FormalPara Key Points

Large randomized trials have shown conflicting evidence regarding the cardiovascular safety of dipeptidyl-peptidase 4 (DPP-4) inhibitors.

This meta-analysis demonstrated that the safety profile of DPP-4 inhibitors is similar to placebo.

Patients and providers can feel reassured that DPP-4 inhibitors are well tolerated and can represent a valuable component in the armamentarium of anti-glycemic therapy.

1 Introduction

In the USA, diabetes mellitus affects approximately 21 million adults [1]. Dipeptidyl-peptidase 4 (DPP-4) inhibitors have emerged as a new class of incretin-based medications for the management of type 2 diabetes mellitus [2]. They improve glucose control without inducing hypoglycemia or weight gain [3]. Some authors had suggested that these medications might exert a cardiovascular protective effect [4]. The American Diabetes Association and the European Association for the Study of Diabetes recommend this class as a second-line agent in patients with type 2 diabetes [5]. In the largest randomized trial designed to explore the cardiovascular safety of this class of medications, there was a 27% relative increased risk for heart failure hospitalizations with saxagliptin compared with placebo [6]. In two other large randomized trials testing alogliptin and sitagliptin, the risk of heart failure was similar compared with placebo [7, 8]. Furthermore, data from real world registries have yielded inconsistent results regarding the risk of heart failure with this class of medications [9–13]. Previous meta-analyses were limited by incomplete evaluation of cardiovascular safety outcomes [14–16], non-comprehensive evaluation of data [17–21], or inclusion of data from observational studies [22]. Given the uncertainty about the cardiovascular safety of this class of medications, we aimed to conduct a comprehensive meta-analysis of placebo-controlled randomized trials to test the cardiovascular safety of this class of medications.

2 Methods

We searched the MEDLINE database without language restriction from inception until August 2015 using the keywords and Medical Subject Headings illustrated in Fig. 1. We also searched the Web of Science and the Cochrane Register of Controlled Trials databases using similar keywords. This meta-analysis was registered at the PROSPERO international prospective register of systematic reviews (CRD42015024674) [23].

Fig. 1
figure 1

Summary of how the systematic search was conducted and eligible studies were identified (PRISMA flow diagram)

Full size image

We selected randomized controlled clinical trials that compared any of the DPP-4 inhibitors (i.e., anagliptin, alogliptin, dutogliptin, linagliptin, omarigliptin, sitagliptin, saxagliptin, teneligliptin, and vildagliptin) with placebo in patients with type 2 diabetes. We required that the published report for the study explicitly reported any cardiovascular outcome (namely, heart failure, all-cause mortality, cardiovascular mortality, myocardial infarction, or ischemic stroke). We excluded trials that compared DPP-4 inhibitors with any other comparator agent (e.g., metformin, sulfonylurea, or thiazolidinediones) in order to test the relative safety of DPP-4 inhibitors with placebo. We also excluded trials that had unequal distribution of a second oral hypoglycemic agent in either arm. For trials with multiple comparison arms, we combined the DPP-4 inhibitor arms irrespective of the doses. For trials that had a second phase with an active agent (i.e., another oral hypoglycemic agent), we reported outcomes only for the placebo-controlled phase. For these trials, we excluded the ones that reported cardiovascular outcomes only at the end of the second phase.

Teams of two paired reviewers extracted data on general study data, study design, sample size, patient characteristics, interventional strategies, and cardiovascular outcomes from the included studies. Two reviewers further reviewed the extracted data to ensure accuracy. Reviewers resolved any discrepancies by discussion. For all clinical outcomes, we tabulated the number of events that occurred in each arm. Since it is mandated that parties submit a summary of the results including serious adverse events to clinicaltrial.gov, we searched this registry for each of the included studies to ensure that we collected any possible cardiovascular outcome. We evaluated the quality of the included trials on the basis of adequate description of treatment allocation, blinded outcome assessment, and description of losses to follow-up [24].

The main outcome evaluated in this analysis was heart failure. We defined heart failure as any reported case of “heart failure,” “cardiac failure,” or “hospitalization for heart failure.” We also evaluated all-cause mortality, cardiovascular mortality, myocardial infarction, and ischemic stroke (defined as ischemic stroke or transient ischemic attack). If a study reported outcomes at different follow-up periods, we preferentially extracted data for the longest reported follow-up.

We followed the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines to conduct a high-quality meta-analysis [25]. We analyzed the outcomes with an intention-to-treat analysis. Since we anticipated that the outcomes are rare [26], we constructed the summary estimate odds ratios (ORs) primarily with Peto’s model [27]. We performed a secondary analysis using random effects summary risk ratios (RRs) with a DerSimonian and Laird model [28]. We performed the overall analysis for each outcome using the data reported from the longest follow-up period. We examined the statistical heterogeneity using I 2 statistic [29]. We evaluated publication bias with Egger’s method [30]. All P values were two-tailed, with statistical significance set at 0.05, and confidence intervals (CIs) were calculated at the 95% level for the overall estimate effect. We conducted all analyses with STATA software version 14 (STATA Corporation; College Station, TX, USA).

