263 patients (median age 65+/-10 years) with multiple myeloma were treated with cyclophosphamide-prednisone. Out of this cohort, 103 patients had progressive disease and were randomly assigned to either VAD (vincristine, doxorubicin, dexamethasone; 50 cases) or VMBCP (vincristine, BCNU, cyclophosphamide, melphalan and prednisone; 53 cases). There were no statistical differences between the two groups with the respect to clinical, biological and radiological parameters. There was no difference in survival between the VAD and VMBCP groups. The 4 months response rate was similar in the two groups (50% VAD, 56% VMBCP). With multivariate analysis for survival (Cox model), two factors had a statistically significant impact: Karnofsky index (> 60) and albuminaemia (< 34 g/l). With both Karnofsky index > 60 and albuminaemia > or = 34 g/l, the median survival was 29 months v 2 months with a Karnofsky index < or = 60 and albuminaemia < 34 g/l (P<0.05). In conclusion, VAD or VMBCP had similar activity for salvage treatment in MM refractory or relapsing to first-line treatment with cyclophosphamide-prednisone.