Cardiovascular safety with linagliptin in patients with type 2 diabetes mellitus: a pre-specified, prospective, and adjudicated meta-analysis of a phase 3 programme

Background This study investigated the cardiovascular (CV) safety profile of the dipeptidyl peptidase (DPP)-4 inhibitor linagliptin versus comparator treatments. Methods This was a pre-specified meta-analysis of CV events in linagliptin or comparator-treated patients with type 2 diabetes mellitus (T2DM) from eight Phase 3 studies. All suspected CV events were prospectively adjudicated by a blinded independent expert committee. The primary endpoint was a composite of CV death, stroke, myocardial infarction, and hospitalization for unstable angina. Three secondary composite endpoints derived from the adjudicated CV events were also pre-specified. Risk estimates were calculated using several statistical methods including Cox regression analysis. Results Of 5239 treated patients (mean ± SD HbA1c 65 ± 10 mmol/mol [8.0 ± 0.9%], age 58 ± 10 years, BMI 29 ± 5 kg/m2), 3319 received linagliptin once daily (5 mg, 3159; 10 mg, 160) and 1920 received comparators (placebo, 977; glimepiride 1-4 mg, 781; voglibose 0.6 mg, 162). Cumulative exposure (patient-years) was 2060 for linagliptin and 1372 for comparators. Primary CV events occurred in 11 (0.3%) patients receiving linagliptin and 23 (1.2%) receiving comparators. The hazard ratio (HR) for the primary endpoint showed significantly lower risk with linagliptin than comparators (HR 0.34 [95% confidence interval (CI) 0.16-0.70]) as did estimates for all secondary endpoints (HR ranging from 0.34 to 0.55 [all upper 95% CIs < 1.0]). Conclusions These results from a large Phase 3 programme support the hypothesis that linagliptin may have CV benefits in patients with T2DM.


Background
Despite continuing medical and pharmacological efforts, patients with type 2 diabetes mellitus (T2DM) still bear a substantial burden of increased cardiovascular (CV) morbidity and premature mortality [1,2]. Although many risk factors are involved, hyperglycaemia remains an important contributor to increased CV disease incidence and seems to potentiate the deleterious effects of lipids and blood pressure elevation [2,3]. Nonetheless, recent large outcomes trials of glycaemic intervention in general and of intensive treatment in particular have shown conflicting results in terms of CV benefits for patients with T2DM [4][5][6][7]. The ambivalence of these findings has led to development of the hypothesis that the effectiveness of intensive glucose control likely depends on individualizing treatment (e.g. treatment modality and glycaemic target) to account for CV risk and other factors [8]. In particular, weight gain and increased hypoglycaemia are often associated with established glucose-lowering treatments that increase insulin secretion (in a glucose-independent manner) or insulin sensitivity and may heighten CV risk. Some agents, such as rosiglitazone, have been shown to increase risk for CV events possibly due to unanticipated pleiotropic CV effects [9]. In light of these concerns, regulatory authorities, including the US Food and Drug Administration (FDA) and the European Medicines Agency, have issued guidance that the development programmes for all new glucose-lowering therapies must show that treatment confers no unacceptable increases in CV risk [10,11].
The need to improve glycaemic control while minimizing harmful side effects has led to interest in therapeutic approaches aimed at avoiding such pitfalls. Dipeptidyl peptidase (DPP)-4 inhibitors, which enhance postprandial levels of the incretin hormones glucagonlike peptide (GLP)-1 and glucose-dependent insulinotropic polypeptide (GIP), have limited side effects [12]. The glucoregulatory actions of incretins include glucosedependent promotion of insulin secretion, glucagon suppression, delayed gastric emptying, and increased satiety.
Linagliptin is a DPP-4 inhibitor that was recently approved as a once-daily oral glucose-lowering drug in the USA, Japan, and Europe. Its molecular structure is xanthine-based, which differs from that of other DPP-4 inhibitors. Linagliptin has pharmacokinetic properties that confer a prolonged terminal half-life (t 1/2 > 100 h), and potent and durable DPP-4 inhibition (maximal inhibition of > 90% and inhibition 24 h after dosing of8 5% with linagliptin 5 mg at steady state); and unlike other DPP-4 inhibitors, it is primarily excreted via bile and the gut [13][14][15]. In Phase 3 trials, linagliptin has demonstrated clinically meaningful glycaemic efficacy and favourable safety/tolerability compared with placebo as monotherapy or in combination with metformin, metformin plus sulphonylurea, or pioglitazone [16][17][18][19].
To thoroughly determine the CV safety of linagliptin, we undertook a meta-analysis of the CV risk associated with linagliptin versus placebo or active comparators in patients with T2DM participating in the linagliptin Phase 3 study programme. This was a pre-specified meta-analysis in which suspected CV events were prospectively captured and adjudicated in a blinded fashion by an independent CV expert committee.

