Insulin resistance (IR), is a well established risk factor for the development of cardiovascular disease (CVD) 1 . The mechanisms of IR are under intense investigation; however, a consistent finding of such research is the close association between IR and inflammation 2 3 4 . Tumour necrosis factor alpha (TNFα), a pro-inflammatory cytokine, is thought to be one of the main mediators of IR 2 . Patients with IR exhibit increased circulating levels of TNFα 5 6 , and administration of TNFα induces IR in healthy individuals 7 . In otherwise healthy individuals, obesity is a significant contributor to IR; obesity is a low-grade inflammatory condition 8 9 and TNFα is also thought to be the link between obesity and insulin resistance 3 .

Rheumatoid arthritis (RA), associates with reduced life expectancy compared to the general population 10 , mainly due to increased prevalence of CVD, and increased morbidity and mortality from CVD compared to the general population 11 12 13 . TNFα is central to the development and progression of RA and a common therapeutic target 14 . Apart from disease activity, treatment with anti-TNFα appears to also improve insulin sensitivity 15 and to reduce CVD risk in RA 16 17 . However, obesity - a potent contributor to IR in the general population - might influence the way anti-TNFα therapy affects IR. Indeed, in the general population, anti-TNFα does not improve IR in obese individuals 18 . The aim of this longitudinal study was to compare the effects of six months of anti-TNFα therapy on IR between normal weight and obese RA patients. Our primary hypothesis was that the possible beneficial effects of anti-TNFα on IR would be limited by the presence of obesity.

At baseline, t-tests did not identify any significant differences between the groups, other than, as expected, in IR (HOMA and QUICKI; Table 1). Following the six-month intervention, repeated measures ANOVA found no changes in BMI in any of the groups (P > 0.05). Inflammation and disease activity were significantly reduced in all groups (P < 0.05) but to a similar extent (P for differences between groups > 0.05 in all cases). Overall, HOMA showed a tendency to reduce (P = 0.088) while QUICKI a tendency to increase (P = 0.092) (Table 2).

When patient grouping was introduced as a factor in the analyses, it was revealed that the treatment resulted in significant decreases in HOMA (Δ6 m = -0.54; P = 0.002) and increases in QUICKI (Δ6 m = 0.046; P = 0.011) in the N+IR, but not any of the other three groups (P > 0.05 in all cases). The magnitude of the changes in IR was significantly different between N+IR and the rest of the groups (N-IR: HOMA, P = 0.008, QUICKI, P = 0.002; O+IR: HOMA, P = 0.019, QUICKI, P = 0.038; O-IR: HOMA, P = 0.000, QUICKI, P = 0.001; Figure 1). No differences in the magnitude of improvements between the other groups were observed (P > 0.05 for both HOMA and QUICKI).

Finally, MANOVA indicated that baseline BMI, baseline RA characteristics, baseline BMI, and the change in BMI did not associate with the change in HOMA or QUICKI in any of the groups (P > 0.05 in all cases). However, the change in ESR and the change in CRP associated with the improvement in HOMA (ESR: F_1-7 = 5.143, P = 0.019; CRP: F_1-7 = 3.122, P = 0.022) and QUICKI (ESR: F_1-7 = 3.814, P = 0.021; CRP: F_1-7 = 2.67; P = 0.041) only in the N+IR group.

Retrospective power calculations (a = 0.05) using the values observed above indicate a power of 0.91 and a power of 0.88 for the changes observed in HOMA (Δ6 m for N+IR = 0.49, for O+IR = 0.11, common St.D. = 0.26) and QUICKI (Δ6 m for N+IR = 0.04, for O+IR = 0.01, common St.D. = 0.02), respectively.

The main aim of this study was to compare the effects of anti-TNFα treatment on insulin sensitivity among normal weight and obese RA patients. To our knowledge, this is the first study to use such an approach and indeed to identify different effects of anti-TNFα treatment in normal-weight vs. obese individuals with RA. Our results indicate a significant improvement of IR only in normal-weight RA patients with IR but not in obese RA patients with IR. Reduction in inflammation was associated with the changes in IR only in the normal weight but not in obese patients.

