Granulocyte-Colony Stimulating Factor for Mobilizing Bone Marrow Stem Cells in Subacute Stroke
The Stem Cell Trial of Recovery Enhancement After Stroke 2 Randomized Controlled Trial
Background and Purpose—Granulocyte-colony stimulating factor (G-CSF) is neuroprotective in experimental stroke and mobilizes CD34+ peripheral blood stem cells into the circulation. We assessed the safety of G-CSF in recent stroke in a phase IIb single-center randomized, controlled trial.
Methods—G-CSF (10 μg/kg) or placebo (ratio 2:1) was given SC for 5 days to 60 patients 3 to 30 days after ischemic or hemorrhagic stroke. The primary outcome was the frequency of serious adverse events. Peripheral blood counts, CD34+ count, and functional outcome were measured. MRI assessed lesion volume, atrophy, and the presence of iron-labeled CD34+ cells reinjected on day 6.
Results—Sixty patients were recruited at mean of 8 days (SD ±5) post ictus, with mean age 71 years (±12 years) and 53% men. The groups were well matched for baseline minimization/prognostic factors. There were no significant differences between groups in the number of participants with serious adverse events: G-CSF 15 (37.5%) of 40 versus placebo 7 (35%) of 20, death or dependency (modified Rankin Score: G-CSF 3.3±1.3, placebo 3.0±1.3) at 90 days, or the number of injections received. G-CSF increased CD34+ and total white cell counts of 9.5- and 4.2-fold, respectively. There was a trend toward reduction in MRI ischemic lesion volume with respect to change from baseline in G-CSF–treated patients (P=0.06). In 1 participant, there was suggestion that labeled CD34+ cells had migrated to the ischemic lesion.
Conclusions—This randomized, double-blind, placebo-controlled trial suggests that G-CSF is safe when administered subacutely. It is feasible to label and readminister iron-labeled CD34+ cells in patients with ischemic stroke.
Few interventions aid recovery for patients with acute stroke, and although earlier trials of putative neuroprotectants failed (eg, NXY-0591), new studies are ongoing, including focusing on colony-stimulating factors, such as granulocyte-colony stimulating factor (G-CSF) and erythropoietin, where promising data exist for preclinical experimental models of stroke.2 G-CSF is a glycoprotein hormone encoded by a single gene located on chromosome 17 q11-223; it plays a key role in the regulation of granulopoiesis and is responsible for the terminal maturation of neutrophils. Recombinant G-CSF is used for treatment of neutropenia and for production of CD34+ hematopoietic stem cells (HSCs) for bone marrow transplantation. When considering G-CSF as a potential treatment for stroke, the mechanisms of action appear to be multimodal. Experimentally, neuroprotection occurs by reducing apoptosis in the ischemic penumbra and attenuating the inflammatory cascade.4 It also appears to be neuroreparative through potentiating angiogenesis5 and neurogenesis,4 in part by activating brain endothelial cells6 and mobilizing HSCs to migrate to the ischemic lesion.7 More contentiously, bone marrow–derived stem cells have been shown to differentiate into neurons and glia,8 and experimental transplants of HSCs have been observed to improve outcome post stroke.9 The purpose of the present trial, Stem Cell Trial of Recovery Enhancement After Stroke 2, was to further evaluate the safety of G-CSF in stroke with emphasis on its effects on bone marrow–derived HSCs and their fate in the brain.
We performed a prospective, single-center, double-blind, randomized, placebo-controlled, phase IIb trial of G-CSF in patients with subacute stroke. The trial was conducted in accordance with the Declaration of Helsinki and the International Conference of Harmonisation of Good Clinical Practice; received authorization from the Medicines and Healthcare Products Regulatory Agency (EudraCT 2006-005345-11) and the local research ethics committee; and was a registered clinical trial (ISRCTN 63336619).
