Background Although recent clinical trials raised concerns about the risk for heart failure (HF) in dipeptidyl peptidase-4 (DPP-4) inhibitor use, data on the cardiovascular risks in the patients with pre-existing HF are still lacking.
Methods We used Taiwan's National Health Insurance Research Database to identify 196 986 patients diagnosed with type 2 diabetes mellitus (T2DM) who had previous history of HF between 2009 and 2013. This population included 30 204 DPP-4 inhibitor users and 166 782 propensity score-matched DPP-4 inhibitor non-users. The outcomes of interest were all-cause mortality, combination of myocardial infarction (MI) and ischaemic stroke, and hospitalisation for HF.
Results The incidence in DPP-4 users compared with non-users was 67.02 vs 102.85 per 1000 person-years for all-cause mortality, 37.89 vs 47.54 per 1000 person-years for the combination of MI and ischaemic stroke, 12.70 vs 16.18 per 1000 person-years for MI and 26.37 vs 32.46 per 1000 person-years for ischaemic stroke. The risk of all-cause mortality was lower in DPP-4 inhibitor users (HR 0.67, 95% CI 0.64 to 0.70), combination of MI and stroke (HR 0.81, 95% CI 0.76 to 0.87), MI (HR 0.80, 95% CI 0.71 to 0.89) and ischaemic stroke (HR 0.83, 95% CI 0.76 to 0.89) than in non-users. Notably, the risk of hospitalisation for HF did not differ significantly between groups. The results were similar after accounting for death as a competing risk.
Conclusions In this nationwide T2DM cohort, the risks of mortality and the combination of MI and ischaemic stroke were lower for patients receiving DPP-4 inhibitors than for those who did not receive such treatment. DPP-4 inhibitor use was not associated with a higher risk of hospitalisation for HF even in patients with pre-existing HF.
Statistics from Altmetric.com
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.
Heart failure (HF) occurs more frequently in patients with type 2 diabetes mellitus (T2DM) than in those without this condition, with a prevalence of about 20–30% compared with 4–6% in non-diabetic subjects.1 ,2 In 2008, the US Food and Drug Administration recommended careful assessment of the cardiovascular risk of new diabetes drugs in preapproval and postapproval outcome studies.3 Dipeptidyl peptidase-4 (DPP-4) inhibitors are a relatively new class of oral hypoglycaemic agents that improve glycaemic control in patients with T2DM, which increases levels of the incretin hormones glucagon-like peptide-1 and glucose-dependent insulinotropic peptide and stimulates insulin secretion in a glucose-dependent fashion.4 ,5 Proving the safety of DPP-4 inhibitors is critically important.
Cardiovascular end point data from the Saxagliptin Assessment of Vascular Outcomes Recorded in Patients with Diabetes Mellitus–Thrombolysis in Myocardial Infarction 53 (SAVOR-TIMI 53) trial suggest that saxagliptin neither increases nor reduces the risks of non-fatal myocardial infarction (MI), non-fatal stroke and cardiovascular death.6 However, saxagliptin was unexpectedly associated with a 27% increased risk of hospitalisation for HF compared with placebo. A subsequent meta-analysis of all registered DPP-4 inhibitor trials showed that the association between the use of this class of drugs and increased risk of HF disappeared when SAVOR-TIMI 53 trial data were excluded.7 The Examination of Cardiovascular Outcomes with Alogliptin versus Standard of Care (EXAMINE) trial,8 Vildagliptin in Ventricular Dysfunction Diabetes (VIVIDD) trial,9 the Trial to Evaluate Cardiovascular Outcomes after treatment with Sitagliptin (TECOS)10 and most recent observational studies have also failed to find an association between DPP-4 inhibitor use and hospitalisation for HF. Post hoc subgroup analysis using SAVOR-TIMI 53 trial data demonstrated an increased probability of hospitalisation for HF among patients with pre-existing HF.11 More importantly, the research has not specifically focused on the pre-existing HF subpopulation to explore the real cardiovascular risks associated with the use of DPP-4 inhibitors.
