A Trial of Intravenous Granulocyte Colony-Stimulating Factor in Acute Ischemic Stroke
Background and Purpose—Granulocyte colony-stimulating factor (G-CSF) is a promising stroke drug candidate. The present phase IIa study assessed safety and tolerability over a broad dose range of G-CSF doses in acute ischemic stroke patients and explored outcome data.
Methods—Four intravenous dose regimens (total cumulative doses of 30–180 μg/kg over the course of 3 days) of G-CSF were tested in 44 patients in a national, multicenter, randomized, placebo-controlled dose escalation study (NCT00132470; www.clinicaltrial.gov). Main inclusion criteria were a 12-hour time window after stroke onset, infarct localization to the middle cerebral artery territory, a baseline National Institutes of Health Stroke Scale range of 4 to 22, and presence of diffusion-weighted imaging/perfusion-weighted imaging mismatch.
Results—Concerning the primary safety end points, we observed no increase of thromboembolic events in the active treatment groups, and no increase in related serious adverse events. G-CSF led to expected increases in neutrophils and monocytes that resolved rapidly after end of treatment. We observed a clinically insignificant drug-related decrease of platelets. As expected from the low number of patients, we did not observe significant differences in clinical outcome in treatment vs. placebo. In exploratory analyses, we observed an interesting dose-dependent beneficial effect of treatment in patients with DWI lesions >14–17 cm3.
Conclusions—We conclude that G-CSF was well-tolerated even at high dosages in patients with acute ischemic stroke, and that a substantial increase in leukocytes appears not problematic in stroke patients. In addition, exploratory analyses suggest treatment effects in patients with larger baseline diffusion-weighted imaging lesions. The obtained data provide the basis for a second trial aimed to demonstrate safety and efficacy of G-CSF on clinical end points.
Granulocyte-colony stimulating factor (G-CSF) is a growth factor of approximately 20 kDa that was initially identified as the main driver for generation of neutrophilic granulocytes and is in widespread clinical use for the treatment of chemotherapy-associated neutropenia. We and others have uncovered that G-CSF also acts as a potent neuronal growth factor. G-CSF acts antiapoptotically on neurons, stimulates neurogenesis, and enhances vessel reformation in the ischemic brain (1). Also, G-CSF is released by neurons in response to cerebral ischemia.1 Systemically administered G-CSF passes the intact blood—brain barrier and decreases infarct size (2). Meta-analyses of the existing data and comparison to other experimental stroke drugs strengthen confidence in the efficacy of G-CSF in rodent models of stroke.3,4 Importantly, G-CSF not only is acutely protective but also improves functional recovery after stroke, even when treatment is initiated at delayed time intervals.5
The main advantages of G-CSF in contrast to previously evaluated neuroprotective drugs in acute ischemic stroke are: (1) its multimodal activity in view of the complex pathophysiology of stroke;6 (2) its broad evidence base for neuroprotective and recovery-enhancing effects; and (3) its long clinical history in other indications with an excellent safety record. Consequently, we have initiated a phase IIa clinical trial in acute ischemic stroke patients with the aim to demonstrate feasibility and safety of intravenous G-CSF treatment in stroke and explore possible hints of efficacy on stroke outcome.
Materials and Methods
Study Type and Randomization
This study was a German multicenter, placebo-controlled, randomized, double-blind dose-escalating study. The study was conducted according to the national law following International Conference on Harmonization Good Clinical Practice and is approved by the ethics committee of Westfalen-Lippe (Münster) and local ethic committees at the study sites. Patients were randomly assigned to either placebo or verum medication in 4 sequentially escalating dose groups in the following manner: dose step 30 μg/kg, 8 verum/6 placebo; dose step 90 μg/kg, 7 verum/3 placebo; dose step 135 μg/kg, 8 verum/3 placebo; and dose 180 μg/kg, 7 verum/2 placebo (Figure 1). The same sample size was initially planned for all 4 dosages (7 patients). One patient in the 30-μg/kg dose group was replaced when severe violation of inclusion criteria was noted. The recruitment of 1 additional patient in the 135-μg/kg dose group was caused by a delay in the central randomization/allocation system.
The strategy of sequentially decreasing the number of placebo patients was chosen to guarantee a sufficient number of placebo patients for analyses, even if the study would have to be terminated for safety reasons during early dose steps. The number of patients and the scheme used ensure that a drug-related event with a frequency of 25% will have an 80% probability to be detected.
