Sulfonylureas are ATP-sensitive potassium (KATP) channel blockers commonly used in the treatment of type 2 diabetes mellitus (T2DM). Activation of KATP channels plays a neuroprotective role in ischemia; thus, whether sulfonylureas affect the outcomes of stroke in patients with T2DM needs to be further studied. In our study, streptozotocin (STZ)-induced diabetic mice subjected to transient middle cerebral artery occlusion (MCAO) showed larger areas of brain damage and poorer behavioral outcomes. Blocking the KATP channel by tolbutamide increased neuronal injury induced by oxygen-glucose deprivation (OGD) in vitro and permanent MCAO (pMCAO) in vivo. Activating the KATP channel by diazoxide reduced the effects of both the OGD and pMCAO. Western blot analysis in STZ mouse brains indicated an early increase in protein levels of N-methyl-d-aspartate receptor 2B and postsynaptic density protein-95, followed by a decrease in phosphorylation of glycogen synthase kinase 3β. Our systematic meta-analysis indicated that patients with T2DM treated with sulfonylureas had a higher odds ratio for stroke morbidity than those who received comparator drugs. Taken together, these results suggest that sulfonylurea treatment in patients with T2DM may inhibit the neuroprotective effects of KATP channels and increase the risk of stroke.
Introduction
Sulfonylureas, as monotherapy or combined with other drugs, have been prescribed to millions of patients with diabetes for more than 60 years. Although the administration of sulfonylureas has decreased in recent years, they are still one of the most widely used oral antidiabetes drugs because of their high efficacy, at least in the short-term, and the low cost (1). In fact, sulfonylureas are still reported in many guidelines and treatment algorithms among possible options as an add-on to metformin in patients in whom monotherapy is failing. Type 2 diabetes mellitus (T2DM) is associated with a two- to threefold increased risk for cardiovascular (CV) disease (2) and is viewed as an independent risk factor for stroke (3). Therefore, prevention of CV morbidity and mortality is an important goal for diabetes treatment.
The ATP-sensitive potassium (KATP) channel is a plasma membrane protein that plays important roles in controlling cellular functions and regulating metabolism. The KATP channel is composed of four pore-forming Kir6.x and four regulatory sulfonylurea receptor SURx subunits that form a hetero-octameric complex (Kir/SUR) (4). Opening and closing of KATP channels are facilitated by ADP and ATP molecules, respectively. The KATP channel was originally found in cardiac myocytes (5) and later in pancreatic β-cells (6) and neuronal cells (7). Sulfonylureas bind to SUR subunits and block the channel (8,9), thus stimulating insulin secretion through the depolarization of pancreatic β-cells (10).
In the brain, the KATP channel is highly expressed in the hippocampus and cortex (7,11–14). Our previous studies demonstrated that focal ischemia (transient middle cerebral artery [MCA] occlusion [tMCAO] model) in the brain led to larger cortical infarcts and more severe neurological deficits in adult Kir6.2 knockout mice (14). Our recent work showed that KATP channels mediated hypoxic preconditioning–induced neuroprotection in a hypoxic-ischemic brain injury model in neonatal mice (15). Inhibition of KATP channels by tolbutamide increased infarct volumes and opening of KATP channels by diazoxide-reduced infarct volumes (14,15). Opening of the KATP channel hyperpolarizes the neurons and stabilizes the resting membrane potential during ischemia when energy failure reduces the ATP-to-ADP ratio (11,14,16). We previously reported that this hypoxia/ischemia-induced membrane hyperpolarization mediated by KATP channels is a cellular mechanism underlying neuroprotection against stroke (14,15). These findings indicate that KATP channels may be considered as a potential therapeutic target for the treatment of stroke. However, the role of KATP channels in cerebral ischemia has not been tested in a diabetic model and a permanent (p)MCAO model in mice.
CV safety is a major issue for glucose-lowering drugs. After meta-analytical data suggesting a possible increase in the risk of myocardial infarction with rosiglitazone (17), the U.S. Food and Drug Administration issued a guideline requiring postmarketing CV safety trials for all novel drugs. However, no formal study on the CV safety of older drugs was required. In particular, the CV safety of sulfonylureas has been widely debated. The use of glibenclamide may counter the effect of preconditioning through the blocking of KATP channel activity (18). The University Group Diabetes Program (UGDP) study reported that the use of tolbutamide increased CV mortality in patients with T2DM compared with other nonsulfonylurea agents (19,20). An increased mortality, despite a similar risk of major CV events, was also found in a meta-analysis of randomized trials with sulfonylureas (21).
Because sulfonylureas prevent KATP channel activation that provides neuroprotection in stroke, their use may potentially increase the incidence of stroke in patients with diabetes (21). Currently, the relationship between sulfonylurea treatment and stroke incidence in patients with diabetes is still unclear. In this study, we present our in vitro and in vivo findings from animal models as well as a systematic meta-analysis of randomized clinical trials.
Research Design and Methods
Animal Studies
Animals
The animal protocol was approved by the University of Toronto Animal Care Committee and conformed to Canadian Council on Animal Care guidelines. Adult male C57BL/6J mice (9–10 weeks old, body weight 23–26 g) were purchased from Charles River Laboratories (Sherbrooke, Quebec, Canada).