Study selection
The current meta-analysis included all randomized, double-blind, placebo-or active-controlled Phase 3 trials of linagliptin of > 12 weeks duration for which the database lock for the interim or final analyses was completed on or before 16 February 2010. This included eight studies that assessed linagliptin 5 mg or 10 mg/day versus placebo, glimepiride 1-4 mg/day, or voglibose 0.6 mg/ day over 18-52 weeks as monotherapy or in combination with various common background therapies (for further details see Additional file 1, or the individual study publications [16][17][18][19]).
All patients from each study provided written informed consent. Local ethics committees/institutional review boards reviewed and approved all study protocols. All studies were conducted within ethical standards and in accordance with the Declaration of Helsinki and any applicable regulatory requirements.

Analysis population
Common criteria across the Phase 3 trials included a diagnosis of inadequately controlled T2DM, age ≥ 18 years and, in most studies, body mass index (BMI) ≤ 40 kg/m 2 . Background medication with metformin was mandatory except where inclusion criteria required treatment-naïve patients, metformin-ineligible patients, or washout of pre-existing oral glucose-lowering drugs (including metformin). In all studies, rescue medication was provided with pioglitazone and/or insulin dose adjustment or supplementation for glycaemic deterioration, triggered by plasma glucose levels measurement on two separate occasions of > 13.3, > 11.1, or > 10.0 mmol/L after overnight fasts during the first 12, 12-24, or > 24 weeks, respectively.

CV event data collection and adjudication
Adverse events (AEs) were captured and collected onsite by the study investigators using electronic case report forms. AEs were then mapped to preferred terms according to the Medical Dictionary for Regulatory Activities (MedDRA). A pre-specified list of trigger events (the Standard MedDRA Queries for ischaemic heart disease and cerebrovascular disorders) and all fatal events were identified for adjudication. In patients with a trigger event, an individual patient data package (patient profile and all available cardiology or neurology tests, laboratory tests, and medical records) was prepared for the adjudication committee. A cardiology or neurology clinical event committee, based on the data package, without knowledge of the treatment allocation, adjudicated on the trigger event and recorded the type of event as appropriate. These adjudicated events were collected and included in the clinical trial database upon completion of database lock for the full or interim analyses.

Study endpoints
The primary endpoint was a composite of CV death (including fatal stroke and fatal myocardial infarction [MI]), non-fatal stroke, non-fatal MI, and hospitalization for unstable angina pectoris (UAP). Secondary endpoints were composites of: (i) CV death, non-fatal stroke, and non-fatal MI; (ii) all adjudicated CV events which included CV death, non-fatal stroke, non-fatal MI, UAP with or without hospitalization, stable angina pectoris (SAP), and transient ischaemic attacks (TIA); and (iii) FDA-defined custom major adverse CV events (MACE) derived from 34 unadjudicated MedDRA preferred terms for stroke and MI. Tertiary endpoints were the individual adjudicated components (as listed above) and total mortality.

