The effectiveness of granulocyte and granulocyte-macrophage colony-stimulating factor (G-CSF and GM-CSF) in the treatment of febrile neutropenic cancer patients remains controversial. To assess their role in this condition, we conducted a systematic review of randomised trials published as full papers. A methodological evaluation using a specifically designed quality scale was performed before meta-analysis. Eleven trials were eligible, 8 of which were meta-analysable. The median quality score for the 11 pooled trials was 58.3% (range: 33.3%-68.8%). No significant quality difference was observed between positive (colony-stimulating factor more effective) and negative trials ( P=0.36). No quality difference was observed between the 8 meta-analysable studies and the 3 others, with respective median scores of 59.3% and 50%. No advantage was detected for the use of CSF in terms of mortality from febrile neutropenia, with a relative risk of 0.71 (95% CI 0.44-1.15). The relative risk was 0.66 (95% CI 0.39-1.13) in the G-CSF subgroup and 0.97 (95% CI 0.34-2.79) in the GM-CSF subgroup. Aggregation of the results on infection-related mortality, length of stay in hospital, fever and of neutropenia duration, antibiotic therapy adaptation and duration, superinfection rate and toxicity was not possible owing to the lack of adequate data in the publications. On the basis of this review, we cannot recommend the routine use of G-CSF or GM-CSF in established febrile neutropenia.