Influence of inflammatory and non-inflammatory rheumatic disorders on the clinical and biological profile of type-2 diabetes

Abstract

Objective

To study the profile of type-2 diabetes (T2D) in patients with RA or OA.

Methods

This observational, multicentre, cross-sectional study included, over a 24-month period, consecutive patients with adult-onset diabetes and RA or OA. We collected demographics, disease activity and severity indices, current treatments for RA and diabetes, history and complications of diabetes. A systematic blood test was performed, assessing inflammatory, immunological and metabolic parameters. The homoeostasis model assessment (HOMA)2-S was used to assess insulin resistance.

Results

We included 167 patients with T2D, 118 with RA and 49 with OA. RA and OA patients had severe T2D with suboptimal metabolic control and a biological profile of insulin resistance. Insulin resistance was significantly higher in RA than in OA patients after stratification on age, BMI and CS use [HOMA2-S: 63.5 (35.6) vs 98.4 (69.2), P < 0.001]. HOMA2-S was independently associated with DAS28 [odds ratio (OR): 4.46, 95% CI: 1.17, 17.08]. T2D metabolic control was not related to disease activity and functional impairment, but HbA1c levels were independently associated with bone erosions (OR: 4.43, 95% CI: 1.18, 16.61). Treatment with low-dose CSs was not associated with decreased insulin sensitivity or increased HbA1c levels. Treatment with TNF-α inhibitors was associated with increased insulin sensitivity compared with patients not receiving biologics [101.3 (58.71) vs 60.0 (32.5), P = 0.001].

Conclusion

RA patients display severe T2D with inflammation-associated insulin resistance. These findings may have therapeutic implications, with the potential targeting of insulin resistance through the treatment of joint and systemic inflammation.

rheumatoid arthritis, osteoarthritis, type-2 diabetes, insulin resistance

Rheumatology key messages

• Patients with rheumatoid arthritis (RA) or osteoarthritis (OA) display severe type-2 diabetes with suboptimal metabolic control.

• Insulin resistance is significantly higher in RA than OA patients after stratification on age, BMI and corticosteroid use.

• RA patients have a profile of insulin resistance associated with joint and systemic inflammation.

Introduction

RA is the most common cause of chronic inflammatory arthritis with a prevalence estimated at 0.5–1% of the adult population around the world. Besides being a source of disability, accumulating evidence has demonstrated an increased cardiovascular (CV) risk in RA compared with the general population [1, 2]. RA leads to accelerated atherosclerosis and is considered an independent CV risk factor [3]. Although a large part of CV risk pertains to the disease per se through chronic inflammation, the weight of traditional cardiovascular risk factors remains substantial [4]. OA is also a major cause of disability, morbidity and healthcare expenditure, affecting ∼15% of the population [5]. Patients with OA are almost three times more likely to develop CV disease or heart failure than those without OA [6]. Beyond obesity-related OA, some epidemiological studies suggest that several CV risk factors may participate in OA pathogenesis [7]. Among these factors, diabetes is the fastest increasing disease worldwide and a substantial threat to human health [8]. Diabetes is a metabolic disease that is heterogeneous with regard to clinical presentation and progression. It is characterized by chronic hyperglycaemia resulting from a variable combination of insulin resistance and defects of insulin secretion [9]. Type-2 diabetes (T2D) is the most frequent form of diabetes; it is characterized by insulin resistance and is thought to concern, 75–85% of the patients with diabetes [10].

RA and T2D share several similarities; they are both frequent and chronic diseases involving sedentary individuals aged over 45 years, and they are characterized by insulin resistance. The frequency of T2D in RA is estimated to be between 10 and 15% [11–13]. Increased risk of T2D in patients with RA can be related to long-term CS therapy, obesity, physical inactivity, lifestyle factors and chronic joint/systemic inflammation [14–17]. Until now, no data are available regarding the profile of adult-onset diabetes in patients with RA and the influence of this chronic inflammatory disease and its treatments on T2D-related complications, metabolic control and insulin resistance. On the other hand, high frequency of OA in patients with T2D has been reported, and the association between both diseases may represent a further step towards the individualization of T2D-related OA within a metabolic OA phenotype [18]. The objective of this study was to compare the profile of T2D in patients with RA or OA, two chronic diseases leading to disability differing by their joint and systemic inflammation levels, and identify the disease characteristics associated with insulin resistance and metabolic control.

