STUDY DESIGN: Experimental.

SAMPLE POPULATION: Fourteen equine cadaver limbs/horses.

METHODS: Simulated fractures were repaired with 2 lag screws under 4-Nm insertion torque (linear repair). Computed tomography (CT) imaging was performed with the leg unloaded and loaded to forces generated while walking. The fracture repair was revised to include 3 lag screws placed with the same insertion torque (triangular repair) prior to CT. The width of the fracture gap was assessed qualitatively by 2 observers and graded on the basis of gap measurements relative to the average voxel size at dorsal, mid, and palmar P1 sites. Interobserver agreement was assessed with Cohen's κ. The effect of repair type, loading condition, and measurement site on fracture gap grades was evaluated by using Kendall's τ-b correlation coefficients and paired nonparametric tests. Significance was set at P ≤ .05.