Materials
Tolbutamide, diazoxide, 2,3,5-triphenyltetrazolium chloride (TTC), 0.25% trypsin-EDTA solution, propidium iodide (PI), and streptozotocin (STZ) were purchased from Sigma-Aldrich Canada; and GlutaMax B-27 supplement, penicillin streptomycin, Neurobasal medium, and Hanks’ balanced salt solution were purchased from Gibco by Life Technologies. Anti–N-methyl-d-aspartate (NMDA) receptor 2B (NR2B) and anti-postsynaptic density protein 95 (PSD95) were from Abcam (Cambridge, MA); anti-phosphorylated glycogen synthase kinase 3β (p-GSK3β), anti-total GSK3β (t-GSK3β), and anti-GAPDH were obtained from Cell Signaling Technology (Danvers, MA).
STZ Administration and tMCAO Animal Model
For the diabetic model, adult male C57BL/6J mice (6 weeks old; body weight, 17–21 g) were fasted before injection for 4 hours. STZ (50 mg/kg) or Na-citrate solution was intraperitoneally administered to the mice once daily for 5 consecutive days. Mice were confirmed hyperglycemic (≥16 mmol/L) (22) at 1, 2, 4, and 5 weeks after injection (Glucometer, Roche Diagnostics GmbH). Body weights were measured at 1, 2, 4, and 5 weeks after injection.
For the tMCAO model (14), mice were anesthetized with 2.5% isoflurane in oxygen. A heating pad was used to maintain body temperature at 37 ± 0.2°C. A 20-mm-long monofilament suture with silicon-coated tip (cat#1620A5; Beijing SUNBIO Biotech Co., Ltd.) was inserted through the right external carotid artery into the internal carotid artery to block the origin of the MCA. After 90 min, the suture was removed to reestablish the perfusion, and animals were allowed to recover for 24 h until histological and behavioral evaluations.
pMCAO
Similar to the tMCAO procedure, the suture was inserted through the right common carotid artery into the internal carotid artery to block the origin of MCA and left in place for 24 h until the next procedure. Because the pMCAO procedure causes larger brain damage and may affect survival after the procedure, it was not performed on mice administered STZ.
Measurement of Infarct Volume
The MCAO brains were sectioned coronally after 24 h into 1-mm slices and incubated in 2% TTC in PBS at 37°C for 30 min. Infarct volumes were calculated by summing the representative areas in all sections and multiplying by the slice thickness. After correcting for edema, the volumes of infarction were calculated: Corrected infarct volume (CIV) (%) = [contralateral hemisphere volume – (ipsilateral hemisphere volume – infarct volume)]/contralateral hemisphere volume × 100 (15).
Neurobehavioral Tests
Behavioral tests were conducted 24 h after pMCAO or tMCAO using a standard 6-point scale (23). Scale of scores: 0–no neurological deficit; 1–retracts left forepaw when lifted by the tail; 2–circles to the left; 3–falls while walking; 4–does not walk spontaneously; 5–dead.
Primary Cortical Culture
Primary cortical cultures were prepared from E16 CD1 mice as previously described (24). Cells were plated on poly-d-lysine coated plates using Neurobasal culture medium at the desired density (1 × 105 per well) and maintained at 37°C in a humidified 5% CO2 atmosphere.
Oxygen and Glucose Deprivation
Cortical cultures were incubated with tolbutamide or diazoxide in oxygen and glucose deprivation (OGD) solution for 30 min, followed by incubation in an anaerobic chamber flushed with 5% CO2 and 95% N2 (v/v) at 37°C for 90 min (25). Cells were then incubated in normoxic conditions for an additional 24 h.
Immunocytochemistry and Confocal Imaging
Cell damage was determined by quantitative measurements of PI fluorescence with a multiwell plate Synergy H1 fluorescence reader (BioTek, Winooski, VT) (25). In brief, cells were incubated with PI (5 μg/mL) for 20 min. Fluorescence intensity was read under an excitation wavelength of 488 nm and an emission wavelength of 630 nm. PI enters the cell and stains DNA in the nucleus when cell membrane integrity is compromised during OGD. Greater fluorescence intensity indicates greater cell damage with OGD. The fraction of OGD-induced dead cells in each culture was calculated as: fraction dead = (Ft – Fo)/(FNMDA – Fo), where Fo is the initial fluorescence reading of the plate before OGD, Ft is the maximum fluorescence reading after OGD, and FNMDA is the PI fluorescence of sister cultures 24 h after a 60 min exposure to NMDA (1 mmol/L) at 37°C (26,27). The relative assessments of neuronal cell death were normalized by comparison with 100% cell death induced by NMDA. The result represents OGD-induced damage as a percentage of NMDA-induced cell death (28).
Images of PI staining were also taken to show damaged cells. In brief, cells were stained with PI for 20 min and then fixed with 4% paraformaldehyde in PBS for 20 min. Images were taken using a Zeiss LSM 710 Confocal microscope, and quantitative analysis was performed by counting PI-positive cells in five random fields with a ×10 objective per coverslip.
Western Blot
To determine NR2B, PSD95, p-GSK3β, and t-GSK3β protein levels in the brains of mice with STZ administration or vehicle treatment, the brain tissues were homogenized in radioimmunoprecipitation assay buffer containing a proteinase inhibitor cocktail. Proteins in equal amounts were separated by SDS-PAGE gels. Protein was transferred to nitrocellulose membranes (Pall, Pensacola, FL). The blots were incubated with appropriate horseradish peroxidase-conjugated secondary antibody accordingly (1:7,500; Jackson ImmunoResearch Laboratories) and analyzed by exposure to film (HyBlot CL).
