Linagliptin and cardiovascular outcomes in type 2 diabetes after acute coronary syndrome or acute ischemic stroke

The National Health Insurance (NHI) program covers more than 99% of the 23 million people in Taiwan. All the submitted standardized information and data on healthcare service are prospectively recorded by the NHI Research Database (NHIRD). The diagnoses are registered using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes. The NHI Bureau routinely and comprehensively performs the validation of accurate records of beneficiaries, including ambulatory visits, inpatient care, disease diagnosis codes and medication prescriptions from the NHIRD data [27‐30]. This kind of nationwide database from the NHIRD is important and contributory to many large population-based studies [31]. The personal informations and records of the patients were de-identified before analysis to ensure patients’ anonymity. This study was approved by the Ethics Institutional Review Board of Chang Gung Memorial Hospital (201701079B1).

This open observational noncrossover retrospective cohort study was derived from the NHIRD. Between June 1, 2012 and December 31, 2013, a total of 1,759,222 T2DM (ICD-9-CM code: 250) patients were initially enrolled and after applying exclusion criteria, a final total of 1203 T2DM patients who were hospitalized for ACS (ICD-9-CM codes: 410–411) or AIS (ICD-9-CM codes: 433–435) were included in our study (Fig. 1). The index hospitalization was defined as the date on which patient was admitted for ACS or AIS. In addition to identifying T2DM patients using ICD-9-CM codes, we defined T2DM patients with at least 90 days of prescribed oral hypoglycemic agents or insulin injection within 1 year of the index hospitalization.

The follow-up period was based upon the index hospitalization to date of death, loss of follow-up, or until December 31, 2013. Patients’ baseline characteristics, comorbidities, medication prescription and previous medical procedures, such as percutaneous coronary intervention (PCI), coronary artery bypass grafting (CABG) and carotid stenting were identified. Patients were excluded if they met any of the following criteria: (1) age below 40 years; (2) use of any DPP-4 inhibitors except for linagliptin or any glucagon-like peptide-1 receptor agonists (GLP-1 RA) before the index hospitalization; (4) use of other DPP-4 inhibitors or any GLP-1 RA after the index hospitalization; (5) expiration during index admission; (6) development of composite primary cardiovascular outcome within 1 month of index hospitalization or (7) follow-up for less than 1 month after index hospitalization. The exclusion criteria are shown in Fig. 1.

In the study period, the exposure to linagliptin was based on a computer-based prescription claims after the index hospitalization. Patients with T2DM after ACS or AIS were classified into the linagliptin group or the control group in which patients did not receive incretin-based therapies (any DPP-4 inhibitors or GLP-1 RA) after ACS or AIS.

Baseline comorbidities were identified using ICD-9-CM diagnosis codes and medications during the index hospitalization (Tables 1, 2). The primary composite outcome included cardiovascular death, non-fatal myocardial infarction and non-fatal ischemic stroke. The definition of cardiovascular death meet the criteria of Standardized Definitions for End Point Events in Cardiovascular Trials draft by the Food and Drug Administration. Death and causes of death were according to the registry data of NHIRD [32]. Secondary outcomes of interest were death due to any cause, hospitalization for heart failure, percutaneous coronary intervention and coronary artery bypass grafting. Safety outcomes were risks of hypoglycemia, diabetic ketoacidosis, hyperosmolar hyperglycemic state, acute pancreatitis, de novo dialysis, acute hepatitis or newly diagnosed malignancy.

We matched the linagliptin cohort with the control group by a 1:2 ratio based on patient’s characteristics, baseline comorbidities, medication prescribed 90 days since indexed hospitalization (listed in Tables 1, 2), and index year and month by propensity score matching (PSM) to minimize potential selection bias for this cohort study. Clinical characteristics between these two study groups were compared using Chi square test for categorical variables and independent sample t test for continuous variables. Differences between these two study groups in time of the first occurrence of a predefined primary or secondary outcome after index hospitalization were determined by Cox proportional hazard models in which the study group (linagliptin group versus control group) was the only explanatory variable. Time-to-event outcomes were analyzed using predefined periods, including 6 months and until the final follow-up for each study group using the Kaplan–Meier method and log-rank test. A P value of less than .05 was considered statistically significant. All data analyses were performed using the SAS version 9.4 (SAS Institute, Cary, NC).

