Diabetes Care 1 Antonios Douros,1,2,3 Hui Yin,1 Pharmacologic Differences of Oriana Hoi Yun Yu,1,4 Kristian B. Filion,1,2,5 Sulfonylureas and the Risk of Laurent Azoulay,1,2,6 and Samy Suissa1,2,5 Adverse Cardiovascular and Hypoglycemic Events https://doi.org/10.2337/dc17-0595 EPIDEMIOLOGY/HEALTH SERVICES RESEARCH OBJECTIVE Sulfonylureas have been associated with an increased risk of cardiovascular adverse events and hypoglycemia, but it is unclear if these risks vary with different agents. We assessed whether the risks of acute myocardial infarction, ischemic stroke, cardiovascular death, all-cause mortality, and severe hypoglycemia differ between sulfonylureas grouped according to pancreas specificity and duration of action. RESEARCH DESIGN AND METHODS Using the U.K. Clinical Practice Research Datalink, linked with the Hospital Episodes Statistics and the Office for National Statistics databases, we conducted a cohort study among patients with type 2 diabetes initiating monotherapy with sulfonyl- ureas between 1998 and 2013. Adjusted hazard ratios (HRs) and 95% CIs were 1Centre for Clinical Epidemiology, Lady Davis estimated using Cox proportional hazards models, comparing use of pancreas- Institute, Jewish General Hospital, Montreal, fi fi Quebec, Canada nonspeci c, long-acting sulfonylureas (glyburide/glimepiride) to pancreas-speci c, 2Department of Epidemiology, Biostatistics, and short-acting sulfonylureas (gliclazide/glipizide/tolbutamide). Occupational Health, McGill University, Mon- treal, Quebec, Canada RESULTS 3Charite´ - Universitatsmedizin¨ Berlin, corporate The cohort included 17,604 sulfonylurea initiators (mean [SD] follow-up 1.2 [1.5] member of Freie Universitat¨ Berlin, Humboldt- fi Universitat¨ zu Berlin, and Berlin Institute of years). Compared with speci c, short-acting sulfonylureas (15,741 initiators), non- Health, Institute of Clinical Pharmacology and specific, long-acting sulfonylureas (1,863 initiators) were not associated with an in- Toxicology, Berlin, Germany creased risk of acute myocardial infarction (HR 0.86; CI 0.55–1.34), ischemic stroke 4Division of Endocrinology, Jewish General Hos- – – pital, Montreal, Quebec, Canada (HR 0.92; CI 0.59 1.45), cardiovascular death (HR 1.01; CI 0.72 1.40), or all-cause 5 – Division of Clinical Epidemiology, Department mortality (HR 0.81; CI 0.66 1.003), but with an increased risk of severe hypoglycemia of Medicine, McGill University, Montreal, Que- (HR 2.83; CI 1.64–4.88). bec, Canada 6Gerald Bronfman Department of Oncology, CONCLUSIONS McGill University, Montreal, Quebec, Canada The nonspecific, long-acting sulfonylureas glyburide and glimepiride do not have an Corresponding author: Samy Suissa, samy. increased risk of cardiovascular adverse events compared with the specific, short- [email protected]. acting sulfonylureas gliclazide, glipizide, and tolbutamide. However, nonspecific, Received 24 March 2017 and accepted 11 August long-acting sulfonylureas glyburide and glimepiride have an increased risk of severe 2017. hypoglycemia. This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/ suppl/doi:10.2337/dc17-0595/-/DC1. Despite the recent approval of several novel groups of antidiabetic drugs, sulfonylureas © 2017 by the American Diabetes Association. remain a cornerstone in the treatment of type 2 diabetes (1). However, concerns Readers may use this article as long as the work is properly cited, the use is educational and not remain regarding their association with cardiovascular and hypoglycemic adverse for profit, and the work is not altered. More infor- events. Although these events have been investigated in several observational studies, mation is available at http://www.diabetesjournals their findings have been contradictory, in part because of important methodological .org/content/license. Diabetes Care Publish Ahead of Print, published online September 1, 2017 2 Safety of Sulfonylureas Diabetes Care shortcomings, including exposure mis- the underlying cause of death (coded us- two successive prescriptions. Termination classification, time-lag bias, and selection ing the ICD-9 and ICD-10 classifications) of treatment was therefore defined by either bias (2). Furthermore, few studies have for all patients who died during follow-up. the absence of a new prescription by the examined the risk of cardiovascular and The linkage of the HES and ONS to the CPRD end of a 30-day grace period, a switch to a hypoglycemic adverse events with individ- is possible from 1 April 1997 onward and is different sulfonylurea group, or an add-on or ual sulfonylureas, as these differ in their limitedtoEnglishgeneralpracticesthathave switch to another antidiabetic drug, which- pharmacodynamic and pharmacokinetic consented to the linkage scheme (currently ever came