The Efficacy of Erythropoietin and Its Analogues in Animal Stroke Models
Background and Purpose— Erythropoietin (EPO) was explored regarding its suitability as a candidate stroke drug in animal experimental studies. We performed a meta-analysis to obtain an overall impression of the efficacy of EPO in published animal experimental stroke studies and for potential guidance of future clinical studies.
Methods— By electronic and manual searches of the literature, we identified studies describing the efficacy of EPO in experimental focal cerebral ischemia. Data on study quality, EPO dose, time of administration, and outcome measured as infarct volume or functional deficit were extracted. Data from all studies were pooled by means of a meta-analysis.
Results— Sixteen studies were included in the meta-analysis. When administered after the onset of ischemia, EPO and its analogues reduced infarct size by 32% and improved neurobehavioral deficits significantly. A meta-regression suggests higher doses of EPO to be associated with smaller infarct volumes. When administered earlier than 6 hours EPO was more effective compared to a later treatment initiation. Both hematopoietic and nonhematopoietic EPO analogues showed efficacy in experimental stroke.
Conclusion— In conclusion, this analysis further strengthens confidence in the efficacy of EPO and its analogues in stroke therapy. Nonhematopoietic EPO analogues which are known to have less systemic adverse effects compared to EPO are also promising candidate stroke drugs. Further experimental studies are required that evaluate the safety of a combination of EPO with thrombolysis and whether EPO is also effective in animals with comorbidity.
EPO is a secreted 30-kDa glycoprotein responsible for proliferation, maturation, and survival of erythroid progenitor cells.1 Recently, biological functions in addition to the regulation of erythropoiesis were uncovered: EPO exerts neuroprotective effects after ischemic and traumatic brain injury.2,3 As main mechanism for this strong neuroprotective effect an autocrine-paracrine function was assumed, and indeed EPO and its receptor were identified and regulated in the central nervous system after brain injury.4 Surprisingly, the EPO receptor that mediates neuroprotection seems to be distinct from those expressed by erythroid precursors.5 Therefore, nonhematopoietic EPO analogues may represent a novel class of drugs for stroke therapy. The advantage of using nonhematopoietic EPO analogues is to avoid the stimulation of hematopoiesis and thereby the prevention of an increased hematocrit with a subsequent procoagulant status or increased blood pressure.
As reviewed by Hasselblatt, EPO was widely explored regarding its suitability as a candidate stroke drug.3 A method to obtain further information on a treatment that was only investigated in single studies with small sample sizes is pooling data by meta-analysis.6 This approach, usually applied to analyze clinical data, was recently used to combine preclinical studies of promising stroke drugs, such as G-CSF or NXY-059.7,8
The aim of the present analysis was to obtain an overall impression of the efficacy of EPO and its analogues in preclinical stroke studies by means of a meta-analysis. Particularly effects of time of first drug administration and drug dose were analyzed and the efficacy of EPO and its hematopoietic analogues were compared with nonhematopoietic EPO analogues for a potential guidance of further clinical studies.
Materials and Methods
Retrieving the Literature
We searched the databases Pubmed, Embase, and Biosis for articles from January 2000 to October 2008. The search strategy used the words “Erythropoietin” or “Epo” AND “ischemia” or “stroke” or “infarct.” The bibliographies of relevant articles were further cross-checked to search for articles not referenced in the above-mentioned databases. We also browsed abstracts of scientific meetings manually and asked senior authors of identified publications for references of other studies.
Selection of Studies and Data Extraction
Studies of EPO and EPO analogues in focal cerebral ischemia in rodents in which the outcome was assessed as infarct volume or as functional deficits were included in the analysis. For comparisons of infarct volumes the EPO dose that was administered not later than 24 hours after the onset of ischemia was considered, and volumes measured at least 24 hours after induction of ischemia were included.