Statistical analysis
Analyses were based on individual patient data in the treated set, which was defined as all patients who were randomized and received at least one dose of study medication, in all Phase 3 studies. Descriptive statistics (incidence and incidence rates per 1000 patient-years) were determined for all endpoints in each of the pooled treatment groups within the treated set.
The primary analyses assessed the CV risk for all primary, secondary, and tertiary endpoints associated with linagliptin versus total comparators. Risk estimates were calculated using several common statistical methods that included: (i) the hazard ratio (HR) for time to first event calculated using the Cox proportional hazards model with adjustments for study and treatment group; (ii) the incidence-rate risk ratio (RR) for time to first event calculated using Poisson regression with adjustment for study and treatment group; (iii) the odds ratio (OR) for occurrence of events calculated using a stratified Exact test; and (iv) the RR for occurrence of events calculated using a stratified Cochran-Mantel-Haenszel (CMH) test with continuity correction for trials with zero events.
Sensitivity analyses included evaluation of the primary endpoint associated with linagliptin versus total comparators in pre-specified subgroups based on age, gender, race, and use of rescue medication, as well as further exploratory subgroups based on occurrence of hypoglycaemia and Framingham 10-year CV risk score. In addition, a post hoc analysis of the primary endpoint assessed events in linagliptin-and placebo-treated patients, taken only from placebo-controlled trials and placebo-controlled periods within trials. In addition, the influence on the primary endpoint of the factors: study, treatment, gender, race and time since diagnosis of diabetes, was investigated using Cox regression.
This combined study analysis was developed to fully adhere to the recent FDA guidance on assessment of CV safety for the development of oral glucose-lowering drugs [11].
The overall mean (± SD) age, BMI, and HbA1c were 58 ± 10 years, 29 ± 5 kg/m 2 , and 64 ± 10 mmol/mol (8.0 ± 0.9%), respectively, and 52.4% of patients had known T2DM for > 5 years. The predominant race was white (60.5%), and there were more males (55.5%) than females. In total, 60.8% of patients had metabolic syndrome (based on the International Diabetes Federation definition), 10.6% coronary artery disease, 2.5% cerebrovascular disease, and 3.3% peripheral artery disease. In addition, 64.6% of patients were hypertensive, and 38.3% were current or ex-smokers. The prevalence of some degree of renal impairment was 24.2% or 44.6% of all patients, according to the Cockcroft-Gault (CG) or Modification of Diet in Renal Diseases (MDRD) formulae, respectively. In general, these baseline demographics and clinical characteristics were comparable between the linagliptin and comparator groups (Tables 1 and 2). Table 3 shows changes from baseline to last measurement for HbA1c, systolic and diastolic blood pressure (BP), total cholesterol, triglycerides and body weight for the pooled linagliptin and the pooled comparator group. Changes were of similar magnitude in both groups for all parameters except for HbA1c, where a meaningful HbA1c reduction was seen for linagliptin, and body weight, where a modest weight increase was seen in the total comparator group. These findings were expected since linagliptin is a glucose-lowering drug and were provided to all patients in the linagliptin group, whereas 50% of patients in the comparator cohort received placebo. Table 4 summarizes the incidence of each CV endpoint. Overall, adjudicated primary CV events occurred in 11 (0.3%) patients receiving linagliptin and 23 (1.2%) receiving comparators (3 on placebo, 20 on glimepiride, and none on voglibose). Notably, the main contributor to the overall differences in incidence of the primary endpoint was the events in the head-to-head study of linagliptin versus glimepiride (mean ± SD dose of glimepiride at week 52: 3.0 ± 1.2 mg).

Adjudicated CV events
Linagliptin treatment versus comparators was associated with reduced CV risk for the primary endpoint. The HR for the primary endpoint indicated a significant risk reduction, as did the OR and RR, for linagliptin versus comparator (i.e. upper bound of 2-sided 95% confidence interval [CI] < 1.0) (Figure 1). The difference in CV risk for the primary endpoint emerged after approximately 8 weeks and did not tend to plateau thereafter ( Figure 2). Further analysis of the primary endpoint for linagliptin against placebo in those patients taken from the same placebo-controlled trials confirmed that linagliptin was associated with no significantly increased risk for the primary endpoint. The HR, OR, and RR with linagliptin versus placebo ranged from 0.69 to 0.90, but all had upper limits of 95% CIs that included 1.0.
The incidence rates for the primary endpoint and the associated CV risk reductions with linagliptin versus comparators in a number of subgroups (based on age, gender, race, rescue medication use, hypoglycaemia occurrence, or Framingham CV risk score) were generally consistent with the results in the overall population (Table 5 and Additional file 2). There were no significant increases in risk based on the HR and RR for the primary endpoint with linagliptin relative to comparators in any subgroup. However, linagliptin did achieve significant risk reductions over comparators in several subgroups, including males, whites, those receiving no rescue medication, those reporting no hypoglycaemia, and those with higher CV risk (Framingham CV risk score > 15%). Further, in the Cox regression analysis the HR was 0.36 (95% CI: 0.17-0.74) in a model where study, treatment, gender, race and time since diagnosis of diabetes were taken into account, i.e., fully in line with the HR seen in the simple Cox model (0.34).
The HRs for all secondary endpoints indicated significantly lower CV risk with linagliptin than comparators ( Figure 3). Similarly, significant ORs and RRs for linagliptin versus total comparators were also observed, with the single exception of the RR for all adjudicated CV events when evaluated with CMH method, where the upper 95% CI was equal to 1.0 (see Additional file 3). Of the tertiary endpoints, most HRs with linagliptin versus comparators showed either a favourable trend for risk reduction, as in the case of CV death, non-fatal MI, UAP, and TIA, or neutrality, as in the case of SAP and total deaths (