Methods

Study design

We conducted an observational, multicentre, cross-sectional usual-care study, including seven rheumatology departments.

Inclusion criteria

We included, over a 24-month period, consecutive patients >18 years of age with RA or OA and already diagnosed or newly diagnosed T2D. T2D was defined according to the American Diabetes Association [19]. Patients with T1D and the following other specific types of diabetes were excluded: latent autoimmune diabetes in adults (LADA), considered according to the Immunology of Diabetes Society, who have established three main criteria including: (i) adult age of onset (>30 years), (ii) presence of glutamic acid decarboxylase antibodies (GADA), and (iii) absence of insulin requirement for at least 6 months after diagnosis [20], monogenic diabetes, disease of the exocrine pancreas, endocrine disorders, drug-induced (excepting CSs), infection-related diabetes and unclassified diabetes.

Patients with RA fulfilled the 2010 ACR/EULAR classification for RA [21]. Control diabetic patients had knee OA or hip OA according to ACR clinical classification criteria for OA of the knee or hip, respectively [22, 23]. All included patients agreed to participate in the study after informed consent, which was recorded in the medical source file. The protocol and the informed consent document have received Institutional Review Board/Independent Ethics Committee (IRB/IEC) approval before initiation of the study (‘Comité de Protection des Personnes’ Paris Ile de France IV, no. 2015/51NI).

Data collection from RA and OA patients

History taking, physical examination and laboratory tests were collected during the inclusion visit of RA and OA patients. CV risk factors (sedentary lifestyle according to the patient's opinion, smoking status, BMI, high blood pressure), acute and chronic complications of diabetes (including severe infections requesting hospitalization microangiopathy and macroangiopathy) and current/past medication use for RA and diabetes were obtained from information provided by patients and from the review of medical records. Waist circumference was measured and considered as increased if >88 cm in women and >102 cm in males. RA disease activity was assessed using the Disease Activity Score based on evaluation of 28 joints (DAS28) [24], using ESR [25]. Health status was measured by the self-administered Stanford HAQ. Systematic hand and foot X-rays were used to measure joint destruction, defined by the presence of erosions.

Laboratory tests

Routine laboratory study tests included Westergren ESR (SEDI 15, Becton Dickinson), CRP concentrations, serum creatinine concentration, fasting glycaemia (all with Cobas 8000, Roche Diagnostics France, Meylan, France) and insulin levels (MAGLUMI Insulin, CLIA, Shenzhen New Industries Biomedical Engineering Co, Shenzhen, CHINA). Metabolic control was reflected by HbA1c, which was measured by capillary electrophoresis (Capillarys 3, Sebia, Evry, France). Rheumatoid factor (RF) and second-generation anti-CCP peptide (anti-CCP2) antibodies were detected by immunoturbidimetry (positive cutoff >14 IU/ml) and electrochemiluminescence (positive cutoff >17 IU/ml), respectively. GADAs were measured with radiobinding assays using ³⁵S-labelled protein-11 (positive cutoff: >0.5 U/ml). The homoeostasis model assessment (HOMA)2 estimates of β-cell function (HOMA2-B), tissue insulin sensitivity (HOMA2-S) and insulin resistance (HOMA2-IR) based on insulin concentration were calculated with the HOMA calculator (Diabetes Trials Unit, University of Oxford, Oxford, UK). Insulin resistance was defined by HOMA2-IR >1.8 [26].

Statistical analysis

All data were expressed as mean (s.d) or number and percentage for continuous and categorical variables, respectively, unless stated otherwise. Statistical analysis was performed using Medcalc v18.9.1 (MedCalc Software Ltd, Ostend, Belgium). Correlations between numeric variables were assessed using Spearman’s rank correlation test. For a two-group comparison, Student’s unpaired t-test was used (continuous variables). The χ² test was used to seek for differences in frequency (binary variables). Three group comparisons were analysed using the Kruskal–Wallis test with Dunn’s correction. A multivariate analysis by logistic regression was also performed to determine the factors independently associated with increased insulin sensitivity or increased HbA1c. All relevant identified covariates with a P-value ≤0.1 in the single variable analysis were then entered in one single step in each model. Odds ratio (OR) and 95% CIs were then calculated. In this model, a P-value <0.05 was considered statistically significant. Discrimination capacities of HOMA2-S to identify patients with a DAS28 > 3.2 were determined by receiver operating characteristic (ROC) curve analysis.

