Article Text
PDF
Abstract
Purpose Rheumatoid arthritis (RA) is associated with a higher risk of diabetes mellitus (DM). Our aim was to determine associations between inflammatory disease activity (including evaluation of specific cytokines and chemokines) and incident DM.
Methods Participants were adults with physician-confirmed RA from Veteran’s Affairs Rheumatoid Arthritis Registry. Disease activity and clinical assessments occur longitudinally as part of clinical care. Thirty cytokines and chemokines were measured in banked serum obtained at the time of enrolment. Cytokine/chemokine values were log-adjusted and standardised (per SD). Incident DM was defined based on validated algorithms using diagnostic codes and medications. Multivariable Cox proportional hazard models evaluated associations between clinical factors and incident DM. Independent associations between cytokines/chemokines and incident DM were assessed adjusting for age, sex, race, smoking, body mass index (BMI) and medication use at baseline.
Results Among 1866 patients with RA without prevalent DM at enrolment, there were 130 incident cases over 9223 person-years of follow-up. High Disease Activity Score (DAS28)-C reactive protein (CRP), obese BMI, older age and male sex were associated with greater risk for incident DM while current smoking and methotrexate use were protective. Patients using methotrexate were at lower risk. Several cytokines/chemokines evaluated were independently associated (per 1 SD) with DM incidence including interleukin(IL)-1, IL-6 and select macrophage-derived cytokines/chemokines (HR range 1.11–1.26). These associations were independent of the DAS28-CRP.
Conclusions Higher disease activity and elevated levels of cytokines/chemokines are associated with a higher risk of incident DM in patients with RA. Future study may help to determine if targeted treatments in at-risk individuals could prevent the development of DM.
- arthritis
- rheumatoid
- epidemiology
- inflammation
Data availability statement
Data may be obtained from a third party and are not publicly available. Data are not publicly available but may be requested from the VA Rheumatoid Arthritis Registry and Biorepository.
Statistics from Altmetric.com
Key messages
What is already known about this subject?
Some studies suggested a greater risk of diabetes in patients with rheumatoid arthritis and some suggest that certain therapies may reduce that risk.
What does this study add?
In this study, elevated disease activity was associated with a greater risk of diabetes.
Elevations in inflammatory cytokines and chemokines were also associated with a greater risk of diabetes.
How might this impact on clinical practice or future developments?
These data support closer attention to the risk of diabetes among patients with elevated disease activity and may support more aggressive treatment to reduce the risk.
Introduction
Patients with rheumatoid arthritis (RA) may be at greater risk of developing diabetes mellitus (DM), though studies are conflicted.1–5 This increase in risk is important since it might contribute to the observed higher risk of cardiovascular disease and premature mortality in this population.6 Some evidence suggests that systemic inflammation might directly lead to insulin resistance and poor insulin production by interfering with cellular functions in the pancreas, liver and skeletal muscle.7 8 Systemic inflammation has been considered to be a potential risk factor and has been associated with incident diabetes in the general population.9
Despite an emerging awareness of a potential link between inflammation and insulin resistance, few studies have evaluated the relationship between disease activity in patients with inflammatory diseases such as RA and the subsequent risk of DM. Some studies have demonstrated that certain treatments are associated with a lower risk of DM in patients with RA and psoriatic arthritis, perhaps suggesting a benefit of superior disease control, but these studies did not directly assess the effect of disease activity itself.10 11
Furthermore, a few studies have evaluated whether there are specific circulating inflammatory mediators that correlate more closely with the risk of diabetes in this population. In the general population, several cytokines and chemokines have been implicated in the development of DM including interleukin(IL)−1, IL-6, tumour necrosis factor (TNF)-α, IL-8, IL-18 and macrophage chemoattractant protein-1 (MCP-1).12–14 The identification of specific circulating cytokines or chemokines that are implicated in the risk of DM would have potential implications for predicting long-term risk and possibly guiding the choice of targeted therapies in patients with RA who are at risk for diabetes.15
Our specific aims were to evaluate associations between clinical disease activity and inflammatory markers, including individual cytokines/chemokines, with incident diabetes in a population of patients with RA. We hypothesised that inflammation related to RA disease activity would be associated with a higher risk of incident DM independent of other clinical factors. We also hypothesised that proinflammatory cytokines and chemokines might contribute to this increase in risk beyond traditional RA disease activity assessment.