Data Analysis
Statistical analysis was performed with SigmaPlot software (Jandel Scientific). Data are presented as mean ± SEM. Differences between experimental groups were evaluated using Student t test for two groups and one-way ANOVA, followed by the Fisher least significant difference method for multiple experimental groups. Value of P < 0.05 was considered statistically significant. All experiments were performed in a blinded manner to all treatment conditions and assessments.
Meta-analysis of Randomized Trials
Search Strategy, Inclusion Criteria, and Data Extraction
The meta-analysis was performed following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (29). We systemically conducted an extensive search in Medline, Web of Science, Library of Congress, Embase, and the Cochrane Library, using the search terms “sulfonylureas” or “glyburide” or “glibenclamide” or “tolbutamide” or “acetohexamide” or “gliclazide” or “glipizide” or “chlorpropamide” or “glimepiride,” in combination with “stroke” or “cerebral infarction” or “cerebral ischemia” (or “ischaemia”) or “cerebrovascular disease” or “cerebrovascular attack” or “brain ischemia” (or “ischaemia”). We collected all randomized controlled trials comparing sulfonylureas with placebo or other antidiabetic drugs up to 9 February 2016. Results of unpublished trials were retrieved if they were available on www.clinicaltrials.gov, www.clinicalstudyresults.org, or www.controlled-trials.com.
Studies on the relationship between oral sulfonylureas and the risk of stroke were selected, and only studies reporting raw data on morbidity during sulfonylureas treatment were included (Table 1). Other entry criteria included 1) publications written in English; 2) studies performed on patients with T2DM; 3) randomized trials with a control group; and 4) follow-up duration of at least 24 weeks.
First author, year (ref.) . | Study duration (week) . | SU . | N . | Comparator . | N . | Stroke (SU/C, n) . | Race (SU/C, % of white) . | Sex (SU/C, % of male) . | Mean . | Adjudication (Yes/No) . | ||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Age (SU/C, year) . | Diabetes duration (SU/C, year) . | HbA1c (SU/C, %) . | BMI (SU/C, kg/m2) . | |||||||||||
Baseline . | End point . | |||||||||||||
Arechavaleta 2011 (52) | 30 | Glimepiride + metformin | 519 | Sitagliptin + metformin | 516 | 2/0 | 57.4/57.6 | 53.8/55.0 | 56.2/56.3 | 6.7/6.8 | 7.5/7.5 | 7.1/7.0 | 30.2/29.7 | No |
Bakris 2006 (53) | 32 | Glyburide + metformin | 180 | Rosiglitazone + metformin | 194 | 1/0 | 76.0/78.0 | 69.0/63.0 | 58.8/60.0 | 7.6/8.0 | 8.3/8.3 | 74/7.8 | 31.8/31.6 | No |
Charbonnel 2005 (54) | 52 | Gliclazide + metformin | 313 | Pioglitazone + metformin | 317 | 3/2 | NR/NR | 49.2/50.8 | 57.0/56.0 | 5.5/5.8 | 8.5/8.7 | 7.9/7.8 | 32.6/32.6 | No |
Ferrannini 2009 (55) | 52 | Glimepiride | 1,393 | Vildagliptin | 1,396 | 7/0 | 85.2/86.3 | 54.1/52.8 | 57.5/57.5 | 5.8/5.7 | 7.3/7.3 | 6.6/6.8 | 31.7/31.8 | Yes |
Gallwitz 2012 (56) | 104 | Glimepiride | 775 | Linagliptin | 776 | 11/3 | 85.0/85.0 | 61.0/60.0 | 59.8/59.8 | NR/NR | 7.7/7.7 | 7.3/7.5 | 30.3/30.2 | Yes |
Gerstein 2010 (57) | 78 | Glipizide | 339 | Rosiglitazone | 333 | 1/5 | NR/NR | 65.8/70.0 | 60.2/61.8 | 4.6/5.0 | 7.2/7.1 | 7.0/6.8 | 29.8/29.3 | Yes |
