Abstract
Objective. In prior reports, individuals with rheumatoid arthritis (RA) exhibited increased insulin resistance. However, those studies were limited by either suboptimal assessment methods for insulin sensitivity or a failure to account for important determinants such as adiposity and lack of physical activity. Our objectives were to carefully assess, compare, and determine predictors of skeletal muscle insulin sensitivity in RA, accounting for adiposity and physical activity.
Methods. Thirty-nine individuals with established (seropositive or erosions) and treated RA and 39 controls matched for age, sex, race, body mass index, and physical activity underwent a frequently sampled intravenous glucose tolerance test to determine insulin sensitivity. Inflammation, body composition, and physical activity were assessed with systemic cytokine measurements, computed tomography scans, and accelerometry, respectively. Exclusions were diabetes, cardiovascular disease, medication changes within 3 months, and prednisone use over 5 mg/day. This investigation was powered to detect a clinically significant, moderate effect size for insulin sensitivity difference.
Results. Despite elevated systemic inflammation [interleukin (IL)-6, IL-18, tumor necrosis factor-α; p < 0.05 for all], persons with RA were not less insulin sensitive [SI geometric mean (SD): RA 4.0 (2.4) vs control 4.9 (2.1)*10−5 min−1/(pmol/l); p = 0.39]. Except for visceral adiposity being slightly greater in controls (p = 0.03), there were no differences in body composition or physical activity. Lower insulin sensitivity was independently associated with increased abdominal and thigh adiposity, but not with cytokines, disease activity, duration, disability, or disease-modifying medication use.
Conclusion. In established and treated RA, traditional risk factors, specifically excess adiposity, play more of a role in predicting skeletal muscle insulin sensitivity than do systemic inflammation or other disease-related factors.
In persons with rheumatoid arthritis (RA), 2-fold increased rates of cardiovascular (CV) disease have been attributed, in part, to insulin resistance1,2. Potential explanations for insulin resistance in RA have included higher systemic inflammation associated with active disease, use of glucocorticoids, and increased traditional risks such as abdominal obesity and inactivity2,3.
Most prior investigations of insulin resistance in RA have relied largely on fasting measures that better reflect hepatic rather than skeletal muscle insulin sensitivity. Assessing insulin sensitivity is critical, because skeletal muscle is responsible for as much as 90% of glucose uptake after a meal. Further, skeletal muscle insulin resistance initiates and perpetuates the development of type 2 diabetes and serves as the central feature of the metabolic syndrome4,5. Of the few investigations in patients with RA that have used glucose challenge-based measures of insulin sensitivity, none have accounted for potential differences in adiposity and physical activity6,7,8,9,10. To improve understanding of how insulin resistance might contribute to increased CV risk in persons with RA, our objectives were to determine (1) whether persons with RA, as compared to controls matched for age, sex, race, body mass index (BMI), and physical activity, exhibit more skeletal muscle insulin resistance as assessed by intravenous (IV) glucose tolerance tests (IVGTT), and (2) in persons with RA, whether disease-specific and/or traditional predictors are related to skeletal muscle insulin resistance.
MATERIALS AND METHODS
Patient population
In this cross-sectional investigation, persons were included if they had (1) RA that met American College of Rheumatology 1987 criteria11, (2) seropositive disease (positive rheumatoid factor or anticyclic citrullinated peptide) or evidence of erosions, and (3) no medication changes within the 3 months prior to study enrollment. Persons using prednisone 5 mg or less daily were included. Any prednisone taper must have been completed at least 3 weeks prior to enrollment. Healthy individuals without a diagnosis of RA or joint pain or swelling lasting more than a week were matched to an RA participant by sex, race, age (within 3 yrs), and BMI (within 3 kg/m2). For 1 African American participant where a race match was not identified, a white person who was otherwise well-matched was included. Exclusion criteria were a history of diabetes or CV disease, current pregnancy, and use of medications known to affect carbohydrate or lipid metabolism (including insulin, oral antidiabetic agents, statins, fibrates, nicotinic acid, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, or β blockers). Participants with RA were recruited from Duke University and Durham Veterans Affairs Medical Center Rheumatology Clinics and from the local community with newspaper and website advertisements. Controls were recruited from the local community with newspaper and Website advertisements. Our study was in compliance with the Helsinki Declaration and was approved by the Duke University Institutional Review Board. All participants gave written informed consent.
