Evidence for the efficacy of statins in animal stroke models: a meta-analysis

Authors

  • Lidia García-Bonilla,

    1. Neurovascular Research Laboratory, Institut de Recerca Vall d’Hebron, Universitat Autònoma de Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Mireia Campos,

    1. Neurovascular Research Laboratory, Institut de Recerca Vall d’Hebron, Universitat Autònoma de Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Dolors Giralt,

    1. Neurovascular Research Laboratory, Institut de Recerca Vall d’Hebron, Universitat Autònoma de Barcelona, Barcelona, Spain
    Search for more papers by this author
  • David Salat,

    1. Neurovascular Unit, Department of Neurology, Universitat Autònoma de Barcelona, Hospital Vall d’Hebron, Barcelona, Spain
    Search for more papers by this author
  • Pilar Chacón,

    1. Lipids Unit, Department of Clinical Biochemistry, Universitat Autònoma de Barcelona, Hospital Vall d’Hebron, Barcelona, Spain
    Search for more papers by this author
  • Mar Hernández-Guillamon,

    1. Neurovascular Research Laboratory, Institut de Recerca Vall d’Hebron, Universitat Autònoma de Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Anna Rosell,

    1. Neurovascular Research Laboratory, Institut de Recerca Vall d’Hebron, Universitat Autònoma de Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Joan Montaner

    1. Neurovascular Research Laboratory, Institut de Recerca Vall d’Hebron, Universitat Autònoma de Barcelona, Barcelona, Spain
    2. Neurovascular Unit, Department of Neurology, Universitat Autònoma de Barcelona, Hospital Vall d’Hebron, Barcelona, Spain
    Search for more papers by this author

Address correspondence and reprint requests to Dr. Joan Montaner, Neurovascular Research Laboratory, Neurovascular Unit, Institut de Recerca, Hospital Vall d’Hebron, Pg. Vall d’Hebron 119-129, 08035 Barcelona, Spain. E-mail: 31862jmv@comb.cat

Abstract

J. Neurochem. (2012) 122, 233–243.

Abstract

Protective effects of statins have been well documented for stroke therapy. Here, we used a systematic review and meta-analysis to assess these evidences. We identified 190 studies using statin treatment in stroke animal models by electronic searching. From those, only studies describing ischemic occlusive stroke and reporting data on infarct volume and/or neurological outcome were included in the analysis (41 publications, 1882 animals). The global estimate effect was assessed by Weighted Mean Difference meta-analysis. Statins reduced infarct volume by 25.12% (20.66%–29.58%, < 0.001) and consistently, induced an improvement on neurological outcome (20.36% (14.17%–26.56%), < 0.001). Stratified analysis showed that simvastatin had the greatest effect on infarct volume reduction (38.18%) and neurological improvement (22.94%), whereas bigger infarct reduction was observed giving the statin as a pre-treatment (33.5%) compared with post-treatment (16.02%). The use of pentobarbital sodium, the timing of statin administration, the statement of conflict of interest and the type of statin studied were found to be independent factors in the meta-regression, indicating their influence on the results of studies examining statin treatment. In conclusion, this meta-analysis provides further evidences of the efficacy of statins, supporting their potential use for human stroke therapy.

Abbreviations used
CI

confidence interval

df

degrees of freedom

eNos

endothelial isoform

WMD

Wekghted Mean Difference

The benefits of statins on ischemic stroke have been well documented since their introduction into clinical therapy in the late 1980s, as an efficacious and safe treatment for cardiovascular disease (Grundy 1988). On average, statins can lower LDL cholesterol by 1.8 mmol/l (70 mg/dl), which translates into a 60% decrease in the number of ischemic heart diseases events and a 17% reduced risk of stroke after long-term treatment (Wald et al. 2003). Aside their lipid-lowering properties (Endo et al. 1976), pleiotropic effects of statins (George et al. 2002) may exert neuroprotection as it has been extensively pointed out in the last two decades by using animal stroke models. On the top of these properties, statins increase the expression of the nitric oxide synthase (NOS), specifically its endothelial isoform (eNOS), which improves endothelial function and increases cerebral perfusion in the ischemic penumbra (Endres & Laufs 2004, Yamada et al. 2000, Endres et al. 1998). Statins can exert neuroprotection by modifying anti-oxidative pathways via inhibition of NADPH oxidase- derived superoxide (Hong et al. 2006, Cui et al. 2009) and additionally, statins may alter inflammatory and immune responses after stroke. These can modulate NF-κB activity and the expression of inflammatory mediators like IL1β, monocyte chemoattractant protein-1 (MCP-1), or tumor necrosis factor alpha (TNF-α) (Sironi et al. 2006, Romano et al. 2000a, Kawai et al. 2011) and attenuate up-regulation of adhesion molecules like P-selectin, ICAM-1, CD18, and CD11b integrins (Romano et al. 2000b, Pruefer et al. 1999, Mayanagi et al. 2008)