Göke 2010 (58) | 52 | Glipizide + metformin | 430 | Saxagliptin + metformin | 428 | 1/0 | 84.2/82.2 | 54.0/49.5 | 57.6/57.5 | 5.4/5.5 | 7.7/7.7 | 6.7/6.7 | 31.3/31.5 | No |
Home 2009 (59) | 286 | SU + metformin | 1,105 | Rosiglitazone + metformin | 1,117 | 32/25 | 98.4/98.9 | 52.9/53.8 | 57.2/57.0 | 6.3/6.1 | 7.8/7.8 | 7.7/7.5 | 32.7/32.8 | Yes |
Hong 2013 (60) | 156 | Glipizide | 148 | Metformin | 156 | 15/10 | NR/NR | 77.0/78.2 | 63.8/62.8 | 5.6/5.6 | 7.6/7.6 | 7.1/7.0 | 25.1/25.2 | Yes |
Kahn 2006 (61) | 260 | Glyburide | 1,441 | Rosiglitazone | 1,456 | 17/16 | 89.0/87.2 | 58.0/55.7 | 56.4/56.3 | NR/NR | 7.4/7.4 | 7.6/7.1 | 32.2/32.2 | Yes |
Glyburide | 1,441 | Metformin | 1,454 | 17/19 | 89.0/89.1 | 58.0/59.4 | 56.4/57.9 | NR/NR | 7.4/7.4 | 7.6/7.3 | 32.2/32.1 | |||
Matthews 2005 (62) | 52 | Gliclazide | 313 | Pioglitazone | 317 | 5/4 | 100.0/99.4 | 49.2/50.8 | 57.0/56.0 | 5.5/5.8 | 8.5/8.7 | 7.3/7.2 | 32.6/32.6 | No |
Mazzone 2006 (63) | 72 | Glimepiride | 228 | Pioglitazone | 230 | 0/1 | 65.4/59.6 | 62.7/63.5 | 59.9/59.3 | 7.5/8.0 | 7.4/7.4 | 7.2/7.2 | 31.9/32.0 | Yes |
Nissen 2008 (64) | 78 | Glimepiride | 273 | Pioglitazone | 270 | 1/0 | 80.6/83.3 | 65.9/68.9 | 59.7/60.0 | 5.9/5.8 | 7.4/7.4 | 7.0/6.9 | 32.0/32.1 | Yes |
Riddle 1998 (65) | 78 | Glimepiride + insulin | 72 | Placebo + insulin | 73 | 1/0 | 57.0/58.0 | 62.5/54.8 | 58.0/58.0 | 7.0/7.0 | 9.7/9.9 | 7.6/7.7 | 32.2/33.7 | Yes |
Sullivan 2011 (66) | 260 | SU | 1,632 | Metformin | 1,746 | 59/45 | 91.0/94.0 | NR/NR | 63.6/61.1 | 5.0/3.0 | 6.7/6.7 | NR/NR | 28.0/31.0 | No |
SU | 1,632 | Diet | 2,627 | 59/63 | 91.0/95.0 | NR/NR | 63.6/62.2 | 5.0/2.0 | 6.7/6.0 | NR/NR | 28.0/29.0 | No | ||
Tolman 2009 (67) | 156 | Glibenclamide | 1,046 | Pioglitazone | 1,051 | 9/10 | 62.1/59.8 | 55.5/57.2 | 55.0/54.0 | 5.6/5.9 | 9.5/9.5 | NR/NR | 32.5/32.5 | No |
UKPDS 1998 (68) | 577 | Chlorpropamide | 619 | Insulin | 911 | 33/42 | 79.0/82.0 | 58.0/72.0 | 54.0/54.0 | NR/NR | 6.3/6.1 | 7.8/7.8 | 27.0/27.0 | Yes |
Chlorpropamide | 619 | Diet | 896 | 33/47 | 79.0/83.0 | 58.0/61.9 | 54.0/54.0 | NR/NR | 6.3/6.2 | 7.8/8.6 | 27.0/27.5 | |||
Glibenclamide | 615 | Insulin | 911 | 45/42 | 84.0/82.0 | 61.9/72.0 | 54.0/54.0 | NR/NR | 6.3/6.1 | 8.4/7.8 | 27.4/27.0 | |||
Glibenclamide | 615 | Diet | 896 | 45/47 | 84.0/83.0 | 61.9/61.9 | 54.0/54.0 | NR/NR | 6.3/6.2 | 8.4/8.6 | 27.4/27.5 |
First author, year (ref.) . | Study duration (week) . | SU . | N . | Comparator . | N . | Stroke (SU/C, n) . | Race (SU/C, % of white) . | Sex (SU/C, % of male) . | Mean . | Adjudication (Yes/No) . | ||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Age (SU/C, year) . | Diabetes duration (SU/C, year) . | HbA1c (SU/C, %) . | BMI (SU/C, kg/m2) . | |||||||||||
Baseline . | End point . | |||||||||||||
Arechavaleta 2011 (52) | 30 | Glimepiride + metformin | 519 | Sitagliptin + metformin | 516 | 2/0 | 57.4/57.6 | 53.8/55.0 | 56.2/56.3 | 6.7/6.8 | 7.5/7.5 | 7.1/7.0 | 30.2/29.7 | No |
Bakris 2006 (53) | 32 | Glyburide + metformin | 180 | Rosiglitazone + metformin | 194 | 1/0 | 76.0/78.0 | 69.0/63.0 | 58.8/60.0 | 7.6/8.0 | 8.3/8.3 | 74/7.8 | 31.8/31.6 | No |
Charbonnel 2005 (54) | 52 | Gliclazide + metformin | 313 | Pioglitazone + metformin | 317 | 3/2 | NR/NR | 49.2/50.8 | 57.0/56.0 | 5.5/5.8 | 8.5/8.7 | 7.9/7.8 | 32.6/32.6 | No |
Ferrannini 2009 (55) | 52 | Glimepiride | 1,393 | Vildagliptin | 1,396 | 7/0 | 85.2/86.3 | 54.1/52.8 | 57.5/57.5 | 5.8/5.7 | 7.3/7.3 | 6.6/6.8 | 31.7/31.8 | Yes |