Drug Dosing and Administration
G-CSF (Filgrastim, Uni-Care, Mainz, Germany) was administered over the course of 3 days via continuous intravenous infusion. One-third of the total dose was administered as a priming dose within 20 minutes, whereas the remainder was administered as continuous infusion at a steady rate over the course of 72 hours. The dose escalation started with 30 μg/kg per body weight G-CSF (cumulative dose over the course of 3 days), followed by 90 μg/kg, 135 μg/kg, and 180 μg/kg per body weight. Between each dose step, the independent data safety monitoring board reviewed the respective safety data and approved the increase. In the highest dose group, the maximal daily dose was 100 μg/kg, which is still less than the dose for which literature safety data exist for humans (115 μg/kg per body weight).
These doses also include the dose ranges used in animal studies that were most effective in reducing infarct size and improving functional recovery after stroke.3 Study medication was provided by the pharmacy of the University of Mainz. For infusion, G-CSF was diluted with 5% dextrose.
Study Design: Inclusion and Exclusion Criteria
The main inclusion criteria were an initial National Institutes of Health Stroke Scale (NIHSS) from 4 to 22, age 18 to 85 years, and stroke localization to the middle cerebral artery territory. Patients were recruited if the onset of the stroke syndrome was within a 12-hour time window. Also recruited were patients who, when waking up with a new stroke syndrome, were last healthy at a time point <12 hours before (eg, when going to sleep or when awakening in the night; so-called wake-up strokes). Presence of mismatch was required and determined at the respective study center by visual inspection of perfusion-weighted imaging diffusion-weighted imaging (DWI) images, with no minimal size of mismatch required. Main exclusion criteria were treatment with recombinant tissue plasminogen activator, signs of very severe strokes on imaging (carotid T occlusion [intracranial carotid bifurcation occlusion with involvement of A1 and M1 segments], likely infarction of more than two-thirds of the middle cerebral artery territory, signs of midline shift), acute hemorrhagic, or lacunar infarction.
Study Logistics and Conduct
The study was conducted at 12 German centers, with a mean recruitment rate of 0.3 patients per center per month. Physicians that examined patients for outcome parameters were not involved in patient care. NIHSS was obtained at baseline and for the next 3 days. Routine laboratory tests, differential blood counts, and cytokine levels were obtained at baseline and then daily for the next 4 days. After 4 days and 4 weeks, the modified Rankin scale and Barthel Index were obtained 3 months after inclusion NIHSS, Barthel Index, and modified Rankin scale were determined. MRI data were obtained at baseline, day 4, and 3 months. MRI readings were performed using a proprietary semiautomatic reading system by 2 blinded readers plus an adjudicator. Perfusion-weighted imaging volumes were derived from relative mean transit time perfusion-weighted imaging maps.
The primary end points of the study were aimed at demonstrating safety. These were the presence of thromboembolic events (related to potentially harmful effects of raised leukocyte counts) and the distribution of serious adverse events (SAE) in placebo vs active arms. Secondary end points examined additional safety parameters, such as effects on platelets, hemorrhagic events, anaphylaxis, mortality, and infections. In addition, we examined secondary efficacy end points. These were changes in modified Rankin scale, NIHSS, and Barthel Index, as well as infarct evolution from baseline to 3 months.
Data Handling and Statistics
Missing values for NIHSS, Barthel Index, and infarct volume at day 90 were imputed by last observation carried forward. This has been performed in 5 cases for NIHSS and in 11 cases for infarct size. Analyses using last observation carried forward or observed data are explicitly identified in the text. Before unblinding, a blinded review was conducted without access to laboratory values and adverse event data, which determined the study sets (safety set, full analysis set according to International Conference of Harmonization E9, and per-protocol set), baseline covariates entered into the multiple regression model, and distribution characteristics of data and transformations used. Statistical analyses were performed using JMP 7.0.1, SAS 9.1, and NCSS 2007 (Statistical Analysis Software, Cary, NC). Efficacy analysis utilized parametric least-squares modeling procedures available in these statistics packages. For frequency analyses, Fisher exact test was used. Significance tests on individual scores (ie, laboratory values or baseline values) were performed using Welch t tests. All analyses refer to the full analysis set if not otherwise stated. Cytokine level time course from day 0 to day 4 was analyzed in JMP 7 (SAS Institute) using the function fit model, with the personality standard least squares, method restricted maximum likelihood with the model effects day, treatment, day*treatment, and subject as random effect. Reported probability values are from the day*dose effect.
Study Sets and Baseline Characteristics
A total of 44 patients were included into the study, representing the safety set (Figure 1). One patient was excluded from the full analysis set because of severe violation of inclusion/exclusion criteria (lacunar stroke and hemorrhage), which made it impossible to assess MRI basal and outcome parameters subsequently. Five patients of the full analysis set were excluded from the per-protocol set because of major protocol violations. Demographic and stroke-specific baseline characteristics were similarly distributed between active and placebo arms with no significant differences (Table 1).