Moreover, the use of statins has widely been explored either as pre-treatment in stroke prevention or as post-treatment after ischemia for neuroprotection of the injured ischemic brain. Their benefits in experimental stroke have been described and compared using several types of statins, such as atorvastatin, pitavastatin, simvastatin, (Hayashi et al. 2005, Laufs et al. 2000), pravastatin (Trinkl et al. 2006, Berger et al. 2008), lovastatin (Endres et al. 1998), or rosuvastatin (Laufs et al. 2002) and reported in both transient or permanent cerebral ischemia models (for details, see table 1).

Table 1.   Pre- or post-treatment with different type of statin in animal models of focal ischemia
 ModelDose (mg/Kg/day)Administration waySpeciesPhaseReference
  1. MCAO, Middle Cerebral Artery; permanent (pMCAO), transient (tMCAO), or embolic (embolic MCAO) stroke model; SHR-SP, Spontaneously Hypertensive Stroke-Prone Rats. Dose (expressed in mg/Kg/day) and way of administration (subcutaneous, sc; oral, po; intraperitoneal, ip). Different administration times are indicated in brackets (hour, h; day, d; week, w; m, month). SD, Sprague-Dowley. Acute, Subacute, or Chronic phases refer the studied time point of the ischemic lesion and/or neurological outcome.