Gallwitz 2012 (56) | 104 | Glimepiride | 775 | Linagliptin | 776 | 11/3 | 85.0/85.0 | 61.0/60.0 | 59.8/59.8 | NR/NR | 7.7/7.7 | 7.3/7.5 | 30.3/30.2 | Yes |
Gerstein 2010 (57) | 78 | Glipizide | 339 | Rosiglitazone | 333 | 1/5 | NR/NR | 65.8/70.0 | 60.2/61.8 | 4.6/5.0 | 7.2/7.1 | 7.0/6.8 | 29.8/29.3 | Yes |
Göke 2010 (58) | 52 | Glipizide + metformin | 430 | Saxagliptin + metformin | 428 | 1/0 | 84.2/82.2 | 54.0/49.5 | 57.6/57.5 | 5.4/5.5 | 7.7/7.7 | 6.7/6.7 | 31.3/31.5 | No |
Home 2009 (59) | 286 | SU + metformin | 1,105 | Rosiglitazone + metformin | 1,117 | 32/25 | 98.4/98.9 | 52.9/53.8 | 57.2/57.0 | 6.3/6.1 | 7.8/7.8 | 7.7/7.5 | 32.7/32.8 | Yes |
Hong 2013 (60) | 156 | Glipizide | 148 | Metformin | 156 | 15/10 | NR/NR | 77.0/78.2 | 63.8/62.8 | 5.6/5.6 | 7.6/7.6 | 7.1/7.0 | 25.1/25.2 | Yes |
Kahn 2006 (61) | 260 | Glyburide | 1,441 | Rosiglitazone | 1,456 | 17/16 | 89.0/87.2 | 58.0/55.7 | 56.4/56.3 | NR/NR | 7.4/7.4 | 7.6/7.1 | 32.2/32.2 | Yes |
Glyburide | 1,441 | Metformin | 1,454 | 17/19 | 89.0/89.1 | 58.0/59.4 | 56.4/57.9 | NR/NR | 7.4/7.4 | 7.6/7.3 | 32.2/32.1 | |||
Matthews 2005 (62) | 52 | Gliclazide | 313 | Pioglitazone | 317 | 5/4 | 100.0/99.4 | 49.2/50.8 | 57.0/56.0 | 5.5/5.8 | 8.5/8.7 | 7.3/7.2 | 32.6/32.6 | No |
Mazzone 2006 (63) | 72 | Glimepiride | 228 | Pioglitazone | 230 | 0/1 | 65.4/59.6 | 62.7/63.5 | 59.9/59.3 | 7.5/8.0 | 7.4/7.4 | 7.2/7.2 | 31.9/32.0 | Yes |
Nissen 2008 (64) | 78 | Glimepiride | 273 | Pioglitazone | 270 | 1/0 | 80.6/83.3 | 65.9/68.9 | 59.7/60.0 | 5.9/5.8 | 7.4/7.4 | 7.0/6.9 | 32.0/32.1 | Yes |
Riddle 1998 (65) | 78 | Glimepiride + insulin | 72 | Placebo + insulin | 73 | 1/0 | 57.0/58.0 | 62.5/54.8 | 58.0/58.0 | 7.0/7.0 | 9.7/9.9 | 7.6/7.7 | 32.2/33.7 | Yes |
Sullivan 2011 (66) | 260 | SU | 1,632 | Metformin | 1,746 | 59/45 | 91.0/94.0 | NR/NR | 63.6/61.1 | 5.0/3.0 | 6.7/6.7 | NR/NR | 28.0/31.0 | No |
SU | 1,632 | Diet | 2,627 | 59/63 | 91.0/95.0 | NR/NR | 63.6/62.2 | 5.0/2.0 | 6.7/6.0 | NR/NR | 28.0/29.0 | No | ||
Tolman 2009 (67) | 156 | Glibenclamide | 1,046 | Pioglitazone | 1,051 | 9/10 | 62.1/59.8 | 55.5/57.2 | 55.0/54.0 | 5.6/5.9 | 9.5/9.5 | NR/NR | 32.5/32.5 | No |
UKPDS 1998 (68) | 577 | Chlorpropamide | 619 | Insulin | 911 | 33/42 | 79.0/82.0 | 58.0/72.0 | 54.0/54.0 | NR/NR | 6.3/6.1 | 7.8/7.8 | 27.0/27.0 | Yes |
Chlorpropamide | 619 | Diet | 896 | 33/47 | 79.0/83.0 | 58.0/61.9 | 54.0/54.0 | NR/NR | 6.3/6.2 | 7.8/8.6 | 27.0/27.5 | |||
Glibenclamide | 615 | Insulin | 911 | 45/42 | 84.0/82.0 | 61.9/72.0 | 54.0/54.0 | NR/NR | 6.3/6.1 | 8.4/7.8 | 27.4/27.0 | |||
Glibenclamide | 615 | Diet | 896 | 45/47 | 84.0/83.0 | 61.9/61.9 | 54.0/54.0 | NR/NR | 6.3/6.2 | 8.4/8.6 | 27.4/27.5 |
Relevant data included the first author’s name, year of publication, follow-up duration, characteristics of the study population (race, sex, mean age, mean duration of diabetes, mean BMI, and initial mean HbA1c). The number of events for the morbidity of stroke was also abstracted. If similar studies were reported in more than one publication, we extracted data from the most recent article. Studies were excluded if the search terms used in the search strategies were not contained in the abstract (filtered using RefWorks). All of the information was summarized in a standardized data extraction table. C, comparator; NR, not reported; SU, sulfonylurea.