Sample size is the main limitation of our study. However, the number of patients that fulfilled the inclusion criteria for this study (that is, insulin resistance, anti-TNFα naive and embarking on anti-TNFα) is limited. This is also reflected by sample size of other studies investigating similar hypotheses 15 . In our study, a larger sample size might have been able to identify improvements in IR in the O+IR group. This group had a slight improvement in IR (as shown in Figure 1). However, this improvement was not statistically significant (mean change in HOMA: -0.11; 95% odds ratio: 0.06 to 0.15; and QUICKI: 0.009; 95% odds ratio: 0.004 to 0.018; P > 0.05 for both) or indeed clinically significant as both variables remained within the "at risk" range. Moreover, the statistical test we used (Repeated measures ANOVA) looks at the differences in the responses between the two groups, not merely the response of each group. We, therefore, doubt that a larger sample size would significantly change the conclusions of the present study. A second limitation of our study is that we did not control for changes in physical activity, which may influence IR. However, any such changes are likely to have occurred throughout all the subgroups, especially since disease activity was reduced equally among them. Additionally, for the duration of the study, no changes in thyroid status were observed (all patients were euthyroid at baseline), there were no changes in other anti-rheumatic medication (including steroids and hydroxychloroquine) or cardio/vasoactive therapy (including angiotensin-converting-enzyme (ACE) inhibitors).

Similar studies in RA patients have shown reduction in IR following anti-TNFα therapy (reviewed in 27 ). BMI, even though assessed, is not reported in some of these studies 28 29 ; in those who report it, observations are predominantly made on normal-weight to slightly overweight patients (BMI: approximately 22 kg/m² 30 to approximately 25 kg/m² 31 32 ). Even in these studies increasing BMI appears to associate with a smaller improvement in IR 31 . Similarly, acute administration of anti-TNFα significantly improved insulin sensitivity in most RA patients; however, this improvement was again minimised with increasing BMI 33 . We have been able to find a single case-study of an obese patient with RA that experienced significantly improved IR following anti-TNFα treatment 34 ; the extremely high baseline HOMA levels (> 25 with a cut-off for IR at 2.5) of this patient should be noted.

The reason for this apparent difference in the responses of lean vs. obese individuals to anti-TNFα is not yet clear. However, our findings for the obese RA patients are in line with data from the general population. In obese individuals with the metabolic syndrome 18 or with type 2 diabetes mellitus 35 anti-TNFα failed to improve IR. It is well established that enlarged adipose tissue is a source of inflammation. Up to 50% of its cell mass is monocytes and macrophages. Adipocytes and macrophages both are a source of inflammatory cytokines reducing insulin sensitivity of muscle and increasing plasma concentrations of these compounds 36 37 . The cytokines TNFα and interleukin-6 (IL-6) in particular, stimulate both the c-Jun amino-terminal kinase (JNK) and the IκB kinase-β (IKK-β)/nuclear factor-κB (NF-κB) pathways, resulting in up-regulation of potential mediators of inflammation and lead to IR 38 . Other potential pathways by which obesity might induce IR include the action of retinol-binding protein-4 (RBP4) which reduces phosphatidylinositol-3-OH kinase (PI(3)K) signalling in muscle and enhances expression of the gluconeogenic enzyme phosphoenolpyruvate carboxykinase in the liver 39 . The role of IL-6 is of particular importance in the context of the present study, as this is not directly targeted by anti-TNFα medication and could thus be important in maintaining IR in obese individuals. A final interesting approach to the potential underlying mechanisms could be the role of the inflammasome and, especially, NLRP3 in IR. This is a cytosolic protein complex leading to the activation of the processing enzyme caspase-1, which is central to the production of Interleukin -1β (IL-1β) 40 . IL-1β has been recently recognized as a potent instigator of the obesity-induced inflammation and as such a contributor to IR 41 . Even though anti-TNFα medication does not directly target IL-1β, it is implicated in the activation of NLRP3 42 , thus indirectly affecting IL-1β production. The associations of obesity, inflammasomes and IL with IR in RA warrant further investigation.