In light of the concerns raised about possible HF risk in patients with pre-existing HF using DPP-4 inhibitors according to the results of post hoc analysis, we performed a nationwide, propensity score-matched, population-based study using data from the entire T2DM population with pre-existing HF in Taiwanese National Health Systems to explore effect of DPP-4 inhibitors on cardiovascular risks in a real-world setting.
Taiwan's National Health Insurance programme, which provides universal coverage, was implemented in 1995. Similar to other nationwide databases, Taiwan's National Health Insurance Research Database (NHIRD) contains comprehensive medical care data (including claims for outpatient and inpatient services, emergency room visits), with detailed relevant diagnosis codes according to the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM); procedure codes; and drug prescription information for 23 million people in Taiwan. To protect patients' privacy, information recorded in the NHIRD was deidentified and encrypted before release for public research purposes. The institutional review board of Taipei City Hospital exempted this study from full review because of the deidentified nature of the data (TCHIRB-10404107-W).
This nationwide population-based study included patients with T2DM and previous HF diagnoses (including those associated with outpatient visits and hospitalisations for HF) registered in Taiwan's NHIRD. The accuracy of T2DM diagnoses in the NHIRD has been validated.12 DPP-4 inhibitors are available in Taiwan's market since March 2009, so we included all patients with the diagnosis of T2DM between March 2009 and June 2013. After inclusion, all patients were followed up until December 2013, allowing at least a half year follow-up to estimate the risk of outcome events.
Detailed information on drug type, prescription date, daily dose and the number of days of drug supply was extracted from the National Health Research Institute prescription database. Based on prescription of DPP-4 inhibitors (including sitagliptin, vildagliptin, saxagliptin and linagliptin) during the observation period, patients with T2DM were classified as DPP-4 inhibitor users and non-users (matched controls). For DPP-4 inhibitor users, the first date of DPP-4 inhibitor prescription was defined as the index date. An index date was assigned randomly to each DPP-4 non-user according to the corresponding index date for a DPP-4 inhibitor user. The mean compliance rate was measured as the proportion of DPP-4 inhibitor use days during the interval from the index date to death or 31 December 2013, whichever came first. Insufficient compliance with DPP-4 inhibitor treatment was deﬁned as <80% of days covered.13
Propensity score matching
We used a propensity score-matching approach to account for baseline differences between DPP-4 inhibitor users and non-users (see online supplementary table S1). Multiple logistic regression was used to generate propensity scores, which predicted the probability of DPP-4 inhibitor use. The following covariates were included in the logistic regression model: year of index date, month of index date, age, sex, monthly income, urbanisation level, Charlson Comorbidity Index score,14 number of outpatient visits to metabolism and endocrinology professionals in the past year, adapted Diabetes Complications Severity Index (aDCSI) score,15 duration of DM diagnosis, other concomitant medications (including antihypertensive drugs, antidiabetic drugs other than DPP-4 inhibitors, antiplatelet drugs, warfarin, statins, non-steroidal anti-inflammatory drugs and antidepressants) and underlying comorbidities. Each DPP-4 inhibitor user was matched with a DPP-4 inhibitor non-user based on the logit of the propensity score using nearest-neighbour matching without replacement, with a calliper width equal to 0.1 of the SD of the logit of the propensity score. After propensity score matching, covariate balance was checked by means of comparing the standardised difference between DPP-4 inhibitor users and nonusers, in which the standardised difference >0.1 represents meaningful imbalance.16
The co-primary outcomes were all-cause mortality, combination of MI (ICD-9-CM code 410.x)) and ischaemic stroke (ICD-9-CM codes 433.x, 434.x or 436) and hospitalisation for HF (ICD-9-CM code 428.x) during follow-up. All patients were followed until death or 31 December 2013.
