Pharmacologic Differences of Sulfonylureas and the Risk of Adverse Cardiovascular and Hypoglycemic Events

Despite the recent approval of several novel groups of antidiabetic drugs, sulfonylureas remain a cornerstone in the treatment of type 2 diabetes (1). However, concerns remain regarding their association with cardiovascular and hypoglycemic adverse events. Although these events have been investigated in several observational studies, their findings have been contradictory, in part because of important methodological shortcomings, including exposure misclassification, time-lag bias, and selection bias (2). Furthermore, few studies have examined the risk of cardiovascular and hypoglycemic adverse events with individual sulfonylureas, as these differ in their pharmacodynamic and pharmacokinetic properties (3). Indeed, some sulfonylureas such as glyburide and glimepiride lack selectivity for pancreatic β-cells, allowing them to bind to other receptors, including those on cardiomyocytes and vascular smooth muscle cells. In the case of glyburide, such binding has been suggested to inhibit ischemic preconditioning (4–7). Moreover, other sulfonylureas such as chlorpropamide, glyburide, and glimepiride yield pharmacologically active metabolites, which could prolong the duration of action and result in an increased risk of hypoglycemia (3).

Thus, the objective of this population-based study was to determine whether the risk of acute myocardial infarction (AMI), ischemic stroke, cardiovascular death, all-cause mortality, and severe hypoglycemia is different among sulfonylureas grouped according to their specificity to the pancreatic β-cells and duration of action.

This study was conducted using the U.K. Clinical Practice Datalink (CPRD) linked to the Hospital Episode Statistics (HES) and Office for National Statistics (ONS) databases. The CPRD is a large primary care database containing the medical records for >14 million people enrolled in >680 general practices (8). Medical diagnoses and procedures are recorded using the Read code classification, and drugs prescribed by general practitioners are coded based on the U.K. Prescription Pricing Authority dictionary. The CPRD contains information on anthropometric variables (such as BMI) and lifestyle variables (such as smoking and alcohol use), and its data have been previously validated and shown to be of high quality (9). The HES contains all inpatient and day-case hospital admission information, including primary and secondary diagnoses (coded using the ICD-10) and hospital-related procedures (coded using the Office of Population Censuses and Surveys classification of interventions and procedures, version 4). The ONS contains the electronic death certificates of all citizens living in the U.K. and was used to identify the underlying cause of death (coded using the ICD-9 and ICD-10 classifications) for all patients who died during follow-up. The linkage of the HES and ONS to the CPRD is possible from 1 April 1997 onward and is limited to English general practices that have consented to the linkage scheme (currently representing 75% of all English practices) (9).

The study protocol was approved by the Independent Scientific Advisory Committee of the CPRD (protocol 14_019AMn) and the Research Ethics Board of the Jewish General Hospital (Montreal, Quebec, Canada).

The cohort consisted of patients newly treated for type 2 diabetes with sulfonylureas in monotherapy between 1 April 1998 and 31 March 2013, with follow-up until 31 March 2014. Cohort entry was defined by the date of the first sulfonylurea prescription. We excluded all patients <40 years of age, as well as those with <1 year of medical history in the CPRD before cohort entry. We also excluded patients prescribed other antidiabetic drugs at any time before cohort entry. Because this study aimed to be representative of real-world practice, patients with a history of cardiovascular disease at cohort entry were not excluded.

Patients meeting the study inclusion criteria were followed until the earliest of the following events: treatment discontinuation, add-on, or switch (defined in detail in exposure definition), occurrence of one of the study outcomes (defined in detail in study outcomes), death from any cause, end of registration with the general practice, or end of the study period (31 March 2014).

