Introduction
The incidence of type 2 diabetes (T2D) is higher in patients with RA (17–20% [
1‐
3]) than in the general population (8% [
4]), independent of glucocorticoid use. Patients with RA also have increased insulin resistance compared with individuals without RA [
3,
5]. RA disease outcomes are poorer in patients with comorbid diabetes, who are also at increased risk for cardiovascular disease relative to patients who have either RA or diabetes only [
1,
2].
Chronic systemic inflammation is implicated in the pathogenesis of both RA and diabetes [
6]. The pro-inflammatory cytokines interleukin (IL)-6, tumour necrosis factor-α (TNFα) and IL-1β play key roles in the synovial inflammation and joint damage associated with RA and also have systemic effects on extra-articular tissues [
7‐
9]. IL-6 can signal through both membrane-bound (
cis-signalling) and soluble (
trans-signalling) IL-6 receptors (IL-6Rs) and therefore has pleiotropic effects on immune/inflammatory and other cell types, such as pancreatic β cells, skeletal muscle, adipose tissue and liver [
10,
11]. Chronically elevated levels of systemic IL-6 have been associated with dysfunctional glucose metabolism and homeostasis and with the induction of insulin resistance in liver and adipose tissue [
11‐
13]. Elevated levels of IL-6 are an independent risk factor for T2D [
13‐
15], and IL-6 alone or in combination with IL-1β inhibits β cell function [
16,
17]. Similarly, effects on glucose metabolism, insulin resistance, pancreatic β cell function and risk of diabetes are attributed to elevations in TNFα and IL-1β [
6,
15,
18,
19], whilst IL-1β antagonism reduces hyperglycaemia and improves pancreatic β cell function in patients with T2D [
20,
21]. Of note, the metabolic effect size of IL-1β antagonism is considerably greater in patients with RA and comorbid T2D [
21], indicating that the efficacy of anticytokine biologics correlates with the inflammatory burden.
Medical management of T2D and RA can be complicated by the potential effects of RA treatments on glucose levels. Oral glucocorticoids increase the risk for diabetes in patients with RA [
22,
23] because of the adverse metabolic actions of these drugs, with higher dose and longer treatment duration increasing the risk [
23]. By contrast, the anti-inflammatory drug hydroxychloroquine reduced the risk of incident diabetes in patients with RA [
24,
25] and was also associated with a favourable effect on glycaemia in patients with RA in the absence of diabetes [
26]. These effects are not necessarily associated with direct actions on insulin resistance and/or pancreatic β cell function [
24,
25,
27], but rather to a reduction of low-grade inflammation by inhibiting the inflammasome [
28]. Methotrexate (MTX) also reduces glycosylated haemoglobin (HbA1c) levels and the risk for diabetes in patients with RA [
29], likely independently of insulin sensitivity [
30].
Because diabetes is a common comorbidity in patients with RA and cytokines are implicated in glucose homeostasis, we conducted post hoc analyses of three sarilumab phase III randomised clinical trials. These analyses aimed to assess the effect of sarilumab (a human monoclonal antibody that blocks the IL-6Rα), as monotherapy or in combination with conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) on HbA1c levels compared with either placebo (+ MTX/csDMARD) or adalimumab monotherapy [
31‐
33]. We also assessed the safety and efficacy of sarilumab in patients with RA with or without comorbid diabetes.
Methods
Study design
Details of the three phase III study designs have been described previously [
31‐
33]. In brief, the MOBILITY trial (NCT01061736) was conducted to investigate the efficacy and safety of up to 52 weeks of sarilumab (or placebo) in combination with MTX in patients with moderate-to-severe active RA and an inadequate response to MTX (MTX-IR); this trial is referred to herein as the MTX-IR sarilumab + MTX study. The TARGET trial (NCT01709578) was conducted to investigate the efficacy and safety of up to 24 weeks of sarilumab (or placebo) in combination with background conventional synthetic DMARDs (csDMARDs) in patients with moderate-to-severe RA who were intolerant of, or who had inadequate response to, TNF inhibitors (TNFi-INT/IR); this study is referred to herein as the TNFi-INT/IR sarilumab + csDMARDs study. The MONARCH trial (NCT02332590) was conducted to compare the 24-week efficacy and safety of sarilumab monotherapy with adalimumab monotherapy in biologic DMARD-naïve patients with moderate-to-severe active RA who were intolerant of, or had an IR to, MTX (MTX-INT/IR); this study is referred to herein as the monotherapy study. Patients with uncontrolled diabetes mellitus (defined by HbA1c ≥ 9% at the screening visit) were excluded from participation. Patients treated with ≤ 10 mg oral prednisone or equivalent at a stable dose for at least 4 weeks prior to the baseline visit were included; however, changes in dose were not permitted during the double-blind treatment periods unless required for treatment of an adverse event (AE) (other than worsening RA).