Pre-treatment
AtorvastatinEmbolic MCAO20sc (14d)SV/129-C57BL/6 miceAcuteAsahi et al. 2005
pMCAO10sc (7d or 14d)C57BL/6J miceAcute/
subacute
Chen et al. 2005
3po (7d)Retired breeder male Wistar ratsAcute/ChronicChen et al. 2006
tMCAO10sc (3d)SD ratsAcuteHong et al. 2006
10sc (14d)SV/129 (wt and eNOS−/−) miceAcuteLaufs et al. 2000
10po (14d)Wistar ratsAcuteHayashi et al. 2005
10sc (14d)129/SV wt miceAcuteGertz et al. 2003
SHR-SP20po (4 w)Spontaneously Hypertensive Stroke-Prone Rats Nagotani et al. 2005
2 or 20po (11w)Spontaneously Hypertensive Stroke-Prone Rats Tanaka et al. 2007
SimvastatinEmbolic MCAO20sc (14d)SV/129-C57BL/6 miceAcuteAsahi et al. 2005
100po (14d)SD rats (hyper- or normotermia)AcuteShabanzadeh et al. 2008
100ip (2w)SD ratsAcuteShabanzadeh et al. 2005
pMCAO20sc (3d)SD ratsAcuteSironi et al. 2003
Sironi et al. 2006
10po (14d)Long Evans ratsAcute/
Chronic
Yrjänheikki et al. 2005
tMCAO20sc (14d)SV/129-C57BL/6
(wt and eNOS–/–) mice
AcuteLaufs et al. 2000
1ip (14d)diabetic swiss albino miceAcuteÇakmak et al. 2007
0.2, 2 or 20
20
sc (14d)
sc (3d)
SV/129, C57BL/6 and eNOS−/− miceAcuteEndres et al. 1998
10po (14d)Wistar ratsAcuteHayashi et al. 2005
Simvastatin
 In combination with Dipyridamole
 and Aspirin
0.1, 1, 10 or 20 (Sim)
60 or 30 (Dip)
10 (ASA)
ip (14d)
po (3d)
po (3d)
C57BL/6 wt and eNOS−/−AcuteKim et al. 2008
RosuvastatintMCAO 10 mg/Kg/d ip (3d)C57BL/6J,
C57BL/6J ob/ob and lean mice
AcuteMayanagi et al. 2008
5mgip (14d)SD ratsAcute/
subacute
Engelhorn et al.
2006
2mgip (14d)129/SV wt miceAcuteLaufs et al.2002
PitavastatintMCAO10po (14d)Wistar ratsAcuteHayashi et al. 2005
SHR-SP10 or 20po (4 w)Spontaneously Hypertensive Stroke-Prone Rats Nagotani et al. 2005
PravastatintMCAO20diet (4w)Wistar ratsAcuteTrinkl et al. 2006
CerivastatinSHR-SP2po (10w)Spontaneously Hypertensive Stroke-Prone Rats Kawashima et al. 2003
MevastatintMCAO2 or 20continuons infusion
(7, 14 or 28d)
129-SV/eVTAcBr miceAcuteAmin-Hanjani et al. 2001
LovastatintMCAO0.2, 2 or 20
20
sc (14d)
sc (3d)
SV/129, C57BL/6 and eNOS–/–) miceAcuteEndres et al. 1998
Post-treatment
AtorvastatintMCAO10sc (2d)SD ratsAcuteSironi et al. 2003
1, 3 or 8po (7d)Wistar ratsAcute/
subacute
Chen et al. 2003
Embolic MCAO20sc (2d)Wistar ratsAcute/
subacute
Zhang et al. 2007
Zhang et al. 2005
20sc (2d)Wistar ratsAcuteDing et al. 2006
Liu et al. 2006
SimvastatinpMCAO10po (7d)Long Evans ratsAcute/
Chronic
Yrjänheikki et al. 2005
20sc (2d)SD ratsAcuteSironi et al. 2003
Embolic MCAO1po (7d)Wistar ratsAcute/
subacute
Zacharek et al. 2009
20 or 40sc (single dose)Wistar ratsAcuteNagaraja et al. 2006
40sc (single dose)SD rats (hyper- or normotermia)Acute/
Chronic
Shabanzadeh et al. 2008
RosuvastatintMCAO0.2 or 2
20
iv (5d q24h)
ip (5d q24h)
129/SV wt miceAcutePrinz et al. 2008
0.5 or 5 or 20ip (single dose)C57BL/6 miceAcuteKilic Ü et al. 2005
20ip (single dose)C57BL/6 and eNOS–/– miceAcuteKilic E et al. 2005
PravastatintMCAO50continuons infusion (6d)Wistar ratsAcute/
subacute
Sugiura et al. 2007
0.1, 0.5, 1 or 2ip (4d)Wistar ratsAcute/
subacute
Berger et al. 2008
FluvastatinpMCAO5po (3m)Wistar ratsChronicShimamura et al. 2007

Here, we conducted a systematic review and meta-analysis to further investigate the described benefits of different statins. We sought to assay whether these drugs globally offer clear advantages in ischemic stroke models that might be translated into human stroke therapy.

Experimental procedures

Search strategy and data extraction

To identify pre-clinical studies investigating the use of statin on animal stroke, we performed an electronic searching in PubMed (all publications until November 30, 2011) using the search terms of [STATIN] AND [ISCHEMIA] OR [ISCHAEMIA] AND [BRAIN]; [STATIN] AND [STROKE] and [STATIN] AND [CEREBRAL INFARCTION], limited to ‘animal species’. Titles and abstracts from identified studies were screened by three independently investigators (DG, DS, and LG-B) to decide whether they were eligible for data extraction. Studies were considered eligible when they described animal models of ischemic occlusive stroke, implying the middle cerebral artery or its branches, and in which data of statin treatment on infarct size and/or on neurological outcome were reported. Discrepancies in extraction were resolved by discussion among the three investigators.

Our pre-determined primary end-points were infarct volume and neurological deficit outcome. From each study, we identified individual comparisons where primary outcomes were measured in a group of animals receiving a specific statin at a specific time(s), and compared with the outcome in a control group. Data regarding the number of animals used, the mean and the standard deviation (SD) or the standard error of mean (SEM) from treated and control groups, were extracted. Where neurological outcome was measured serially, only the final measure was used. Data from studies with multiple treatment groups investigating dose response or time course, were extracted from each group as individual values. Where data required for meta-analysis were missing or unpublished, we contacted authors to request additional information. Data presented graphically rather than reported in the text, were obtained using digital ruler software (Universal Desktop Ruler). Studies in which data required for meta-analysis were not presented or obtainable were excluded from the analysis.