Results

Study population

We initially included 176 patients with a clinical profile of T2D. Among these patients, nine were further excluded because of positive GADA (five with OA and four with RA) and a diagnosis of LADA.

We finally considered 118 RA patients: 75% women, mean age 63.8 (11.7) years, mean disease duration 15.6 (11.2) years, 75% with positive ACPA antibodies and 63% with erosive disease. The proportion of RA patients long-term treated with oral CSs was 64% (70/110 patients with available data): 60 patients were treated with a dose <10 mg/day, eight patients received a dose of 10 mg/day, and two patients a dose >10 mg/day (12.5 and 15 mg/day, respectively). Methotrexate and other conventional synthetic DMARDs were used in 75/115 (65%) and 23/114 (20%) patients with RA, respectively and 59/110 (54%) were treated with targeted biologic therapies. We recruited 49 patients with T2D (73% women) with OA (30 knee OA and 19 hip OA), who were significantly older [69.6 (10.9) years, P = 0.003] and had a significantly higher BMI [31.8 (6.3) vs 27.7 (5.5) kg/m², P < 0.001] than patients with RA. Detailed characteristics of our study sample are provided in Table 1.

Table 1

Disease characteristics of patients with RA and OA

Characteristic	RA + type-2 diabetes	OA + type-2 diabetes	P-value
	(n = 118)	(n = 49)
Age, mean (s.d.), years	63.8 (11.7)	69.6 (10.9)	0.003
Females, n (%)	88/118 (75)	36/49 (73)	0.788
BMI, mean (s.d.), kg/m2	27.7 (5.5)	31.8 (6.3)	<0.001
Smokers, n (%)	18/97 (19)	5/46 (11)	0.229
High blood pressure, n (%)	69/99 (70)	37/49 (76)	0.446
Dyslipidaemia, n (%)	45/98 (46)	26/45 (58)	0.144
Sedentary lifestyle, n (%)	55/96 (57)	34/47 (72)	0.084
Waist circumference, mean (s.d.), cm	101 (15)	105 (17)	0.133
Disease duration, mean (s.d.), years	15.6 (11.2)
Erosions, n (%)	67/107 (63)
Positive anti-CCP2 antibodies, n (%)	85/114 (75)
Positive rheumatoid factor, n (%)	85/113 (75)
DAS28, mean (s.d.)	2.99 (1.35)
HAQ, mean (s.d.)	0.97 (0.82)
CRP, mean (s.d.), mg/l	9.7 (16.2)
ESR, mean (s.d.), mm H1	24 (20)
Current CS use, n (%)	70/110 (64)
CS dose, mean (s.d.), mg/day	3.8 (3.5)
Current methotrexate use, n (%)	75/115 (65)
Methotrexate dose, mean (s.d.), mg/week	17.5 (5.0)
Other csDMARDs	23/114 (20)
Leflunomide	13/114 (12)
Salazopyrine	2/114 (2)
Hydroxychloroquine	8/114 (7)
Current targeted biologic treatment, n (%)	59/110 (54)
TNF-α inhibitors, n (%)	21 (36)
Rituximab, n (%)	19 (33)
Tocilizumab, n (%)	14 (24)
Abatacept, n (%)	5 (9)

csDMARD: conventional synthetic DMARD; DAS28, Disease Activity Score based on evaluation of 28 joints.

Characteristics of T2D in patients with osteoarthritis and rheumatoid arthritis

The characteristics of T2D in OA patients were as follow: mean age >65 years, high BMI >30 kg/m², 26.5% of insulin requirement, high frequency of cardiovascular risk factors (Table 1), macroangiopathy found in 57.8% of patients, biological criteria of insulin resistance found in 65% of patients (HOMA2-IR > 1.8) and HOMA-IR was 2.5 (1.9) (Table 2).