For the outcome of heart failure, we further performed a pre-specified sensitivity analysis including only the high-quality trials, and another sensitivity analysis including only trials that enrolled >1000 subjects. We conducted subgroup analyses according to the follow-up time (i.e., ≤24 weeks, 24–52 weeks, >52–104 weeks, and >104 weeks). Random effects meta-regression analyses were pre-specified for the outcome of heart failure with the type of DPP-4 inhibitor, age, male gender, hemoglobin A1c (HbA1c) level, duration of diabetes, and body mass index (BMI).

3 Results

The electronic search yielded 1040 articles, which we screened by reviewing the title and/or abstract; 137 articles were deemed potentially eligible. Upon further review of the full article, 90 trials with 66,730 patients were included in the final analysis (Fig. 1) [6–8, 31–117]. Table 1 summarizes the baseline characteristics of the included trials. Sitagliptin was the most frequent medication tested (i.e., in 24 trials), while 15 trials evaluated saxagliptin. All of the included studies were multicenter and double blinded. The follow-up time ranged from 2 to 156 weeks. Supplemental Table 1 reports the quality of the included studies (see the electronic supplementary material, online resource 1). The primary outcome for most of the included studies was the change in the HbA1c level at the end of the follow-up period. Only three trials evaluated cardiovascular outcomes as the primary outcome [6, 8, 50].

Table 1 Baseline characteristics of the included trials
Full size table

Twenty-five trials evaluated the outcome of heart failure (nine of these studies had zero events in both arms). The mean follow-up was 108 ± 45 weeks (median follow-up time 109 weeks). We utilized data regarding heart failure events for the Examination of Cardiovascular Outcomes with Alogliptin versus Standard of Care (EXAMINE) trial from the pre-specified analysis that evaluated heart failure [7]. Compared with placebo, DPP-4 inhibitors were associated with a non-significant increase in the risk of heart failure using both Peto’s method (OR 1.11, 95% CI 0.99–1.25, P = 0.07) (Fig. 2) and the DerSimonian and Laird model (RR 1.11, 95% CI 0.99–1.24, P = 0.09, I 2 = 0%). There was no evidence of publication bias with Egger’s test (P = 0.23). The pre-specified sensitivity analysis including only high-quality studies yielded similar results (OR 1.13, 95% CI 0.99–1.28, P = 0.07, I 2 = 27%), as well as the pre-specified analysis limited to trials that enrolled >1000 subjects (OR 1.12, 95% CI 1.00–1.26, P = 0.06, I 2 = 42%) (Supplemental Figure 1). Subgroup analyses according to the follow-up time showed that the risk of heart failure was comparable to placebo at ≤24 weeks (OR 0.48, 95% CI 0.13–1.73), at 24–52 weeks (OR 2.64, 95% CI 0.54–12.87), and at >52–104 weeks (OR 0.35, 95% CI 0.02–7.60), but a non-significant increase in heart failure at >104 weeks (OR 1.12, 95% CI 0.99–1.26) (P for interaction = 0.27). Meta-regression analysis did not identify a difference in treatment effect based on the type of DPP-4 inhibitors, age, male gender, HbA1c level, duration of diabetes, and BMI (P = 0.76, 0.34, 0.24, 0.23, 0.66, and 0.10, respectively).

Fig. 2
figure 2

Summary plot for heart failure. Trials were listed in the forest plot only if there was at least one event in either arm. The relative size of the data markers indicates the weight of the sample size from each study. CI confidence interval, DPP-4 dipeptidyl-peptidase 4, OR odds ratio

Full size image

Compared with placebo, DPP-4 inhibitors were associated with a similar risk of all-cause mortality (OR 1.03, 95% CI 0.94–1.12, P = 0.53, I 2 = 0%) (Fig. 3), cardiovascular mortality (OR 1.02, 95% CI 0.92–1.14, P = 0.72, I 2 = 0%) (Supplemental Figure 2), myocardial infarction (OR 0.98, 95% CI 0.88–1.09, P = 0.69, I 2 = 10%) (Supplemental Figure 3), and ischemic stroke (OR 0.99, 95% CI 0.85–1.15, P = 0.92, I 2 = 20%) (Supplemental Figure 4). There was no evidence of publication bias for the secondary outcomes. In Table 2, we summarize the summary estimates for the outcomes assessed in this meta-analysis.