Questionnaires and physical examination
Participants completed questionnaires for medications, disability (Health Assessment Questionnaire Disability Index)12, comorbidities13, a visual analog scale for health rating, and the Stanford Brief Activity Survey for physical activity14. Each completed a visual analog scale (VAS) for health, and underwent anthropometric measures, a 28-joint examination, and fasting blood collection for glucose, insulin, and erythrocyte sedimentation rate (ESR). The VAS, joint examination, and ESR were used to compute a Disease Activity Score (DAS28)15.
Accelerometry
Participants wore an RT3 tri-axial accelerometer (Stayhealthy Inc.) for 7 days. Data were obtained from accelerometers as activity calories per min (Stayhealthy RT3 Assist Version 1.0.6). Ninety consecutive min of no measured activity was used to indicate and eliminate periods of time when individuals were not wearing the device. Also, if there were fewer than 10 h of measured activity in a single day, all data from that day were excluded from analysis. Similarly, we excluded data from participants (n = 7) who had fewer than 4 days of valid data (each of 10 h or more of wear time). Two participants did not return the device.
Systemic inflammatory measures
A panel of proinflammatory cytokines including interleukin (IL)-1β, IL-6, IL-8, and tumor necrosis factor (TNF)-α were measured in plasma samples using a multiarray quantitative immunoassay (Meso Scale Discovery). IL-18 and high-sensitivity C-reactive protein (hsCRP) were measured using sandwich ELISA (R&D, and MP Biomedicals, respectively). Mean intraassay (within plate) coefficients of variation were as follows: IL-1β 9.3%, IL-6 3.4%, IL-8 3.9%, TNF-α 2.9%, IL-18 2.4%, and hsCRP 3.5%. A human plasma control was run on each plate, and mean interassay (between plates) coefficients of variation were as follows: IL-1β 5.5%, IL-6 3.0%, IL-8 7.8%, TNF-α 7.7%, IL-18 6.0%, and hsCRP 3.9%.
Insulin sensitivity and fatty acids
Insulin sensitivity was determined using a frequently sampled IVGTT. Glucose (50%) was injected through a catheter at 0.3 g/kg body mass. Insulin was injected at min 20 at 0.025 units/kg body mass. Twenty-nine blood samples were obtained over 3 hours, centrifuged, and stored at −80°C. Glucose was measured on a Beckman-CoulterDxC600 analyzer, and insulin by electrochemiluminescent assay from Meso Scale Discovery. Insulin sensitivity index was calculated using Bergman’s minimal model16. Nonesterified free fatty acids (NEFA) were measured by colorimetric enzymatic assay using a Beckman-Coulter DxC600 analyzer, with reagents from Wako.
Abdomen and thigh computerized tomography (CT) scans
Single 10-mm-thick axial sections at the liver and mid-thigh were performed using a General Electric CT/I scanner (GE Medical Systems). Cross-sectional areas for adipose tissue depots and liver and muscle density were measured using OsiriX (Pixmeo). CT scan determinations of adipose tissue depots are very accurate, precise, and reliable17,18.
Statistical analyses
Cytokines, NEFA, and insulin sensitivity index were logarithmically transformed prior to group comparisons. Comparisons between patients with RA and controls were performed with mixed models, which accounted for the repeated measure of matched participants. Bivariate associations were assessed with Spearman correlations. Multivariable modeling for insulin sensitivity index (log) was performed using linear models with backward stepwise variable selection. Initial variable inclusion was based on conceptual hypotheses regarding traditional risk factors and disease-associated factors affecting insulin sensitivity index and included age, sex, waist circumference, physical activity, disease activity (DAS28), disease-modifying agent use, biologic agent use, and prednisone use. Waist circumference was selected rather than BMI based on the high levels of correlation between the 2 (r = 0.87) and because waist circumference was more strongly correlated with insulin sensitivity index. For the laboratory model (based on results from bivariate analyses), visceral adiposity and thigh intermuscular adiposity were included rather than waist circumference, and IL-6 was included rather than disease activity; all other variables from the clinical model were included.
Statistical power
Our investigation was designed to detect a difference in insulin sensitivity that corresponded to a moderate effect size (0.4–0.5)19. Based on 39 matched pairs and an α of 0.05, we had 80% power to detect a standardized difference in insulin sensitivity index of 0.46.
RESULTS
Patients with RA were well matched to controls in age, sex, and BMI (Table 1). While patients with RA had slightly less abdominal visceral adiposity (p = 0.03), otherwise there were similar amounts of fasting NEFA, abdominal and thigh adipose tissue depots, and thigh muscle area and density (p > 0.05 for all). There were no differences in physical activity, measured as total energy expended (p > 0.05). In the context of a wide range of disease activity (DAS28 0.78–6.4) and duration (5–506 mos), individuals with RA exhibited a proinflammatory profile, as demonstrated by elevated TNF-α, IL-6, and IL-18 (p < 0.05 for all). Despite elevated systemic inflammation, persons with RA were not significantly less insulin-sensitive than controls [insulin sensitivity index (SI) geometric mean (SD) = 4.0 (2.4) vs 4.9 (2.0)*10−5 min−1/(pmol/l); mean SD for log SI = 0.23; p = 0.39].