Table 2

Characteristics of type-2 diabetes in both populations

Characteristic	RA + type-2 diabetes	OA + type-2 diabetes	P-value
	(n = 118)	(n = 49)
Disease duration, mean (s.d.), years	11.2 (10.5)	10.6 (7.3)	0.716
Specialized follow-up in diabetology, n (%)	36/76 (47.4)	22/37 (59.5)	0.229
Insulin requirement, n (%)	33/118 (27.9)	13/49 (26.5)	0.854
Diabetic retinopathy, n (%)	14/77 (19.2)	6/41 (14.6)	0.729
Diabetic neuropathy, n (%)	14/73 (19.2)	4/34 (11.7)	0.336
Diabetic nephropathy, n (%)	8/89 (8.9)	6/45 (13.3)	0.432
Creatinin levels, mean (s.d.), µmol/l	77 (75)	90 (38)	0.251
Macroangiopathya	39/90 (43.3)	26/45 (57.8)	0.113
Severe infectionb	14/118 (11.9)	6/49 (12.2)	0.957
HbA1c, mean (s.d.)	7.0 (1.19)	7.3 (1.29)	0.150
HOMA2-Bc, mean (s.d.)	86.5 (65.9)	83.4 (65.9)	0.782
HOMA2-Sc, mean (s.d.)	75.1 (64.3)	73.5 (65.9)	0.885
HOMA2-IRc, mean (s.d.)	2.1 (1.4)	2.5 (1.9)	0.134
Treatment of type-2 diabetes
Biguanides
Metformin, n (%)	65/118 (55.0)	29/49 (59.1)	0.628
Sulfonylureas
Gliclazide, n (%)	17/118 (14.4)	9/49 (18.4)	0.518
Glimepiride, n (%)	5/118 (4.2)	5/49 (10.2)	0.137
Glibenclamide, n (%)	2/118 (1.7)	0/49 (0.0)	0.360
Incretin mimetics
GLP-1 analogues
Liraglutide, n (%)	5/118 (4.2)	2/49 (4.1)	0.977
Dulaglutide, n (%)	1/118 (0.8)	4/49 (8.2)	0.011
DPP-4 inhibitors
Sitagliptin, n (%)	4/118 (3.4)	4/49 (8.2)	0.188
Vildagliptin, n (%)	1/118 (0.8)	0/49 (0.0)	0.531
Saxagliptin, n (%)	1/118 (0.8)	0/49 (0.0)	0.531
Meglitinides
Repaglinide, n (%)	5/118 (4.2)	7/49 (14.3)	0.022
α-Glucosidase inhibitors
Acarbose, n (%)	1/118 (0.8)	0/49 (0.0)	0.531
Combination of oral treatments
Metformin + sitagliptin, n (%)	10/118 (8.5)	4/49 (8.2)	0.949
Metformin + vildagliptin, n (%)	2/118 (1.7)	0/49 (0.0)	0.360
Metformin + saxagliptin, n (%)	2/118 (1.7)	0/49 (0.0)	0.360
Insulin
Long-acting insulin, n (%)	22/118 (19.0)	11/49 (22.4)	0.618
Intermediate-acting insulin, n (%)	12/118 (10.2)	6/49 (12.2)	0.705
Short-acting insulin, n (%)	11/118 (9.3)	5/49 (10.2)	0.857
Combination of insulin and oral treatment
Insulin + liraglutide, n (%)	2/118 (1.7)	0/49 (0.0)	0.360

a

Coronary arterial disease, history of myocardial infarction, history of ischemic stroke, peripheral artery occlusive disease.

b

Infection requiring hospitalization.

c

Evaluated after excluding type-2 diabetes patients treated with insulin. DPP-4: dipeptidyl peptidase-4; GLP-1: glucagon-like peptide 1; HOMA2: homoeostasis model assessment 2; HOMA2-B: HOMA2 estimate of β-cell function; HOMA2-IR: HOMA2 estimate of insulin resistance; HOMA2-S: HOMA2 estimate of insulin sensitivity.