Moreover, information on study design including species, age, weight, sex, dose, method of ischemia induction, type of statin, dose, time and route of administration, time of assessment of outcome, methods of both neurological and infarct volume evaluation and quality of the study, were collected from each study.

Study quality

Study quality was assessed based on a quality 10 points-checklist modified from the CAMARADES study (Macleod et al. 2004), comprising (1) peer review publication, (2) randomized assignment to study treatment, (3) blinded treatment administration, (4) blinded assessment of outcome, (5) statement of control of temperature, (6) sample size calculation, (7) statement of compliance with animal welfare regulations, (8) avoidance of anesthetics with marked intrinsic neuroprotective properties (ketamine), (9) statement of potential conflicts of interest, and (10) the use of accurate/suitable/adequate animal models (i.e. use of aged or co-morbid animals).

Data Analysis

Extracted data were analyzed as previously described (Sena et al. 2010a) using MS access from CAMARADES group. From each comparison, the effect size was calculated by applying the formula: Effect (%) = 100 x ((outcome control) – (outcome treatment))/(outcome control) on infarct volume or neurological outcome of extracted data. For the purpose of meta-analysis, we considered neuro-behavioral scales as continuous scales even when they were ordinal. Weighted Mean Difference meta-analysis (WMD) and its 95% confidence interval (CI) with random effects model to avoid heterogeneity (DerSimonian R and Laird N), were used to assay differences of the global estimate effect on infarct volume and neurological outcome. Stratified analysis was conducted to explore the influence of the quality of the study, types of stroke models (permanent or transient), dose, timing of first dose (pre-treatment or post-treatment), or type of statin on estimates of effect size. Differences of mean effect sizes were assessed partitioning heterogeneity using χ2 distribution with n-1 degrees of freedom (df). P-values <0.003 were considered significant using Bonferroni correction. Correlation between drug administration timing and statin effect on infarct volume reduction was assessed using Spearman’s rho correlation coefficient.

Multivariable meta-regression and related biases were performed using STATA software as describe elsewhere (Frantzias et al. 2011). Meta-regression was conducted to explore potential confounders among the following co-variates: number of dose, timing of first dose (pre-treatment or post-treatment), type of statin, effect of anesthesia, co-morbidity, and statement of conflict interest. Funnel plots were used to assess publication bias. Asymmetry was detected in the plots by visual inspection and when any bias was observed, a new corrected effect size was calculated using “trim and fill” by the Egger regression method (Sterne et al. 2001).

Results

Design of the study

Forty-seven of 190 total identified studies investigating the statin effects on animal stroke models met our inclusion criteria of occlusive ischemia model and examination of infarct volume and/or neurological outcome to be analyzed (fig. 1). Six of those were excluded from the analysis because data were neither reported nor obtainable. Then, 136 comparisons of the final 41 included publications were extracted in the analysis (supplementary table 1); 99 in 1882 animals reported data as infarct volume and 37 in 796 animals as neurological score. All the studies were in mice or rats, except for one performed in rabbit (supplementary table 1).

Figure 1.

 Flow diagram of the study selection.

Meta-analysis

Treatment with statins reduced the infarct volume by 25.12% (95% confidence interval (CI), 20.66%–29.58%, < 0.001) as compared with the control group. Consistent with this, a considerable improvement in the neurological outcome was observed (20.36% (14.17%–26.56%), < 0.001), (fig. 2).

Figure 2.

 Meta-analysis of effect of statins on infarct volume reduction (a) and neurological improvement (b). The horizontal lines represent the mean estimated effect size (MES) and the 95% confidence intervals (CI) for each individual comparison according to their effect on infarct volume (a) and neurological score (b). The mean difference (MD) and the 95% CI of the global estimate are represented as solid and dashed vertical lines, respectively.

Reported study quality

Overall, the median quality (range scale 0-10) of the 41 included studies was modest (4, interquartile range, 4–6) with a score that ranged from 1 to 6. No studies scored 0 or high quality rating (7–10 points). Twenty-seven studies (65.9%) reported temperature controlling during surgical procedure, but only 17 studies (17.5%) reported randomization of the assignment to study treatment and 12 studies (29.3%) reported that the treatment was concealed to investigators during outcome assessment. Moreover, any study reported blinded treatment administration.

