The properties of impacted morsellised bone graft (MBG) in revision total knee arthroplasty (TKA) were studied in 12 horses. The left hind metatarsophalangeal joint was replaced by a human TKA. The horses were then randomly divided into graft and control groups. In the graft group, a unicondylar, lateral uncontained defect was created in the third metatarsal bone and reconstructed using autologous MBG before cementing the TKA. In the control group, a cemented TKA was implanted without the bone resection and grafting procedure. After four to eight months, the animals were killed and a biomechanical loading test was performed with a cyclic load equivalent to the horse's body-weight to study mechanical stability. After removal of the prosthesis, the distal third metatarsal bone was studied radiologically, histologically and by quantitative and micro CT. Biomechanical testing showed that the differences in deformation between the graft and the control condyles were not significant for either elastic or time-dependent deformations. The differences in bone mineral density (BMD) between the graft and the control condyles were not significant. The BMD of the MBG was significantly lower than that in the other regions in the same limb. Micro CT showed a significant difference in the degree of anisotropy between the graft and host bone, even although the structure of the area of the MBG had trabecular orientation in the direction of the axial load. Histological analysis revealed that all the grafts were revascularised and completely incorporated into a new trabecular structure with few or no remnants of graft. Our study provides a basis for the clinical application of this technique with MBG in revision TKA.