J. Neurochem. (2010) 115, 845–853.
Thiazolidinediones (TZDs) may prevent or attenuate CNS injury arising from an ischemic event. We performed meta-analysis of experimental studies in which a TZD (either rosiglitazone or pioglitazone) was administered in a rodent model of focal or global cerebral ischemia. Infarct volume was the primary endpoint for analysis of drug efficacy, and neurological outcome was also assessed. We identified 31 studies through the use of PubMed and Embase, 22 of which met our pre-specified inclusion criteria and were analyzed with the Cochrane Review Manager software. Treatment with TZDs decreased infarct volume and improved neurological outcome regardless of study quality, dose timing, or ischemia model (transient or permanent). Rosiglitazone and pioglitazone were similarly effective in reducing infarct volume and protecting neurologic function. Importantly, the collective data suggest that pre-treatment with a TZD is not required for neuroprotection, although additional studies are clearly needed to define the breadth of the therapeutic window. The data warrant further studies into the potential acute use of TZDs for ischemic stroke therapy in the general population.
middle carotid artery occlusion
peroxisome proliferator-activated receptor gamma
Thiazolidinediones (TZDs), including rosiglitazone and pioglitazone, are approved by the Food and Drug Administration as insulin-sensitizing agents for the treatment of type 2 diabetes (Yki-Järvinen 2004; Barnett 2009). These agents were developed as agonists of peroxisome proliferator-activated receptor gamma (PPARγ), a transcription factor involved in regulating cell differentiation and metabolism (Michalik et al. 2006). Patients with diabetes are at high risk of macrovascular events and epidemiological data indicate that pioglitazone reduces the number of macrovascular events including stroke in type 2 diabetic patients (Dormandy et al., 2005). In contrast, similar data on chronic administration of rosiglitazone suggest a modest increase in cardiovascular risk (Nissen and Wolski, 2007; Graham et al. 2010) sufficient to prompt additional safety warnings that are the topic of ongoing controversy (http://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationforPatientsandProviders/ucm143349.htm). TZDs have also been observed to inhibit inflammation resulting from ischemia and decrease various inflammatory markers independently of their effects on glycemic control (Sundararajan et al. 2005; Luo et al. 2006). In vitro studies have shown that neuronal PPARγ can contribute to neuroprotection and that TZDs can prevent neuronal apoptosis induced by lipopolysaccharide (Luna-Medina et al. 2005). In response to oxidative injury, pioglitazone decreases cyclooxygenase-2 expression and prevents neuronal death (Zhao 2006). PPARγ ligands reduce the secretion of neurotoxic molecules (i.e. nitric oxide, tumor necrosis factor-α, interleukin-6) that follows macrophage and microglia activation resulting from ischemia (Lee 2009; Luna-Medina 2005). The in vitro neuroprotective effects of TZDs are linked to increased PPARγ DNA binding and PPARγ antagonists reverse these effects (Zhao 2006).
Pioglitazone and rosiglitazone are currently being investigated in ongoing as well as closed clinical trials for neuroprotection in patients with type 2 diabetes or metabolic syndrome. In these studies, stroke and cognition are among the endpoint measures (see http://clinicaltrials.gov: clinical trial numbers NCT00095654, DREAM Trial Investigators et al. 2008; NCT00636766, Kadoglou et al. 2010; NCT00212290, NCT00736996; NCT00879970; NCT00242593). In addition, there is an ongoing study of the neuroprotective efficacy of pioglitazone in non-diabetic patients in hemorrhagic stroke (NCT00827892). There is an increased incidence of Alzheimer’s Disease in individuals with type 2 diabetes (Craft 2006). This observation has led to pilot studies for the use of pioglitazone in patients with type 2 diabetes and Alzheimer’s Disease or mild cognitive impairment (Hanyu et al. 2009). Additional trials are investigating pioglitazone administration in normoglycemic patients with mild cognitive impairment (NCT00242593) and in depressive disorders in individuals with or without diabetes (McIntyre et al. 2007; NCT01109030; NCT00671515, NCT00835120, Kemp et al. 2009; NCT00242619, Rasgon et al. 2010). The anti-inflammatory properties of pioglitazone prompted a pilot study that has reported promising results for its use in multiple sclerosis patients (Shukla et al. 2010; NCT00242177). Further, pioglitazone is currently being tested in a small cohort of Friedreich’s Ataxia patients (NCT00811681). However, a trial in amyotrophic lateral sclerosis (NCT00690118) was halted early as a result of a lack of efficacy at interim analysis.
