The effect of granulocyte colony-stimulating factor (G-CSF) on neutropenia, infection, and cytotoxic chemotherapy administration was studied in a randomized trial in patients receiving intensive weekly chemotherapy for non-Hodgkin's lymphoma (NHL). Eighty patients (aged 16 to 71 years) with high-grade NHL (Kiel) of any stage were randomized to receive VAPEC-B chemotherapy alone (39 patients) or with G-CSF administered as a daily subcutaneous dose of 230 micrograms/m2 (41 patients). Prophylactic ketoconazole and cotrimoxazole were administered to all patients throughout treatment. The protocol specified identical dose modification and antibiotic treatment criteria bor both groups. Neutropenia (absolute neutrophil count [ANC] less than 1.0 x 10(9)/L) occurred in 15 of 41 (37%) of the G-CSF-treated patients and in 33 of 39 (85%) of the controls, giving a relative risk for control patients of 2.31 (95% confidence interval [CI], [1.51, 3.54]; P = .00001). Fever (greater than or equal to 37.5 degrees C) with neutropenia (ANC less than 1.0 x 10(9)/L) occurred in 9 of 41 (22%) of the G-CSF group and in 17 of 39 (44%) of the controls (relative risk for control, 2.26; 95% CI [1.01, 5.06]; P = .04). There were fewer treatment delays, with shorter duration (P = .01) in patients receiving G-CSF. Chemotherapy doses were reduced in 4 of 41 (10%) of the G-CSF patients and 13 of 39 (33%) of the controls (P = .01). The dose intensity of cytotoxic chemotherapy was significantly increased in patients receiving G-CSF (median of 95% in G-CSF group compared with 83% in control patients). Three vascular deaths occurred in the G-CSF group. Delays in the control group most commonly resulted from neutropenia (19 patients, compared with 2 patients in the G-CSF-treated group, P = .000007). Severe mucositis was the major dose-limiting toxicity in G-CSF-treated patients, but did not occur more frequently than in controls (15 patients in each group). Overall, patients randomized to receive G-CSF achieved a greater dose intensity than control patients, but this did not result in significant differences in drug toxicity (other than neutropenia), intravenous antibiotic usage, or hospitalization between the two groups.