Patients were classified into two groups based on the sulfonylurea they initiated: pancreas nonspecific, long-acting sulfonylureas (glyburide and glimepiride), and pancreas-specific, short-acting sulfonylureas (tolbutamide, gliclazide, and glipizide; reference group) (3,5–7). Other compounds such as chlorpropamide or gliquidone were excluded becauase of the low number of exposed patients (less than five). We used an as-treated exposure definition in which patients were considered continuously exposed if the duration of one prescription overlapped with the date of the subsequent prescription. In the event of nonoverlap, we allowed for a 30-day grace period between two successive prescriptions. Termination of treatment was therefore defined by either the absence of a new prescription by the end of a 30-day grace period, a switch to a different sulfonylurea group, or an add-on or switch to another antidiabetic drug, whichever came first. Switches within the same group of sulfonylureas (e.g., glyburide to glimepiride or vice versa) were permitted.

We considered five outcomes: hospitalization for AMI, hospitalization for ischemic stroke, cardiovascular death, all-cause mortality, and hospitalization for hypoglycemia. Hospitalization for AMI and ischemic stroke were identified using the HES and the ONS databases (AMI ICD-9 codes: 410.x; ICD-10 codes: I21.x; ischemic stroke ICD-9 codes: 433.x, 434.x, and 436.x; ICD-10 codes: I63.x and I64.x; in primary or secondary position). The diagnostic codes to identify AMI in HES have been shown to be highly valid (>90%) (10), whereas the validity of stroke diagnoses in administrative data is also high (>80%) (11). Cardiovascular death was identified in the ONS (underlying cause of death; ICD-9 codes: 390.x–398.x, 401.x–405.x, 410.x–417.x, 420.x–429.x [except 427.5], 430.x–438.x, 440.x–447.x; and ICD-10 codes: I00.x–I77.x [except I46.9]), and all-cause mortality was identified from all three databases. Hospitalization for hypoglycemia was identified in HES (ICD-10 code: E16.2; in primary or secondary position).

For each patient, we used multivariable logistic regression to calculate a high-dimensional propensity score (hdPS); this method empirically selects covariates based on their prevalence and potential for confounding (12). The hdPS defined the probability of being exposed to nonspecific, long-acting sulfonylureas as compared with specific, short-acting sulfonylureas, conditional on several predefined and 500 empirically identified covariates measured at cohort entry.

We identified the empirical covariates from seven data dimensions (five dimensions from the CPRD: drug prescriptions, procedures, diagnoses, disease history, and administrative information; and two dimensions form the HES: diagnoses and procedures), whereas the predefined covariates were age, sex, calendar year, duration of diabetes prior to cohort entry (defined as the time since a first diagnosis of type 2 diabetes or an elevated glycated hemoglobin A_1c [HbA_1c] level [>7.0% or 53 mmol/mol], whichever appeared first in the medical record), alcohol-related disorders (based on diagnoses for alcohol-related disorders such as alcoholism, alcoholic cirrhosis of the liver, alcoholic hepatitis, and failure and other related disorders), smoking status (ever, never, or unknown), BMI category (<25, 25–30, or ≥30 kg/m² or unknown), HbA_1c level (≤7% or ≤53 mmol/mol, 7.1–8.0% or 54–64 mmol/mol, >8% or >64 mmol/mol, or unknown), heart failure, arterial hypertension, thyroid disease, cerebrovascular disease including transient ischemic attack, atrial fibrillation or flutter, cancer, chronic obstructive pulmonary disease, coronary artery disease, hyperlipidemia, previous coronary revascularization, previous myocardial infarction, angina pectoris, anemia, and diabetic complications (neuropathy, peripheral vascular disease, nephropathy, and retinopathy). We also included the number of hospitalizations in the year prior to cohort entry (0, 1, 2, 3, or ≥4).

In addition, prescriptions for angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, β-blockers, calcium-channel blockers, thiazide, loop or other diuretics, digoxin, statins, fibrates, clopidogrel, warfarin, acetylsalicylic acid, nonsteroidal anti-inflammatory drugs, opioid analgesics, acetaminophen, and nitrates in the year prior to cohort entry were assessed. We also assessed the overall number of nonantidiabetic drugs in the year prior to cohort entry (0, 1, 2, 3, and ≥4). The hdPS procedure was repeated for each outcome, because this method calculates a bias term that accounts for the association with a specific outcome. Patients in nonoverlapping regions of the hdPS were trimmed from the analysis.