Despite significantly younger age and lower BMI, patients with RA displayed similar T2D characteristics, with 27.9% of insulin requirement, 43.3% of macroangiopathy and HOMA-IR similar to that of OA. The frequency of microvascular complications (diabetic retinopathy, nephropathy and neuropathy) was not different between OA and RA patients (Table 2). The metabolic control of patients with OA and RA was comparable, with a mean HbA1c of 7.3% (1.29%) and 7.0% (1.19%), respectively. The two main oral treatments used for T2D in OA and RA patients were metformin (59.1% and 55.0%, respectively) and gliclazide (18.4% and 14.4%, respectively) (Table 2). Dulaglutide (glucagon-like peptide 1 analogue) and repaglinide (meglitinides) were more frequently prescribed in OA patients, but they concerned a limited number of patients only (Table 2).

Insulin resistance profile of T2D patients with osteoarthritis and rheumatoid arthritis

The insulin resistance profile of OA and RA patients not treated with insulin was similar (Table 2). However, after adjusting for age, BMI and CS use, RA patients had a significantly higher insulin secretion [HOMA2-B: 91.8 (68.3) vs 49.3 (25.7), P < 0.001] and insulin resistance [HOMA2-IR: 1.94 (1.35) vs 1.10 (0.71), P < 0.001] compared with OA patients, as well as a significant reduction of insulin sensitivity [HOMA2-S: 63.5 (35.6) vs 98.4 (69.2), P < 0.001].

Factors associated with insulin resistance and metabolic control in patients with osteoarthritis

As expected, in the population of OA patients with T2D not treated with insulin, negative correlations were observed between insulin sensitivity (HOMA2-S) and age (r = −0.41, P = 0.037), BMI (r = −0.50, P < 0.001) as well as abdominal perimeter (r = −0.45, P = 0.001). HbA1c did not correlate with age (r = 0.03, P = 0.853), BMI (r = −0.16, P = 0.284), abdominal perimeter (r = −0.03, P = 0.864) and HOMA2-S (r = −0.08, P = 0.634), but was negatively correlated with β-cell function, reflected by the HOMA2-B (r = −0.45, P = 0.005). HbA1c was also significantly higher in patients with insulin requirement [8.06% (1.53%) vs 6.96% (0.99%), P = 0.004].

Factors associated with insulin resistance in patients with rheumatoid arthritis

In RA patients with T2D not treated with insulin, HOMA2-S did not correlate with age (r = −0.19, P = 0.174), BMI (r = −0.15, P = 0.165) and abdominal perimeter (r = −0.26, P = 0.064). Decreased insulin sensitivity was associated with RA disease activity, with a negative correlation between the DAS28 and HOMA2-S (r = −0.29, P = 0.011). Indeed, patients with moderate (3.2 ≤ DAS28 ≤ 5.1) or high (DAS28 > 5.1) disease activity, had significantly decreased HOMA2-S when compared with patients with moderate or low disease activity (Fig. 1). Insulin sensitivity was not modified by RA-related physical incapacity assessed by the HAQ (r = −0.11, P = 0.411) or the presence of erosions [67.9 (50.2) vs 68.4 (46.3), P = 0.968]. No association was observed between disease activity or severity indices and HOMA2-B.

Fig. 1

Insulin sensitivity assessed by the HOMA2-S according to DAS28

HOMA2-S was evaluated in the subgroups shown after excluding type-2 diabetes patients treated with insulin. Statistical test: Kruskal–Wallis test with Dunn’s correction, *P < 0.05. DAS28: Disease Activity Score based on evaluation of 28 joints; HOMA2-S: homoeostasis model assessment 2 estimate of insulin sensitivity.

A ROC curve analysis was then carried out to analyse the discrimination capacity of HOMA2-S to identify patients with a DAS28 > 3.2. This analysis provided an area under the ROC curve of 0.70 (P = 0.010) for a threshold of HOMA2-S < 42. To determine whether decreased insulin sensitivity was independently associated with disease activity, we created logistic regression models, including HOMA2-S (<42) as the dependent variable and other relevant factors known to influence insulin resistance as independent variables. This analysis confirmed an independent association between the HOMA2-S index and a DAS28 > 3.2 (OR: 4.46, 95% CI: 1.17, 17.08), and revealed a second independent association with increased CRP levels (OR: 6.92, 95% CI: 1.61, 29.71) (Table 3).