Because a meta-analysis requires a reasonable number of studies, only functional tests that were frequently used in the different studies were included in the analysis. We categorized neurobehavioral tests in 3 clinically meaningful groups: (1) Neuroscore as gross neurological deficit score, (2) the Morris water maze test and the passive avoidance test for memory function, and (3) foot fault test, cylinder test, skilled reaching test, and limb placing test for limb function. When neurobehavioral deficits were assessed at different times, only the last time-point was included. For comparison of functional outcome data the cumulative EPO dose was calculated. Because neuroprotectants were assumed to be beneficial when administered before the onset of ischemia, eg, before procedures with a high risk for ischemic stroke such as carotid endarterectomy or heart surgery, we separately investigated the effect of EPO and EPO analogues on infarct size reduction after preischemic treatment initiation in animal models. When values for data were expressed graphically only, values were read off the graphics using Corel Draw version 12 (Corel Corporate Communications).
We evaluated the methodological quality of the included studies according to a previously published 11-item quality scale and categorized the studies in 3 categories (category I: 8 to 11 items, category II: 4 to 7 items, category III: 0 to 3 items).8 To analyze the impact of each quality item on the efficacy, the outcome parameters of treated animals were plotted against those in the corresponding control group (L’Abbé plots) separately for each of the quality items. In addition, we evaluated whether the preclinical EPO studies met the currently updated STAIR recommendations.9
The published results from the studies were displayed graphically and summarized by means of meta-analysis techniques separately for each of the end points infarct volume and the 3 types of functional tests.
Each trial comprises 1 control and 1 or more treatment groups. For a simple comparison of treated and untreated animals, results from all treatment groups of each study were averaged and their standard errors estimated by the inverse variance method.10 We quantified the effect of treatment by the ratio of the mean outcome in a treatment group to the mean outcome in the corresponding control group. Thus, changes in results attributable to treatment are expressed as percentages. This approach amounts to an analysis of differences of logarithmized responses,11–13 whereby individual outcomes or averaged outcomes across several treatment groups within 1 trial are logarithmized and their variances are approximated by standard techniques (delta method).
The ratio of the mean outcome in treatment groups to the mean outcome in the control group for each trial is presented in forest plots. The average ratios and their 95% confidence intervals that are shown in the forest plots are based on meta-analytic methods applied to the logarithmized ratios for a random-effects model.13,14 The results for infarct volume are additionally displayed as L’Abbé plot.
Finally, percentage changes in the outcomes attributable to dosage and timing of treatment was estimated by applying random effects meta-regression to the logarithms of the results of each treatment and control group.15 Here, only EPO and a systemic EPO administration after the onset of ischemia were considered, because EPO analogues and local administration are not comparable. Dosage was used as continuous explanatory variable, whereby administration of EPO within the first 6 hours after infarction and later administration were considered to be 2 distinguished treatments.
All analyses were carried out with SAS version 9.1 (SAS Inc). Probability values below 0.05 were considered to be significant.
Electronic search identified 37 studies (Figure 1). Nineteen studies using an ischemia model in neonatal rats were excluded. This model does not reflect human stroke, which usually occurs in the elderly. One study was excluded because infarct sizes were measured as areas and not as volumes,16 which is an unusual way for infarct size calculation and not comparable with infarct size measurements in all other studies that were included. A study using a rabbit clot embolic stroke model17 was not incorporated because the outcome was measured as weight of clot producing neurological dysfunction in 50% of the rabbits. This outcome is not comparable with outcome measurement in all other studies. Overall, the meta-analysis is based on the data of 16 articles (Table 1). In 12 studies treatment was initiated after the onset of ischemia, in 3 studies treatment was initiated before ischemia induction, and 1 study investigated pre- and postischemic treatment. Infarct size calculation was assessed in more than 340 animals, and neurobehavioral deficits were evaluated in more than 348 animals. In 2 studies the number of animals used was not mentioned. The meta-regression analysis which estimates the impact of EPO dose and time of treatment initiation on infarct volume and limb function is based on the data of 6 studies and 4 studies (only postischemic treatment initiation), respectively.