Patients 3 to 30 days after ischemic and hemorrhagic stroke were recruited from the stroke service at Nottingham Universities Hospital National Health Service Trust from July 2007 to January 2010 (for inclusion and exclusion criteria, online-only Supplemental Table 1; http://stroke.ahajournals.org). Written informed consent was obtained from each patient. If the patient was unable to consent (eg, because of confusion or dysphasia), proxy consent was obtained from a relative or carer.
Participants were randomly assigned 2:1 to receive subcutaneous human recombinant G-CSF (1×106 units/kg, equivalent to 10 μg/kg, Neupogen, Amgen, the maximum dose in Stem Cell Trial of Recovery Enhancement After Stroke 110) or matching subcutaneous saline once per day for 5 days. Treatment was prepared centrally and both the participants and those administering the injections were blinded to treatment and its assignment. Randomization involved computerized minimization (online-only Supplemental Table 1), and randomized treatment was administered in addition to best medical care.
The primary outcome measure was safety assessed as the number of participants having a serious adverse event (SAE) by day 90. Clinical secondary outcomes included tolerability, feasibility, impairment (National Institutes of Health Stroke Scale), grip strength, dependency (modified Rankin Score), disability (Nottingham Extended Activities of Daily Living and Barthel Index), cognition (Mini-Mental State Examination), and mood (Zung depression score); measurements were made at days 10 and 90. All of the assessments were blinded to treatment assignment. The Data Monitoring Committee assessed unblinded safety data after recruiting 20 and 40 patients into the trial.
Peripheral blood CD34+ cell count was measured at day 5 using flow cytometry (FACScalibur, Becton Dickinson, Oxford, UK) and complied with International Society for Hematotherapy and Graft Engineering guidelines.11 Full blood counts and putative markers of neuroprotection efficacy (d-dimer, B-natriuretic peptide, matrix metalloproteinase 9, and protein S-100β; Triage Stroke Panel, Biosite Inc, San Diego, CA) were assessed on days 0, 5, and 10.
MRI of the brain was performed at baseline and day 90 (±7) in patients with ischemic stroke, except in those who were intolerant, had contraindications to MRI, or refused imaging. The volume in acutely cerebral infarcted tissue was calculated at baseline based on diffusion-weighted imaging hyperintense lesion outlines at day 0 and at day 90, at which time point intermediately T2-weighted scans were used (online-only, Supplemental Table 2). Contralateral ventricular volume on diffusion-weighted images (including the lateral ventricles to the level of the third ventricle) was also assessed as an indirect measure of cerebral atrophy. Two assessors blinded to treatment allocation (T.J.E., M.A.) were trained by an experienced neuroradiologist (D.P.A.) and measured volumes independently.
CD34+ Cell Labeling
All of the patients were offered the opportunity to take part in the HSC labeling substudy. For this procedure, 150 mL of blood was collected on day 6 and processed in an HTA licensed clean room. CD34+ cells were immunomagnetically separated from the blood using a CliniMACS Separator (Miltenyi Biotec, Bergisch Gladbach, Germany). The CD34+ antibody used for immunoseparation contained a dextran-coated iron-oxide nanobead,12 effectively labeling the cells and allowing visualization with gradient echo (T2*-weighted) imaging.7 After labeling, the CD34+ cells were reinjected intravenously back into the donor patient on the same day and tracked with MR T2* imaging on day 10 (echo time: 16.11 ms; repetition time: shortest; slice thickness: 3 mm). These images are highly sensitive to iron and, therefore, to iron-labeled cells. A neuroradiologist blinded to treatment allocation and cell counts interpreted prelabeling and postlabeling scans.
The primary outcome was the number of participants with ≥1 SAE; this was chosen because G-CSF was associated with a nonsignificant increase in SAEs in our earlier phase IIa trial.10 The sample size was calculated as 60 assuming α5%, power 80%, SAE rate G-CSF 50% versus placebo 20%, ratio G-CSF:placebo 2:1, and losses 5%. Data were analyzed using Fisher exact test, Mann-Whitney U test, repeated-measures ANOVA, or ANCOVA as appropriate. Data are presented as number (percentage), mean (±SD), or median (interquartile range). Stroke lesion volume data were positively skewed; it was, therefore, transformed using log(10) to normalize the distribution and antilogged to present the geometric mean. All of the analyses were performed using PASW statistics data editor (version 18.0) and MediStat; analysis was by intention to treat (according to randomization), and statistical significance was taken at P<0.05.