Thus, the breadth of indications in which TZDs are being tested in humans has broadened beyond insulin resistant, hyperlipidemic conditions. The purpose of this review is to evaluate the pre-clinical data on the use of rosiglitazone and pioglitazone in stroke models to collectively assess and compare their ability to reduce infarct volume and improve neurological outcome in normoglycemic animals. In particular, we analyze the available pre-clinical data on the ability of TZDs to provide protection without pre-treatment, to more aptly address the potential for their acute use in stroke therapy.
To identify pre-clinical studies investigating the use of TZDs in focal or global cerebral ischemia, literature searches were conducted in PubMed and Embase for all published articles up until 1 July 2010. Keywords included combinations of TZD, rosiglitazone or pioglitazone with stroke, ischemic brain injury, cerebral ischemia, or middle carotid artery occlusion (MCAO). Reference lists from the resulting publications and reviews were used to identify further relevant publications. To prevent bias, inclusion criteria were pre-specified as the following: (i) experimental ischemic stroke was induced in rodents by transient stroke (transient MCAO or embolic stroke), permanent MCAO, or transient global ischemia (bilateral CAO); (ii) rosiglitazone or pioglitazone was administered; (iii) there was no administration of other agents with potentially neuroprotective effects; (iv) infarct volume was assessed; and (v) a control group was included in the study design. Unlike infarct volume, assessment of neurological outcome was not required for inclusion in this study, but was evaluated as an outcome measure in studies also reporting infarct volume. We chose to only evaluate studies of rosiglitazone and pioglitazone, as these are currently the only TZDs on the market approved by the Food and Drug Administration.
Data on infarct size and neurological outcomes from vehicle- and TZD-treated animals were extracted from all included studies. The effects of rosiglitazone and pioglitazone treatment on infarct volume and neurological deficits were compared with control groups of animals treated with vehicle prior to, during, or after induction of cerebral ischemia. Extracted data included the mean infarct size (expressed as mm3 or percent of normal brain) or mean neurological score. In addition to the mean values, the standard deviation and the number of animals per treatment group were also extracted. Many studies had multiple groups to facilitate dose responses or time course evaluations; the data from these experiments were extracted from each group as individual values. Data presented graphically rather than reported in the text were obtained from enlarged versions of the graphs. In all included studies, edema was measured and was reported as not significant or was used to correct for infarct volume. Data for neurological outcome were extracted from various neurological tests including: (i) neurological score (subjective sensory motor assessment); (ii) modified limb placement test; (iii) rotarod latency; (iv) tightrope test for forelimb and hindlimb grasping; (v) paw placement cylinder test; (vi) wire hang test; (vii) adhesive tab removal; and (viii) incline test, as documented in Table 1. If ischemic duration was insufficient to induce a loss in neurologic function in a particular test, the data were not included. Data were not extracted if mean neurologic scores were not reported (i.e. if only median and confidence intervals were given).