Vital Signs and Laboratory Parameters
There was no treatment influence on temperature or blood pressure. G-CSF serum levels were determined at baseline (day 0) and during the next 4 days (Figure 2A). In the highest dose arm, levels increased from 107±47 pg/mL at baseline to 249 752±140 899 pg/mL on day 1, but declined rapidly after discontinuation (to 1977±1334 pg/mL at day 4).
As expected, leukocyte counts significantly increased in the 4 verum arms (Figure 2B). The mean peak was reached at day 3 and declined notably at day 4. In the 30-μg/kg dose arm, leukocyte counts increased from 12.1±6.9 nL at day 0 to 35.4±10.2/nL at day 3, and declined to 21.5±6.6/nL at day 4. At the highest dose (180 μg/kg over the course of 72 hours), counts increased from 9.5±1.9/nL to 49.4±13.5/nL and declined to 41.2±22.8/nL at day 4. Peak values reached were 84.3/nL in the highest dose arm. White blood cell increase included neutrophilic granulocytes, monocytes, and lymphocytes (Supplemental Figure I available online at http://stroke.ahajournals.org). There was no change of hematocrit or hemoglobin related to treatment (Supplemental Figure I).
We noticed a slight decrease in platelet counts in the active arms (Figure 2C). In the highest dose arm, platelet counts decreased from 244±46/nL at baseline to 195±44/nL at day 3. Platelet aggregation was not different between placebo and verum treatment.
We measured serum levels of a number of cytokines, IL-1, IL-2, IL-4, IL-6, IL-10, IL-12, tumor necrosis factor-α, and interferon-γ from day 0 to day 4. Data from 35 patients are available. Interestingly, G-CSF treatment significantly (P=0.041 for effect time*treatment by linear regression analysis) lowered interferon-γ levels in the serum and showed a visible trend toward lowering tumor necrosis factor-α and IL-12 levels over the time course of 4 days (Figure 3).
There were no drug-related effects in a number of blood chemistry parameters obtained, such as electrolytes (sodium, potassium, calcium), activated partial thromboplastin time, international normalized ratio, creatinine, or creatine kinase. Lactate dehydrogenase levels showed an increase from 207.7±69.3 U/L to 411.9±140.6 U/L at day 4, which was expected.7
Tolerability and Safety
Thromboembolic events were similarly distributed in placebo and active arms. There were 2 central nervous system thromboembolic events in the placebo group, and there were 3 in the active arms (eg, re-infarction or TIA) as reported by the study physicians. Two central nervous system thromboembolic events happened in the 30-μg/kg dose group (lowest dose group), and 1 happened in the 135-μg/kg dose group (second-highest dose group). There was 1 cardiac thromboembolic event in the placebo arm and 2 in the active arms. By analyzing presence and degree of occlusion and stenoses on MRA at baseline and day 3, we observed no difference between treatment arms.
Eight SAE were noted in the placebo arm and 34 were noted in the active arms, with several patients who accumulated multiple-linked SAE during 1 incident (Table 2). None was judged as related to drug treatment. There was no statistically significant influence of dose or treatment on the number of patients who had any SAE (Table 2, last line), or on the total number of SAE per patient. In contrast, age significantly influenced the probability of having SAE (P=0.0416 by logistic fit). SAE are listed in Supplemental Table I (available online at http://stroke.ahajournals.org). During the observation period, 1 patient died in the placebo arm (14 patients), and 3 died in the active arms (30 patients). These events occurred at least 14 days after the end of study treatment. Causes of death were cardiopulmonary and judged as not related to study medication. Mean age of patients with fatal outcome was 80 years.
A total of 196 adverse events (AE) was noted, 42 in the placebo group and 156 in the active arms. If the category “disorders of the blood and lymphoid system” is omitted, which notes all leukocytoses (32 AE), then 2.9 AE per patient are noted in the placebo groups and 4.3 AE per patient are noted in the active arms. The number of AE in the highest (180 μg/kg) dose group is 3.4 AE per patient without the category “disorders of the blood and lymphoid system” and 4.6 AE per patient overall. Known adverse effects of G-CSF that could account for the higher AE rate include bone pain, headache, and gastrointestinal symptoms.