Of 205 screened participants who were eligible, 60 were included, and the groups (40 G-CSF and 20 placebo) were well matched at baseline (Figure 1 and Table 1). The first dose of trial drug was received, on average, 10 days after stroke onset. The mean number of treatment doses received was 4.5 (90.5% in total) in participants randomized to G-CSF and 5.0 (100%) for placebo; 2 participants randomized to G-CSF refused any treatment (but agreed to continued follow-up) before they received their first dose. Other participants stopped treatment after receiving G-CSF because of lower back pain (n=1, 2 of 5 doses received), acute illness not related to the trial drug (n=1), and administration error (n=2, missed doses). One patient withdrew from the trial; although he or she refused follow-up, he or she allowed information on vital status and modified Rankin Score to be obtained from the general practitioner.
No significant difference in the number of patients with SAE (the primary outcome) or in the mean number of SAEs and adverse events per patient was observed between treatment groups (Table 2 and online-only Supplemental Table 3). There were 3 deaths in the G-CSF group (pulmonary embolus, ischemic bowel, and community pneumonia), but all occurred after the end of treatment phase; no deaths occurred in the placebo group. There were no differences in vascular events (nonfatal stroke, nonfatal myocardial infarction, pulmonary embolus, and deep vein thrombosis) or infection rates by day 90 (Table 2). Fifteen percent (6 G-CSF and 3 placebo) of the participants were included with a primary intracerebral hemorrhage, and no safety concerns were present in this subgroup (online-only Supplemental Table 3).
No significant difference emerged between treatment and placebo groups with respect to measurements of dependency (modified Rankin Score), disability (Barthel Index and Nottingham Extended Activities of Daily Living), impairment (National Institutes of Health Stroke Scale, Motoricity Index, and grip strength), cognition (Mini-Mental State Examination), or mood (Zung depression score; Table 2). When modified Rankin Score is dichotomized (poor outcome >2), no difference is observed between groups (G-CSF 28 of 40 versus placebo 14 of 20; P=1.0).
Stroke lesion volumes at baseline and day 90 were analyzed in 20 participants (14 G-CSF and 6 placebo; Table 2). Scans were not indicated in 9 recruits (primary hemorrhage) and not performed because of patient refusal (n=17), contraindication (n=4), scan intolerance (n=1), and severe illness or death (n=5). Four scans were not analyzable (eg, because of poor quality or the acute lesion becoming embedded in chronic ischemic change). Stroke lesion territory and size varied considerably, but there was no significant difference in baseline diffusion-weighted imaging lesion volume (Table 2). When adjusted for baseline diffusion-weighted imaging lesion volume, a trend toward reduced lesion volume at day 90 was present in G-CSF–treated patients (ANCOVA, P=0.06). With respect to change in ventricular volume over 90 days, no significant difference was seen between groups (Table 2).
In comparison with placebo, G-CSF significantly elevated CD34+ cell count (peak: 31.4 versus 3.3 cells per microliter), white cell count (peak: 40.2×109 versus 9.5×109 cells per liter), and white cell components, including neutrophils (peak: 34.1×109 versus 6.9×109 cells per liter), at day 5 (online-only Supplemental Table 4 and Figure 2). By day 10, the counts were returning to normal. There were apparent absolute differences between groups in baseline values in B-natriuretic peptide and d-dimer, but they did not reach statistical significance (P=0.08 and 0.07, respectively). Change in hemoglobin, platelet count, red cell count, hematocrit, d-dimer, B-natriuretic peptide, matrix metalloproteinase 9, and protein S100-β did not differ significantly between the treatment groups.