Table 3

Factors independently associated with decreased insulin sensitivity in multivariate logistic regression analyses

Modela	Univariate P-value	OR (95% CI)	P-value
Model 1
Age	0.486	0.99 (0.93, 1.05)	0.784
BMI > 30	0.621	0.56 (0.14, 2.20)	0.408
DAS28 > 3.2	0.042	4.35 (1.27, 14.94)	0.019
Model 2
Age	0.486	0.99 (0.93, 1.06)	0.920
BMI > 30	0.621	0.59 (0.13, 2.52)	0.474
DAS28 > 3.2	0.042	4.46 (1.17, 17.08)	0.029
Sedentary lifestyle	0.093	1.64 (0.39, 6.79)	0.494
CS use	0.100	1.17 (0.25, 5.51)	0.846
Model 3
Age	0.486	0.97 (0.91, 1.04)	0.439
BMI > 30	0.621	0.58 (0.12, 2.86)	0.508
Sedentary lifestyle	0.093	1.59 (0.37, 6.85)	0.534
CS use	0.100	0.61 (0.11, 3.41)	0.576
CRP >10 mg/l	0.011	6.92 (1.61, 29.71)	0.009

a

These multivariate models included variables with P-value <0.1, as well as age and BMI given that these factors are well known risk factors of insulin resistance and were inversely correlated with HOMA2-S in our series of OA patients with T2D. DAS28: Disease Activity Score based on evaluation of 28 joints; HOMA2-S: homoeostasis model assessment 2 estimate of insulin sensitivity; OR: odds ratio.

Factors associated with metabolic control of T2D in patients with rheumatoid arthritis

HbA1c correlated with diabetes duration (r = 0.36, P = 0.004), was significantly higher in patients with insulin requirement [7.45 (0.85)% vs 6.92 (1.17)%, P = 0.037] and negatively correlated with HOMA2-B (r = −0.38, P < 0.001). No correlation was observed between HbA1c and age, BMI (r = −0.05, P = 0.607), abdominal perimeter (r = −0.03, P = 0.859) and HOMA2-S (r = −0.11, P = 0.379).

Regarding the potential link between metabolic control and disease activity, no correlation was observed between HbA1c levels and the DAS28 (r = −0.03, P = 0.771) or CRP levels (r = −0.029, P = 0.776) in patients with RA and T2D.

Regarding the link between metabolic control and functional impairment, no correlation was detected between HbA1c levels and HAQ (r = −0.14, P = 0.243) and HbA1c was not modified according to the presence of a sedentary lifestyle [7.15% (1.06%) vs 6.85% (1.43%), P = 0.252]. However, HbA1c levels were significantly higher in patients with bone erosions [7.32% (1.13%) vs 6.77% (0.91%), P = 0.019]. In a multivariate logistic regression model, the presence of erosions was independently associated with HbA1c >7%, taken as the dependent variable (OR: 4.43, 95% CI: 1.18, 16.61) (Table 4)

Table 4

Factors independently associated with HbA1c >7% in multivariate logistic regression analyses

Variable	Univariate P-value	OR (95% CI)	P-value
Agea	0.101	1.02 (0.95, 1.11)	0.498
Diabetes durationa	0.042	1.09 (0.99, 1.20)	0.072
Insulin requirementa	0.008	1.33 (0.26, 6.89)	0.727
Bone erosiona	0.091	4.43 (1.18, 16.61)	0.027
Sedentary lifestyle	0.215
BMI > 30	0.385
DAS28 > 3.2	0.677
CRP >10 mg/l	0.768
HAQ >1.5	0.485
CS use	0.431

a

Variables included in the multivariate analysis. DAS28: Disease Activity Score based on evaluation of 28 joints; OR: odds ratio.