Descriptive statistics were calculated for demographic and clinical variables for DPP-4 inhibitor users and non-users in the cohort. Standardised differences were used to check for balance in the matched characteristics. Incidence rates of outcomes of interest were compared between DPP-4 inhibitor users and non-users based on Poisson distributions. HRs and 95% CIs of the main outcomes of interest were modelled using Cox proportional-hazards regression. As the outcomes of interest and death have important competing effects in patients with T2DM, the risk of dying before the outcomes happen must be taken into account. We thus used Fine and Gray competing-risks proportional-hazards regression for competing risks analyses.17 Schoenfeld residuals were used to test the assumption of Cox models (all-cause mortality (p=0.051), combination of MI and ischaemic stroke (p=0.235), MI (p=0.370), stroke (p=0.515) and HF (p=0.121)) and competing risk models (see online supplementary figures S1–S4).18 ,19
We also chose cancer as a negative control outcome to detect residual confounding.20 Subgroup Cox regression analyses were performed and the likelihood ratio test was used to examine interactions between the DPP-4 inhibitor use and the following covariates: age, sex, hypertension, MI, cerebrovascular disease, chronic kidney disease, insulin use and history of hospitalisation for HF. Linkage, processing and sampling of NHIRD data were performed using the 2012 SQL Server (Microsoft Corporation, Redmond, Washington, USA). All statistical analyses were conducted with STATA statistical software (V.12.0; StataCorp, College Station, Texas, USA), and propensity scores were calculated using SAS V.9.3 (SAS Institute, Cary, North Carolina, USA). Statistical significance was defined as p<0.05.
Characteristics of the study population
A total of 196 986 patients with T2DM and pre-existing HF recorded between March 2009 and June 2013 were included in the analysis. The mean age of the population was 67.9±12.0 years. The median duration of DM diagnosis was 101 (IQR 61–133) months, and the median aDCSI score was 4. In our study, 30 204 (15.3%) patients were treated with DPP-4 inhibitors. After propensity score matching, 22 510 pairs of DPP-4 inhibitor users and non-users with comparable baseline characteristics were included in analyses. The demographic and clinical characteristics of these cohorts are summarised in table 1.
Risks of mortality and combination of MI and ischaemic stroke
Compared with the DPP-4 inhibitor non-users, DPP-4 inhibitor users had lower risks of all-cause mortality (HR 0.67, 95% CI 0.64 to 0.70), combination of MI and ischaemic stroke (HR 0.81, 95% CI 0.76 to 0.87), MI (HR 0.80, 95% CI 0.71 to 0.89), ischaemic stroke (HR 0.83, 95% CI 0.76 to 0.89) and HF (HR 0.96, 95% CI 0.88 to 1.04; table 2). After accounting for death as a competing risk, DPP-4 inhibitor users still had lower risks of combination of MI and ischaemic stroke (HR 0.84, 95% CI 0.79 to 0.90), MI (HR 0.83, 95% CI 0.74 to 0.93), ischaemic stroke (HR 0.86, 95% CI 0.79 to 0.93) and HF (HR 1.00, 95% CI 0.92 to 1.08). We detected confounding of associations between DPP-4 inhibitor use and outcomes by using cancer as a negative control outcome. We found no difference in cancer risk between DPP-4 inhibitor users and non-users (table 2). Subgroup analyses produced almost consistent results (figures 1 and 2 and online supplementary tables S2 and S3). Notably, a signiﬁcant interaction (pinteraction<0.001) was noted for using insulin and all-cause mortality in DPP-4 inhibitor users versus non-users (figure 1 and online supplementary table S2).
To our knowledge, this study examined the effects of DPP-4 inhibitor users versus non-users on clinical outcomes in patients with pre-existing HF in the largest real-world population to date. We found no difference in the risk of readmission for HF between DPP-4 inhibitors and control. By accounting for death as a competing risk, our study showed significantly reduced risks of combination of MI and stroke and in DPP-4 inhibitor users. Our study findings may be considered to constitute the information currently available regarding the safety of these drugs in this population.