Methods
Study setting
The Veterans’ Affairs Rheumatoid Arthritis (VARA) study is an ongoing national repository and multicentre RA registry that has been active for more than 17 years (initiated 2003).16–23 At the time this study was conducted, 13 VA sites had contributed data. Veterans with RA are identified during routine care by the treating rheumatologist at individual sites and consented for enrolment. All Veterans who fulfill the 1987 American College of Rheumatology classification criteria for RA and are over 18 years of age are eligible.24 Physician investigators at each site record clinical data at enrolment and at routine follow-up visits as part of normal clinical care. All study patients provided informed written consent. Patients and the public were not involved in the design of the VARA registry or the current study.
RA disease activity
The results of clinical testing of C reactive protein (CRP, mg/dL), clinical joint counts (0–28) and patient/physician global scores were extracted from the registry and from the VA electronic medical record by querying data in the Corporate Data Warehouse (CDW). Our primary disease activity measure was the Disease Activity Score in 28 Joints with the CRP (DAS28(CRP)) and was categorised as remission (<2.6); low disease activity (2.61–3.2); moderate activity (3.21–5.09) and high activity (>5.1).25 Missing components for the DAS28(CRP) were imputed by carrying forward from the prior visit.
Serologies, inflammatory markers, cytokine and chemokine assays
Cytokine and chemokines levels were determined by the V-PLEX multiplex panel from Meso Scale Discovery (Rockville, Massachusetts USA). These analytes were measured from serum obtained at the time of registry enrolment, the only time point for which samples are routinely banked for these study participants. Following sample collection, specimens were processed and stored at −70°C until time of measurement. Thirty analytes were examined at enrolment: IL-1α, IL‐1β, IL-1 receptor antagonist (RA), IL‐2, IL-3, IL‐4, IL‐5, IL‐6, IL‐7, IL‐8, IL-9, IL‐10, IL‐12p70, IL‐13, IL-15, IL-16, IL-17α, IL-23, IL-27, interferon (IFN)-γ, granulocyte‐macrophage colony‐stimulating factor, macrophage-derived chemokine (MDC), MCP-1, MCP-4, macrophage inflammatory protein (MIP)−1α, MIP‐1β, MIP-3α, Vascular Endothelial Growth Factor, eotaxin and TNF-α. Assays were performed as per manufacture protocols and analysed on the MESO QuickPlex SQ 120 imager (Meso Scale Discovery). From banked serum, a second-generation commercial anti-cyclic citrullinated peptide antibody and high sensitivity CRP (hsCRP) were also measured from banked enrolment serum as previously reported.26
DM outcome definition
We identified incident DM by querying inpatient and outpatient diagnoses as well as antidiabetic medications within the VA CDW. Incident diabetes was defined as fulfilling one of the following: (1) two or more outpatient diagnosis codes (International Classification of Diseases (ICD)−9 clinical modification (CM) or ICD-10-CM (online supplemental table 1), (2) one or more discharge diagnosis codes or (3) one outpatient diagnosis code and one or more antidiabetic medication (VA or non-VA medications). Similar approaches have demonstrated excellent sensitivity and specificity (83.2%/99.2%) as well as high positive and negative predictive value in administrative data (92.5%/98.1%).27 Any patient with a medication or diagnosis code occurring within 180 days after enrolment in the registry was considered a prevalent case.