As expected with the small number of patients included, no effect on outcome parameters was detected when simply comparing mean scores and distributions after 3 months for the final modified Rankin scale (2.57 for placebo vs 2.76 in the active arms), for the NIHSS (4.3 for placebo; 5.7 for active arms), the Barthel Index (72.1 for placebo; 67.8 for active arms), or infarct volumes (41.6 vs 43.3 cm3).
We conducted exploratory outcome analyses using a model based on strongly influential factors identified at blinded data review (NIHSS at baseline, age, and diffusion volume at baseline). The first 2 factors have been identified before as predictors of stroke outcome in large cohorts.8 Diffusion volume at baseline is a strongly influential variable on final lesion volume. In this model, we detected that the interaction term of dose*diffusion volume at baseline significantly influenced all clinical outcome parameters at 3 months (modified Rankin scale, NIHSS, Barthel Index; P<0.01) and had a trend influence on infarct evolution (P=0.11), implying that the dose-response of drug is dependent on the volume of the initial diffusion deficit. In patients with infarct volumes more than ≈14 to 17 cm3 G-CSF appears to improve final clinical outcome (or 10 to 12 cm3 when using measured serum levels as cofactor in our statistical model instead of treatment). Figure 4 shows the dose-dependent effect of G-CSF in our model in comparison to the effects of initial NIHSS and age at a DWI volume of ≈55 cm3. Bootstrap analyses confirmed the statistical stability of the model.
A potential explanation for the dependence of drug action on DWI volume may lie in the highly significant increase in cortical volume fraction with increasing infarct volume (P=0.016). Of note, G-CSF appeared to have a much stronger effect at cortical volumes in contrast to subcortical volumes (P=0.0634 for the full analysis set last observation carried forward dataset; P=0.0236 for the actually obtained MRI data [n=32] at day 90; not significant for subcortex). In animal models, effects of G-CSF on cortical volumes are stronger than those on subcortical volumes.5
Two pilot studies have been published exploring low doses of G-CSF in acute9 or subacute stroke patients.10 The current trial is the first systematic assessment of high intravenous doses of G-CSF in a dose-escalation study.
The major concern with G-CSF therapy in stroke is the strong hematologic effects of the drug. Elevation of leukocytes may impact negatively on brain inflammation or blood flow. In addition, effects of G-CSF on platelets might influence coagulation or alter bleeding risks.
The dosing scheme chosen here led to marked increases of white blood cells. However, we never reached a prespecified level of 85/nL, which would have led to discontinuation of the drug. Also, leukocyte counts did rapidly decrease after end of therapy. The decrease in platelet counts with G-CSF treatment was clinically negligible. Higher leukocyte counts in our study were not associated with increased risk of SAE or worse outcome after stroke.
An additional concern that guided our safety end points was potential direct influences of G-CSF on vessel walls. We did not observe an increase of thromboembolic events at high doses of G-CSF. Also, careful statistical analyses could not detect a negative influence on the development of vessel occlusions/stenoses from baseline to day 3.
To obtain data on the systemic inflammatory status, we monitored levels of a number of cytokines, most of which were not significantly altered by G-CSF treatment. However, we detected a lowering of interferon-γ levels with verum treatment. Interferon-γ is a proinflammatory cytokine and part of the Th1 pattern. Interferon-γ appears to have a detrimental role in experimental stroke models.11 The decrease associated with verum treatment might be caused by direct central nervous system effects of G-CSF (eg, decrease in inflammatory responses and less interferon-γ production, as suggested by animal work12) or by systemic immunomodulatory effects by changing leukocyte immunoprofiles.
Therefore, it appears that G-CSF–induced leukocytosis is not harmful in stroke patients. An increase in cellular immunocompetence may even be beneficial in stroke.13 Moreover, G-CSF may change the systemic inflammatory state after stroke to a more beneficial profile.
We could not detect clear efficacy signals from the low number of patients. However, we did observe an interesting dependence of drug action on DWI volume in an exploratory model, with an apparent benefit for patients with volumes >14 to 17 cm3. Although exploratory in nature, and based on only 43 patients, the observed interaction of drug effect with DWI volume at baseline on functional outcome after 3 months was statistically stable for all clinical outcome scores examined (and different transformations) and confirmed by bootstrap analyses. A speculative but attractive explanation for the dependence of the action of G-CSF and infarct size may lie in the greater effects of G-CSF on cortical volumes and the increasing content in cortex of larger infarcts.
A common recommendation in the stroke field for clinical studies now is to increase the lower inclusion limits as measured by the NIHSS. Effects of all lytic or brain-protective stroke drugs are primarily targeted at rescuing brain tissue, which is not strongly correlated with NIHSS. Therefore, it may be principally advantageous to use the size of biological damage as an additional entry criterion in acute stroke trials.