When the effect size on infarct volume reduction and neurological improvement were examined relative to study quality score, a significant heterogeneity over the scoring range was observed (fig. 3). Although no significant differences were found in theses parameters between low-scored studies and higher-scored studied, the effect size on infarct volume decreased by 12% in comparisons with maximun score (6). Similarly, comparisons with a low quality score of 2 and 3 showed higher efficacy on neurological outcome (23% and 36% effect, respectively) than comparisons scoring 4-5 (15%) and 6 (13%).

Figure 3.

 Effect on infarct volume reduction (a) and neurological score improvement (b) regarding reported quality score. The horizontal lines represent the mean (solid line) and the 95% confidence intervals (dashed lines) of the global estimate. Each bar represents the mean +/- SEM for the individual estimates. The number of both studies and animals contributing to each individual estimate are indicated in brackets.

Stratified meta-analysis

Effect size on infarct volume was significantly higher when the statin was administered as a pre-treatment (33.57%; 95% CI, 28.47%–38.53%) compared with post-treatment (16.02% (11.63%–20.42%); χ= 408, df=1, < 0.001) and it was accompanied by a better neurological improvement (26.52% (15.05%–37.99%) vs. 14.37% (7.26%–21.48%),; χ= 17, df=1, < 0.001), (fig. 4). The median and the interquartil range timing for statin pre- or post-treatment administration were 14days (5–14d) and 4h (1.13–12h), respectively. Statin efficacy was correlated with the timing of administration (r= 0.334, < 0.001), being more effective as soon as it was administered. There was a decrease of the efficacy to reduce infarct volume of 0.5% for every day delay to pre-treatment and of 0.06% for every hour (1.4% per day) delay to post-treatment.

Figure 4.

 Effect size on infarct volume reduction (a) and neurological score improvement (b) stratified by statin pre- (blue color) or post-treatment (red color). The horizontal lines represent the mean estimated effect size (MES) and the 95% confidence intervals (CI) for each individual comparison. The mean difference (MD) and the 95% CI of the global estimate are represented as solid and dashed vertical lines, respectively.

The analysis according to the route of administration showed better results when the drug was administered orally (infarct volume reduction of 34.50% (26.65%–42.35%)) as compared to intraperitoneally (16.96% (10.62%–23-31%)), whereas the effect was similar in those animals receiving the statin subcutaneously (28.17% (20.39%–35-95%); χ= 219, df=2, < 0.001) (fig. 5a–c). In terms of effect size on neurological outcome, oral administration led to higher favorable improvement (28.35% (22.74%–33.97%) than either subcutaneous (9.76% (-0.91%– 20.42%)) or intraperitoneal administration (19.66% (3.32%–36.00%), (χ= 21, df=2, < 0.001), (fig. 5d–f).

Figure 5.

 Effect size on infarct volume reduction (a–c) and neurological score improvement (d–f) stratified by route of statin administration. The horizontal lines represent the mean estimated effect size (MES) and the 95% confidence intervals (CI) for each individual comparison according to their effect on infarct volume (a–c) and neurological score (d–f). The mean difference (MD) and the 95% CI of the global estimate are represented as solid and dashed vertical lines, respectively.

The global estimated benefits of statins on infarct volume were similar when comparing permanent [25.48% (17.66%–33.29%)] or transient (24.70% (20.38%–29.02%)) cerebral ischemia. However, there was a high statistical significant heterogeneity between studies (χ= 283, df=1, < 0.001). Regarding neurological improvement, better outcome was seen in animals that underwent transient ischemia (9.11% (-4.86%–23.08%) vs. 23.05% (16.63%–39.46%), < 0.001), (fig. 6). Significant inter-study heterogeneity was also observed (χ= 77, df=1, < 0.001).

Figure 6.

 Effect size on infarct volume reduction (a) and neurological score improvement (b) stratified by permanent (blue color) or transient (red color) cerebral ischemia model. The horizontal lines represent the mean estimated effect size (MES) and the 95% confidence intervals (CI) for each individual comparison. The mean difference (MD) and the 95% CI of the global estimate are represented as solid and dashed vertical lines, respectively.

Analysis of effect size on infarct volume stratifying by the number of statin dose, repeated or single dose, showed significantly larger reduction in animals receiving multiple dose (30.08% (25.49%-34.68%) vs. 13.72%(8.76%–18.69%), P < 0.001), (fig. 7).

Figure 7.

