Objective Dipeptidyl peptidase-4 (DPP4), also known as CD26, is a transmembrane glycoprotein that has a costimulatory function in the immune response. DPP4 inhibitors (DPP4i) are oral glucose-lowering drugs for type 2 diabetes mellitus (T2DM). This study evaluated the risk of incident rheumatoid arthritis (RA) and other autoimmune diseases (AD) such as systemic lupus erythematosus, psoriasis, multiple sclerosis and inflammatory bowel disease, associated with DPP4i in patients with T2DM.
Methods Using US insurance claims data (2005–2012), we conducted a population-based cohort study that included initiators of combination therapy with DPP4i (DPP4i plus metformin) and non-DPP4i (non-DPP4i plus metformin). RA and other AD were identified with ≥2 diagnoses and ≥1 dispensing for AD-specific immunomodulating drugs or steroids. Composite AD includes RA or other AD. Propensity score (PS)-stratified Cox proportional hazards models compared the risk of AD in DPP4i initiators versus non-DPP4i, controlling for potential confounders.
Results After asymmetric trimming on the PS, 73 928 patients with T2DM starting DPP4i combination therapy and 163 062 starting non-DPP4i combination therapy were selected. Risks of incident RA and composite AD were lower in the DPP4i group versus non-DPP4i with the PS-stratified HR of 0.66 (95% CI 0.44 to 0.99) for RA, 0.73 (0.51 to 1.03) for other AD and 0.68 (95% CI 0.52 to 0.89) for composite AD.
Conclusions In this large cohort of diabetic patients, those initiating DPP4i combination therapy appear to have a decreased risk of incident AD including RA compared with those initiating non-DPP4i combination therapy. These results may suggest possible pharmacological pathways for prevention or treatment of AD.
- Autoimmune Diseases
- Rheumatoid Arthritis
- Systemic Lupus Erythematosus
Statistics from Altmetric.com
Dipeptidyl peptidase-4 (DPP4) inhibitors, such as sitagliptin, saxagliptin and linagliptin, are oral glucose-lowering drugs that can be used as monotherapy or combination therapy with other oral hypoglycaemic agents for type 2 diabetes mellitus (T2DM).1–4 Sitagliptin was the first DPP4 inhibitor (DPP4i) approved by the US Food and Drug Administration (FDA) for adults with type 2 diabetes in October 2006, followed by saxagliptin, FDA-approved in July 2009, and linagliptin in May 2011. These drugs are generally well tolerated without a specific contraindication.
DPP4i have their hypoglycaemic effect by acting through increasing glucagon-like peptide-1 and glucose-dependent insulinotrophic polypeptide, subsequently leading to increases in insulin and C-peptide, decreases in glucagon and improvements in oral glucose tolerance.5 However, DPP4 is a transmembrane glycoprotein, also known as CD26, widely expressed in various cell types such as fibroblast, endothelial and epithelial cells, T lymphocytes and macrophages, and thus has many biological functions beyond glucose metabolism, including chemotaxis, signal transduction, as well as T cell activation.5–8
While DPP4 has biological functions in proinflammatory pathways, the effects of DPP4i on the immune system, particularly in the pathogenesis of autoimmune diseases (AD), are not well known. On the one hand, a number of studies reported decreased levels of DPP4 activity in patients with rheumatoid arthritis (RA),9 ,10 systemic lupus erythematosus,11 inflammatory bowel disease12 ,13 and psoriasis.14 On the other hand, several studies noted upregulation of DPP4 expression in psoriasis15 ,16 and multiple sclerosis,17 as well as elevated numbers of CD26-positive T cells in RA18 ,19 and multiple sclerosis.20 While studies suggested a potential role of DPP4i as a novel therapy for several inflammatory diseases by inhibiting T cell proliferation and cytokine production,6 ,7 ,21–27 a few cases of inflammatory arthritis potentially related to use of DPP4i have been reported.28
The objective of this study was to estimate the risk of incident systemic AD including RA, systemic lupus erythematosus, psoriasis, psoriatic arthritis, multiple sclerosis and inflammatory bowel disease in patients with diabetes starting a DPP4i drug compared with those starting non-DPP4i oral hypoglycaemic agents. We hypothesised that patients starting a DPP4i would have a reduced risk of incident RA and other AD compared with those starting non-DPP4i drugs only.
