Ceramic/polymer composite bone grafts offer the potential advantage of combining the osteoconductivity of ceramic component with the ductility of polymeric component, resulting in a graft that meets many of the desired properties for bone void fillers (BVF). However, the relative contributions of the polymer and ceramic components to bone healing are not well understood. In this study, we compared remodeling of low-viscosity (LV) ceramic/poly(ester urethane) composites to a ceramic BVF control in a sheep femoral condyle plug defect model. LV composites incorporating either ceramic (LV/CM) or allograft bone (LV/A) particles were evaluated. We hypothesized that LV/CM composites which have the advantageous handling properties of injectability, flowability, and settability would heal comparably to the CM control, which was evaluated for up to 2 years to study its long-term degradation properties. Remodeling of LV/CM was comparable to that observed for the CM control, as evidenced by new bone formation on the surface of the ceramic particles. At early time points (4 months), LV/CM composites healed similar to the ceramic clinical control, while LV/A components showed more variable healing due to osteoclast-mediated resorption of the allograft particles. At longer time points (12-15 months), healing of LV/CM composites was more variable due to the nonhomogeneous distribution and lower concentration of the ceramic particles compared to the ceramic clinical control. Resorption of the ceramic particles was almost complete at 2 years. This study highlights the importance of optimizing the loading and distribution of ceramic particles in polymer/ceramic composites to maximize bone healing. © 2016 Wiley Periodicals, Inc. J Biomed Mater Res Part B: Appl Biomater, 105B: 2333-2343, 2017.