Supplemental material
Covariables
Demographics and disease-specific characteristics at baseline and during follow-up were obtained from the VARA registry database. Current smoking was considered time-invariant (presence or absence of the exposure reported at baseline). Body mass index (BMI) was extracted from the vital sign packages available in the CDW and the closest BMI value (within 30 days) to the visit date was used. Observations with missing BMI data were imputed by carrying forward from the prior observation. BMI categories were defined as (underweight, <20 kg/m2; normal weight, ≥20–25 kg/m2; overweight, ≥25–30 kg/m2; obese, ≥30–35 kg/m2 and severely obese (≥35 kg/m2).
RA treatments were extracted from VA pharmacy databases. Each prescription fill of a drug was defined as a dispensing episode.19 For each episode, the amount of the drug dispensed and the expected duration of the treatment episode were determined. The expected days of supply were determined based on the dosing instructions. A drug course was defined as a period of continuous treatment consisting of one or more dispensing episodes without a gap of ≥90 days between the expected end of the days of supply for that episode and the start of the subsequent dispensing episode. Participants were considered exposed to the therapy if the current visit occurred during a defined medication course. Active glucocorticoid use was physician-reported and extracted from the registry database.
Statistical analysis
Characteristics of the study population were described among participants with prevalent and non-prevalent DM at enrolment. The primary analyses used Cox proportional hazards models to assess associations between baseline characteristics and the time to the development of incident DM among participants without prevalent DM at baseline, clustering on study site. Secondary analyses also incorporated time-varying assessments of disease activity, BMI and RA treatments (including glucocorticoids). Time-varying models provide an opportunity to assess the association with the most recently collected measure of the exposure rather than focusing on the baseline assessment. We focused on methotrexate, hydroxychloroquine, TNF-inhibitors and abatacept as potential confounders given prior data demonstrating potential reductions in risk with these therapies.11 Other hypothesised confounders included demographics, smoking, BMI, disease duration, anti-citrullinated peptide antibody (ACPA) status (positive vs negative) and calendar year.
Cytokines and chemokines were log-adjusted to approximate a normal distribution and standardised so that a 1-unit difference in the value represented a 1 SD difference for all individual analytes. Separate multivariable Cox proportional hazards models evaluated associations between each individual cytokine/chemokine and the time to the development of DM. Each of these models was adjusted for the factors defined above. Primary models did not adjust for clinical disease activity, however, we also explored models adjusting for disease activity. We performed Simes-Benjamini-Hochberg adjustment for multiple comparisons and noted cytokines and chemokines that remained significant after adjustment (p<0.008). We also assessed the improvement in model fit with the inclusion of key cytokines that remained significant after adjustment for multiple comparisons.
Results
A total of 2541 registry participants were evaluated at baseline. Of these, 2341 had cytokines and chemokine data available and 667 (26%) had prevalent DM. The enrolment characteristics of participants with and without prevalent DM are shown in table 1. Participants with diabetes were older, were less likely to be female, had higher BMI and had higher DAS28(CRP). Diabetics had lower levels of IL-4, IL-12, TNF-α and MCP-1 and higher levels of IL-1RA and IL-17α. There were no other significant differences in circulating cytokines or chemokines between diabetics and non-diabetics at enrollment.
Enrolment characteristics of those with prevalent diabetes compared with those without prevalent diabetes at baseline in the VARA registry
There were 1866 participants without DM at baseline that were included in longitudinal analyses. There were 130 RA patients who developed incident diabetes over 9223 person-years of follow-up, a rate of 1.41 cases per 100 person-years. Among those that developed diabetes, the median time to DM diagnosis was 4.7 (3.3) years. Higher disease activity was independently associated with a greater risk of DM in a dose-dependent manner, with high disease activity at baseline being associated with a significant increase in risk compared with those in remission (HR: 2.07 (95% CI 1.34 to 2.85) p<0.001) after adjustment for confounders (figure 1, table 2). A test for trend was significant (p<0.001).
Proportion of remaining subjects with diabetes by baseline disease activity category (unadjusted).
