This study was designed to test the hypothesis that administration of granulocyte colony-stimulating factor (G-CSF; filgrastim) during induction chemotherapy with CHOP (cyclophosphamide, vincristine, doxorubicin, prednisone) or CNOP (doxorubicin replaced with mitoxantrone) in elderly patients with aggressive non-Hodgkin lymphoma (NHL) improves time to treatment failure (TTF), complete remission (CR) rate, and overall survival (OS). Furthermore, the efficacy of CHOP versus CNOP chemotherapy was compared. A total of 455 previously untreated patients older than 60 years with stages II to IV aggressive NHL were included in the analysis. Patients (median age, 71 years; range, 60-86 years) were randomized to receive CHOP (doxorubicin 50 mg/m(2)) or CNOP (mitoxantrone 10 mg/m(2)) with or without G-CSF (5 microg/kg from day 2 until day 10-14 of each cycle every 3 weeks; 8 cycles). Forty-seven patients previously hospitalized for class I to II congestive heart failure were randomized to receive CNOP with or without G-CSF (not included in the CHOP versus CNOP analysis). The CR rates in the CHOP/CNOP plus G-CSF and CHOP/CNOP groups were the same, 52%, and in the CHOP with or without G-CSF and CNOP with or without G-CSF groups, 60% and 43% (P <.001), respectively. No benefit of G-CSF in terms of TTF and OS could be shown (P =.96 and P =.22, respectively), whereas CHOP was superior to CNOP (TTF/OS P <.001). The incidences of severe granulocytopenia (World Health Organization grade IV) and granulocytopenic infections were higher in patients not receiving G-CSF. The cumulative proportion of patients receiving 90% or more of allocated chemotherapy was higher (P <.05) in patients receiving G-CSF. Concomitant G-CSF treatment did not improve CR rate, TTF, or OS. Patients receiving CHOP fared better than those given CNOP chemotherapy. The addition of G-CSF reduces the incidence of severe granulocytopenia and infections in elderly patients with aggressive NHL receiving CHOP or CNOP chemotherapy.