Participant characteristics. Data are presented as means (SD) for continuous variables and number (%) of participants for dichotomous variables. Data that were not normally distributed (systemic inflammatory markers, cytokines, and insulin sensitivity) are presented as geometric means (SD).
Table 2 shows correlations of key traditional and disease-specific risks with insulin sensitivity. In persons with RA, lower insulin sensitivity index was associated with a larger BMI (r = −0.45, p < 0.002), a greater waist circumference (r = −0.47, p < 0.001), and increased amounts of total abdominal adipose tissue, visceral adiposity, total thigh area, thigh intermuscular adiposity, and thigh muscle area (r = −0.3 to 0.5, p < −0.05 for all).
Spearman correlations for insulin sensitivity, n = 44.
Multivariable modeling was performed using variables available in a clinic setting: age, sex, waist circumference, physical activity, disease activity, biologic use, use of disease-modifying antirheumatic drugs, and prednisone use. After backward variable selection, the model retained only waist circumference and accounted for 26% of the variance in insulin sensitivity index (Table 3, p < 0.0005). To better understand the contribution of inflammation and adiposity, a “laboratory” model was constructed. For this model, initial variables included visceral adiposity and thigh intermuscular adiposity rather than waist circumference, IL-6 rather than disease activity, and all of the other variables in the initial clinical model. In the laboratory model, visceral adiposity, thigh intermuscular adiposity, and IL-6 accounted for 46% of the variance in insulin sensitivity index (Table 3; p < 0.0005). Visceral adiposity (p < 0.005) and thigh intermuscular adiposity (p < 0.02) were each independently related to insulin sensitivity index.
Multivariable model for insulin sensitivity index (log) in persons with rheumatoid arthritis.
DISCUSSION
In 39 persons with established RA (well controlled and reflective of many clinic cohorts), insulin sensitivity index, as measured with an IVGTT, was not significantly lower than in matched controls. Also, for persons with RA, insulin sensitivity was related to traditional risk factors of large adipose tissue depots. Other than concentrations of IL-6, disease-specific factors, including disease activity, biologic agent use, and disease-modifying agent use had little apparent influence on this outcome.
To our knowledge, this is the first well-controlled comparison of insulin sensitivity index for persons with established and treated RA. Previously, most studies addressing insulin sensitivity in RA used fasting glucose-derived and insulin-derived indices such as homeostasis model assessment and the quantitative insulin sensitivity check index2; however, those fasting-derived indices reflect insulin sensitivity mainly in the liver. In contrast, glucose tolerance tests better reflect insulin sensitivity in skeletal muscle. Because skeletal muscle is responsible for the majority (up to 90%) of systemic glucose uptake after a glucose load such as during a meal, assessing insulin sensitivity index with glucose challenge tests is critical for understanding insulin action.
Despite the importance outlined above, glucose challenge tests, including oral and IV glucose loads as well as euglycemic clamps, have been used to compare insulin sensitivity between RA and controls in a relatively small number of investigations6,7,8,9,10,20. Findings from these have been conflicting, likely because of a failure to adequately assess and account for the significant contributors to insulin resistance, adiposity, and physical inactivity. Specifically, several investigations demonstrated that persons with untreated RA had greater insulin resistance, but none included BMI-matched controls or assessed physical activity8,9,20. Also, in 2 investigations, persons with RA had insulin sensitivity similar to that of controls, but in one, the comparison was confounded by the control group having more males and higher BMI6. The second investigation was limited to normal weight, premenopausal females, limiting generalizability. Our study shows definitively that despite elevated amounts of systemic inflammation when compared to controls matched for adiposity and physical inactivity, persons with RA have similar insulin sensitivity.
Thus, in typical clinical cohorts of established and well-controlled RA, the average reduction in insulin sensitivity index imposed by RA-associated factors appears both statistically and clinically insignificant. Statistical power for this investigation aimed to detect a difference in insulin sensitivity corresponding to a moderate effect size, so the observed small difference was not statistically significant. Clearly, a larger sample might have been able to detect the small difference as statistically significant, but the critical issue is the clinical relevance of such a small difference in insulin sensitivity.