We conducted a cohort study using the claims data for the period 1 January 2005 to 31 December 2012 from a commercial US health plan that insures primarily working adults and their family members. This database contains longitudinal claims information including medical diagnoses, procedures, hospitalisations, physician visits and pharmacy dispensing on its approximately 14 million subscribers across the USA on a yearly basis. Results for outpatient laboratory tests, including glycated haemoglobin (HbA1c), were available on a subset of beneficiaries. The distribution of race and ethnicity is representative of the US general population.29 The quality of these data on medical diagnoses, procedures, healthcare use and drug dispensing is also known to be high.29 Patient informed consent was not required as the dataset was de-identified to protect subject confidentiality. The study protocol was approved by the Institutional Review Board of the Brigham and Women's Hospital.
Patients who had at least one dispensing for an oral hypoglycaemic agent any time during the study period were first identified. To avoid selecting patients with type 1 DM, we selected patients aged 40 years and older with a visit coded with the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD 9-CM) code, 250.xx, for DM for the study cohort. As DPP4i drugs are more commonly used as a second or third agent for T2DM, two mutually exclusive exposure groups were defined: (1) initiators of DPP4i combination therapy (DPP4i plus metformin) and (2) initiators of non-DPP4i oral combination therapy (metformin plus another non-DPP4i drug). DPP4i drugs include linagliptin, saxagliptin and sitagliptin. Non-DPP4i drugs include metformin, sulfonylureas, thiazolidinediones (TZD) and meglitinides.
The index date was defined as the earliest date of starting a DPP4i drug with concurrent use of metformin for the DPP4i combination therapy group and the earliest date of adding a second non-DPP4i drug including metformin for the non-DPP4i combination therapy group. Patients were required to have at least 365 days of continuous health plan eligibility before the index date. Therefore, the earliest index date that patients can have is 1 January 2006. Patients with a prior diagnosis of autoimmune disease (RA, systemic lupus erythematosus, psoriasis, psoriatic arthritis, multiple sclerosis and inflammatory bowel disease), HIV, cancer and use of insulin-containing drugs, glucagon-like peptide 1 agonists and immunomodulating drugs including disease-modifying antirheumatic drugs in the 365 days prior to the index date were excluded. Patients were required to be naive to DPP4i in the 180 days prior to the index date. For the non-DPP4i combination therapy group, patients were required to have at least 180 days without using multiple oral hypoglycaemic drugs prior to their index date (figure 1).
For the subgroup analysis, two additional comparator groups—TZD combination therapy and sulfonylurea combination therapy—were selected. For the comparison between DPP4i and TZD combination therapies, the index date was defined as the earliest date of adding a DPP4i to metformin or a TZD drug plus metformin. For the comparison between DPP4i and sulfonylurea combination therapies, the index date was defined as the earliest date of adding a DPP4i to metformin or a sulfonylurea drug plus metformin. In the subgroup analysis, patients were required to be naive to both drug categories in the 180 days prior to the index date.
Follow-up began on the day after the index date. In the primary analysis, patients were followed up to the first of any of the following censoring events: discontinuation or switching of study drugs (‘as treated’), occurrence of RA or other AD, loss of health plan eligibility, end of study database, death or 365 days. Patients were allowed to have gaps of up to 30 days between prescription fill dates in the calculation of continuous therapy. In the case of drug discontinuation or switching, the exposure risk window for each patient treatment episode extended until 30 days after the expiration of the supply of the last fill. Patients were allowed to enter the study cohort one time only.
Outcomes of interest were a new diagnosis of RA or other AD including systemic lupus erythematosus, psoriasis, psoriatic arthritis, multiple sclerosis and inflammatory bowel disease, defined with at least two visits, which were at least 7 days apart, with a disease-specific diagnosis code and at least one dispensing for disease-specific immunomodulating drugs or steroids (see online supplementary table S1).30–36 The date of outcome occurrence was defined as the latest date of either diagnoses or a drug dispensing. The composite outcome, a new diagnosis of RA or other AD, was also assessed.