We used descriptive statistics to summarize the characteristics of the patients in the cohort. Crude incidence rates of each outcome were calculated with 95% CIs based on the Poisson distribution and expressed as number of events per 1,000 patients per year. Moreover, we conducted five separate analyses using Cox proportional hazards regression models that provided estimates of the hazard ratio (HR) and the 95% CIs for each outcome. The models for AMI, ischemic stroke, and cardiovascular death were adjusted for hdPS deciles as well as for history of the respective outcome (or, in the case of cardiovascular death, history of AMI or ischemic stroke). The model for all-cause mortality was adjusted for hdPS deciles. The model for severe hypoglycemia was adjusted for hdPS quintiles. We also conducted a post hoc secondary analysis assessing the risks of all study outcomes separately for each of the two pancreas nonspecific, long-acting sulfonylureas (i.e., glyburide and glimepiride), given the previously reported pharmacologic differences between the two compounds regarding their effect on ischemic preconditioning and their risk of hypoglycemia (13).

We conducted four sensitivity analyses to assess the robustness of our findings. First, to assess possible exposure misclassification, we repeated the analyses using a 60-day grace period between nonoverlapping successive prescriptions. Second, to assess potential informative censoring in the as-treated approach (i.e., to account for the possibility that censoring was related to the outcome), we repeated the analyses using an intention-to-treat approach with a maximum follow-up of 1 year. In the intention-to-treat approach, patients are assumed to be continuously exposed to their initial treatment for the whole duration of follow-up, regardless of any switching or discontinuation that occurs. Third, a sensitivity analysis was conducted after excluding patients receiving tolbutamide, because this first-generation sulfonylurea was shown to be associated with increased mortality (14). Finally, the analyses for AMI, ischemic stroke, cardiovascular death, and severe hypoglycemia were repeated after excluding patients with a history of these events. All statistical analyses were conducted using SAS, version 9.4 (SAS Institute, Inc., Cary, NC) and R (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org/).

The cohort included 1,863 patients newly treated with nonspecific, long-acting sulfonylureas (917 glyburide and 946 glimepiride users) and 15,741 patients newly treated with specific, short-acting sulfonylureas (14,412 gliclazide, 775 glipizide, and 554 tolbutamide users) (Fig. 1). The mean (SD) follow-up was 1.2 (1.6) years, generating a total of 104,011 patient-years. During this follow-up, there were 245 AMI events (incidence rate: 11.8 per 1,000/year; 95% CI 10.4–13.4), 237 ischemic stroke events (incidence rate: 11.4 per 1,000/year; 95% CI 10.0–13.0), 458 cardiovascular deaths (incidence rate: 21.9 per 1,000/year; 95% CI 19.9–24.0), 1,332 deaths from any cause (incidence rate: 64.2 per 1,000/year; 95% CI 60.8–67.7), and 87 severe hypoglycemic events (incidence rate: 4.2 per 1,000/year; 95% CI 3.4–5.2). The most frequent causes of death were cancer (34%), cardiovascular diseases (34%), respiratory diseases (14%), and gastrointestinal diseases (5%).

Table 1 presents the baseline characteristics of the two treatment groups prior to trimming nonoverlapping hdPS distributions. Patients newly treated with nonspecific, long-acting sulfonylureas were similar compared with patients newly treated with specific, short-acting sulfonylureas. However, they were less likely to have a history of nephropathy or to have been hospitalized prior to cohort entry. Overall, initiators of sulfonylurea treatment were predominantly male (57%), had a mean age of 68 years, and a mean duration of nonpharmacologically treated diabetes of 1.6 years. The most common comorbidities were arterial hypertension (56%), coronary artery disease (29%), and hyperlipidemia (20%), whereas the most common comedications were acetaminophen (34%), statins (33%), and acetylsalicylic acid (31%).