Clinical relevance for specific insulin sensitivity values is difficult to determine, but some investigations linking insulin sensitivities to outcomes can provide insight into this issue. When persons were followed longitudinally for 25 years for diabetes development, the difference was much larger than ours between those who developed type 2 diabetes and those who did not [3.2 ± 2.4 vs 8.1 ± 6.7 10−5 min−1/(pmol/l)]21. Also, when comparing tertiles of insulin sensitivity, the upper tertile (≥ 2.39*10−5 min−1 /[pmol/l]) was associated with an incidence of progression of < 10%22. This low incidence suggests that the effect of the small difference observed between persons with well-controlled RA and matched controls is clinically negligible.
As a means of confirming that risk of insulin sensitivity in well-controlled RA is similar to that of matched controls, we observed that predictors of insulin sensitivity in RA are largely traditional cardiometabolic risk factors rather than RA-specific ones. Prior predictors of RA-associated insulin resistance include both altered body composition and inflammation2,3,6,9. Covering a wide range of insulin sensitivities and disease activities, our sample provided a rich source to identify potential unique mediators of insulin resistance development in persons with RA. Nonetheless, except for IL-6 concentrations, all correlates of insulin sensitivity were to body composition. These findings imply that in persons with established and treated RA, traditional risk factors, specifically excess adiposity, play more of a role in predicting insulin sensitivity index than do inflammation or medication use.
A wide range of hypotheses and evidence link adipose tissue and skeletal muscle insulin resistance. Adipose tissue contributes to skeletal muscle insulin resistance by providing a source of free fatty acids, which inhibit insulin-stimulated glucose uptake in skeletal muscle both directly and indirectly (by promoting production of acyl-CoA and reactive oxygen species)23. Additionally, adipose tissue triggers skeletal muscle insulin resistance through the production of a number of insulin resistance-promoting inflammatory cytokines and adipokines24.
While adipose tissue clearly contributed, skeletal muscle insulin resistance in RA appeared to have little relation to disease-specific factors. Of these, only IL-6 was related. Specifically, higher IL-6 was related to poorer insulin sensitivity, with a p value of 0.05 in bivariate analyses and a trend toward statistical significance in a multivariable model. In persons without systemic inflammatory disease, IL-6 has shown a complex relationship with insulin sensitivity25. Acutely, increases in IL-6 associated with exercise have been shown to improve insulin sensitivity, but chronic elevations appear to worsen insulin sensitivity25. Here, in persons with elevated systemic concentrations of IL-6, this cytokine was related to poorer insulin sensitivity in contrast to other disease-related variables.
We are aware that this investigation has limitations. One of the main limitations is a small sample size, in turn reducing study power and increasing the likelihood of a Type II statistical error. That, and the heterogeneity of our population, may have contributed to our lack of statistical significance in the difference in insulin sensitivity between RA and matched controls. However, heterogeneity provided a valuable opportunity to determine predictors of insulin sensitivity in persons with RA. Nonetheless, we recognize that the predictive capability of the models presented is relatively modest. However, developing models as tools for predicting insulin sensitivity was not the study goal, but rather the objective was to determine the relative contribution of disease-related and traditional risk factors for insulin resistance in RA. Also, we believe this sample of persons with established and treated RA reflects what is seen in many rheumatology clinic cohorts, thus allowing generalizability of our findings regarding risks for insulin sensitivity in RA. One of the main strengths is using IVGTT to assess insulin sensitivity index in RA, thus emphasizing that stimulated tolerance tests allow a more complete assessment of insulin action.
Thus, in a population of patients with RA reflective of typical clinical cohorts, as compared to well-matched controls, insulin sensitivity index was not significantly lower in those with RA. Increased abdominal and thigh adiposity contributed to poorer insulin sensitivity but not disease activity or medication use. These findings imply that in established and treated RA, adipose depots, not disease-related factors, account for insulin sensitivity.
Acknowledgment
We thank the participants in this investigation as well as the Duke University Division of Rheumatology staff members who referred patients for this investigation. We appreciate helpful discussions with career award mentors, Drs. Gregory Samsa and Deborah Muoio, and assistance from the Department of Radiology (Dr. Rendon Nelson and Carolyn Lowery).
Footnotes
-
Supported by US National Institutes of Health/National Institute of Arthritis and Musculoskeletal and Skin Diseases K23AR054904, an American College of Rheumatology-Rheumatology Research Foundation (ACR-RRF)/Association for Specialty Professors (ASP) Junior Career Development Award in Geriatric Medicine funded by Atlantic Philanthropies, ACR-RRF, John A. Hartford Foundation and ASP, and an Early Career Development Award from the Central Society for Clinical Research.
- Accepted for publication April 11, 2014.
REFERENCES
In this issue