 Effect size on infarct volume reduction stratified by single or multiple doses of statin administration. The horizontal lines represent the mean estimated effect size (MES) and the 95% confidence intervals (CI) for each individual comparison according to their effect on infarct volume. The mean difference (MD) and the 95% CI of the global estimate are represented as solid and dashed vertical lines, respectively.

By type of statin, simvastatin was significantly associated with a greater infarct size reduction [38.18% (33-23%–43-14%)] with respect to atorvastatin [23.71% (17.05%–30.38%)] and rosuvastatin [13.88% (6.54–21.21%); χ=345, df=2, < 0.001], (fig. 8a–c). Neurological improvement was similar in animals receiving simvastatin [22.94% (12.16%–33.73%)] and artovastatin [19.24% (10.52%–27.96%)] and, in contrast, no effect size was observed in rosuvastatin treated-animals [-0.27% (-18.9%–18.35%)], (data not shown). The heterogeneity was high in that analysis (χ= 25, df=2, < 0.001). Furthermore, we sought to analyze the efficiency on the infarct volume reduction regarding administration of low versus high-dose of different statins. Rosuvastatin, atorvastatin, and simvastatin triggered higher effects at high-dose (20.13% (9.25%–31.01%), 27.40%(18.07%–36.73%) and 40.46% (33.4%–47.52%), respectively) compared with low-dose (11.36% (2.13%–20.59%), 16.55% (1.93%–31.17%), and 36.10% (28.85%–43.35%; (χ= 632, df=1, < 0.001), (χ= 489, df=1, < 0.001), (χ= 637, df=1, < 0.001); p < 0.001) (fig. 8d–f).

Figure 8.

 Effect size on infarct volume reduction stratified by type of statin: simvastatin (a), atorvastatin (b), and rosuvastatin (c). The horizontal lines represent the mean estimated effect size (MES) and the 95% confidence intervals (CI) for each individual comparison. The mean difference (MD) and the 95% CI of the global estimate are represented as solid and dashed vertical lines, respectively. Bar plots represent the mean +/- SEM for the individual estimates comparing low and high-dose of simvastatin (d), atorvastatin (e), or rosuvastatin (f).

Meta-regression analysis

Meta-regression was conducted for infarct volume including data that assayed the effect of simvastatin, atorvastatin, or rosuvastatin, to further explore meta-analysis heterogeneity. The analysis identified four factors: type of anesthetics (pentobarbital sodium vs. others), timing of first dose (pre-treatment or post-treatment), statement of conflict of interest, and type of statin to account for 60.9% of between-study variance (τ= 98.5, adjusted r= 0.609). Studies that used pentobarbirtal sodium as anesthetic reported 18.4% (4.6%-32.1%) lower efficacy than those studies that used a different anesthetic. Pre-treatment with statin reduced the infarct volume, an additional 9% (0.3%-17.7%) compared with post-treatment. Studies that stated conflict of interest showed an efficacy of 12.6% (0.6%-24.6%) less than those that did not. Relative to type of statin, administration of rosuvastatin reduced the infarct volume by 11.7% (25%-20.8%) less compared with either atorvastatin or rosuvastatin.

Bias publication

Finally, we sought to identify whether small studies’ effects may contribute to publication bias in those analysis performed by statin-type (fig. 9). Funnel plot evidenced asymmetry in atorvastatin data (Egger regression, < 0.001) while visual inspection of both simvastatin and rosuvastatin funnel plots indicated obvious symmetric inverted funnel shapes (Egger regression, = 0.939 and = 0.679, respectively).

Figure 9.

 Funnel plot for simvastatin (a), atorvastatin (b), and rosuvastatin (c).

Discussion

Our systematic review and meta-analysis reinforces the value of statin treatment in cerebral ischemia and points to simvastatin as the statin that provides the highest neuroprotective effect to the injured brain. The estimated effect was higher when statins were given as pre-treatment rather than post-treatment. Effectiveness was greater when drug was given orally, and multiple-dose instead of a single-dose treatment provided better neurological outcomes. However, statin treatment led to similar neuroprotection in either transient or permanent cerebral ischemia.

Concerning the study quality, we did not find differences on statin efficacy across the range of scores, but studies with a lower quality showed a trend toward better outcomes. Therefore, the global estimated effect may be overstated in low-scored studies. Another point of note is the reduction of the neuroprotective effect, by more than 10%, when titles stated conflict of interest, or the fact that the articles did not report blinding administration of the treatment, revealing the importance of these study-quality issues in their contribution to publication bias.