|Study||Parameters assessed||TZD (dose)||First dose timing||Species||Model||Route or drug administration4|
|Allahtavakoli et al. (2006)||Infarct volume, % (48 h)||Rosi (0.033, 0.1, 0.3, 1 mg/kg)||−1 h, 0 h, or +4 h||Male WR||Embolic stroke||i.p.|
|Allahtavakoli et al. (2007)||Infarct volume, % (48 h)||Rosi (0.1 mg/kg)||0 h||Male WR||Embolic stroke||i.p.|
|Allahtavakoli et al. (2009)||Infarct volume, % (72 h)||Rosi (5 mg/kg)||+24 h||Male WR||Embolic stroke||i.p.|
|Chu et al. (2006)||Infarct volume, %(24 h); MLPT (1, 7, 14, 21, 28, 35 days)||Rosi (0.3, 3 mg/kg)||−7 days||Male SD||MCAO (t) 90 min||p.o.|
|Fatehi-Hassanabad and Tasker (2010)||Infarct volume, % (l, 3, 7, 14 days); TT (6, 24, 72 h), PPT & RR (6 h, l, 3, 7, 14 days)||Rosi (5, 10 mg/kg)||−10 min||Male CD1 mice||BCAO (t) 45 min||i.p.|
|Glatz et al. (2010)||Infarct volume, mm3 (24,48 h); SSMA (24, 48 h)||Pio (3 nmol/h)||−6 days||Male WR||MCAO (t) 90 min||i.c.v.|
|Ji et al. (2009)||Infarct volume, mm3(72 h)||Pio (20 mg/kg)||0 h or +1 days||129/SV mice||MCAO (t)||i.p.|
|Lee et al. (2009)||Infarct volume, mm3(24 h)||Pio (40 mg/kg/day)||−3 days||Male C57BL/6 mice||MCAO (t)||p.o.|
|Luo et al. (2006)||Infarct volume, mm3 (2, 7 days); SSMA & CT (7 days)||Rosi (0.75–12 mg/kg)||−1 h||Male C57BL/6 mice||MCAO (t) 90 min||i.p.|
|Nategh et al. (2010)||infarct volume, % (24 h)||Rosi (1 mg/kg)||0 h||Male WR||Embolic stroke||i.p.|
|Pereira et al. (2006)||Infarct volume, mm3 (48 h); SSMA (48 h)||Rosi (1–3 mg/kg)||+10 min or +2 h||Male FR||MCAO (p)||i.p.|
|Schmerbach et al. (2008)||Infarct volume, mm3 (48 h); SSMA (24, 48 h)||Pio (2, 20 mg/kg/day)||−5 days||Male WR||MCAO (t) 90 min||p.o.|
|Schock et al. (2008)||Infarct volume, % (72 h)||Rosi (6 mg/kg)||0 h||Male mice||MCAO (t) 60 min||i.p.|
|Shimazu et al. (2005)||Infarct volume, mm3 (24 h); SSMA-U (24 h)|
infarct volume, mm3 (48 h & 7 days); SSMA (48 h), WH (48 h)
|Pio (20 mg/kg/day)||−72 h||Male SD|
Male FR, B6;129F2/J
|MCAO (t) 90 min, MCAO (p)||feed|
|Sobrado et al. (2009)||Infarct volume, % (24 h & 22 days); SSMA (l–22 days), ATR||Rosi (1, 3 mg/kg)||+10 min||mice||MCAO (p)||i.p.|
|Sundararajan et al. (2005)||(l–22 days), IT(l–22 days)||Pio (1 mg/kg)||−24 h||Male WR|
Male SD, SHR, C57BL/6, SV129, db/db, S2/SvPasCrl
|MCAO (t) 2 h||i.p.|
|Tureyen et al. (2007)||Infarct volume, mm3(72 h), SSMA (72 h)||Rosi, Pio (0.5–6 mg/kg)||−4 h through +6 h||mice||MCAO (t) 2 h||i.p.|
|Victor et al. (2006)||Infarct volume, % (24 h)||Pio, Rosi (1 & 0.1 mg/kg||−24 h||Male WR||MCAO (t) 2 h||i.p.|
|Wang et al. (2009)|
|Experiment 1||Infarct volume, % (48 h); SSMA-U||Rosi (2, 10 mg/kg/day)||−2 weeks||Male WR||Embolic stroke||p.o.|
|Experiment 2||Infarct volume, % (48 h)||Rosi (2 mg/kg/day)||+1, 3, or6 h||Male WR||Embolic stroke||p.o.|
|Wu et al. (2009)|
|Experiment 1||Infarct volume, mm3(24 h)||Rosi (0.2–500 ng)||0 h||Long Evans||BCAO (t) 30 min||i.c.v.|
|Experiment 2 and 3||Infarct volume, mm3(24 h)||Rosi (50 ng)||+2 h||Long Evans||BCAO (t) 30 min||i.c.v.|
|Zhao et al. (2005)||Infarct volume, mm3 (48 h); SSMA (24, 48 h)||Pio (3 nmol/h)||−5 days||Male WR||MCAO (t) 90 min||i.c.v.|
|Zhao et al. (2006)||Infarct volume, mm3(48 h)||Pio (3 nmol/h)||−5 days||Male WR||MCAO (t) 90 min||i.c.v.|
An 8-point scale was used to determine the quality of each of the included studies based on their stated methodology (Gibson et al. 2008). Written evidence was required in order for a study to receive one point in each of the following categories: randomization of treatment groups, assessment of physiological measurements, dose–response evaluation, time course for treatment, investigator blinding, outcome measurements at days 1–3, outcome measurements beyond day 3, and measurement of both infarct size and neurological outcome.
