Eight RCTs were included in meta-analyses (n=385). Sample sizes ranged from 18 to 114. For study quality, concealment of allocation, baseline comparability and intention-to-treat analysis were adequate in all studies. Randomisation was adequately described in most studies. Four RCTs failed to blind participants and caregivers to study allocation; all studies had adequate blinding for outcome assessment. The loss to follow-up rate ranged from 0% to 16%. Duration of follow-up ranged from one to 12 months.
There were no significant differences between the G-CSF therapy and placebo group in terms of outcomes of left ventricular ejection fraction, infarct scar size, left ventricular end-diastolic volume and left ventricular end-systolic volume.
Subgroup analyses showed that G-CSF therapy was associated with a significant increase in left ventricular ejection fraction compared with placebo (WMD 4.73, 95% CI 2.67 to 6.79; seven RCTs) in studies with patients with mean ejection fraction less than 50% at baseline, but not in those studies with patients with mean ejection fraction at least 50% at baseline.
Subgroup analyses revealed a significant increase in ejection fraction when G-CSF therapy was initiated less than 37 hours after the acute event (WMD 4.65, 95% CI 2.51 to 6.80; five RCTs), but not when G-CSF therapy was initiated more than 37 hours after the acute event.
For the subgroup of patients with mean ejection fraction less than 50% at baseline, there were significant improvements in left ventricular end-diastolic volume (WMD -7.82, 95% CI -14.68 to -0.96) and left ventricular end-systolic volume (WMD -10.07, 95% CI -18.88 to -1.26). No significant results were found in other subgroup analyses.
There were no significant differences in rate of death, recurrent myocardial infarction and in-stent restenosis between the G-CSF therapy and placebo group either in the main analysis or in any subgroup.
A significant interaction was observed between the baseline left ventricular ejection fraction and the change in left ventricular ejection fraction (p<0.0001).
The authors stated that analyses using both fixed-effects and random-effects models yielded similar results for the major outcomes. Significant heterogeneity was observed only in the outcome of left ventricular ejection fraction when pooling all studies (I2=79.1%) and left ventricular ejection fraction for the subgroup of patients with mean ejection fraction less than 50% at baseline (I2=66.3%) and left ventricular ejection fraction for the subgroup of patients with initiation of G-CSF therapy less than 37 hours after the acute event (I2=75.2%).
The risk of publication bias was inconclusive based on the funnel plots.