Previous clinical trials have produced conflicting results, with no clear consensus on drug safety in patients with pre-existing HF. The SAVOR-TIMI 53 trial showed an unexpected 27% increased risk of hospitalisation for HF in saxagliptin users compared with placebo users.6 Post hoc analysis revealed that patients with pre-existing HF were more likely to be hospitalised for HF after DPP-4 inhibitor use. The VIVIDD trial, a study of the safety of DPP-4 inhibitors in patients with pre-existing HF, has shown no difference in the left ventricular contractile or emptying function, or B-type natriuretic peptide level, from baseline to 52 weeks.9 Unexpectedly, this trial showed increased left ventricular end-diastolic volumes in vildagliptin users, which may be implicated in increasing the risk of HF. The EXAMINE trial showed a non-significant numerical increase in risk for hospitalisation for HF (HR 1.19) between alogliptin and placebo users8 ,21 Notably, the EXAMINE trial included twice the proportion of patients with histories of HF as did the SAVOR-TIMI 53 trial (28% vs 14.8%) and may be non-significant due to the relatively small sample size and short follow-up period. On the other hand, the most recent TECOS trial showed no difference in the risk for HF between sitagliptin and placebo users (HR 1.00 (CI 0.83 to 1.20)), similar to the findings of our study. Thus, clinical trials to date have produced conflicting results, with no clear consensus on the safety of these drugs in patients with T2DM and pre-existing HF.
Previous meta-analyses of recent randomised controlled trials found an increased risk of HF in DPP-4 inhibitor users;7 ,22 however, the principal findings were based mainly on data from the SAVOR-TIMI 53 trial, which did not focus specifically on patients with pre-existing HF. Without the inclusion of data from this trial, data on increased risks associated with DPP-4 inhibitor use are insufficient, and real-world observational studies are needed to address this knowledge gap. Some previous observational studies have suggested the patients with T2DM receiving incretin-based regimens were at increased risk of hospitalisation for HF,23 ,24 whereas other studies have found no such association25–27 or even a decreased risk of hospitalisation.28 ,29 However, only one of these studies focused on patients with pre-existing HF.24 That nested case–control study employing data from 7620 patients with incident HF recorded in a US claims database found that sitagliptin was associated with a 16% increased risk of rehospitalisation for HF.24 However, the results were limited by the small sample (only 4137 sitagliptin users) and non-inclusion of other classes of DPP-4 inhibitor. In addition, the definition of incident HF was based on the lack of a history of HF diagnosis within 1 year before the outcome of an HF event; thus, patients who had histories of HF >1 year before the outcome were misclassified as control subjects. In contrast, our study found a neutral effect of DPP-4 inhibitor use on the risk of HF.
The SAVOR-TIMI 53, EXAMINE and TECOS trials found that DPP-4 inhibitors had a neutral effect on the composite cardiovascular outcome. In the present study, the combination of MI and stroke occurred in fewer DPP-4 inhibitors. There are several possible explanations for the different ﬁndings between previous trials (the SAVOR-TIMI 53, EXAMINE and TECOS trials) and our study. First, our study sample size may provide sufficient power to detect a significantly lower risk of the combination of MI and stroke. Similarly, a previous large meta-analysis that included a total of 70 trials with 41 959 patients also found a lower risk of cardiovascular outcomes among DPP-4 inhibitor users and non-users.30 Second, the enrolment criteria and clinical characteristics of participants in these clinical trials differed substantially from those of patients encountered in our study. For example, the SAVOR-TIMI 53 trial included 66.6% of male, the EXAMINE trial included 68% of male, the TECOS trial included 70.9% of male in DPP-4 inhibitor users, whereas our study includes <50% of male. Second, most patients in this study were treated with sitagliptin and therefore the comparison to previous clinical trials using other DPP-4 inhibitors is somewhat problematic. Third, previous trials included a minority of Asians, but our study includes only Asians. Thus, the results of this study may not necessarily be extrapolated to other populations or studies.
