We evaluated the efficacy of Escherichia coli-derived recombinant human bone morphogenetic protein-2 (E-BMP-2) in a mini-pig model of spinal anterior interbody fusion. A total of 14 male mini-pigs underwent three-level anterior lumbar interbody fusion using polyether etherketone (PEEK) cages containing porous hydroxyapatite (HA). Four groups of cages were prepared: 1) control (n = 10 segments); 2) 50 μg E-BMP-2 (n = 9); 3) 200 μg E-BMP-2 (n = 10); and 4) 800 μg E-BMP-2 (n = 9). At eight weeks after surgery the mini-pigs were killed and the specimens were evaluated by gross inspection and manual palpation, radiological evaluation including plain radiographs and micro-CT scans, and histological analysis. Rates of fusion within PEEK cages and overall union rates were calculated, and bone formation outside vertebrae was evaluated. One animal died post-operatively and was excluded, and one section was lost and also excluded, leaving 38 sites for assessment. This rate of fusion within cages was 30.0% (three of ten) in the control group, 44.4% (four of nine) in the 50 μg E-BMP-2 group, 60.0% (six of ten) in the 200 μg E-BMP-2 group, and 77.8% (seven of nine) in the 800 μg E-BMP-2 group. Fusion rate was significantly increased by the addition of E-BMP-2 and with increasing E-BMP-2 dose (p = 0.046). In a mini-pig spinal anterior interbody fusion model using porous HA as a carrier, the implantation of E-BMP-2-loaded PEEK cages improved the fusion rate compared with PEEK cages alone, an effect that was significantly increased with increasing E-BMP-2 dosage.