We studied platelet recovery in relation to graft content in CFUs and CD34+ cells in 31 patients with multiple myeloma (21) or non-Hodgkin lymphoma (10) receiving marrow-ablative therapy followed by autologous transplantation with G-CSF mobilized CD34+ cells purified from leukapheresis products. Twelve patients had prolonged post-transplantation thrombopenia (> or = 14 days): their graft contents in CD34+ cells, CFU-GM and BFU-E were significantly inferior to those of patients with rapid platelet recovery. Although numbers of infused CD34+ cells and CFU-GM or BFU-E were well correlated, the graft content in CD34+ cells was the only parameter predictive of platelet recovery (r = -0.38, p = 0.04), with a threshold of 2.5 x 10(6) CD34+ cells/kg. However, because rapid platelet reconstitution was obtained for 4 of 16 patients re-infused with < 2.5 x 10(6) CD34+ cells/kg, we investigated whether the graft CFU-MK content might be a better predictor of platelet reconstitution than the CD34+ cell content. Eighteen CD34 grafts were studied for CFU-MK content: CD34 and CFU-MK contents were weakly correlated (r = 0.52, p = 0.03), but there was no correlation between numbers of infused CFU-MK and time to platelet recovery. We conclude that, for autologous CD34 grafts, CFU-MK assays, like CFU-GM or BFU-E assays, cannot be used to predict platelet recovery. A CD34+ cell content > or = 2.5 x 10(6)/kg remains the only reliable indicator of the platelet reconstitution capacity of a CD34 graft.