The three protocols were approved by the appropriate ethics committees/institutional review boards, and each patient provided written informed consent before participation in the study. The studies were conducted in compliance with institutional review board regulations, the International Conference on Harmonisation Guidelines for Good Clinical Practice and the Declaration of Helsinki.
For the purposes of these post hoc analyses, patients were classified as having diabetes if they reported either a medical history of the disease or baseline use of medication to treat diabetes (e.g. metformin, sulfonylureas, dipeptidyl peptidase 4 inhibitors or insulins).
Assessments
HbA1c was measured at baseline in all three studies and also at weeks 12 and 24 in the TNFi-INT/IR sarilumab + csDMARD and monotherapy studies. In the MTX-IR sarilumab + MTX study, HbA1c was assessed after baseline at the investigators’ discretion.
Safety assessments included the incidence of treatment-emergent adverse events (TEAEs), serious TEAEs (SAEs), serious infections and specific abnormalities in laboratory tests. AEs were described at the Medical Dictionary for Regulatory Activities (version 16.0) preferred-term level.
Efficacy assessments included the American College of Rheumatology 20% (ACR20) response rate and the change from baseline in the Health Assessment Questionnaire-Disability Index (HAQ-DI), Disease Activity Score (28 joints) using C-reactive protein (DAS28-CRP) and Clinical Disease Activity Index (CDAI) at week 24.
Statistical methods
The analyses of patients with and without diabetes were conducted post hoc; therefore, p values should be considered nominal.
Changes in HbA1c were analysed for subgroups defined by a medical history of diabetes or baseline use of an antidiabetic medication. To determine whether changes in HbA1c might be therapeutically relevant for patients with marginal glycaemic control, the analyses were repeated in a subgroup of patients who had a baseline HbA1c value ≥ 7.0% [
34]. To investigate potential modulators of HbA1c, other than the investigational treatment, patients were also classified by oral glucocorticoid use and type of treatment for diabetes, and Spearman’s rank correlation coefficients (
rS) were calculated for baseline HbA1c vs. baseline CRP, DAS28-CRP and haemoglobin and for changes in HbA1c vs. changes in CRP, DAS28-CRP and haemoglobin.
Values for observed cases were used without imputation for analyses of safety and laboratory parameters. No formal statistical analysis of AEs was conducted for patients with RA in the presence vs. absence of diabetes (comparisons were descriptive).
Efficacy was assessed in the intent-to-treat population. For efficacy analysis, data from the MTX-IR sarilumab + MTX study and the TNFi-INT/IR sarilumab + csDMARD studies were pooled. For categorical efficacy variables, patients were considered nonresponders from the time they started rescue therapy or discontinued the study medication. For continuous efficacy variables, assessments were set to missing from the time a patient received rescue therapy or discontinued study medication before the end of the study. Missing values were not imputed. Changes in efficacy variables were modelled using a repeated-measures-mixed-effect model, assuming unstructured covariance with treatment, region, visit, subgroup, treatment-by-visit interaction, treatment-by-subgroup interaction and treatment-by-visit-by-subgroup interaction included in the model. A nominal p value < 0.05 was considered significant.
Discussion
Diabetes is a common comorbid condition among patients with RA [
1,
2,
35]. Medical management of both conditions can be complicated, because RA disease activity and the use of glucocorticoids, csDMARDs and bDMARDs (often in combination) to treat RA may affect glucose levels through their effects on glucose metabolism, insulin sensitivity and pancreatic β cell function [
3,
24,
25,
36].