Study Quality and Impact of Quality Criteria on Efficacy
In studies where treatment was initiated after the onset of ischemia the median quality checklist score was 5 (range 3 to 8). Ranking studies by quality category did not reveal a relationship between study quality and infarct volume reduction, limb function improvement, and memory function improvement, whereas the neuroscore might be biased by study quality with lower efficacy in high-quality studies (Figures 3 and 4⇓). When plotting the various outcome parameters of treated animals against those in the corresponding control group (L'Abbé plots) separately for each of the quality items no substantial effect of a particular quality item was identified (not shown).
An analysis of how far preclinical EPO studies followed the STAIR recommendations is presented in Table 2.
Efficacy of EPO and EPO Analogues
The distribution of the circles in the lower right half of the L'Abbé plot (Figure 2) indicates a superiority of the verum treatment with respect to infarct volume. Compared with placebo, EPO and its analogues reduced infarct volumes by 32% (95% CI, 16% to 44%, Figure 3). Limb function deficits were reduced by 38% (95% CI, 27% to 47%, Figure 4). EPO and EPO analogues improved the neuroscore by 37% (95% CI, 14% to 54%) and enhanced the memory function by 37% (95% CI, 6% to 59%, Figure 4). When administered before the onset of ischemia, EPO reduced infarct volumes by 36% (95% CI, 8% to 56%).
The efficacy of EPO and hematopoietic EPO analogues were compared with EPO analogues which do not stimulate erythropoiesis (only postischemic treatment initiation). EPO and hematopoietic EPO analogues reduced infarct volumes by 22% (95% CI, −4% to 42%) and nonhematopoietic EPO analogues reduced infarct volumes by 38% (95% CI, 14% to 55%). Limb function was improved by 31% (95% CI, 14% to 44%) by EPO and hematopoietic EPO analogues and by 47% (95% CI, 25% to 62%) by nonhematopoietic EPO analogues. There were no significant differences between the different EPO derivates regarding infarct size reduction (P=0.3706) and limb function improvement (P=0.0713). Studies which assessed memory function and the neuroscore only used EPO and therefore no comparisons with nonhematopoietic EPO analogues could be performed.
To analyze the effects of EPO dose and time of first drug administration, we subdivided the studies into those where treatment was initiated within 6 hours after the onset of ischemia and those where treatment started later. Separate regression lines for the effect of dose were fitted for both groups in a joint meta-regression model. Negative slopes of the regression lines reflect a beneficial effect of EPO. For the analysis of infarct volume, EPO dose varies between 50 and 10000 IU/kg body weight. The effect of EPO is expressed as reduction of infarct volumes by 5.4% (P<0.0001) and 0.1% (P=0.9324) per 1000 IU/kg body weight of EPO dose for treatment initiation within the first 6 hours and later than 6 hours after the onset of ischemia, respectively. Analogously, an increase in EPO dose of 1000 IU/kg body weight for cumulative doses between 150 and 70 000 IU/kg improved limb function by 2.8% (P=0.0067) when applied within 6 hours after stroke. Administration of EPO later than 6 hours after stroke showed no effect (0.1% increase in limb function per 1000 IU/kg body weight, P=0.6242). A delay of treatment initiation after the first 6 hours reduced the efficacy of EPO regarding limb function improvement and infarct size reduction significantly (P=0.006 and P=0.0027, respectively). Because of the small number of studies which explored the neuroscore and the memory function a meta-regression analysis for these 2 outcome parameters was not performed.
Efficacy of EPO and EPO Analogues in Experimental Stroke
This meta-analysis demonstrates that EPO and its analogues substantially reduce infarct size and improve neurobehavioral deficits in animal models of focal cerebral ischemia. Infarct volumes were reduced by 32% and neurobehavioral deficits, which were categorized into limb function, neuroscore, and memory function, were improved by 38%, 37%, and 37%, respectively. A separate analysis of the efficacy of EPO when administered before the onset of ischemia revealed that verum treatment reduced infarct sizes by 36% compared to placebo treatment. While comparing nonhematopoietic EPO analogues with EPO and hematopoietic EPO analogues, both substance groups exhibit infarct-reducing and functional recovery–enhancing properties (38% versus 22% infarct size reduction, 47% versus 31% limb function improvement, statistically not significant). The latter conclusion is, however, limited, because it is based on an indirect comparison and because of the fact that nonhematopoietic EPO analogues were only investigated in a few studies.