It is also interesting to note the finding that the use of the anesthetic pentobarbital markedly decreased (18%) the neuroprotective effect of statins. The influence of the use of pentobarbital on the outcome of animal stroke models has been recently described in a systematic review and meta-analysis on the efficacy of tPA treatment (Sena et al. 2010a). However, in contrast with our findings, this study showed a better neurological outcome when pentobarbital was used. No interactions between statin and pentobarbital have been described so far, but since pentobarbital is a strong cytochrome P450 inducer (Waxman & Azaroff 1992) and statins are predominantly metabolized by the cytochrome P450 system (Bellosta et al. 2004), statin plasma concentrations might be reduced.

Experiments with animal models usually start with a pre-treatment protocol owing to the fact that therapeutic manipulations generally work better when administered before or immediately after the ischemic insult (Danton & Dietrich 2004). Nevertheless, that is not the best approach when we aim to look for a neuroprotectant drug to be administered for acute stroke patients that arrive to the Emergency Departments after the brain artery occlusion. In accordance, we found that statin pre-treatment yielded a 2-fold significant reduction on the infarct lesion (33.57%; 52 comparison) as compared with post-treatment (16.02%; 53 comparison). Reliable clinical evidences have demonstrated that the use of statin preventive therapy reduces the stroke risk (Prinz & Endres 2011, Amarenco & Labreuche 2009) and improves the functional outcome after ischemic stroke (Biffi et al. 2011). On the other hand, clinical studies using statin therapy in the acute phase of the ischemic stroke are scarce. A recent systematic review concluded that insufficient data are available from randomized trials to establish whether statins are safe and effective in cases of acute ischemic stroke (Squizzato et al. 2011). Indeed, to our best knowledge, the MISTICS is the only pilot clinical trial, up to date, examining the simvastatin effect on neurological outcome (Montaner et al. 2008), which showed an improvement among simvastatin-treated patients by the third day of stroke onset (46.4% vs. 17.9%, = 0.022). The post-treatment effectiveness found in this study has been similarly reported in a recently published meta-analysis ( Baryan et al. 2012). The post-treatment effect, albeit being smaller than the pre-treatment effect, supports the therapeutic potential of statin treatment in the acute stroke.

To compare the efficacy by type of statin, we examined the neuroprotective effect associated with simvastatin, atorvastatin, or rosuvastatin administration. We did not include pravastatin, pitavastatin, and perivastatin in the analysis because of the small group data. We observed a higher estimate of efficacy for simvastatin, followed by atorvastatin and rosuvastatin. The least benefit found for rosuvastatin treatment could be explained by its selective uptake into hepatocytes (Nezasa et al. 2003), reducing the systemic bioavailability and its effective dose.

In spite of the well-known efficacy of atorvastatin, we detected a substantial publication bias that was not observed for either simvastatin or rosuvastatin. For atorvastatin, the estimated effect of 23.71% on infarct volume including 32 comparisons, decreased to 4.82% when it was adjusted for publication bias with an input of 15 studies missing. Therefore, the existence of negative or less efficient atorvastatin-treatment unpublished results might be implied.

As expected, multiple doses of statins, well tolerated and safe in long-term exposure (Maron et al. 2000), triggered higher positive effects than a single-dose treatment. Of note, however, in all the studies in which a multiple-dose regimen was used, statins were administered before the ischemic insult. In keeping with this, the increase in statin efficacy might be related not to the dose regimen treatment, but to the use of a pre-treatment protocol. As we found that pre-treatment with statins additionally reduced the infarct volume by 9% and that the factor ‘single vs. multiple-dose’ was not independently identified in the regression analysis, this assumption may be plausible.

To clarify the influence of the dosage on the effect of each statin, we sought to analyze the estimate by high versus low-dose of statin given. All the statins evaluated (simvastatin, atorvastatin, and rosuvastatin), had individually bigger effects on the reduction of the infarct volume at higher dose, as shown by the global estimated effect. Nevertheless, the eventual use of higher-statin doses in stroke therapy needs to be considered because of potential toxic side effects (STAIR 1999). In this regard, hydrophilic statins such as rosuvastatin, are less toxic than lipophilic statins such as atorvastatin and simvastatin, owing to the already mentioned first-pass effect into the liver (Nezasa et al. 2003). Moreover, although lipophilic statins can easily cross the blood-brain barrier and potentially exert neuroprotection, the administration of high doses might trigger harmful effects (Wood et al. 2010).