Multivariable associations between disease activity and other clinical factors and the risk of incident diabetes among participants without diabetes at enrolment (all variables included in a single model)
In models including time-varying measures of disease activity and covariates (including glucocorticoids), higher disease activity was again observed to be independently associated with a higher risk of DM (test for trend: p<0.001). Specifically, in time-varying models, moderate disease activity was associated with a 58% higher risk (HR 1.58 (95% CI 1.26 to 1.90) p<0.001). High disease activity was not significantly associated, though with low precision due to a low number of observations (HR 1.52 (95% CI 0.77 to 2.99) p=0.23). The average DAS28(CRP) over all prior observations was also associated with incident DM (HR (per one unit): 1.16 (95% CI 1.03 to 1.30) p=0.02). Low disease activity on average over all prior observations was not associated with a higher risk compared with remission (HR 1.00 (95% CI 0.75 to 1.34) p=0.99). In models incorporating all components of time-varying disease activity separately (swollen joint count, tender joint count, patient global score and CRP), only CRP was independently associated with the risk of DM (HR (per 1 mg/dL): 1.05 (95% CI 1.00 to 1.09) p=0.03 (full model not shown).
Prior to imputation for longitudinal measures, BMI was missing in 12% of observations and DAS28(CRP) was missing in 23%. In regression models that used only unimputed data, results were highly similar to the primary analysis, though high disease activity was significantly associated with the risk of DM (HR 1.97 (95% CI 1.01 to 3.87) p=0.04) (full model not shown).
Methotrexate use at baseline and as a time-varying covariate was associated with a lower risk of DM in these models. Baseline and time-varying prednisone use was not significantly associated with a higher risk of incident DM independent of measures of disease activity, BMI and other factors (table 2). Hydroxychloroquine and TNFi were not associated with diabetes independent of other factors. Female sex and current smoking were also associated with a substantially lower risk of DM. Higher BMI was the strongest risk factor for DM in these models, with severely obese participants at baseline having a sevenfold higher risk compared with those in the normal weight category (HR 6.80 (95% CI 3.27 to 14.10) p<0.001) (table 2).
There was substantial intercorrelation between cytokines and chemokines studied (online supplemental table 2). Levels of a number of specific individual cytokines and chemokines were associated with a higher risk of DM (figure 2) after adjustment for multiple potential confounders. Of the cytokines and chemokines analysed in this study, IL-1 RA, IL-1α, IL-1β, IL-4, IL-6, IL-12p70, IL-15, MDC, MCP-1, MCP-4, MIP-1β, MIP-3α and eotaxin were each significantly associated with the risk of DM (HR per SD range: 1.11–1.26; p<0.05) (figure 2). These associations were independent of clinical disease activity (online supplemental figure 1). Associations with IL-6 were not attenuated after excluding two participants receiving anti-IL-6 therapies at baseline. TNF-α was associated with the development of DM only after adjusting for DAS28(CRP) (HR (per 1 SD): 1.07 (95% CI 1.01 to 1.15) p=0.03). The effect was numerically similar but not statistically significant when excluding TNF-inhibitor treated patients at baseline ((N=1348) HR 1.06 (0.90, 1.24) p=0.50). hsCRP measured at enrolment was not significantly associated with the risk of DM.
Adjusted HRs (aHR) for the risk of incident diabetes among patients with rheumatoid arthritis and no history of diabetes (n=1635, events=119) by individual cytokine concentrations (per 1 SD) measured on banked serum from enrolment. Each point estimate and CI shown represents results of the individual analyte examined in a separate multivariable model adjusting for age, sex, race, current smoking, baseline body mass index and baseline use of methotrexate, TNF-biologics and prednisone. *P<0.05, **p<0.008 (Simes-Benjamini-Hochberg adjustment for multiple comparisons). IL, interleukin; IFN, interferon; GMCSF, granulocyte macrophage colony stimulating factor; hs-CRP, high-sensitivity C reactive protein; MCP, macrophage chemoattractant protein; MDC, macrophage-derived chemokine; MIP, macrophage inflammatory protein; TNF, tumour necrosis factor; VEGF, vascular endothelial growth factor.
