We examined the effect of rhBMP-2 (recombinant human bone morphogenetic protein-2), delivered in a porous poly(DL-lactic acid) implant, on bone formation in a critical-sized defect in the radial diaphysis in rabbits. A unilateral segmental defect, twenty millimeters long, was created in the radius in ninety-six skeletally mature New Zealand White rabbits. Forty-eight rabbits were evaluated at four weeks and forty-eight, at eight weeks. Six groups were studied at each time-period. The defect was left empty in one group (control), the defect was filled with an autogenous corticocancellous bone graft in one group, and the defect was filled with a porous poly(DL-lactic acid) implant containing zero, seventeen, thirty-five, or seventy micrograms of rhBMP-2 (one group each). Radiographs of the defects were made every two weeks. The percentage of the total area of the defect that was radiopaque was determined with use of computerized radiomorphometry, and this percentage was used as a quantitative measure of the extent of new-bone formation in the defect. There were time and dose-dependent responses to rhBMP-2 for as long as four weeks; thereafter, the effects of seventeen, thirty-five, and seventy micrograms of rhBMP-2 were independent of dose and time (p < or = 0.05). The defects that had been treated with either thirty-five or seventy micrograms of rhBMP-2 had a significantly greater (p < or = 0.05) area of radiopacity than the defects that had been treated with either zero or seventeen micrograms of rhBMP-2. No significant difference could be found between the defects treated with thirty-five or seventy micrograms of rhBMP-2 and the defects filled with an autogenous graft. Healing and bone formation were examined histologically and histomorphometrically as well. At four weeks, polarized light microscopy revealed remnants of poly(DL-lactic acid) only in the defects that had been filled with an implant containing zero micrograms of rhBMP-2. At eight weeks, regardless of the dose of rhBMP-2, poly(DL-lactic acid) was not visible on histological examination. The presence of multinucleated giant cells was the hallmark of the inflammatory response elicited by poly(DL-lactic acid). At four and eight weeks, macrophages and lymphocytes were also present. The intensity of the cellular response at four weeks suggested an inverse relationship between these cells and the dose of rhBMP-2 -- that is, there appeared to be more multinucleated giant cells in defects treated with zero micrograms of rhBMP-2 than in defects treated with seventy micrograms of rhBMP-2. At eight weeks, multinucleated giant cells were rare in the defects treated with seventeen, thirty-five, or seventy micrograms of rhBMP-2. Histomorphometric data at four and eight weeks indicated that the amount of bone formation in the defects treated with seventeen, thirty-five, or seventy micrograms of rhBMP-2 was equivalent to the amount in the defects treated with an autogenous graft and was significantly less (p < or = 0.05) in the untreated defects and the defects treated with zero micrograms of rhBMP-2 (p < or = 0.05). By eight weeks, only thirty-five and seventy micrograms of rhBMP-2 had restored cortices and marrow elements.