To identify an ideal carrier for rhBMP-2 during spinal fusion surgery, we produced a posterolateral rats lumbar fusion model to compare the bone induction potentials of rhBMP in collagen sponge(CS) and rhBMP-2 mixed with insoluble porcine collagen(PC) gel. Forty Sprague-Dawley rats(350-400 grams) subjected to bilateral lumbar intertransverse process fusion were allocated to CS Groups(Groups 1 to 4) or PC gel Groups(Group 5 to 8). Group 1 animals were implanted with CS and injected with 1 ?g rhBMP-2; Group 2 with CS plus 3 ?g rhBMP-2; Group 3 with CS plus 5 ?g rhBMP-2; Group 4 with CS plus 10 ?g rhBMP-2; Group 5 with PC gel containing 1 ?g rhBMP-2; Group 6 with PC gel plus 3 ?g rhBMP-2; Group 7 with PC gel plus 5 ?g rhBMP-2; and Group 8 with PC gel plus 10 ?g rhBMP-2. Spinal fusion masses were evaluated by manual palpation, radiographic assessment, microcomputer tomographic scanning, and histologically at 8 weeks after surgery. Bone fusion was radiologically confirmed in 50% in Group 3, 66% in Group 4, 16.7% in Group 5, 20% in Group 6, 80% in Group 7, and 100% in Group 8. Treatment with PC gel plus rhBMP-2 resulted in higher spinal fusion rates than CS plus rhBMP-2 based on manual palpation, microCT, and histologic examinations(P < 0.01). In this study, the mixing of gelatin gel and rhBMP-2 resulted in more effective lumbar intertransverse process fusion than injecting collagen containing rhBMP-2.