Many materials have been used experimentally as carriers of osteoinductive growth factors. However, there is some doubt about whether the biomechanical strength of the materials affects spinal fusion from early stages of recovery. The aim of this study was to clarify which carrier was biomechanically more effective for bone morphogenetic proteins in spinal fusion. Three biomaterials, each having a different structure and biomechanical strength, were selected as carriers of recombinant human bone morphogenetic protein-2: (1) alpha-tricalcium phosphate cement, which has sufficient biomechanical strength; (2) sintered bovine bone (True Bone Ceramics) coated by type I collagen, which is similar to artificial hydroxyapatite; and (3) type I collagen sheet. Bilateral lumbar intertransverse process arthrodeses were designed in a rabbit model. Spinal fusions were evaluated by radiographic analysis, manual palpation, biomechanics (uniaxial tensile test), and histologic analysis (hematoxylin and eosin, and Villanueva-Goldner's trichrome stains) 3 and 6 weeks after surgery; they were then compared for the three carriers. For achieving the earliest solid spinal fusion, alpha-tricalcium phosphate cement (which has good inherent strength) and True Bone Ceramics (which has good porosity to allow bone penetration) did better than plain collagen (the commonly used carrier).