The meta-regression analysis suggests higher doses of EPO to be associated with smaller infarct volumes and a significantly improved limb function. The finding of an increasing efficacy with higher doses in the meta-regression analysis strengthens the results of a single study,18 which reported an improved infarct reducing effect with higher doses of EPO. In addition to a dose–response relation, we found an effect of time of treatment: An EPO administration later than 6 hours after stroke had virtually no effect on infarct volume and limb function.
A lack of strength of a meta-analysis of experimental studies might be caused by a potential publication bias attributable to unpublished neutral or negative studies. In our meta-analysis, however, some of the included studies report nonsuperiority of EPO (eg, Esneault and Brines). In addition, a graphical analysis did not suggest the presence of a publication bias. The impression that studies with higher quality overestimate the efficacy of EPO regarding neuroscore improvement fails, because in studies with lower quality treatment was initiated earlier and at higher doses, both sufficiently explains the higher efficacy in these studies. The possibility that some negative or neutral data were not reported cannot be ruled out, particularly because a request to a senior author of an identified publication revealed that unpublished data regarding the efficacy of EPO in stroke models exist. The author was unable to disclose these data because of confidentiality reasons.
In a clinical pilot trial 40 patients were randomized to receive either EPO (33 000 IU daily for 3 days) or saline.19 Functional recovery measured by the NIH Stroke Scale and the Scandinavian Stroke Scale at day 30 was improved by ≈18% (P<0.09) and ≈14% (P<0.03), respectively. Several reasons might explain the lower efficacy in this clinical study. First, the assessment of functional recovery in animals with subtle tests, such as the skilled reaching test or the Morris water maze test, differs from the relatively global neurological outcome scales, such as the NIS Stroke Scale used in the clinical trial. Moreover, neglected quality criteria in the experimental studies as discussed below may contribute to differences between the clinical trial and preclinical studies regarding the efficacy.
Implication for Further Studies
As recently shown by a meta–meta-analysis the efficacy of a therapy can be overestimated in preclinical studies when quality items are omitted.20 In our meta-analysis a graphical analysis did not suggest the presence of a substantial impact of the various quality items on the estimate of efficacy. A reason therefore might be the relatively good overall median quality score of the analyzed studies. However, the STAIR recommendations were only incompletely followed by preclinical EPO stroke studies. Only one of the included studies investigated the efficacy of EPO in animals with comorbidity (hypertension). In future stroke studies EPO should certainly be tested in aged and diabetic animals, particularly because there is evidence that studies using healthy animals may overestimate the effectiveness of an intervention.20
The combination of EPO or an EPO analogue with thrombolysis was not investigated in experimental stroke studies so far. Because of their overlapping therapeutic time window the combination of EPO and thrombolysis is reasonable, and therefore interactions of these 2 drugs should be determined in animals. This is particularly important, because the recent clinical phase III trial obviously showed a negative interaction of both drugs with an increased mortality in the combination group compared to placebo.21,22 Death from intracranial hemorrhage occurred in approximately 4% of patients who received EPO compared to 1% of patients who received placebo.21 In contrast to these findings, which suggest a negative effect of EPO on intracerebral hemorrhage, EPO was found to be safe in hemorrhagic stroke models. An increased blood pressure or other side effects were not observed in animal models of intracerebral hemorrhage and ischemic stroke.23,24 Moreover, EPO was found to improve functional recovery in experimental studies of intracerebral and subarachnoid hemorrhage.23,25 However, more experimental studies are needed that investigate the safety and efficacy of EPO in hemorrhagic stroke.
This meta-analysis confirms efficacy of EPO and EPO analogues regarding infarct size reduction and functional recovery improvement poststroke. Nonhematopoietic EPO analogues might be as effective as hematopoietic EPO analogues. A meta-regression showed that higher EPO doses presumably exert a higher efficacy and that treatment initiation within the first 6 hours after stroke onset is more effective than a delayed treatment. However, before further clinical trials will be performed, experimental studies are required that investigate the efficacy of EPO in animal stroke models with comorbidity and the safety of a combination with thrombolysis.