CD34+ Cell Labeling
A phantom model was used to demonstrate that iron-oxide microbead-labeled CD34+ cells (harvested from a volunteer) dispersed in 0.5% agar (4×103 cells per mL) were visible on 3T T2* magnetic resonance scanning. Agar alone and unlabeled CD34 cells were not detectable (Figure 3).
Eight participants (G-CSF 6 and placebo 2) with an ischemic event underwent CD34+ cell labeling. There was marked variation in each participant's response to G-CSF; after 5 days of treatment, between 50 and 430×104 cells were harvested from 150 mL of whole blood. Participants receiving placebo yielded between 2 and 7×104 CD34+ cells. Labeled cells were then introduced intravenously on day 6. One G-CSF-treated participant developed a hypodensity compatible with iron deposition within his or her left gangliocapsular infarct that was evident on both day 10 and 90 T2* scans (Figure 3). The hypodensity was not present at baseline, and this participant received the highest number (430×104) of labeled CD34+ cells with administration 14 days after stroke onset. Three participants (G-CSF 2 and placebo 1) had evidence of hypodensities compatible with hemorrhagic transformation at days 0, 10, and 90 (T2*-weighted images). The remaining 4 participants receiving labeled cells had no hypodensities in their infarct zones at any time point.
The Stem Cell Trial of Recovery Enhancement After Stroke 2 was designed to test the safety of G-CSF and explore mechanisms by which it might work in participants with subacute stroke. There were no significant differences in SAEs (primary outcome) between 2 well-matched groups, supporting that treatment with G-CSF is safe. The inclusion of 9 hemorrhagic strokes also provides initial safety data, although further phase II trials designed specifically in this subgroup are warranted. The majority of subcutaneous injections were received suggesting that the treatment is also feasible to administer, at least when given subacutely. Adverse events experienced that were probably or definitely related to the study drug were typical expected adverse effects of G-CSF (musculoskeletal pain and dizziness, see online-only Supplemental Table 3). Despite significant increases in white cell count, no difference in the frequency of vascular events (arterial and venous) or infection was seen. Two of 3 deaths in the G-CSF group were vascular in nature and did not appear to be secondary to active treatment; the overall vascular event rates were identical in each group (15%).
A secondary aim of the trial was designed to establish whether G-CSF mobilized iron-labeled CD34+ cells migrate to the site of the stroke lesion and, therefore, contribute to neurorepair, as seen experimentally.13 The complexity of recovering, labeling, and readministering CD34+ cells meant that we only studied 8 participants. Of these, T2* MRI suggested the presence of migrated-labeled cells in 1 participant, as demonstrated by new signal loss within the infarct on days 10 (4 days after reinjection) and 90 and without evidence of baseline hemorrhagic transformation (at day 14 after stroke). This followed intravenous injection of 430×104-labeled CD34+ cells (the highest number recovered from any participant). There was no evidence of T2* signal loss in 4 participants, and baseline hemorrhagic transformation occurred in 3. Bone marrow–derived stem cells have been tracked successfully in animal models of stroke using extracellular and intracellular iron labels,7,14 a method that confers visibility on gradient echo MRI. Nevertheless, qualitative interpretation of T2* images was limited by scan quality and preexisting signal loss (eg, because of hemosiderin deposition from previous bleeds). New petechial or parenchymal bleeding from hemorrhagic transformation further limited specificity of interpretation of new signal loss. Future mechanistic cell-labeling trials should exclude patients with any degree of baseline hemorrhage and consider further the dose of cells administered an alternative route (eg, intra-arterial) and using an intracellular cell label.15
Although no significant differences between groups were seen with respect to functional outcome (and the trial was not powered to detect this), a trend toward improvement was observed in change in the National Institutes of Health Stroke Scale over 90 days in G-CSF–treated subjects. This is consistent with other measured clinical parameters (Motoricity Index, grip strength, and Nottingham Extended Activities of Daily Living), whereby the mean value in the G-CSF group was nonsignificantly better when compared with placebo. Further phase III trials are, therefore, required to test efficacy. Similarly, a trend to reduced stroke lesion volume and growth in the G-CSF group was observed. This is unexpected considering that G-CSF administration in the subacute phase is aimed to improve outcome via promotion of neurogenesis4 and angiogenesis5 and not through neuroprotection. This should be interpreted with caution because of the small sample size and heterogeneity of case mix, although this was accounted for, in part, by adjusting for baseline stroke volume. In addition, preclinical concern on the effects of G-CSF on cerebral atrophy16 was not demonstrated here; subgroup analysis revealed no significant difference between groups in change in contralateral ventricular volume, although the numbers are small and 90 days may be an insufficient follow-up period.
