotinamide (29). Pharmacologic inhibition of visfatin led to reduced levels of intracellular NAD in inflammatory cells and decreased production of TNFα and IL-6, with clinical effects comparable to those of a TNFα inhibitor in a murine arthritis model (30). Taken together, these findings suggest that visfatin may be an independent contributor to the pathogenesis of RA and associated radiographic joint damage and may represent a novel therapeutic target.
As has been the case in some (16, 17, 31), but not all (18, 19, 32, 33), studies in RA, leptin concentrations were higher in patients with RA and were associated with some measures of inflammation. The association between increased concentrations of leptin and RA was not affected by statistical adjustment for BMI, but after adjustment for inflammation, it was no longer significant. Thus, the higher concentrations of leptin in patients with RA than in controls can be attributed to differences in inflammation rather than BMI. In addition to its association with inflammation, leptin may also modulate joint damage.
Considering that obesity, despite inducing low-grade inflammation, is associated with reduced levels of radiographic joint damage in RA (1–3), it is particularly important to define the potential role of the adipocytokines most closely associated with obesity: leptin and adiponectin. In one study (34), serum leptin concentrations were higher in patients with erosive RA. Another study showed that concentrations of leptin in synovial fluid were lower than those in plasma in patients with nonerosive RA, but not in patients with erosive RA (31). This was thought to indicate that consumption of leptin in the joint was associated with protection against erosions and that leptin therefore protected against bone erosions (31). We found that higher serum leptin concentrations were associated with reduced radiographic joint damage and that this relationship was attenuated by adjustment for BMI but enhanced by adjustment for inflammation. This finding suggests that the beneficial effects of leptin and BMI on radiographic joint damage may be related, and it may provide an explanation that contributes to understanding the observation that obesity is associated with less radiographic joint damage.
The mechanism whereby higher concentrations of leptin, which is generally considered to be proinflammatory, might reduce radiographic joint damage is not clear. Leptin does, however, also have anabolic effects, such as stimulating the expression of cartilage growth factors and increasing proteoglycan synthesis, that could protect against radiographic joint damage (35). Animal studies have not clarified whether leptin is likely to promote or reduce radiographic joint damage; in different murine models of arthritis, leptin deficiency produced different responses, aggravating (36) or ameliorating (37) arthritis. Extrapolation of the findings in leptin-deficient animal models to humans has been problematic (38). For example, leptin-deficient mice are extremely obese, whereas obese humans have paradoxically high, rather than low, leptin concentrations. This is thought to reflect a state of leptin resistance in obese humans (39). If a similar condition of leptin resistance exists in other tissues, then higher leptin concentrations, which would usually be expected to be proinflammatory, may not show these effects, and the reduced radiographic joint damage associated with obesity could be explained by such functional leptin deficiency. The observation that CRP directly inhibits the binding of leptin to its receptors and blocks its ability to signal in vitro (40) suggests that CRP may play a role in inducing leptin resistance. Thus, increased CRP concentrations could represent 1 mechanism responsible for inducing leptin resistance in RA. Indeed, concentrations of leptin and CRP were significantly correlated in our study.
Another explanation for the inverse association between leptin and radiographic joint damage is that high concentrations of leptin in RA patients do indeed act to promote radiographic joint damage, but these effects are overshadowed by another mediator associated with obesity that protects against radiographic joint damage. A potential candidate would be adiponectin (1, 3), an adipocytokine that has antiinflammatory properties (5).
Adiponectin, like leptin, is strongly correlated with BMI, but unlike leptin, the relationship is inverse, and its transcription is suppressed by inflammatory cytokines such as TNFα and IL-6 (4). Therefore, the higher concentrations of adiponectin in patients with RA in our study are counterintuitive; however, this observation has also been made by other investigators (15, 16). Despite the higher concentrations of adiponectin in patients with RA, the inverse correlation with BMI remained. Thus, patients with a higher BMI had lower concentrations of the antiinflammatory adipocytokine, making it unlikely that adiponectin is related to the decreased radiographic joint damage in obese RA patients. Indeed, adiponectin was weakly associated with higher, not lower, Larsen scores.
Resistin induces inflammatory cytokines and causes arthritis when injected into the joints of mice (8). It is present in rheumatoid synovial tissue and synovial fluid (15, 41), and in a previous study, resistin was associated with higher Larsen scores despite the absence of higher serum concentrations in patients with RA than in controls (42). We found that serum resistin concentrations were higher in RA patients, likely occurring through inflammatory mechanisms, since after adjustment for inflammation, the statistical significance was lost. Resistin concentrations were weakly associated with the Larsen score, and this was attenuated after adjustment for inflammation, suggesting that similar mechanisms are operant.
Limitations of our study include the cross-sectional design and the unavailability of radiographs for determining the Larsen score in some patients. However, apart from differences in age and BMI (variables that were adjusted for in all models), the demographic characteristics of patients with and without radiographs did not differ significantly. The Larsen score could be affected by many variables, but a sensitivity analysis that adjusted for a range of such variables, including the time between the date of enrollment and the date of the radiograph, yielded results that were essentially unchanged. Radiographs were not obtained at the same time that disease activity and adipocytokines were measured. Thus, the estimate of the relationship between adipocytokine concentrations and radiographic joint damage may be less precise than would be the case if the radiographs and blood samples had been obtained on the same date. Future longitudinal studies examining the relationship between visfatin concentrations and radiographic joint damage and studies to define the effects of inhibiting visfatin on arthritis in humans, as has been done in animal models (30), will provide important information.
In summary, we found that concentrations of adipocytokines are increased in patients with RA and modulate radiographic joint damage. Visfatin is associated with RA, RF, and increased radiographic joint damage. Leptin is associated with RA, but with decreased radiographic joint damage.
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. Stein 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. Rho, Solus, Sokka, Oeser, Chung, Gebretsadik, Shintani, Stein.
Acquisition of data. Solus, Sokka, Oeser, Pincus, Stein.
Analysis and interpretation of data. Rho, Solus, Sokka, Oeser, Chung, Gebretsadik, Shintani, Stein.