Table 2 shows the results for the five outcomes. Compared with the use of specific, short-acting sulfonylureas, use of nonspecific, long-acting sulfonylureas was not associated with an increased risk of AMI (9.1 vs. 12.1 per 1,000/year; adjusted HR 0.86; 95% CI 0.55–1.34), ischemic stroke (9.1 vs. 11.7 per 1,000/year; adjusted HR 0.92; 95% CI 0.59–1.45), cardiovascular death (16.8 vs. 22.5 per 1,000/year; adjusted HR 1.01; 95% CI 0.72–1.40), or all-cause mortality (40.6 vs. 67.3 per 1,000/year; adjusted HR 0.81; 95% CI 0.66–1.003). However, use of nonspecific, long-acting sulfonylureas was associated with an increased risk of severe hypoglycemia (7.4 vs. 3.8 per 1,000/year; adjusted HR 2.83; 95% CI 1.64–4.88). In the secondary analysis assessing the risks of the study outcomes separately for each of the two pancreas-nonspecific, long-acting sulfonylureas (i.e., glyburide and glimepiride), we observed no major differences between the two compounds (Supplementary Table 1).

The results of the primary analysis remained consistent in the sensitivity analyses (summarized in Fig. 2 and presented in Supplementary Tables 2–5). Indeed, no associations were observed with AMI, ischemic stroke, cardiovascular death, and all-cause mortality, whereas an increased risk was observed for severe hypoglycemia (Supplementary Tables 2–4). For the severe hypoglycemia outcome, the intention-to-treat approach led to a dilution of the HR, resulting in a nonstatistically significant association (HR 1.51; 95% CI 0.73–3.12) (Supplementary Table 5).

To our knowledge, this is the first observational study comparing major cardiovascular adverse events and severe hypoglycemia among different sulfonylureas grouped based on their specificity to pancreatic β-cells and duration of action. In comparison with the use of specific, short-acting sulfonylureas, the use of nonspecific, long-acting sulfonylureas was not associated with an increased risk of AMI, ischemic stroke, cardiovascular death, and all-cause mortality. In contrast, these sulfonylureas were associated with an increased risk of severe hypoglycemia. Moreover, a secondary analysis showed no major differences in the risks of the study outcomes between the two pancreas nonspecific, long-acting sulfonylureas (i.e., glyburide and glimepiride), suggesting a similar risk profile. Overall, the findings of the primary analysis remained consistent in several sensitivity analyses.

Our findings on AMI, ischemic stroke, cardiovascular death, and all-cause mortality add new information to a field characterized by an abundance of studies that have generated contradicting results (13,15–17). To date, several observational studies have suggested that individual sulfonylureas, and particularly glyburide, possess an increased cardiovascular risk as compared with others (15,17,18). However, methodological limitations, including exposure misclassification, time-lag bias, and selection bias, could affect the validity of these results (2). Our findings indicate that the binding of glyburide and glimepiride to cardiac and vascular structures, as reported in several preclinical studies (5), does not necessarily translate into an increased risk of ischemic adverse events or cardiovascular death as compared with other pancreas specific sulfonylureas. Possible explanations include reduced sulfonylurea concentrations on site (i.e., in cardiomyocytes or vascular smooth cells), thus alleviating the consequences of receptor binding or the existence of competing factors in the cardiovascular system, such as the impact of certain sulfonylureas on atherosclerosis or weight gain (5,19,20).

Our findings on hypoglycemia are concordant with those of previous clinical trials and observational studies (21,22). They also support the hypothesis that pharmacologically active metabolites resulting in longer durations of action can increase the hypoglycemic risk of sulfonylureas (3). A recent CPRD-based study by van Dalem et al. (23) compared long-acting with short-acting sulfonylureas but did not show a difference in hypoglycemic risk. However, in contrast to our study, their outcome definition was based on outpatient diagnoses rather than events leading to hospitalization. As their outcome definition likely included a mix of both mild and severe cases of hypoglycemia, direct comparison between the two studies is challenging. Moreover, the inclusion of mild hypoglycemia as their outcome definition may have introduced misclassification in the timing of the outcome, resulting in a dilution of the association (24). Indeed, as mild hypoglycemia is generally self-managed by patients, it is possible that the date of the hypoglycemic event recorded is inaccurate.

