A meta-analysis was used to determine whether administering recombinant granulocyte colony-stimulating factor (rG-CSF) to neonates with bacterial septicemia reduces mortality. Five studies were identified, involving 73 rG-CSF recipients and 82 control subjects. Mortality was lower among the rG-CSF recipients (odds ratio, 0.17; CI, 0.03-0.70; P <.05). However, when the non-randomized studies were excluded, the P value was.13. For the subgroups "<2000 g" or "neutropenia," the P value was <.02. Thus the routine use of rG-CSF cannot be recommended for all neonates with sepsis.