Adipose tissue also secretes adipokines, such as adiponectin and resistin 43 , reduced levels of adiponectin and increased levels of resistin associate with IR 44 45 . IR and other features of the metabolic syndrome independently associate with atherosclerosis in RA 46 . Adiponectin stimulates fatty acid oxidation via AMP-activated protein kinase (AMPK) and peroxisome proliferator activated receptor-α-dependent pathways 38 .

Anti-TNFα treatment may affect levels of these adipokines. In patients with severe RA, refractory to conventional disease-modifying antirheumatic drug (DMARD) therapy, periodical treatment with infliximab (an anti-TNFα blocker) results in a rapid reduction of serum resistin levels 47 . Serum resistin levels, in these patients, associate with laboratory markers of inflammation, particularly CRP, but not with BMI 47 . Moreover, in the same cohort of patients, an independent negative correlation of high-grade inflammation with circulating adiponectin concentrations has been observed 48 . Low adiponectin concentrations further correlate with atherogenic dyslipidemia and high plasma glucose 48 .

Moreover, anti-TNFα treatment may reduce lipolysis within the muscle by inhibiting TNFα action locally 43 - obese individuals have high fat intramuscular content. This may lead to further increases in intramuscular concentration of fatty acid metabolites, such as diacylglycerol, fatty acyl-coenzyme A, and ceramides, as a result of increased release of non-esterified fatty acids 39 . This leads to serine/threonine phosphorylation of insulin receptor substrate-1 and -2, and eventually to insulin resistance 49 .

From a clinical point of view, our findings would suggest that additional measures to anti-TNFα therapy should be employed in order to combat IR in obese RA patients. Weight-loss is probably the primary such measure. However, due to the lack of weight-loss studies in RA 50 and the high prevalence of muscle wasting among these patients 51 , any such interventions should be employed with great care. Exercise, which is known to benefit RA patients in several ways 52 , could possibly prove a valuable tool in reducing body weight while maintaining muscle mass and improving insulin sensitivity, but further research is needed. In addition to IR, such interventions could improve other CVD risk factors and reduce the high CVD risk of RA patients.

From the above results, it seems that the effects of anti-TNFα therapy on IR differ significantly between normal-weight and obese patients with RA. Such therapy seems to improve IR in normal-weight but not in the obese RA patients. In the latter group, obesity-related processes seem to counteract the potential benefits of anti-TNFα. Further studies looking at the specific mechanisms by which anti-TNFα therapy affects IR and the mechanisms by which obesity interacts in this process should be pursued.

In conclusion, anti-TNFα therapy seems to improve insulin sensitivity in normal weight RA patients with IR but not in obese RA patients with IR. It is possible that the mechanisms leading to insulin resistance are partly different in normal-weight and obese patients with RA.

ACE inhibitor: angiotensin-converting-enzyme inhibitor; AMPK: AMP-activated protein kinase; ANOVA: analysis of variance; BMI: body mass index; CRP: C-reactive protein; CVD: cardiovascular disease; DAS28: Disease Activity Score-28; DMARD: disease-modifying antirheumatic drug; ESR: erythrocyte sedimentation rate; HAQ: Health Assessment Questionnaire; HOMA: Homeostasis Model Assessment of Insulin Resistance; IKK-β: IκB kinase-β; IL-6: interleukin 6; IR: insulin resistance; +IR: normal-weight with insulin resistance; JNK: c-Jun amino-terminal kinase; MANOVA: multivariate analysis of variance; NF-κB: nuclear factor-κB; N-IR: normal-weight without insulin resistance; NO+IR: obese with insulin resistance; O-IR: obese without insulin resistance; PI(3)K: phosphatidylinositol-3-OH kinase; QUICKI: Quantitative Insulin sensitivity Check Index; RA: rheumatoid arthritis; RBP4: retinol-binding protein-4; TNFα: tumor necrosis factor alpha.

The authors declare that they have no competing interests.

AS was involved in the conception and design of the study, and in data collection, data analyses and the writing of the manuscript. GM participated in the data collection and manuscript writing. VP was involved in data collection and data analyses. PN is the statistics expert of the team and led data analyses. YK was involved in the design of the study and drafting of the manuscript. GK was involved in the conception and design of the study, drafting of the manuscript, and acted as the study supervisor. All authors have read and approved the manuscript for publication.