These post hoc analyses of three phase III clinical studies in patients with RA show that sarilumab as monotherapy or in combination with MTX/csDMARDs is associated with a greater reduction in HbA1c than adalimumab monotherapy or placebo + MTX/csDMARDs, particularly in patients with diabetes. Reductions in HbA1c were more prominent in patients treated with sarilumab compared with either MTX/csDMARDs or adalimumab in patients whose baseline HbA1c was ≥ 7%, a level that exceeds the target HbA1c of < 7% recommended by the American Diabetes Association, supporting the possibility of improving, as well as maintaining, glucose homeostasis.
It is well recognised that oral glucocorticoids, in addition to immune-suppressive actions, also affect glucose homeostasis. In RA, oral glucocorticoids are commonly used and it is important to notice that the effects of sarilumab on HbA1c were not attenuated by concomitant glucocorticoid treatment (≤ 10 mg prednisone equivalent). In addition, the observed reduction in HbA1c could not be explained by the expected increase in haemoglobin associated with IL-6R blockade, which indicates that its effect on HbA1c did not reflect an indirect effect of a change in haemoglobin levels. Whilst the current analyses cannot exclude the general influence of systemic inflammation as a cause of changes in glycaemia, a correlation between changes in HbA1c and changes in CRP was not observed in the TNFi-INT/IR sarilumab + csDMARDs study and correlations were minimal when sarilumab was used as monotherapy (Spearman’s rank correlations < 0.24). Similar results were observed with three models incorporating different measures of change in disease activity and inflammation (CRP and CDAI; DAS28-CRP; and tender and swollen 28 joint counts, patient and physician global assessments, and CRP). These results support an effect of sarilumab in reducing HbA1c that is independent of anti-inflammatory effects, although some degree of association between general effects on inflammation and changes in glycaemia cannot be ruled out.
These analyses also showed that sarilumab + MTX/csDMARDs or as monotherapy is efficacious in patients with or without diabetes; this is consistent with the overall efficacy findings from all three phase III studies [
31‐
33]. Furthermore, there were no major differences in the safety profile of sarilumab in patients with RA in the presence or absence of diabetes, although patients with uncontrolled diabetes and therefore potentially greater risk were excluded from these studies. These findings are reassuring given the possible vulnerabilities associated with diabetes (e.g. patients with poor glycaemic control have greater susceptibility to developing infections, especially if they are older or more likely to be receiving oral glucocorticoids [
37]). In this analysis, the rate of serious infections, opportunistic infections and infections leading to treatment discontinuation was similar in patients with or without diabetes, despite a numerically higher rate of decreased ANC in patients with diabetes. Patients with treated diabetes are also at risk for hypoglycaemia, especially those whose treatment includes insulin or an insulin secretagogue [
38]. There were no reports of symptomatic or biochemical hypoglycaemia in any of the studies, even though approximately 17% of the patients with diabetes were taking ≥ 2 noninsulin blood glucose-lowering medications and approximately 8% were using insulin or an insulin secretagogue. Modelling to assess the effect of concomitant use of hydroxychloroquine, known to induce symptomatic hypoglycaemia, by diabetic status showed no significant interaction at week 12 or 24 (data not shown). Changes in lipid parameters were generally comparable in the presence or absence of diabetes, and no interaction between diabetes and weight gain was observed. This finding may be particularly important, given the association between RA and comorbid diabetes and cardiovascular risk [
39].
These analyses have some notable limitations: they were conducted post hoc; and none of the studies included were designed specifically to evaluate HbA1c levels. The patients who had diabetes were selected on the basis of previous clinical history or current antidiabetic medication use, and patients with uncontrolled diabetes, defined as HbA1c ≥ 9.0%, were excluded. Whilst this exclusion is standard in trials, it does introduce a bias in the study population when compared with the general population. In addition, diet and exercise were not monitored systematically during the study, and there were no specific recommendations to maintain dietary or exercise habits in the individual study protocols. In the monotherapy study, the number of patients with diabetes in the two treatment groups was small. The proportion of non-white patients, in whom the prevalence of diabetes may differ, was low in the pooled studies. No analyses were performed to investigate any differences between such populations. Despite these limitations, findings across the three studies were consistent.