A total of 1,759,222 T2DM patients were initially enrolled between June 1, 2012 and December 31, 2013, among whom 30,115 T2DM patients were admitted for ACS or AIS. After applying the exclusion criteria, a total of 17,628 T2DM patients aged ≥ 40 years who were hospitalized for ACS or AIS were eligible for our study cohort. After PSM was used to reduce potential confounding and selection bias, the data of 1203 patients were finally included for analyses (Fig. 1).

Among the 1203 included patients, 401 (33.3%) were in the linagliptin group and 802 matched patients (66.7%) were in the control group. The mean age for the overall cohort was 69.3 years (standard deviation [SD] = 11.2 years). The mean follow-up period was 4.7 months (SD = 2.7 months) and the maximum follow-up time was 15 months. After PSM, the two study groups were well matched in terms of baseline characteristics, comorbidities and non-study medications (Tables 1 and 2). The most common co-morbidity was hypertension (84% vs. 85%), followed by chronic kidney disease (48.4% vs. 44.6%) and dyslipidemia (46.4% vs. 43.6%) in the linagliptin and the control groups, respectively. In addition, patients with old myocardial infarction, heart failure and old ischemic stroke in the linagliptin group were 11.2, 17.5 and 25.2%, respectively; in the control group, those with old myocardial infarction, heart failure and old ischemic stroke were 10.3, 16.1 and 28.6%, respectively (Table 1).

Based on the summary of matched nonstudy medications of anti-diabetic agents (Table 2), both groups had high prevalence of insulin use (linagliptin group: 60.8% and control group: 60.3%), sulfonylurea use (linagliptin group: 50.9% and control group: 55.5%), and alpha-glucosidase inhibitors (linagliptin group: 20.2% and control group: 20.6%).

The event-free survival curves of primary composite outcome and each component in these two study groups were plotted (Fig. 2a–d). Events of primary composite outcome occurred in 28 patients (7%) in the linagliptin group and in 49 patients (6.1%) in the control group (HR 1.06; 95% CI .66–1.68) at the final follow-up (Fig. 3). With regard to the individual composite outcome, there was no significant differences in the risks of cardiovascular death (HR 1.12; 95% CI .55–2.29), non-fatal myocardial infarction (HR 1.53; 95% CI .64–3.70) and non-fatal ischemic stroke (HR .96; 95% CI .45–2.07) (Fig. 3). In the subgroup analysis, there were no significant interactions in the prespecified primary cardiovascular outcomes (Fig. 4).

Tables 3 and 4 show primary outcomes in T2DM patients with ACS or AIS, respectively and reveal linagliptin had a neutral effect on cardiovascular death, non-fatal myocardial infarction and non-fatal stroke in either ACS cohort or AIS cohort.

For secondary outcomes, no significant differences occurred between the linagliptin group and the control group in the respective incidence of all-cause mortality (5.5 and 5.9%; P = .550), hospitalization for heart failure (5.2 and 5.1%; P = .964), percutaneous coronary intervention (7.2 and 4.6%; P = .098) and coronary artery bypass grafting (1.0 and .9%; P = .868) (Fig. 5). Tables 3 and 4 show secondary outcomes in T2DM patients with ACS or AIS, respectively and reveal no significant differences in either ACS cohort or AIS cohort.

The linagliptin and control groups did not differ significantly with respect to the incidences of hypoglycemia (4.2 and 3.7%; P = .765), diabetic ketoacidosis/hyperosmolar hyperglycemic state (.7 and 1.1%; P = .502), acute pancreatitis (0 and .4%; P = not applicable), de novo dialysis (4.7 and 5.9%; P = .600), acute hepatitis (.5 and 0%; P = not applicable), or newly diagnosed malignancy (2.0 and 2.7%; P = .633) (Fig. 5).