Cochrane Review Manager (version 5, http://www.cochrane.org/) and the statistical package therein (fixed effect model) was used to analyze the extracted data. A one-way anova with a post hoc Bonferroni test was used to determine statistically significant effects among the routes of drug administration. The standardized mean difference was calculated and applied to assess the effect of rosiglitazone and pioglitazone on infarct volume or neurological outcome compared with vehicle-treated controls. By this method, the difference between each control and treatment group is divided by the standard deviation of the data. This accounts for the fact that measurements may have been taken differently in each experiment (i.e. mm3 vs. percent for infarct volume and neurological score vs. other tests for neurological outcome) and that different animal models may have been employed. Statistical heterogeneity was accounted for through the use of the DerSimonian and Laird (1986) pooling model of random effects. The experiments were grouped and analyzed based on: (i) outcome (infarct volume or neurological deficits), (ii) methodological quality, (iii) type of TZD administered (rosiglitazone or pioglitazone), (iv) timing of first dose (prior to or following ischemia), or (v) type of model (transient or permanent). For those tests in which an increase in nominal value represented an improvement in outcome (for instance, in the wire hang test), the inverse of the extracted data was used for data comparisons.
Design of studies
Based on our search criteria, we identified 31 studies that investigated the use of TZDs in an animal model of cerebral ischemia. The earliest study was published in 2005 (Shimazu et al. 2005). Seven of these studies were excluded because they did not measure infarct volume and one study was excluded because the injury model was traumatic brain injury rather than focal or global ischemia (Yi et al., 2008). Of the final 22 included studies (Table 1; Shimazu et al. 2005; Sundararajan et al. 2005; Zhao et al. 2005, 2006; Allahtavakoli et al. 2006, 2007, 2009; Luo et al. 2006; Pereira et al. 2006; Victor et al. 2006; Tureyen et al. 2007; Nategh et al. 2010; Schmerbach et al. 2008; Schock et al. 2008; Wu et al., 2009; Ji et al. 2009; Lee et al. 2009; Sobrado et al. 2009; Wang et al. 2009; Fatehi-Hassanabad and Tasker 2010; Glatz et al. 2010), 15 employed the use of an animal model of transient MCAO, five performed studies of embolic stroke (Allahtavakoli et al. 2006, 2007, 2009; Chu et al. 2006; Nategh et al. 2010; Wang et al. 2009), and two used transient bilateral CAO (Wu et al. 2009; Fatehi-Hassanabad and Tasker 2010).
All studies compared administration of either rosiglitazone or pioglitazone with a vehicle-treated group of animals and measured mean infarct volume in mm3 or as a percent of either total brain volume or the control hemisphere. The included studies came from 19 independent research groups and data were analyzed from 1348 experimental subjects for infarct volume measurements. Ten of the included studies, from nine independent groups, included measurements of neurological function in 939 experimental subjects. Experimental models for assessment of neurological outcome included subjective neurological scoring as well as objective tests for motor and sensory performance (Table 1).
Male rat models (Wistar, Sprague-Dawley, Fisher, Long Evans) were used in 17 of the studies, whereas the remaining five used male mouse models (C57BL/6, CD1, 129/SV), although primary data on infarct volume and neurological outcome may not have been measured or reported from all of the included animal models (Table 1). In 18 of the studies, the TZD was administered either orally or intraperitoneally. Four studies administered the TZD intracerebroventricularly using pumps while one mixed the drug into peanut butter (Shimazu et al. 2005). The timing of the first dose ranged from 2 weeks prior to ischemic insult to 1 day following injury. Fourteen studies began administration of the drug prior to injury, whereas 10 started treatment with the TZD either at the time of ischemia or thereafter. Rosiglitazone was administered in 13 of the studies and pioglitazone was used in the remaining nine studies (Table 1).
Overall efficacy of drug treatment
There was a significant protective effect of TZD treatment on infarct size (1.54, 1.40–1.68, p < 0.00001) and neurological outcome (1.19, 1.04–1.34, p < 0.00001) across all models (Fig. 1). The nature of this statistical approach is to compare the standardized mean difference between drug treatment and vehicle across all studies. A standardized mean difference of zero represents a lack of drug effect, whereas a positive value indicates a beneficial effect, and a negative mean difference reflects a detrimental effect of the drug.