Based on the data of this trial, a large phase II trial (AXIS-2) has been initiated aimed at confirming safety of intravenous G-CSF in a larger population of stroke patients and demonstrating efficacy on clinical end points. This study will include 328 patients with an initial DWI >15 cm3 and test a cumulated dose of 135 μg/kg per body weight against placebo.
The authors thank the patients who were recruited and screened for this study. The authors thank all participating staff of the hospitals for their help in conducting the study, especially the study nurses, D. Schulte-Buehne, D. Urban, P. Schnitzer, D. Otto, and S. Schulte.
Sources of Funding
The study was funded by SYGNIS Bioscience.
W.R.S., R.K., A.S., and S.S. are inventors on patent applications claiming the use of granulocyte colony-stimulating factor for the treatment of stroke. W.R.S. has received financial compensation as principal investigator. S.S., M.F., W.R.S., W.H., and J.G. received financial compensation from SYGNIS Bioscience for consulting.
AXIS study organization. Data safety monitoring board (DSMB): Ingeborg Walter-Sack, Gerlinde Egerer, Darius Günther Nabavi. Safety officer: Carl Kirchmaier. Clinical advisory board: Marc Fisher, James Grotta, Werner Hacke. Principal investigator (LKP): Wolf-Rüdiger Schäbitz. Monitoring and data collection: KKS Heidelberg, Germany. Statistical analysis: HAAPACS, Germany. Image acquisition procedures and analysis: Theralys, Lyon (Luc Bracoud, Chahin Pachai). MRI reading group: Jochen Fiebach, Olaf Jansen, Marc Hermier. Study medication: University Pharmacy Mainz, Germany (Judith Thiesen). Participating centers and investigators: see Supplemental Table II available online at http://stroke.ahajournals.org.
The online-only Data Supplement is available at http://stroke.ahajournals.org/cgi/content/full/STROKEAHA.110.579508/DC1.
- Received January 19, 2010.
- Accepted July 27, 2010.
Schneider A, Kruger C, Steigleder T, Weber D, Pitzer C, Laage R, Aronowski J, Maurer MH, Gassler N, Mier W, Hasselblatt M, Kollmar R, Schwab S, Sommer C, Bach A, Kuhn HG, Schabitz WR. The hematopoietic factor G-CSF is a neuronal ligand that counteracts programmed cell death and drives neurogenesis. J Clin Invest. 2005; 115: 2083–2098.
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.
Philip M, Benatar M, Fisher M, Savitz SI. Methodological quality of animal studies of neuroprotective agents currently in phase II/III acute ischemic stroke trials. Stroke. 2009; 40: 577–581.
Rogalewski A, Schneider A, Ringelstein EB, Schabitz WR. Toward a multimodal neuroprotective treatment of stroke. Stroke. 2006; 37: 1129–1136.
Fischmeister G, Kurz M, Haas OA, Micksche M, Buchinger P, Printz D, Ressmann G, Stroebel T, Peters C, Fritsch G, Gadner H. G-CSF versus GM-CSF for stimulation of peripheral blood progenitor cells (PBPC) and leukocytes in healthy volunteers: Comparison of efficacy and tolerability. Ann Hematol. 1999; 78: 117–123.
Weimar C, Konig IR, Kraywinkel K, Ziegler A, Diener HC, German Stroke Study Collaboration. Age and National Institutes of Health Stroke Scale Score within 6 hours after onset are accurate predictors of outcome after cerebral ischemia: Development and external validation of prognostic models. Stroke. 2004; 35: 158–162.
Shyu WC, Lin SZ, Lee CC, Liu DD, Li H. Granulocyte colony-stimulating factor for acute ischemic stroke: A randomized controlled trial. CMAJ. 2006; 174: 927–933.
Sprigg N, Bath PM, Zhao L, Willmot MR, Gray LJ, Walker MF, Dennis MS, Russell N. Granulocyte-colony-stimulating factor mobilizes bone marrow stem cells in patients with subacute ischemic stroke: The stem cell trial of recovery enhancement after stroke (STEMS) pilot randomized, controlled trial (ISRCTN 16784092). Stroke. 2006; 37: 2979–2983.
Yilmaz G, Arumugam TV, Stokes KY, Granger DN. Role of T lymphocytes and interferon-gamma in ischemic stroke. Circulation. 2006; 113: 2105–2112.
Dirnagl U, Klehmet J, Braun JS, Harms H, Meisel C, Ziemssen T, Prass K, Meisel A. Stroke-induced immunodepression: Experimental evidence and clinical relevance. Stroke. 2007; 38: 770–773.