The major strength of the present study lies in the use of data from unselected patients with T2DM from a large real-world national population, which provided greater power to detect small but clinically significant differences in the risk of readmission for HF compared with individual randomised controlled trials. In addition, analyses of the cardiovascular safety of DPP-4 inhibitors in patients with T2DM and pre-existing HF in previous clinical trials were post hoc, and the trials were not designed primarily to investigate this specific subpopulation. This study also has potential limitations. First, some laboratory data and imaging findings related to the likelihood of subsequent hospitalisation for HF, such as circulating levels of B-type natriuretic peptide, ejection fraction measurements and imaging of heart valve defects, were not available in our data set. However, as the same was true for the control group, this limitation did not result in a biased estimation of the effects of DPP-4 inhibitors on clinical outcomes in favour of the control group. Second, our analysis did not include examination of haemoglobin A1c levels, which are correlated with the severity of diabetes. However, propensity score matching accounted for the duration of DM and the use of hypoglycaemic agents. In addition, the severity of diabetic complications, based on aDCSI scores, was comparable in the DPP-4 inhibitor and control groups. The performance of aDCSI scores recorded in Taiwan's NHIRD in predicting the risk of hospitalisation has been validated and found to be comparable to the performance of other nationally representative claims databases.15 Third, the subgroup analyses of our study found although risks of all-cause mortality were lower in DPP-4 inhibitor users compared with non-users, the protective effects may be greater among those taking DPP-4 inhibitors in combination with insulin. These data might implicate that DPP-4 inhibitors have helped in patients with T2DM who have more severe diabetes and need a combination of insulin therapy, yet the topic is beyond the scope of our study and further investigation in this regard is warranted.
In this real-world nationwide cohort of patients with T2DM and pre-existing HF, we found that DPP-4 inhibitor use did not increase the risk of readmission for HF and was associated with lower risks of all-cause mortality and the combination of MI and ischaemic stroke. These findings extend those of previous research on the safety of DPP-4 use in this population, which may aid treatment decisions in routine clinical practice.
What is already known on this subject?
Recently, the Saxagliptin Assessment of Vascular Outcomes Recorded in Patients with Diabetes Mellitus–Thrombolysis in Myocardial Infarction 53 (SAVOR-TIMI 53) study raised concerns about the risk of heart failure (HF) in patients with type 2 diabetes mellitus (T2DM) treated with dipeptidyl peptidase-4 (DPP-4) inhibitors. Post hoc subgroup analysis using SAVOR-TIMI 53 trial data demonstrated an increased probability of hospitalisation for HF among patients with pre-existing HF.
What might this study add?
This is the first nationwide cohort study comparing the effects of DPP-4 inhibitors and non-users in patients with T2DM and pre-existing HF, which provides several new insights: (1) DPP-4 inhibitor users had a significantly lower risk of all-cause mortality (HR 0.67); (2) DPP-4 inhibitor users had a significantly lower risk of the combination of myocardial infarction and ischaemic stroke (HR 0.81), even after accounting for the competing risk of death compared with DPP-4 inhibitor non-users (HR 0.84); and (3) the risk of hospitalisation for HF did not differ significantly between DPP-4 inhibitor users and non-users, even those with pre-existing HF.
How might this impact on clinical practice?
These findings extend those of previous research on the safety of DPP-4 use in pre-existing HF population, which may provide clinical evidence of treatment choice in our daily practice.
This study was based in part on data from the NHIRD provided by the BNHI of the Department of Health and managed by the National Health Research Institute.
Contributors S-MO and H-TC contributed equally to the article. C-JS and Y-TC are joint senior authors. Conceived and designed the research: S-MO, H-TC, C-JS and Y-TC. Performed statistical analysis: S-MO, H-TC, S-CK, T-JC, C-JS and Y-TC. Handled funding and supervision: S-MO, C-JS and Y-TC. Acquired the data: S-MO, T-JC, C-JS and Y-TC. Drafted the manuscript: S-MO, H-TC, S-CK, T-JC, C-JS and Y-TC.
Funding This work was supported in part by grants from Taipei Veterans General Hospital (V104A-003; V104E4-003; V105A-003), and Taipei Veterans General Hospital-National Yang-Ming University Excellent Physician Scientists Cultivation Program (no. 104-V-B-044).
Competing interests None declared.
Ethics approval TCHIRB-10404107-W.
Provenance and peer review Not commissioned; externally peer reviewed.