Reported study quality
The quality scores ranged from 1 to 8 with a mean of 4.3. Analysis of studies based upon individual quality scores was not very meaningful as scores were unevenly distributed over the scoring range; for instance, no studies had a quality score of 3 or 7. Therefore, data from experiments with a study quality of 1–4 were grouped together and compared to those with a study quality of 5–8 (Fig. 2). Infarct volume was reduced to a similar degree regardless of study quality (χ2 = 1.91, df = 1, p = 0.17). However, studies with a quality rating of 1–4 were less likely to show efficacy of TZD treatment for improvements in neurological outcome than those with a study quality of 5–8 (χ2 = 14.01, df = 1, p < 0.0002).
Benefits of rosiglitazone vs. pioglitazone
Rosiglitazone and pioglitazone were analyzed separately for their effects on infarct volume and neurological deficits (Fig. 3). Rosiglitazone significantly decreased lesion volume (effect size = 1.55, CI = 1.39–1.72, p < 0.00001) and improved neurological outcome (1.19, 0.99–1.38, p < 0.00001). Pioglitazone also reduced lesion size (1.50, 1.24–1.76, p < 0.00001) and improved neurological outcome (1.2, 0.95–1.44, p < 0.00001). There was no difference in the effectiveness of the two different drugs with regard to either infarct volume (χ2 = 0.12, df = 1, p = 0.73) or neurological deficits (χ2 = 0.00, df = 1, p = 0.95).
Efficacy of TZD depending on first dose timing
Thiazolidinedione treatment was effective at reducing lesion volume and improving neurological outcomes when dosed prior to or after the induction of focal or global ischemia (Fig. 4). When dosed before ischemic injury, TZDs decreased infarct size (1.56, 1.36–1.76, p < 0.00001) and improved neurological outcomes (1.12, 0.96–1.29, p < 0.00001). Administering TZDs at or after the induction of ischemic stroke was also effective in reducing lesion volume (1.52, 1.32–1.72, p < 0.00001) and improving neurological outcomes (1.73, 1.26–2.19, p < 0.00001). There was no difference in lesion volume reduction because of dose timing (χ2 = 0.08, df = 1, p = 0.78). Treatment with a TZD after the onset of ischemic injury resulted in a greater improvement in neurological outcomes than did dosing prior to ischemia (χ2 = 5.82, df = 1, p = 0.02).
Effect of TZD on infarct size depending on drug and timing of first dose
Although there were no statistically significant differences in anatomical outcome measures based on the timing of the first dose or the type of TZD administered (rosiglitazone or pioglitazone), there was a difference in effect when the data was analyzed separately for each of the two drugs based on dose timing (Fig. 5). The effects of rosiglitazone on the reduction of lesion size were similar whether it was administered before or after ischemia (χ2 = 1.33, df = 1, p = 0.25). Pioglitazone was more effective at reducing lesion volume when administered before the induction of focal or global ischemia (χ2 = 5.93, df = 1, p = 0.01).
Drug effects based on model
All of the included studies employed a model of either transient or permanent ischemia. TZD treatment in transient ischemia models led to improvements in lesion size (1.51, 1.36–1.65, p < 0.00001) and neurological outcome (1.17, 1.01–1.34, p < 0.00001) (Fig. 6). The administration of rosiglitazone or pioglitazone was also effective in permanent ischemia models at reducing infarct volume (2.00, 1.40–2.60, p < 0.00001) and neurological impairments (1.36, 0.84–1.87, p < 0.00001). There were no significant differences in either outcome measure comparing permanent versus transient ischemia models (infarct volume: χ2 = 2.43, df = 1, p = 0.12; neurological outcome: χ2 = 2.09, df = 1, p = 0.15).
Effect size according to route of drug administration
The effects of TZDs on anatomic and neurological outcomes significantly differed based on the route of drug administration (Fig. 7). Dosing the TZD by oral gavage tended to be less effective at reducing infarct volume than intraperitoneal (p < 0.05), intracerebroventricular (p < 0.05), or food-based administration (NS). Conversely, for neurologic outcome, the efficacy according to route of administration was highest with peroral, followed by intraperitoneal and intracerebroventricular delivery (p < 0.05 between all groups).