Variables potentially related to development of AD were assessed using data from the 365-day baseline period before the index date. These variables included age, sex, smoking, comorbidities such as obesity, thyroid disease and other cardiovascular diseases, medications including calcium channel blockers, β-blockers, anticonvulsants, antipsychotics, procainamide, quinidine, hydralazine, methyldopa, thiazides, systemic steroids and non-steroidal anti-inflammatory drugs and healthcare use factors including visits to various specialists (see table 1). To further quantify patients’ comorbidities at baseline, we also calculated a comorbidity score that combined 20 medical conditions included in both the Charlson Index and the Elixhauser system based on ICD-9.37 To characterise diabetes treatment intensity, the number of oral hypoglycaemic drugs taken at the index date was also determined. Baseline HbA1c levels were available in a subgroup of the study cohort.
We compared the baseline characteristics between DPP4i and non-DPP4i groups. To control for potential confounders, we decided to use the propensity score (PS) method—PS-stratified and PS-matched analysis.38 Multivariable logistic regression including all the baseline covariates listed in table 1 was used to estimate the PS, defined as the predicted probability of a patient receiving combination therapy with DPP4i versus non-DPP4i. For the PS-stratified analysis, patients were grouped into PS deciles after excluding those in the non-overlapping tails of the PS distribution. We used asymmetric trimming with the cutpoint at the 2.5th centiles and 97.5th centiles of the PS distribution in the exposed and unexposed for each comparison.39 For PS-matched analysis, we used nearest neighbour matching within a ‘caliper’ of 0.025 on the PS at a fixed ratio of 1:1.40 ,41 In both trimmed and matched cohorts, incidence rates and HR of RA and other AD with 95% CIs were calculated in DPP4i initiators versus non-DPP4i. Kaplan–Meier curves were plotted for the cumulative incidence of each outcome in the PS-matched DPP4i and non-DPP4i cohorts. All these analyses were repeated for the subgroup analyses comparing DPP4i versus TZD initiators, and DPP4i versus sulfonylureas initiators. The proportional hazards assumption was assessed by testing the significance of the interaction term between exposure and time and was not violated except the Cox model for other AD comparing DPP4i versus sulfonylureas.42 We therefore further stratified analyses of the risk of other AD in DPP4i versus sulfonylureas by follow-up days 0 to 180, and 181 to 365. All analyses were done using SAS V.9.2 Statistical Software (SAS Institute Inc., Cary, North Carolina, USA).
We identified 1 140 060 patients who had at least one dispensing for a DPP4i or non-DPP4i drug in the study database. After applying the inclusion and exclusion criteria, the cohort included 75 893 diabetic patients who started a combination therapy with a DPP4i drug and 167 260 diabetic patients who started a combination therapy with non-DPP4i drugs only. After the 2.5th and 97.5th asymmetric trimming based on the PS distribution, 73 928 DPP4i and 163 062 non-DPP4i initiators were included. Matching on PS with a 1:1 ratio further selected a total of 47 884 pairs of DPP4i and non-DPP4i initiators (see online supplementary figure S1).
After asymmetric trimming, the mean age of patients was 55.5 years. 40% of DPP4i group and 39% of non-DPP4i group were female (see online supplementary tables 1 and S2). Overall, comorbidities, diabetic medications and other drugs, and healthcare use including number of total physician visits and proportions of patients with specialty clinic visits were slightly more common in the DPP4i group. The mean (SD) number of days on metformin in the 365-day baseline period was 207 (130) days for DPP4i and 162 (136) days for non-DPP4i. The mean proportion of days covered by metformin during the 365-day baseline period was 57 (36)% for DPP4i and 44 (37)% for non-DPP4i. Also, 32% of DPP4i and 27% of non-DPP4i had a baseline HbA1c level measured. The mean HbA1c was 8.1% for both groups. In both DPP4i and non-DPP4i groups, most patients started a combination therapy with two or three oral hypoglycaemic agents at the index date. The mean (SD) follow-up was 0.74 (0.86) years for DPP4i and 0.72 (0.91) years for non-DPP4i. The baseline characteristics of the DPP4i and non-DPP4i combination therapy were well balanced after PS matching (see online supplementary table S3).