Altogether, 4 trials have evaluated G-CSF in ischemic stroke, this and the first Stem Cell Trial of Recovery Enhancement After Stroke, as well as 2 others.17,18 Of these, the first assessed 10 patients (7 in the G-CSF arm dosed at 15 μg/kg OD for 5 days, 3 in the control arm) randomized within 7 days of ictus and no safety concerns were raised.17 In the Treatment with AX200 for Acute Ischemic Stroke Trial (n=44), where treatment was instituted with 12 hours of stroke onset at a cumulative dose range of 30 to 180 μg/kg,18 there was no significant difference between treatment groups with respect to SAEs and thromboembolic complications. Although G-CSF transiently increases total white cell count, therefore potentially leading to vascular complications, there is currently no evidence to suggest that G-CSF causes thromboembolic events or aggravates stroke symptoms.19 All 4 of the trials were too small individually to assess the effect of G-CSF on functional outcome; when data from the 4 trials are aggregated, no safety concern is obvious with G-CSF (odds ratio: 1.11 [95% CI: 0.38–3.25]; Figure 4). In light of the above evidence and a recent negative trial assessing administration of erythropoietin (another colony-stimulating factor20) within 6 hours of ischemic stroke,21 further phase II/III trials are required evaluating the safety and efficacy of G-CSF.
In summary, this randomized, double-blind, placebo-controlled trial suggests that G-CSF is safe when administered subacutely. It is feasible to label and readminister iron-labeled CD34+ cells in patients with ischemic stroke.
Sources of Funding
The trial, including T.J.E. and part of M.A., were funded by the Medical Research Council (grant G0501997). G.S. was funded by the British Heart Foundation (grant PG/05/137). P.M.W.B. is the Stroke Association Professor of Stroke Medicine.
We thank Sally Anderson, Lucy Kerr, research nurses funded by the Stroke Research Network, and Naranjargal Dashdorj for assisting in the trial. The Data Monitoring Committee was composed of Prof Martin Dennis (Edinburgh) and Assoc Prof Nikola Sprigg.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.636449/-/DC1.
P.M.W.B. was the chief investigator for this study.
- Received August 19, 2011.
- Revision received August 19, 2011.
- Accepted September 16, 2011.
- © 2012 American Heart Association, Inc.
- Simmers RN,
- Webber LM,
- Shannon MF,
- Garson OM,
- Wong G,
- Vadas MA,
- et al
- Chen J,
- Li Y,
- Wang L,
- Zhang Z,
- Lu D,
- Lu M,
- et al
- Sprigg N,
- Bath PM,
- Zhao L,
- Willmot MR,
- Gray LJ,
- Walker MF,
- et al
- Bornhauser M,
- Theuser C,
- Soucek S,
- Holig K,
- Klingebiel T,
- Blau W,
- et al
- Shyu WC,
- Lin SZ,
- Lee CC,
- Liu DD,
- Li H
- Schabitz WR,
- Laage R,
- Schwab S,
- Schneider D,
- Hamann G,
- et al
- Ehrenreich H,
- Weissenborn K,
- Prange H,
- Schneider D,
- Weimar C,
- Wartenberg K,
- et al