The quality of the extracted studies was assessed by the two experimenters (Y.L. and H.F.) independently using the criteria for judging risk of bias as suggested by the Cochrane Handbook of Reviews of Assessing Risk of Bias. Disagreements were resolved by discussion between the two experimenters or the third person (R.L.) by referencing the original reports.
Outcome Measures and Data Analysis
The outcome of the meta-analysis was the effect of sulfonylureas, compared with other antidiabetes drugs, on the incidence of stroke in T2DM; the odds ratio (OR) and 95% CI were used as the operational measures. All ORs were pooled using random-effects models.
A sensitivity analysis was performed including only trials in which stroke was part of the primary end point. Further sensitivity analyses were performed: 1) including only trials in which sulfonylureas was used as monotherapy; 2) excluding trials in which the quality estimation was “unclear”; 3) including only trials in which events of stroke were formally adjudicated, irrespective of the fact that they were part of the principal end points.
Funnel plots (30) and the Harbord modified test (31) were used to estimate possible publication bias. Funnel plots provide a graphic representation of treatment effect, plotted versus trial sample size; an asymmetry in favor of a more beneficial effect of the experimental treatment in smaller trials suggests a publication bias, based on the assumption that larger studies have a greater chance of being published even when the results are unfavorable. The Harbord test provides a statistical support to data reported in the funnel plot, based on the same principle. Subgroup analyses were performed for different comparators (diet, metformin, dipeptidyl peptidase 4 inhibitor [plus metformin], thiazolidinediones [plus metformin]), and insulin [plus placebo]), follow-up duration (<72 or ≥72 weeks), mean age (<60 or ≥60 years), percentage of men (<60 or ≥60%), mean diabetes duration (<5, ≥5 years, or not reported), mean baseline HbA1c (<7.5 or ≥7.5%), mean baseline BMI (<30 or ≥30 kg/m2), and generation of sulfonylureas (first, second, third, or not reported). All of those analyses were performed using Stata 12.0 and RevMan 5.2 software.
Results
Effect of STZ-Induced Diabetes in Stroke
The group with STZ-induced diabetes showed larger infarct volumes compared with the vehicle-treated control group (Fig. 1). Infarct volume in the STZ group (50.39 ± 7.66%, n = 8) was significantly larger than that of the vehicle-treated group (32.44 ± 2.93%, n = 8; P = 0.046) (Fig. 1B). The neurological deficit score in STZ mice (4.13 ± 0.30, n = 8) was significantly higher than that of the vehicle-treated group (3.00 ± 0.42, n = 8; P = 0.047) (Fig. 1C). These results indicate that stroke is more severe in diabetic mice than in vehicle-treated mice. (STZ-induced diabetes was confirmed by lower body weights and higher blood glucose levels shown in Fig. 1D and E.)
Diabetes Upregulated the Protein Levels of NR2B and PSD95 but Decreased p-GSk3β in Diabetes
To explore the role of stroke-related proteins in the mouse brain of STZ-induced diabetes, we compared the protein levels of p-GSK3β, NR2B, and PSD95 in STZ and vehicle groups by Western blot. We found that levels of p-GSK3β, rather than the t-GSK3β, were significantly decreased in the STZ group (P < 0.05) (Fig. 2A and B). We also found that protein levels of NR2B and PSD95, which mediate brain damage in stroke, were upregulated in mice with STZ-induced diabetes (P < 0.05) (Fig. 2A–D). These results were consistent with our findings that stroke is more severe in the group with STZ-induced diabetes. These findings are likely to underlie a potential mechanism for higher incidence of stroke in diabetes.
KATP Channels Protect Against OGD-Induced Primary Cortical Neuronal Injury In Vitro
At 24 h after cortical cultures were exposed to 90 min of OGD, cells treated with 500 μmol/L tolbutamide showed significantly higher neuronal cell death (61.93 ± 3.15%, n = 5) compared with vehicle-treated cells (39.78 ± 1.74%, n = 5; P < 0.001) (Fig. 3A). Conversely, neurons treated with 750 μmol/L diazoxide showed less cell death (21.78 ± 3.38%, n = 5) than vehicle-treated neurons (38.07 ± 4.68%, n = 5; P = 0.004) (Fig. 4A).
PI fluorescence images also indicated that PI-positive cell counts increased significantly in the tolbutamide group (313.8 ± 34.77, n = 5; P < 0.05) (Fig. 3B and C), but decreased significantly in the diazoxide group (79.92 ± 12.71, n = 5; P < 0.05) (Fig. 4B and C) compared with vehicle-control groups (175.52 ± 24.42 and 162.44 ± 14.67, respectively, n = 5 per group). These results indicate that activation of the KATP channel protects against OGD-induced neuronal cell injury in vitro.
KATP Channels Provide Neuroprotection Against Stroke In Vivo
To investigate the effect of KATP channels in neuroprotection, the KATP channel blocker, tolbutamide (100 mg/kg, i.p.), KATP channel opener, diazoxide (20 mg/kg, i.p.), or vehicle (DMSO) was administrated 20 min before pMCAO (in mice not administered STZ). The infarct volume was increased significantly in the tolbutamide group (52.99 ± 4.46%, n = 5; P < 0.05) (Fig. 5A and B) but was decreased significantly in the diazoxide group (25.01 ± 2.39%, n = 5; P < 0.05) (Fig. 6A and B), compared with the vehicle-control groups (38.12 ± 3.85% and 42.43 ± 3.78%, respectively, n = 4 per group). These results indicate that activation of neuronal KATP channels provides neuroprotection to stroke in vivo.