Fig. 3
figure 3

Summary plot for all-cause mortality. Trials were listed in the forest plot only if there was at least one event in either arm. The relative size of the data markers indicates the weight of the sample size from each study. CI confidence interval, DPP-4 dipeptidyl-peptidase 4, OR odds ratio

Full size image
Table 2 Summary estimates for the outcomes assessed
Full size table

4 Discussion

In this comprehensive meta-analysis of 90 double-blind, multicenter, placebo-controlled randomized clinical trials with 66,730 patients; we demonstrated that DPP-4 inhibitors were associated with a non-significant increase in the risk of heart failure at a mean of 108 weeks. We performed various sensitivity and meta-regression analyses to further explore any potential explanation for this observed finding. Our results suggested that any potential increase in the risk of heart failure was driven by one large trial [6]. Reassuringly, we also demonstrated that DPP-4 inhibitors were associated with a similar risk of all-cause mortality, cardiovascular mortality, myocardial infarction, and ischemic stroke compared with placebo.

In the Saxagliptin Assessment of Vascular Outcomes Recorded in Patients with Diabetes Mellitus (SAVOR)–Thrombolysis in Myocardial Infarction (TIMI) 53 trial, saxagliptin neither increased nor decreased the composite of cardiovascular death, myocardial infarction, or ischemic stroke compared with placebo [6]. There was an unexpected 27% increased relative risk of hospitalization for heart failure in the saxagliptin arm. A post hoc analysis of this trial revealed that the risk of hospitalization for heart failure was increased among patients with elevated levels of natriuretic peptides at baseline, previous heart failure, or chronic kidney disease [118]. It remains unclear how saxagliptin might predispose to heart failure; a pooled analysis of 20 trials suggested that there was no evidence of fluid retention or weight gain with saxagliptin [21]. One plausible explanation for the increased heart failure risk in the SAVOR-TIMI 53 trial was the relatively large number of subjects with a prior history of heart failure at baseline (~13%). Most of the large randomized trials evaluating oral hypoglycemic agents in general had a lower number of subjects with previous heart failure history [119]. In the two other large randomized trials evaluating the cardiovascular outcomes with DPP-4 inhibitors [i.e., Trial Evaluating Cardiovascular Outcomes with Sitagliptin (TECOS) and EXAMINE], the risk of hospitalization for heart failure was not increased with either sitagliptin or alogliptin, respectively [7, 8, 50]. We performed a sensitivity analysis for these three large trials and found that there was a non-significant increase in the risk of heart failure, again driven by the results of the SAVOR-TIMI 53 trial. In a large multicentre cohort of 1,499,650 diabetic patients, incretin-based drugs [i.e., DPP-4 inhibitors and glucagon-like peptide-1 (GLP-1) analogs] were not associated with an increased risk of hospitalization for heart failure, as compared with oral antidiabetic drugs [120]. Our analysis of placebo-controlled randomized trials further supports that DPP-4 inhibitors as a class have only weak evidence for an increased risk of heart failure.

A recent systematic review evaluated the risk of heart failure with DPP-4 inhibitors and concluded that the risk of heart failure is uncertain with DPP-4 inhibitors [22]. However, that analysis was limited by the inclusion of observational studies that can be prone to bias. In addition, the authors used placebo and active agents in the comparator arm, which could have affected their results. In the present analysis, we included only placebo-controlled randomized trials in order to conduct a robust analysis. Furthermore, we assessed a wide spectrum of cardiovascular outcomes besides heart failure (i.e., all-cause mortality, cardiovascular mortality, myocardial infarction, and ischemic stroke), in order to provide a comprehensive analysis on the cardiovascular safety of DPP-4 inhibitors.

The present analysis has some limitations. First, the follow-up duration was variable among the included studies; thus we performed several subgroup analyses according to the follow-up time and demonstrated that the results were almost similar among these subgroups. Second, we performed our primary analysis with a fixed effects model (i.e., Peto’s). We determined that Peto’s model would be a good model for this particular analysis, given that the outcomes that we assessed were rare [26]. Furthermore, a secondary analysis with a DerSimonian and Laird model showed that the results were fairly robust irrespective of the methodology used. Third, the definition of heart failure was variable among the included studies; however, we observed no heterogeneity with statistical testing. Fourth, most of the included studies were small and not designed to address cardiovascular outcomes; however, all the included studies were designed to test the safety of the medication. In addition, we performed a sensitivity analysis limited to the three trials that tested cardiovascular outcomes as the primary outcome, which yielded similar results. Fifth, a considerable number of the studies had a significant drop-out rate at the end of the follow-up period; therefore we performed a sensitivity analysis excluding these low-quality studies. Finally, a lack of access to patient level data precluded a full evaluation to identify patient characteristics (e.g., renal disease, prior history of heart failure) associated with the potential risk for heart failure; however, we performed multiple meta-regression analyses using the available study-level data and found that none of the tested demographics were significant.

5 Conclusion

In patients with type 2 diabetes, DPP-4 inhibitors are relatively well tolerated compared with placebo. As a class, there is only weak evidence for an increased risk of heart failure.