Discussion
This CV meta-analysis indicates that linagliptin may have a beneficial or neutral impact on CV outcomes in a large population of patients with T2DM compared with control treatments. Furthermore, the risk for CV events was unchanged, or lowered, across a number of pre-specified subgroups based on key demographic and clinical characteristics. These results include comparisons with placebo, as well as two active comparators, namely glimepiride (a second-generation sulphonylurea frequently used as second-line therapy in the USA and Europe) and voglibose (an α-glucosidase inhibitor commonly used in Asia), either as monotherapy or in combination with common oral glucose-lowering drugs. The clinical characteristics of the overall study population were generally comparable to those reported in general T2DM populations. In this study, mean age was 58 years, BMI was 29 kg/m 2 , and 44.5% were females compared with age of 60 years, BMI of 32 kg/m 2 , and 52.4% females in the US National Health and Nutrition Examination Survey (NHANES) in 2003-2004 [20]. In this study versus the 2002 Cost of Diabetes in Europe (CODE)-2 study, the prevalence of previous MI was 10.6% versus 9.0% and of previous stroke was 2.5% versus 5% [21]. Furthermore, in the current study population, 44.6% had some degree of renal impairment versus 43.8% of those with self-reported T2DM in the NHANES population in 2009 (based on the MDRD equation) [22].
The incidence rates for CV events in this meta-analysis of linagliptin Phase 3 trials were relatively consistent  Values are mean ± SD. BP, blood pressure.
with those observed in previous CV meta-analyses of other DPP-4 inhibitors' clinical trial programmes. The incidence rates (per 1000 patient-years) for the primary CV endpoint were 5.3 for linagliptin versus 16.8 for total comparators. In comparison, other CV meta-analyses reported incidence rates for custom MACE ranging from 5.8 to 14.6 with sitagliptin, saxagliptin, or vildagliptin and 9.0 to 14.1 with comparators [23][24][25].   Although the results of the different meta-analyses of DPP-4 inhibitors are not entirely comparable (due to differences in primary composite endpoints and CV adjudication methods), all are supportive of the hypothesis that, in general, DPP-4 inhibitor treatment does not have a deleterious impact on the incidence of CV events. The present analysis shows that linagliptin treatment does not increase CV risk and may even yield CV benefits in patients with T2DM. Meta-analyses of other DPP-4 inhibitors were frequently retrospective in nature. However, the pre-specified design of the present metaanalysis involved prospective and blinded adjudication of CV events, which should strengthen the validity of the current findings. In addition, this meta-analysis was based on individual patient data from a consistently designed, large clinical development programme; this allows consistent derivation of endpoints and extensive subgroup analyses and minimizes between-study heterogeneity that can confound analyses of unrelated studies.
There are several mechanisms that could underlie potential CV benefits for linagliptin. First, linagliptin may confer the beneficial effects of improved glycaemic control, including the lowering of postprandial glucose, without the potentially harmful effects of weight gain or increased hypoglycaemia [12,26]. Second, linagliptin increases GLP-1 and GIP levels which may provide beneficial cardioprotection; experimental and clinical data suggest that GLP-1 elevation can positively modulate lipid metabolism [26], reduce infarct size and improve cardiac function [26,27]. Third, DPP-4 substrates include not only incretins but also vasoactive peptides involved in inflammation, immunity, and CV function; some evidence, mainly from preclinical studies, indicates that reduced DPP-4 activity can lessen inflammation, stimulate endothelial repair, and blunt ischaemic injury [28]. Finally, linagliptin holds inherent anti-oxidative properties, most likely due to its xanthine-based molecular structure [29]. These properties, both directly through reduction of reactive oxygen species and indirectly through beneficial effects on inflammatory mediators and endothelial function, could reduce the atherosclerotic burden [30].
This analysis has several potential limitations. First, despite a large total patient exposure of 3432 years, individual patient exposure was of a maximum duration of 1.7 years; further longer-term data are needed to confirm the current findings. Second, the low incidence of CV events, low rates of triple oral therapy, and lack of insulin treatment all suggest that a large proportion of patients had less advanced T2DM, and thus a lower CV risk than those with more advanced T2DM. However, around 30% of patients had a baseline Framingham 10year CV risk score of > 15% and more than a half also had > 5 years' known disease duration, which indicates a proportion of the population were at increased CV risk. Finally, the observed CV risk reductions for the primary and secondary endpoints were influenced by the differences in CV events in one study with linagliptin versus glimepiride. Despite this, it is important to note that glimepiride is an established and recommended second-line therapy with a well-characterized safety profile, which has not been directly linked to increased CV risk either as part of intensive treatment regimens or when compared with other conventional treatments [31]. Moreover, analysis of the pooled placebo studies alone confirmed that linagliptin did not increase CV risk against placebo.

Conclusions
In summary, this pre-specified CV meta-analysis of a large Phase 3 programme that involved prospective and independent adjudication of CV events provides valuable new insights on the CV safety profile of linagliptin.