Effects of KATP Channel Modulators in Behavioral Outcomes
Meta-analysis Showed That Use of Sulfonylureas Poses a Greater Risk of Stroke for Patients With T2DM
Figure 7A shows the study selection process and the criteria for exclusion. We selected 17 studies (Table 1) from the 1,954 articles that were retrieved. The total number of patients was 27,705, including 11,441 receiving sulfonylurea treatment (11 trials with monotherapy and 6 with combination therapy) and 16,264 receiving comparator drugs or placebo. The baseline characteristics and the number of stroke events recorded are reported in Table 1. The length of follow-up ranged from 30 to 577 weeks, mean age ranged from 54.0 to 63.8 years, and diabetes duration ranged from 2.0 to 8.0 years. All of the studies provided the number of dropouts and the reasons as a total or the number of individuals per reason. No major asymmetry appeared in the funnel plot (Fig. 7B1), and the Harbord modified test showed that there was no publication bias (P = 0.31) (Fig. 7B2). Among the 17 trials, all accepted blinded method (UK Prospective Diabetes Study [UKPDS] blinded outcome assessment method) and 11 accepted double-blinded (blinding of participants and personnel as well as outcome assessment). We used the Cochrane system to estimate the quality of trials, which is shown in Fig. 7C. The evaluation contained three levels: high risk, unclear, and low risk. Among the 17 studies, 41.18% were in the low-risk level, 58.82% were in the unclear level, and no studies were high risk. The total quality of the included articles was high.
Data were combined using random-effects models. Figure 7D shows the OR for stroke morbidity for the 17 included trials. Patients who received sulfonylurea treatment had a higher OR for stroke morbidity of 1.39 (95% CI 1.16–1.65) than those who received comparator drugs. The I2 statistic for heterogeneity between trials was 0.0% (P = 0.66), suggesting that there was no obvious heterogeneity. In a sensitivity analysis, we included only trials in which stroke events were formally adjudicated. This did not reduce the between-study heterogeneity (I2 = 24.5%) and did not change the OR estimation (1.35, 95% CI 1.08–1.69) (Fig. 7E1) for stroke morbidity of those patients using sulfonylurea treatment. In addition, we assessed monotherapy only. This did not reduce the between-study heterogeneity (I2 = 14.3%) and did not change the OR estimation (1.37, 95% CI 1.14–1.66) for stroke morbidity of those patients using sulfonylurea treatment (Fig. 7E2). In another sensitivity analysis, after excluding the studies in which quality assessments were “unclear,” the heterogeneity and OR estimation (1.41, 95% CI 1.09–1.81) did not vary to a relevant degree (Fig. 7E3).
Subgroup analysis was performed for further investigation of the effect of sulfonylureas on stroke incidence in patients with T2DM with different characteristics. The OR did not differ significantly by sex, age, duration of diabetes, BMI, or the HbA1c level (Fig. 7F). In direct comparisons, the increase of risk with sulfonylureas reached statistical significance versus dipeptidyl peptidase 4 inhibitors.
Discussion
With the increasing diabetes epidemic around the world, a thorough knowledge of the effects of individual glucose-lowering medications on the incidence of CV diseases is critical. The main focus is on CV risk assessment rather than extension of glucose control before new drug approval (32). Some studies have explored the CV risk of treatment with sulfonylureas in T2DM (33,34), but very few have assessed the risk of stroke. Our results suggest that sulfonylureas may increase the risk of stroke in patients with T2DM.
Diabetes may increase risk of stroke. In our in vivo study, STZ-induced diabetic mice with hyperglycemia and weight loss showed exacerbated brain damage in the stroke model, suggesting that diabetes could have a negative effect on the outcome of stroke. This result is in line with findings from epidemiological studies reporting that diabetes is associated with a greater lethality of stroke (35). An increased morbidity and mortality for stroke in patients with T2DM could be a result of the effect of hyperglycemia (21,36–38) or of the treatment of diabetes (39), or both. In particular, sulfonylureas could enhance anatomical and functional damage caused by cerebral ischemia, with a potential effect on stroke incidence (39). Experimental data also support the notion that inhibition of the KATP channel increases ischemic brain damage and that activation of the KATP channel confers neuroprotection in cerebral ischemia (14,15).
Excessive calcium entry into neurons through NMDA receptors is thought to be one of the triggers of neuronal injury, where the NR2A subunit mediates cell survival signaling and the NR2B subunit links cell death signaling (40). PSD95 is a postsynaptic protein that interacts with NR2B. Disruption of the NR2B/PSD95 complex inhibits NMDA receptor–mediated neuronal death after stroke (41). Thus, increased levels of NR2B or PSD95 may predict a higher risk of stroke and poorer neurological outcomes. The current study mainly focused on protein expression of these two molecules, which may provide an insight into the reason patients with diabetes show a greater risk of stroke. Our results confirm our hypothesis that the NR2B and PSD95 proteins were both upregulated in the brain after STZ-induced diabetes.