Although the analyses were conducted post hoc, data were collected in a prospective and blinded manner, and blood samples were analysed by a central laboratory, reducing assay variability. Although the numbers of diabetic patients in the sarilumab studies were small, they were larger than those reporting reductions in HbA1c in previous studies of tocilizumab, in which patients were treated openly (
n = 10 [
40] and
n = 34 [
41]) or only evaluated HbA1c after 104 weeks [
42]. In none of the aforementioned tocilizumab studies nor during the course of randomised, well-controlled, multinational studies designed to support marketing authorisation of a therapy for RA was HbA1c collected, nor were patients as thoroughly characterised with respect to concomitant glucocorticoid use, treatment for diabetes, or safety of the treatments. Other studies of the effects of IL-6 receptor blockade on insulin sensitivity/resistance in patients with RA excluded patients with diabetes [
24,
43].
Although chronic inflammation has long been implicated as a mediator of insulin resistance and β cell failure in patients with diabetes, the literature provides no clear guidance on the use of bDMARDs in patients with RA comorbid with diabetes. Studies of acute cytokine infusion on glucose metabolism in healthy volunteers may be inappropriate models for chronic inflammatory disease, given that they have produced contradictory results: acute infusion of human recombinant IL-6 has resulted in both an increase in fasting glucose concentrations [
44] and a decrease in postprandial glucose concentrations with increased insulin sensitivity [
45].
To date, the impact of TNFi on HbA1c, insulin sensitivity/resistance or pancreatic β cell function is unclear [
46‐
52]. An infusion of human recombinant TNFα has been shown to increase [
53], as well as decrease [
54], insulin sensitivity in healthy volunteers. Symptomatic hypoglycaemia has been reported in patients with diabetes and RA or psoriasis who were treated with etanercept [
55‐
58]. Properly designed clinical trials testing the effect of biologic anti-inflammatory drugs inhibiting TNF (e.g. CDP-571, etanercept) are lacking.
The evidence for IL-1 signalling involvement in glucose regulation is more supportive. A recent meta-analysis of > 2900 patients T2D treated with biologics that block IL-1 signalling (anakinra, canakinumab, gevokizumab, LY2189102) demonstrated a significant overall reduction in HbA1c of 0.32%; this included a study of patients with RA and T2D that showed a reduction of > 0.8% [
21].
However, no bDMARD has yet been recommended or approved for the treatment of diabetes, although a clinical trial to assess the potential efficacy of anti-IL-6 therapy (tocilizumab) in patients with type 1 diabetes is in progress [
59]. A better understanding of the potential differences in the effect of IL-6R blockade vs. IL-1β/TNFα antagonism on common comorbidities in RA should lead to more informed and individually tailored choices in RA disease management.
Competing interests
MCG has received research grants and consulting fees or other remuneration (payment) from Sanofi Genzyme, Genentech, Roche and R-Pharm.
GRB has received research grants from AbbVie, Pfizer, UCB and Roche and has received consulting fees or other remuneration (payment) or participated in speakers’ bureaus from/for AbbVie, Lilly, Merck Sharp & Dohme, Pfizer, Roche, Sanofi and UCB.
TM-P has received consulting fees from Sanofi.
OH, MI-R and KT are employees of Sanofi Genzyme and may hold stock and/or stock options in the company.
GSJ is an employee of Regeneron Pharmaceuticals, Inc. and may hold stock and/or stock options in the company.
MAG-G has received research grants and consulting fees from AbbVie, Roche, Sanofi, Eli Lilly and Novartis.
RF has received research grants from AbbVie, Acea, Amgen, Bristol-Myers Squibb, Celgene, Celltrion, Genentech, Eli Lilly, GlaxoSmithKline, Pfizer, Roche, Sanofi and UCB and has received consulting fees from AbbVie, Akros Pharma, Amgen, Bristol-Myers Squibb, Eli Lilly, Pfizer, Samsung, Taiho and UCB.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.