Statin effectiveness exerted by oral route was superior compared to either intraperitoneal or subcutaneous administration. This difference might be related to the first-pass hepatic metabolism, after absorption from the gastrointestinal tract, where several bioactive metabolites may be produced. However, this finding should be interpreted with caution, because the regression analysis did not identify the route of administration as an independent variable.

In addition, we found that statin neuroprotective effect was present regardless of the cerebral ischemia model used. Restoration of cerebral blood flow achieved after transient ischemia allows a better delivery of the drug to the injured area, as well as increasing the possibility of rescuing the ischemic penumbra (Nagahiro et al. 1998). In that sense, we expected to find more favorable outcomes in transient than in permanent cerebral ischemia. On the contrary, the statin treatment has been extensively described in both models (table 1 and supplementary table 1). It has been well demonstrated that statins can act at several different levels (i.e. improving cerebral perfusion, anti-oxidant, and anti-inflammatory properties) and exert neuroprotection in both models. Nevertheless, as our analysis showed a high inter-study heterogeneity, and the method of ischemia induction was not found to be independent in the meta-regression, this result must be taken with caution and we cannot make any inferences on the effect of statin by transient or permanent ischemic model.

Extrapolating the findings from these preclinical studies to clinical ones is difficult. For example, a small fraction of the studies reported detrimental effect of statin on stroke outcomes in our analysis: 17% and 9.8% of the studies reported a negative effect size on infarct volume reduction and neurological deficit, respectively. Such an effect is usually caused by the phenomenon of “file drawer problem” (Sena et al. 2010b) by which negative results, as concluded to be of no interest, are unpublished. Therefore, statin effectiveness maybe be overstated in our meta-analysis. In fact, the funnel plot analysis revealed atorvastatin efficacy was overstated. Other limitations of inferring clinical conclusions from this analysis are the fact that the included studies were performed on rodent stroke models, which do not recapitulate all the features of human stroke and may introduce important confounders, such as anesthesia and surgical trauma (Iadecola & Anrather 2011). The lack of co-morbidities in animal models represents another way in which it fails to mimic human stroke. In our meta-analysis, most of the studies were performed in young healthy rodents and only five studies used animals with comorbidities, whereas stroke patients are old population with secondary complications (hypertension, high cholesterol). Thus, a great effort is needed to improve the quality of neuroprotection experimental research and extrapolating data from meta-analysis of animal model must be taken with caution.

The main limitation of our analysis is that we only performed an electronic searching strategy in PubMed. Even though it represents an effective strategy to search for studies, combination of other electronic sources, such as EMBASE and BIOSIS, and handsearching should be performed to adequately identify all reports. Thus, we cannot discard the fact that we have missed unpublished or awaiting publication reports in the meta-analysis, which might contribute to publication bias on the reported effectiveness of statins.

Overall, there is enough evidence to confirm the benefits of statins in ameliorating brain damage after cerebral ischemia. In addition, our meta-analysis supports a neuroprotective effect when statins are administered either as pre-treatment or post-treatment. According to this data, the ideal clinical trial should be conducted with simvastatin, given orally at high and repeated doses, as soon as possible following the stroke onset. A clinical trial designed with these characteristics is ongoing (STARS trial) and might help the translation of this information into the acute clinical practice.

Acknowledgements

L.G-B is recipient of a Sara Borrell grant, M.C holds a BEFI grant and A.R. is supported by the Miguel Servet programme (CP09/00265) from the Spanish Ministry of Health (Instituto de Salud Carlos III). This work has been supported by research grants from the Fondo de Investigaciones Sanitarias (FIS 08/0481 and EC07/90195) and the Spanish stroke research network RENEVAS (RD06/0026/0010). The research leading to these results has received funding from the European Union’s Seventh Framework Programme (FP7/2007-2013) under grant agreements n° 201024 and n° 202213 (European Stroke Network). The authors would like to thank Malcolm Macleod and Emily Sena from CAMARADES group for their advice and for helping the set up the database. The authors declare no conflict of interest.

Ancillary