The results of this meta-analysis suggest that rosiglitazone and pioglitazone are effective agents for reducing infarct volume and improving neurologic outcome in rodent models of stroke. In the available data, the two drugs were similarly neuroprotective, and more importantly, the protective effects were retained regardless of the timing of the first dose of drug (Table 1).
Whereas the effects of TZD treatment on infarct volume were independent of first dose timing, the improvement in neurological outcome was greater when drug was given during or following the ischemic insult (Fig. 4). Although there may be multiple explanations, it is possible that TZDs independently affect sensory/motor coordination in a manner that is influenced by the timing of the dose and/or the cumulative dose. At the time of behavioral assessment, the cumulative dose in the periphery would likely be lower with post-ischemic administration than with pre-treatment. The possibility that effects in the periphery are influencing performance in the behavioral tests is supported by the data in Fig. 7 that documents the diminished benefit to neurologic outcome when drug is administered intracerebroventricularly.
An advantage of considering TZDs for acute stroke therapy is that these are currently clinically used drugs for which the pharmacokinetics and toxicity are generally well studied. With longer term use, there are issues with these compounds with regard to tolerability, and repeated use is associated with significant undesirable side effects, including weight gain and edema. These shortcomings would unlikely compromise the ability to utilize the compounds acutely for the treatment of stroke, as only short-term dosing might be required. The issue with regard to potential cardiovascular risk of rosiglitazone might also be obviated, but this would require further consideration.
Mechanistically, the efficacy of TZDs in stroke models has been attributed to a variety of effects, including their broad anti-inflammatory properties involving induction of antioxidant enzymes, suppression of proinflammatory cytokines, inducible nitric oxide synthase, cyclooxygenase-2, and adhesion molecules, decreased macrophage infiltration, and promotion of neovascularization (Chu et al. 2006; Culman et al. 2007). There is significant evidence that protection involves activation of PPARγ, at least in part (Lin et al., 2006; Pereira et al. 2006; Victor et al. 2006; Zhao et al. 2006; Tureyen et al. 2007; Schock et al. 2008; Sobrado et al. 2009; Wu et al. 2009; Glatz et al. 2010; Luo et al. 2006), but there is also data suggesting that non-PPARγ-mediated mechanisms specifically at mitochondria may be involved in the mechanism of action of TZDs (Feinstein et al. 2005). Whether the higher affinity of rosiglitazone for PPARγ (Feinstein et al. 2005) is what allows for administration of generally lower concentrations than pioglitazone (Table 1) remains unclear.
Although promising, the data indicate that additional studies are required to determine the efficacy of delayed administration of TZDs (i.e. 6 h or longer) and assess outcome measures at delayed time points (i.e. longer than 72 h, as reported in Sundararajan et al. 2005; Chu et al. 2006; Luo et al. 2006; Fatehi-Hassanabad and Tasker 2010). Additional studies should take into consideration other variables as suggested by Howells et al. (2010) that may strengthen the rationale and guide plans for testing in the clinical setting. Some of these variables were broached in the current studies, such as testing in the context of co-morbidities, including hypertension and diabetes (Tureyen et al. 2007) or tissue plasminogen activator treatment (Wang et al. 2009) and elevated temperature (Nategh et al. 2010). In each of these studies, TZD treatment was neuroprotective. Of note, most studies that tested delayed administration of TZD from 4 to 24 h (Allahtavakoli et al. 2006, 2009; Tureyen et al. 2007; Ji et al. 2009; Wang et al. 2009) found protective effects of rosiglitazone. In the single study that tested delayed (24 h) administration of pioglitazone with assessment at 6 weeks (Ji et al. 2009), treatment was no longer as effective as it had been at an earlier point of evaluation, but these observations need to be strengthened with further studies.
Overall, current data are supportive that TZDs could be effective agents in the acute treatment of ischemic stroke.
Note added in proof
During review of the current work, new data emerged addressing the post-ischemic therapeutic window for pioglitazone: Gamboa J, Blankenship DA, Niemi JP, Landreth GE, Karl M, Hilow E, Sundararajan S. (2010) Extension of the neuroprotective time window for thiazolidinediones in ischemic stroke is dependent on time of reperfusion. Neuroscience. 170:846–857.
Support for this study is provided by the American Diabetes Association 0-08-RA-139, NIH-NIDDK 1R42DK081298, NIH-NIDDK 2P01 DK054441 to A.N.M.