Effects of rheumatoid arthritis therapies on insulin resistance profile and metabolic control of T2D

The frequency of microvascular and macrovascular complications, as well as severe infections, was not different according to the use of CSs (Supplementary Table S1, available at Rheumatology online). Chronic treatment with CSs <10 mg/day was not associated with decreased insulin sensitivity [77.4 (73.4) vs 69.9 (47.4), P = 0.627], altered β-cell function [88.6 (71.1) vs 78.9 (61.1), P = 0.547] or increased HbA1c [7.02 (1.10) vs 7.13 (1.41), P = 0.666] in patients with T2D and RA. Treatment with methotrexate and hydroxychloroquine did not significantly modify insulin sensitivity [78.1 (69.3) vs 70.9 (55.5), P = 0.629 and 67.5 (67.1) vs 77.5 (65.2), P = 0.699, respectively], β-cell function [85.3 (68.1) vs 85.6 (67.6), P = 0.986 and 88.2 (53.1) vs 86.7 (69.6), P = 0.957, respectively] and HbA1c levels [7.13  (1.26) vs 6.89 (1.07), P = 0.359 and 6.77 (1.34) vs 7.07 (1.20), P = 0.497, respectively].

Regarding targeted biologic therapies, patients with RA and T2D treated with TNF-α inhibitors were more likely to have decreased DAS28 [2.38 (1.03) vs 3.28 (1.29), P = 0.012] and increased insulin sensitivity [101.3 (58.71) vs 60.0 (32.5), P = 0.001] compared with patients not receiving biologics (Fig. 2). However, treatment with TNF-α inhibitors was not associated with better β-cell function [84.7 (67.4) vs 86.3 (69.3), P = 0.934] and metabolic control [HbA1c: 7.22 (2.34) vs 7.07 (1.12), P = 0.717]. Other targeted biologic treatments (tocilizumab, abatacept and rituximab) did not modify insulin sensitivity (Fig. 2), β-cell function and HbA1c.

Fig. 2

Effects of targeted biologic therapies on insulin sensitivity (HOMA2-S)

HOMA2-S was evaluated in these subgroups after excluding type-2 diabetes patients treated with insulin. Statistical test: Kruskal–Wallis test with Dunn’s correction. ABA: abatacept; HOMA2-S: homoeostasis model assessment 2 estimate of insulin sensitivity; RTX: rituximab; TBT: targeted biologic therapies; TNF-I: TNF-α inhibitors; TOCI: tocilizumab.

Discussion

We defined the clinical and biological profile of RA and OA patients with T2D recruited from seven rheumatology departments in Paris. OA and RA patients frequently displayed poorly controlled diabetes, highlighting the burden of this comorbidity in patients hospitalized in rheumatology departments [27]. The clinical–biological profile of diabetic OA patients was that of severe insulin-resistant diabetes correlated to age and obesity [9]. RA patients with T2D displayed a similar biological profile of insulin resistance, which was, however, not related to age and BMI unlike OA patients, but was related to the joint and systemic inflammatory activity of the disease. These results are consistent with those previously reported in the population of RA patients without diabetes. Indeed, RA patients without T2D exhibit higher insulin resistance compared with the general population, whose severity is associated with RA disease activity and systemic inflammation [28–30].

Insulin sensitivity and insulin secretion are linked by a negative feedback loop. In this regard, when an individual becomes more insulin resistant, β-cell function is enhanced. In our population of RA patients, inflammatory status was associated with decreased insulin sensitivity (HOMA2-S) but not with enhanced β-cell function (HOMA2-B), which may be a consequence of the chronic hyperglycaemia-related decline of β-cell function and mass [31].

Interestingly, disease activity, despite more severe insulin resistance, was not associated with poorer metabolic control of T2D. This finding may be explained by the fact that metabolic control was rather related to β-cell function (which mainly depends on the degree of hyperglycaemia) than insulin resistance [31].

We report for the first time an independent association between HbA1c and bone erosions. Indeed, we observed that patients with bone erosions were more likely to have poorer metabolic control, not related to CS use, physical incapacity assessed by the HAQ nor sedentary lifestyle. Other contributors to chronic disease-associated bad prognosis should be studied to explain this potential association, including poor socio-economic status or absence of regular medical follow-up.