Several experimental and clinical researches revealed that T2DM patients who were treated with DPP-4 inhibitors had lower risks for cardiovascular diseases as compared to those treated without DPP-4 inhibitors, except for metformin users [33], and DPP-4 inhibitors did not increase the risk of heart failure compared with sulfonylurea [34]. With regard to linagliptin, several studies with limitations of pooled analysis showed that linagliptin in patients with T2DM is not associated with an increase in cardiovascular events [35‐38]. The results of our study for primary composite outcome (cardiovascular death, non-fatal myocardial infarction and non-fatal ischemic stroke) are compatible with those of previous randomized controlled trials (i.e., SAVOR [12], EXAMINE [13], and TECOS [14] trials), which indicated that short-term use of DPP-4 inhibitors have a neutral effect on cardiovascular death, myocardial infarction and stroke. Prespecified subgroup analyses in our study suggested that no significant interactions were observed, even in chronic kidney disease (CKD) with or without dialysis. Compared to the results of subgroup analyses in the EXAMINE study which excluded end-stage renal disease (ESRD) patients, the CKD subgroup analyses revealed that alogliptin had an increased primary cardiovascular risk trend in the moderate or severe renal impairment group when compared to the normal or mild renal insufficiency group (P value for interaction with treatment = .046) [13].

For issues of hospitalization for heart failure, the SAVOR-TIMI 53 study indicated a 27% increase in hospitalization for heart failure among T2DM patients receiving saxagliptin as compared with those receiving placebo (HR 1.27; 95% CI 1.07–1.51; P = .007) [12]. The main predictors of hospitalization for heart failure were previous history of heart failure, elevated brain natriuretic peptide (BNP) and CKD [39]. Compared to patients in the SAVOR study, our patients had more previous history of heart failure (17.5% in the linagliptin group and 16.1% in the control group) than did those with previous history of heart failure in the SAVOR study (12.8% in the saxagliptin group and 12.8% in the placebo group). Besides, unlike the SAVOR study which excluded patients with ESRD receiving long-term dialysis, our study enrolled those with ESRD on dialysis (8.2% in the linagliptin group and 7.5% in the control group). As a result, our result showing that linagliptin had a neutral effect on hospitalization for heart failure could be more convincing and supported that it should not be a class effect of all DPP-4 inhibitors in regards to heart failure.

In preclinical studies, linagliptin could prevent female mice from western diet-induced vascular abnormalities [40], and even reverse western diet-induced diastolic dysfunction possibly by targeting TRAF3IP2 expression which is associated with downstream inflammatory signaling [41]. DPP-4 inhibitor-medicated increased GLP-1 may have a direct effect at the neuronal level of brain which was suspected to be associated with the neuroprotective effect on the stroke mice [42].

In clinical studies, the initial combination of linagliptin and metformin substantially enhanced glycemic control without weight gain and with infrequent hypoglycemia [43] and also significantly improved microvascular function in the fasting state [44]. Besides, the special issues of linagliptin is about neuroprotective effects based on some evidences showing that patients treated with linagliptin were significantly associated with fewer events of stroke [25, 26]. Patients with diabetes have a twofold excess risks for ischemic stroke compared with those without diabetes [2] and acute stroke could lead to stress hyperglycemia with increased mortality and poor prognosis [45]. Moreover, stroke-mediated damage could increase the permeability of the BBB which may have a influence on DPP-4 inhibitors for neuroprotection [46]. In the present study, we included patients with recent AIS as a part of the study population. Nevertheless, our results did not find a significant anti-stroke effect with linagliptin treatment at the final follow-up. A 2-year, randomized, double-blind, non-inferiority trial including 1,552 patients by Gallwitz et al. indicated that patients treated with linagliptin had significantly fewer non-fatal stroke than those treated with glimepiride (3 vs. 11 patients; HR .27; 95% CI .08–.97; P = .03) [25]. The actual reason for this discrepancy is unclear. However, any of the following explanations may apply. First, the patients in our study conducted in Taiwan were mostly an Asian population. By comparison, the study by Gallwitz et al. only enrolled 12% Asian patients. Because intracranial atherosclerosis is relatively common in Asia [47], the clinical influence of linagliptin may be different in Asian subjects compared to Western subjects. Second, the etiology of ischemic stroke is heterogeneous with large vessel disease, small vessel disease, and embolism which may lead to different effects of linagliptin according to the different etiologies. Third, in our study, we enrolled patients with old ischemic stroke (28.1% in the linagliptin group and 29.5% in the control group) after recent AIS; in contrast, the study by Gallwitz et al. excluded patients with stroke or transient ischemic attack within 6 months before enrollment [25]. Because of the obvious differences in disease severity of the populations, the overall rate of non-fatal ischemic stroke in our study was 2.4% at 15 months but the overall non-fatal stroke rate of the study by Gallwitz et al. was only .9% at 2 years.