Risk of autoimmune diseases
In the PS-trimmed cohorts, there were a total of 179 patients newly diagnosed with RA, 249 with other AD and 424 with composite AD after the initiation of either DPP4i or non-DPP4i combination therapies. The incidence rate was 1.26 per 1000 person-years for RA and 1.78 per 1000 person-years for other AD in the DPP4i group, and 1.64 per 1000 person-years for RA and 2.26 per 1000 person-years for other AD in the non-DPP4i group. In the PS decile-stratified analysis, the risk of incident RA (HR 0.66, 95% CI 0.44 to 0.99) and composite AD (HR 0.68, 95% CI 0.52 to 0.89) within 365 days of follow-up was decreased for DPP4i initiators compared with non-DPP4i (table 2).
In the PS-matched cohorts, the incidence rates of RA or other AD were similar as were HRs for all outcomes with wide CIs due to smaller sample sizes compared with those in the PS-trimmed cohorts. Figure 2 displays the Kaplan–Meier curves comparing the cumulative incidence of AD between the PS-matched DPP4i and non-DPP4i groups.
Baseline characteristics of the study subgroups after asymmetric trimming also showed slightly more common comorbidities, other medications and healthcare use in the DPP4i group compared with the TZD (see online supplementary table S4) and sulfonylureas groups (see online supplementary table S5). In the PS-trimmed subgroups of DPP4i and TZD, and DPP4i and sulfonylureas, overall incidence rates of RA and other AD were low as seen in the main cohorts (tables 3 and 4). The risk of incident RA was not decreased for DPP4i compared with TZD (HR 1.04, 95% CI 0.51 to 2.12) or sulfonylureas (HR 0.66, 95% CI 0.38 to 1.15), while the risk of other AD (HR 0.49, 95% CI 0.30 to 0.80) and composite AD (HR 0.53, 95% CI 0.36 to 0.77) remained reduced for DPP4i compared to sulfonylureas. In the subgroup analysis comparing DPP4i to sulfonylureas for other AD that violated the proportionality of hazards, the HR for other AD was 0.42 (95% CI 0.21 to 0.82) in DPP4i during the first 180 days and 0.90 (95% CI 0.36 to 2.28) for the follow-up days 181 to 365.
To date, the epidemiological effect of DPP4i on AD has not been examined despite biological mechanisms and case reports that suggest a possible relationship. In a large population-based cohort of T2DM patients, we found a decreased risk for RA and composite AD among initiators of DPP4i combination therapy compared with initiators of non-DPP4i combination therapy. Subgroup analysis comparing DPP4i to sulfonylureas also showed a decreased risk for other AD and composite AD, although the risk for RA was not significantly reduced. While it is possible that DPP4i does not change a risk of AD but sulfonylureas increase a risk of AD, there are currently no data that suggest such association between sulfonylureas and AD. When comparing DPP4i to TZD combination therapy initiators, no difference in the risk of RA was seen. This might be related to TZD's immunomodulating or anti-inflammatory action as suggested by the beneficial effects of these agents on disease activity observed in several clinical trials in T2DM patients with RA, psoriatic arthritis or inflammatory bowel disease.43–46
This study may have important implications for better understanding the pathogenesis of AD. To date, there are no proven strategies for disease prevention in any of the AD studied. While the current study did not investigate mechanisms of AD pathogenesis, it seems likely that DPP4 (CD26) may play a role in the development of AD. It is known that DPP4 (CD26) is present in various tissues and cells including lymphocytes and monocytes as a transmembrane glycoprotein and is associated with immunoregulatory functions.5–8 DPP4i inhibits T cell proliferation and cytokine production,6 ,7 ,21–27 both known to be involved in AD pathogenesis. DPP4i is generally well tolerated: recent clinical trials of T2DM patients who were at high risk for cardiovascular events showed that DPP4i did not increase the rate of ischaemic cardiovascular events.47 ,48 As diabetes is fairly common in patients with pre-existing AD,49–51 it might be worth considering a study that examines a role of DPP4i as a novel treatment of AD in patients with T2DM. This line of study is supported by both animal studies showing partial improvement of inflammatory bowel disease with DPP4i 7 ,52 ,53 and improvement of psoriatic skin lesions after the initiating of a DPP4i.26