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
J.M. and J.H. contributed equally to this study.
- Received April 19, 2009.
- Revision received May 3, 2009.
- Accepted May 8, 2009.
Fisher JW. Erythropoietin: physiology and pharmacology update. Exp Biol Med (Maywood). 2003; 228: 1–14.
Brines ML, Ghezzi P, Keenan S, Agnello D, de Lanerolle NC, Cerami C, Itri LM, Cerami A. Erythropoietin crosses the blood-brain barrier to protect against experimental brain injury. Proc Natl Acad Sci U S A. 2000; 97: 10526–10531.
Brines M, Grasso G, Fiordaliso F, Sfacteria A, Ghezzi P, Fratelli M, Latini R, Xie QW, Smart J, Su-Rick CJ, Pobre E, Diaz D, Gomez D, Hand C, Coleman T, Cerami A. Erythropoietin mediates tissue protection through an erythropoietin and common beta-subunit heteroreceptor. Proc Natl Acad Sci U S A. 2004; 101: 14907–14912.
Macleod MR, van der Worp HB, Sena ES, Howells DW, Dirnagl U, Donnan GA. Evidence for the efficacy of NXY-059 in experimental focal cerebral ischaemia is confounded by study quality. Stroke. 2008; 39: 2824–2829.
Minnerup J, Heidrich J, Wellmann J, Rogalewski A, Schneider A, Schabitz WR. Meta-analysis of the efficacy of granulocyte-colony stimulating factor in animal models of focal cerebral ischemia. Stroke. 2008; 39: 1855–1861.
Fisher M, Feuerstein G, Howells DW, Hurn PD, Kent TA, Savitz SI, Lo EH. Update of the Stroke Therapy Academic Industry Roundtable Preclinical Recommendations. Stroke. In press.
Egger M, Davey Smith G, Altman DG Systematic Reviews in Health Care: Meta-Analysis in Context. London: BMJ Books; 2001.
Hartung J, Knapp G, Sinha BK. Statistical Meta-Analysis With Applications. Wiley Series in Probability and Statistics. Hoboken, NJ, USA: John Wiley & Sons; 2008.
Wang Y, Zhang ZG, Rhodes K, Renzi M, Zhang RL, Kapke A, Lu M, Pool C, Heavner G, Chopp M. Post-ischemic treatment with erythropoietin or carbamylated erythropoietin reduces infarction and improves neurological outcome in a rat model of focal cerebral ischemia. Br J Pharmacol. 2007; 151: 1377–1384.
Ehrenreich H, Hasselblatt M, Dembowski C, Cepek L, Lewczuk P, Stiefel M, Rustenbeck HH, Breiter N, Jacob S, Knerlich F, Bohn M, Poser W, Ruther E, Kochen M, Gefeller O, Gleiter C, Wessel TC, De Ryck M, Itri L, Prange H, Cerami A, Brines M, Siren AL. Erythropoietin therapy for acute stroke is both safe and beneficial. Mol Med. 2002; 8: 495–505.
Crossley NA, Sena E, Goehler J, Horn J, van der WB, Bath PM, Macleod M, Dirnagl U. Empirical evidence of bias in the design of experimental stroke studies: a metaepidemiologic approach. Stroke. 2008; 39: 929–934.
Early Communication about an Ongoing Safety Review Epoetin alpha. U.S. Food and Drug Administration; 2008.
Epoetin alpha: Todesfälle in deutscher Schlaganfall-Studie. Ärzteblatt.de; 2008.
Grasso G, Passalacqua M, Sfacteria A, Conti A, Morabito A, Mazzullo G, De VG, Buemi M, Macri B, Tomasello F. Does administration of recombinant human erythropoietin attenuate the increase of S-100 protein observed in cerebrospinal fluid after experimental subarachnoid hemorrhage? J Neurosurg. 2002; 96: 565–570.