After inclusion of all cytokines/chemokines separately associated with DM (after accounting for multiple comparisons), only IL-6 and IL-1α remained independently associated with DM (p<0.05; full model not shown). Inclusion of IL-6 and IL-1α in the model improved model fit compared with model 1, table 2 (p=0.03).
Discussion
This cohort study demonstrates strong associations between clinical disease activity and the risk of incident DM in patients with RA. This association does not appear to be explained by differences in age, BMI or the use of RA treatments, including glucocorticoids. Overall, the findings support the hypothesis that systemic inflammation promotes insulin resistance in patients with RA and supports efforts to prevent diabetes through optimal control of disease activity. Further evidence of a relationship between inflammation and DM is the associations observed between several circulating proinflammatory cytokines and chemokines and the development of DM in this population. These associations might suggest that preventative approaches targeting these pathways could be of value or that measurement of inflammatory mediators may help predict populations at risk for this condition.
While evidence has suggested that RA is associated with a higher risk of DM3–5 and that therapies for the disease appear to reduce that risk,10 this study is among the first to demonstrate an important impact of the clinical disease activity itself. The immediate clinical implication of this work is the support of efforts to optimally control disease activity in order to reduce a patient’s risk of DM. The observations from this study also provide insight into understanding the relationship between inflammation and the development of DM in RA and in other populations. While moderate and high disease activity were associated with an increased risk, we found no clear evidence that patients with low disease activity were at greater risk compared with those that were in remission. Thus, these data do not support a more aggressive approach to treatment beyond achievement of low disease activity in order to lower the risk of DM, though a small benefit of a more aggressive treatment target is not ruled out.
This study identified several cytokines and chemokines associated with DM risk of potential interest including IL-1α, IL-1β and IL-1β RA. IL-1β has been shown to be associated with DM in the general population and has been considered a potential target of therapy in DM.13 However, a recent large randomised trial performed in the general population, did not demonstrate a reduced risk of DM among participants receiving canakinumab, an anti-IL-1β therapy.28 Thus, while the current study supports upregulation of IL-1 and its RA as an important marker of DM risk in this population, it is not sufficient to support the use of anti-IL-1β therapy with the goal of preventing this outcome.
IL-6 was also demonstrated to be independently associated with incident diabetes as has been shown in other populations.13 It has been proposed that obesity-related inflammation may result in high levels of IL-6.29 In our study, elevated IL-6 levels were associated with DM independent of BMI, but we were not able to directly assess fat in the visceral compartment. Whether obesity-related inflammation is a primary driver of the metabolic complications of obesity or simply a marker of the severity of visceral adiposity is not fully elucidated. Further study with longitudinal IL-6 measurement is needed in order to determine if IL-6 is in the causal pathway leading from obesity to diabetes. It therefore remains unknown whether blocking IL-6 might prevent the onset of DM in the general population and in patients with RA.
Several MDCs were identified in this study that were associated with the risk of DM. MCP-1 is a chemokine that recruits monocytes and other inflammatory cells to the site of inflammation in response to inflammatory stimuli. MCP-1 may be produced in visceral fat and higher levels have been linked to a greater risk of DM in other settings.14 30 Gene polymorphisms resulting in lower levels appear protective, supporting MCP-1 as a potential target of intervention.31 32 Prior studies in the general population have similarly demonstrated expression of MIP-1β in visceral fat of diabetic patients.33 34 It remains unclear if elevations in MDCs are a consequence of RA-related inflammation or reflect inflammation related to obesity and visceral adiposity.
In this study, prednisone use was not significantly associated with a higher risk of DM independent of the effect of disease activity. The effect of prednisone on the risk of DM is well known. It is possible that the tendency for providers to use low-dose prednisone in patients with RA and to avoid its use in obese patients and in patients with pre-diabetes explains why there appears to be no significant increase in risk. However, the current study is limited in the accurate assessment of dose to further delineate ‘safe’ doses of prednisone. Furthermore, the tendency for providers to use of prednisone in patients with the most active disease and those with contraindications to other therapies suggests that confounding may limit our ability to make causal inferences in this study that was not aimed at this question. Non-differential misclassification of prednisone use is also possible, given the reliance on physician and this may have biased results to the null.