GSK3β is also highly expressed in the brain and regulates a variety of neuronal functions, including regulation of glucose metabolism and neuronal apoptosis. Our research confirms that p-GSK3β was decreased in diabetes. Because GSK3β is negatively regulated by phosphorylation at Ser9 (42), inhibition of GSK3β was neuroprotective and ameliorated stroke-induced neurological impairments. Together with the increased NR2B and PSD95 expressions in the diabetic mouse brain, targeting NR2B/PSD95 and GSK3β may be a potential strategy to reduce the risk of stroke and improve stroke outcomes.
Previous studies showed that KATP channels are expressed in CA1, CA3, and cortical neurons (14,43), interneurons, and granule cells (7). The neuronal KATP channels are important in neuroprotection against ischemic injury, in focal and global ischemia (14,44,45), and in neonatal hypoxic-ischemic brain injury (15). Neuronal KATP channels are activated during hypoxia and ischemia when there is a dramatic decrease in the ATP-to-ADP ratio in the neurons. Opening of the KATP channels hyperpolarizes the cell membrane and suppresses neuronal excitability, providing cytoprotection (11,14–16).
In this study, we further confirm the neuroprotective role of KATP channels in vivo and in vitro by using KATP blocker and opener. Brain damage in neuronal injury induced by pMCAO and OGD is increased by blocking the KATP channel by tolbutamide and decreased by opening the KATP channel by diazoxide. KATP channels in neurons and pancreatic β-cells share the same isoform, Kir6.2/SUR1 (10). It is important to note that the STZ-induced diabetes model in the animals may not fully resemble the T2DM model; however, the STZ-induced diabetes model shares the essential common feature, hyperglycemia, of the T2DM model and is a commonly used animal model. Future testing of the brain damage in response to sulfonylureas treatment in mouse models of diabetes driven by diet or genetic defect will be desirable.
It is possible that sulfonylureas block the neuronal KATP channels and thus increase the likelihood of stroke in patients with diabetes. To test our hypothesis, we further performed a meta-analysis using clinical data to evaluate the effect of sulfonylureas on the risk of stroke in diabetes.
Our systematic meta-analysis comparing the effect of sulfonylurea and nonsulfonylurea treatments on the risk of stroke in patients with T2DM found that patients who used sulfonylureas had a higher risk of stroke. Notably, this result was obtained by summarizing available randomized trials, where there are no risk-confounding factors. In addition, the increased risk was confirmed by several sensitivity analyses with different criteria for trial inclusion. This strengthens the result, which cannot be attributed to the choice of any particular set of inclusion or exclusion criteria.
Although 20% of strokes may be associated with diabetes (46), the mechanisms linking stroke and diabetes are not clearly understood. Epidemiological studies have showed that diabetes is a critical risk factor for ischemic stroke and that stroke is the second leading cause of death and disability in humans (47–49). Moreover, ischemic injury as a complication of diabetes leads to increased neuronal damage and poor functional recovery (50,51); therefore, exploring the mechanisms of ischemic brain injury and developing better neuroprotective measures in the population with diabetes will be a major focus of medical research.
Taken together, this study in our mouse stroke model showed, first, that STZ administration and blocking KATP channels increased brain damage. It is very possible that sulfonylureas block the neuronal KATP channels in the brain and thus increase the likelihood of stroke in patients with diabetes. Second, we further tested our hypothesis by performing a meta-analysis using clinical data to evaluate the effect of sulfonylureas on the risk of stroke in diabetes. Our meta-analysis also showed that patients with diabetes receiving sulfonylurea treatment have a higher relative risk for stroke morbidity than those receiving comparator drugs.
In summary, our study shows:
STZ-induced diabetes increased brain damage after stroke. STZ-administered mouse brains also showed increased expression of NR2B/PSD95 proteins, indicating potentially further increase stroke risk in diabetes.
The KATP channel blocker tolbutamide increased brain and neuronal injury in vivo and in vitro. The KATP channel opener diazoxide reduced brain and neuronal injury in vivo and in vitro.
The meta-analysis indicates patients with T2DM receiving sulfonylurea treatment have a higher relative risk for stroke morbidity than those receiving comparator drugs.
Our findings suggest that the KATP channel is a potential target for neuroprotection against stroke and that antidiabetic therapy using sulfonylureas should be considered carefully or even avoided in the long-term management of T2DM.
Article Information
Funding. This research was supported by China Scholarship Council Fellowships for R.L. and B.X., a Canadian Institutes of Health Research (CIHR) Studentship for E.T. (CIHR-CGS-M), Ontario Graduate Studentships for E.T., N.D., C.L.F.S., and A.B., and operating grants to Z.-P.F. from the National Sciences and Engineering Research Council of Canada (NSERC-249962-09) and to H.-S.S. from the Heart and Stroke Foundation of Canada (G-13-0003069) and Canadian Institutes of Health Research China–Canada Joint Health Research Initiative (CIHR, FRN #132571).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. R.L., H.W., and B.X. performed experiments. R.L., Y.L., and H.F. performed the meta-analysis. H.W., B.X., W.C., E.T., N.D., C.L.F.S., and A.B. performed experimental data analysis. All authors contributed to manuscript preparation, discussed the results, analyzed data, and commented on the manuscript. Z.-P.F. and H.-S.S. developed the concepts and designed the study. Z.-P.F. and H.-S.S. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
See accompanying article, p. 2479.