Hypoglycemia has been implicated as a risk factor for cardiovascular adverse events and all-cause mortality in patients with diabetes (25). The Veterans Affairs Diabetes Trial (VADT) found severe hypoglycemia to be a predictor of both cardiovascular death and all-cause mortality (26). The Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation (ADVANCE) trial also indicated that severe hypoglycemia could contribute to cardiovascular adverse events (27). In our study, the increased hypoglycemic risk associated with pancreas-nonspecific, long-acting sulfonylureas was not accompanied by an increased cardiovascular or mortality risk, arguing against a major contribution of hypoglycemia in sulfonylurea-related cardiovascular disease or mortality. However, the much lower incidence rate of hypoglycemia in our study cohort relative to the rates of the other outcomes could have masked its true impact, thus hampering definite conclusions.

Our study has a number of strengths. First, it is one of the first studies to group sulfonylureas based on important pharmacodynamic (i.e., specificity to pancreatic β-cells) and pharmacokinetic (i.e., duration of action) properties, thus accounting, at least partly, for the high pharmacologic heterogeneity of this drug class. This classification does not consider the varying effects of the two pancreas-nonspecific sulfonylureas (i.e., glyburide and glimepiride) on ischemic preconditioning (13). However, we did consider this phenomenon in a secondary analysis, in which the risks of all study outcomes were assessed separately for glyburide and glimepiride, and found a similar risk profile. Second, the population-based design, the inclusion of patients with previous events, and the few exclusion criteria make its results highly generalizable. Third, the definition of hypoglycemia was based on solely hospitalization-associated events, which likely maximized the specificity of this outcome definition.

Our study also has some limitations. First, because of its observational nature, there is potential for residual confounding. However, the application of a new user design with an active comparator and the use of robust statistical adjustment such as the hdPS likely minimized this potential bias. Second, the result for hypoglycemia in our sensitivity analysis using an intention-to-treat approach was attenuated and no longer statistically significant. Although a dilution of the effect is a potential limitation of such analyses, the existence of informative censoring in our primary analysis cannot be ruled out. Third, the results of this study are generalizable to the use of sulfonylureas as first-line treatment; further research is needed to evaluate the safety of sulfonylureas when used as second- or third-line treatment. Fourth, because of the relatively short follow-up of the study, it is not possible to exclude long-term differences in risks in cardiovascular and mortality outcomes between the two groups.

In summary, our population-based cohort study showed no difference in the risk of major cardiovascular adverse events, cardiovascular death, and all-cause mortality between pancreas-nonspecific and pancreas-specific sulfonylureas, thereby contradicting previous studies and arguing that the clinical implications of the lack of pancreas specificity of certain sulfonylureas may have been overstated (28,29). Finally, it corroborates the previously reported increased risk of severe hypoglycemia associated with long-acting sulfonylureas as compared with short-acting compounds.

Funding. A.D. is the recipient of a Research Fellowship from the German Research Foundation (Deutsche Forschungsgemeinschaft). K.B.F. holds a Canadian Institutes of Health Research New Investigator Award. L.A. holds a Chercheur Boursier award from Fonds de recherche du Québec–Santé and is the recipient of a William Dawson Scholar award. S.S. is the recipient of the James McGill Professorship award.

Duality of Interest. This research was funded in part by grants from the Canadian Institutes of Health Research, the Canadian Foundation for Innovation, and Boehringer Ingelheim. S.S. has received research grants and has participated in advisory board meetings or as a speaker at conferences for AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Merck, and Novartis. No other potential conflicts of interest relevant to this article were reported.

The sponsors were not directly involved in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript.

Author Contributions. A.D., H.Y., O.H.Y.Y., K.B.F., L.A., and S.S. contributed to study concept and design, analyzed and interpreted the data, and critically revised the manuscript. A.D. drafted the manuscript. A.D., H.Y., and L.A. conducted the statistical analysis. S.S. acquired the data, obtained funding, and supervised the study. S.S. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.