first. Switches within the same properties (3). Indeed, some sulfonylureas representing 75% of all English practices) (9). group of sulfonylureas (e.g., glyburide to such as glyburide and glimepiride lack se- The study protocol was approved by the glimepiride or vice versa) were permitted. lectivity for pancreatic b-cells, allowing Independent Scientific Advisory Committee them to bind to other receptors, including of the CPRD (protocol 14_019AMn) and the Study Outcomes We considered five outcomes: hospitaliza- those on cardiomyocytes and vascular Research Ethics Board of the Jewish Gen- tion for AMI, hospitalization for ischemic smooth muscle cells. In the case of gly- eral Hospital (Montreal, Quebec, Canada). stroke, cardiovascular death, all-cause buride, such binding has been suggested mortality, and hospitalization for hypogly- to inhibit ischemic preconditioning (4–7). Study Population The cohort consisted of patients newly cemia. Hospitalization for AMI and ische- Moreover, other sulfonylureas such as treated for type 2 diabetes with sulfonyl- mic stroke were identified using the HES chlorpropamide, glyburide, and glimepir- ureas in monotherapy between 1 April and the ONS databases (AMI ICD-9 codes: ide yield pharmacologically active metab- 1998 and 31 March 2013, with follow-up 410.x; ICD-10 codes: I21.x; ischemic stroke olites, which could prolong the duration until 31 March 2014. Cohort entry was ICD-9 codes: 433.x, 434.x, and 436.x; ICD- of action and result in an increased risk of defined by the date of the first sulfonyl- 10 codes: I63.x and I64.x; in primary or hypoglycemia (3). urea prescription. We excluded all pa- secondary position). The diagnostic codes Thus, the objective of this population- tients ,40yearsofage,aswellasthose to identify AMI in HES have been shown to based study was to determine whether with ,1 year of medical history in the be highly valid (.90%) (10), whereas the the risk of acute myocardial infarction CPRD before cohort entry. We also ex- validity of stroke diagnoses in administra- (AMI), ischemic stroke, cardiovascular death, cluded patients prescribed other antidia- tive data is also high (.80%) (11). Car- all-cause mortality, and severe hypoglycemia betic drugs at any time before cohort diovascular death was identified in the is different among sulfonylureas grouped ac- entry. Because this study aimed to be rep- ONS (underlying cause of death; ICD-9 cording to their specificity to the pancreatic resentative of real-world practice, patients codes: 390.x–398.x, 401.x–405.x, 410.x– b-cells and duration of action. with a history of cardiovascular disease at 417.x, 420.x–429.x [except 427.5], 430. – – RESEARCH DESIGN AND METHODS cohort entry were not excluded. x 438.x, 440.x 447.x; and ICD-10 codes: Patients meeting the study inclusion I00.x–I77.x [except I46.9]), and all-cause Data Sources criteria were followed until the earliest mortality was identified from all three da- This study was conducted using the U.K. of the following events: treatment dis- tabases. Hospitalization for hypoglycemia Clinical Practice Datalink (CPRD) linked to continuation, add-on, or switch (defined was identified in HES (ICD-10 code: E16.2; the Hospital Episode Statistics (HES) and in detail in the section “Exposure Defini- in primary or secondary position). Office for National Statistics (ONS) data- tion”), occurrence of one of the study out- bases. The CPRD is a large primary care comes (defined in detail in the section High-Dimensional Propensity Score database containing the medical records For each patient, we used multivariable “Study Outcomes”), death from any for .14 million people enrolled in .680 logistic regression to calculate a high- cause, end of registration with the gen- general practices (8). Medical diagnoses dimensional propensity score (hdPS); eral practice, or end of the study period and procedures are recorded using the this method empirically selects covariates (31 March 2014). Read code classification, and drugs pre- based on their prevalence and potential scribed by general practitioners are coded Exposure Definition for confounding (12). The hdPS defined based on the U.K. Prescription Pricing Au- Patients were classified into two groups the probability of being exposed to non- thority dictionary. The CPRD contains in- based on the sulfonylurea they initiated: specific, long-acting sulfonylureas as formation on anthropometric variables pancreas nonspecific, long-acting sulfo- compared with specific, short-acting sul- (such as BMI) and lifestyle variables nylureas (glyburide and glimepiride), fonylureas, conditional on several prede- (such as smoking and alcohol use), and and pancreas-specific, short-acting sul- fined and 500 empirically identified its data have been previously