Methotrexate use was associated with a lower risk of DM. It is tempting to attribute this association to causal reduction in risk that is perhaps independent of the effects of the drug on measures of disease activity. Few clinical studies have addressed this question and have yielded conflicting results.35 36 Despite a desire to attribute causal benefit, there may be residual confounding at play given that patients with hepatic steatosis may simultaneously be both less likely to use methotrexate and also are at higher risk of developing DM. Further study is needed to determine if there are drug-specific benefits of methotrexate or whether this noted association is the result of channelling. We did not identify significant associations with other RA therapies such as hydroxychloroquine (HCQ), TNFi or abatacept independent of other factors such as disease activity, though the study was not designed with the primary aim of characterising DM risk with these therapies.
Further study with alternative designs is needed to better determine if particular therapies might have a greater impact on reducing the risk of DM. The nature of the registry does not provide an opportunity to study regional changes in adiposity, such as visceral deposition, nor behavioural risk factors such as diet and physical activity, which may play a role in understanding the risk of DM in RA. The size of this study limits the ability to identify weaker relationships between specific cytokines and chemokines with a relatively uncommon outcome. In addition, the observation of strong interanalyte correlation limits our ability to identify key cytokines that may play a pathogenic role. Furthermore, cytokines and chemokines were only available at enrolment and the impact of changes in these measures over time was not possible to assess. We also present results of multiple comparisons and some significant findings may have occurred by chance. While we found no significant differences between men and women in this study, the VARA registry, which is predominantly male, may not be entirely generalisable to other populations with RA. Finally, misclassification between type 1 and type 2 DM, though in this age group the incidence of type 1 DM is expected to be quite rare. Notable strengths of the study include the long-term follow-up of patients with rheumatologist-diagnosed RA, the availability of longitudinal clinical assessments, and the direct assessment of inflammatory cytokines and chemokines.
In conclusion, greater disease activity is associated with a greater risk of DM among patients with RA. Circulating levels of cytokines and chemokines are independently associated with the development of DM supporting the hypothesis that inflammation is an important factor in risk. These data may also support the targeting of specific inflammatory pathways for intervention, though further study is needed. Better control of disease activity and related systemic inflammation may help to reduce the long-term risks of metabolic complications of RA such as the observed increase in risk of DM.
Data availability statement
Data may be obtained from a third party and are not publicly available. Data are not publicly available but may be requested from the VA Rheumatoid Arthritis Registry and Biorepository.
Ethics statements
Patient consent for publication
Ethics approval
Each individual site has institutional review board approval.
Acknowledgments
JFB would like to acknowledge funding through a Veterans Affairs Clinical Science Research and Development Career Merit Award (I01 CX001703).
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Handling editor Josef S Smolen
Presented at This work was presented at the 2019 American College of Rheumatology Annual meeting in Atlanta, GA (Abstract #839).
Correction notice This article has been corrected since it published Online First. The provenance and peer review statement has been included.
Contributors JFB was responsible for study concept, data acquisition, study design, analysis, data interpretation and scientific writing. BRE, GC and TRM were responsible for data acquisition, study design, data interpretation and scientific writing. MG, BS, AO, BCH, CH, MJD and GT were responsible for data interpretation, scientific writing. All authors had final approval of the manuscript.
Funding JFB is supported by a Veterans Affairs VA Merit Award (I01 CX001703). BRE is supported by the Rheumatology Research Foundation and the NIGMS (U54GM115458). TRM is supported by a VA Merit Award (I01 BX0046000) and grants from NIAAA (R25AA020818), NIGMS (U54GM115458) and NIAMS (P50AR60772).
Disclaimer The contents of this work do not represent the views of the Department of the Veterans Affairs or the US Government.
Competing interests JFB has received consulting fees from Bristol-Myers Squibb and Gilead.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.