Although it is not known whether any past exposure to DPP4i has an effect on the development of AD, we restricted the stud cohort to ‘new users’ of combination therapy to reduce biases such as survivor bias and time-varying confounding.54 ,55 In addition, to minimise confounding by indication inherent in observational studies, we used rigorous pharmacoepidemiologic approaches in the study design and analysis including active comparator, and PS- stratified and PS-matched analyses. Multivariable logistic models for PS estimation incorporated a comprehensive list of potential confounders including age, sex, calendar year, comorbidities, medications and healthcare use patterns. Nonetheless, residual confounding by indication or by obesity, smoking, family history of AD and socioeconomic status might be still an issue in this study. However, it is unlikely that physicians who treat patients with T2DM choose oral hypoglycaemic drugs based on the future risk of AD in patients with T2DM. Surveillance bias can play a role in diagnosing more or less AD in patients with DPP4i versus non-DPP4i. Prior to PS matching, the DPP4i group had a greater number of physician visits and higher proportions of visits to specialists, which would likely bias the results to the opposite direction (ie, more AD diagnoses in DPP4i vs non-DPP4i); we then included various healthcare use factors in the PS estimation and achieved balance in these variables between the groups. Furthermore, we conducted a subgroup analysis comparing DPP4i to TZD initiators, as both DPP4i and TZD are relatively newer drugs and frequently used with other oral hypoglycaemic drugs, and found similar HRs, although with wide CIs including the null, for other and composite AD.
This study has limitations. First, we assessed a number of variables (eg, smoking, obesity, periodontal disease, infectious mononucleosis and use of various drugs) potentially related to development of AD using the claims data from the 12 months prior to the index date; however, it is possible that the 12-month baseline period was not long enough to capture all the information on potential confounders and that there was incomplete ascertainment of those variables in the claims data. Second, the requirement of ≥180 days free of DPP4i or non-DPP4i combination therapy may not be long enough to differentiate new users from intermittent users. We assumed that a washout period of 180 days would be sufficient for patients who received a DPP4i or non-DPP4i combination therapy on and off. Third, in this study, we mainly relied on diagnosis codes and drug dispensing for outcome ascertainment. A prior validation study using the same data source reported that there was 96% agreement between a claim-based medical diagnosis and the medical record or the patient survey.56 To further minimise the potential for outcome misclassification, all AD outcomes were defined with at least two diagnosis codes and at least one dispensing for disease-specific immunomodulating drugs.30–36 The incidence rates for RA from this study are slightly higher than the incidence rate of RA from the Rochester Epidemiology Project in the USA.57
In conclusion, initiating DPP4i combination therapy appears to be associated with a decreased risk of AD including RA compared with initiating non-DPP4i combination therapy. These results may suggest new mechanistic pathways for preventing or delaying the onset of AD and could lead to a potential new therapeutic approach for patients with pre-existing AD. If other studies find DPP4i also effective in prevention of autoimmune disease, future research would be needed to determine the effect and safety of DPP4i in the non-diabetic population.
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.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Handling editor Tore K Kvien
The abstract of this study was presented as an oral presentation at the American College of Rheumatology 2013 meeting in San Diego, CA on 29 October 2013.
Contributors SCK takes responsibility for the integrity of the data and the accuracy of the data analysis. She is the guarantor for the study. All authors conceived and designed the study, analysed and interpreted the data, and critically revised the manuscript for important intellectual content. SCK drafted the paper.
Competing interests SCK receives research support from Pfizer and tuition support for the Pharmacoepidemiology Program at the Harvard School of Public Health partially funded by the Pharmaceutical Research and Manufacturers of America (PhRMA) foundation. SS is consultant to WHISCON, LLC and to Aetion, Inc. of which he also owns shares. He is principal investigator of investigator-initiated grants to the Brigham and Women's Hospital from Novartis, and Boehringer-Ingelheim unrelated to the topic of this study. RJG received research grants from AstraZeneca and Novartis. DHS receives research support from Amgen, Lilly, Pfizer, and CORRONA and serves in unpaid roles on studies sponsored by Pfizer, Novartis, Lilly and Bristol Myers Squibb. DHS also receives royalties from UpToDate.
Ethics approval Brigham and Women's Hospital IRB.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Owing to the data use agreement with the data provider, we are unable to provide raw data, but please contact the corresponding author for additional unpublished data if needed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.