CSs reduce insulin sensitivity, can impair glucose tolerance and increase the risk of T2D [32, 33]. In our population, 64% of RA patients with diabetes were, however, treated with CSs, which was not associated with higher frequency of diabetes-related complications. Moreover, insulin sensitivity, β-cell function and metabolic control did not differ between RA patients with T2D treated or not with CSs. A limitation of our study is the absence of data collection regarding cumulative dose of CSs, which may have affected our results. However, the average corticoid dose was actually relatively low [3.8 (3.5) mg/day] and only two patients received a daily dose >10 mg.

The inflammatory component of T2D has been increasingly recognized [34, 35]. Several inflammatory mediators including IL-1β, IL-6 and TNFα, released from adipose tissue favour the development of insulin resistance and impaired insulin secretion, thus promoting initiation and progression to T2D [36, 37]. Clinically, although conducted with small sample sizes, two randomized trials of the IL-1 receptor antagonist anakinra, including one performed on patients with RA and T2D, showed improvements in patients with established T2D [38, 39]. More recently, treatment over a median period of 3.7 years with canakinumab reduced HbA1c during the first 6–9 months of treatment but failed to reduce incident T2D [40]. TNFα is also able to promote insulin resistance by inducing the synthesis of SOCS3, an inhibitor of the insulin signalling pathway, decreasing the expression of the insulin receptor IRS-1 and of the glucose transporter Glut4 and inhibiting the synthesis of peroxisome proliferator-activated receptor γ, which plays a key role in adipogenesis and, therefore, in insulin sensitivity [41]. In our cohort, treatment with TNF-α inhibitors was associated with reduced disease activity and increased insulin sensitivity. This observation may be related to a better disease control in patients with TNF-α inhibitors, with a mean DAS28 of 2.38 (1.03), indicating remission, compared with rituximab, tocilizumab and abatacept with respective DAS28 of 2.86 (1.57), 2.82 (1.31) and 2.77 (1.08), indicating low disease activity. This result is consistent with the results of a prospective study of 61 non-diabetic RA patients with active RA treated with TNF-α antagonists for 12 weeks, which showed that patients with high insulin resistance at baseline had a significant improvement of insulin sensitivity. However, TNF-α inhibitors were not associated with increased β-cell function and better metabolic control, and further prospective studies are necessary to determine whether this therapeutic class may improve the outcome of T2D. We did not observe increased insulin sensitivity in patients treated with other targeted biologic therapies (tocilizumab, abatacept and rituximab), conversely to preliminary previous studies, which have suggested that tocilizumab and abatacept may improve insulin sensitivity and decrease HbA1c [42, 43]. The low number of patients treated with these drugs may explain this result.

Our study included consecutive longstanding patients who were carefully assessed and phenotyped in tertiary centres with a long-lasting experience in RA evaluation and care. However, our study is limited by its observational design, and any pathogenic link emerging from this type of study should be taken very cautiously.

In conclusion, OA and RA patients with adult-onset diabetes display a clinical profile of severe T2D with suboptimal metabolic control. RA patients displayed a biological profile of insulin resistance associated with joint and systemic inflammation. These findings may have therapeutic implications, with the potential targeting of insulin resistance through the treatment of joint and systemic inflammation.

Acknowledgements

Prof. Laure Gossec, Sorbonne Université, Service de Rhumatologie, Hôpital Pitié Salpêtrière, AP-HP, Paris, France. Dr Camille Deprouw, Sorbonne Université, Service de Rhumatologie Hôpital Saint Antoine, AP-HP, Paris, France. Ms Sabrina Leveau and Ms Alexandra Bruneau (URC Cochin). Ms Carole Desbas and Ms Caude Lecoeur for expert secretariat assistance. Prof. Laure Gossec and Dr. Camille Deprouw for helping in the recruitment and the inclusion of patients. Ms Sabrina Leveau and Ms Alexandra Bruneau for the technical supervision of the study.

Funding: This work was supported by the Société Française de Rhumatologie and a Research grant from Bristol Myers Squibb. Bristol-Myers Squibb was not involved in the study design, data acquisition, data analysis, or writing of the manuscript.

Disclosure statement: J.A. has received research funding for this study from Bristol Myers Squibb. The other authors have declared no conflicts of interest.

Data availability statement

The authors declare that all data supporting the findings of this study are available within the paper and its supplementary material.

Supplementary data

Supplementary data are available at Rheumatology online.