To compare the accuracy of radiography and CT to diagnose anatomical surgical reduction of lateral humeral condylar fractures using an ex vivo canine model. Ten right and left pairs of thoracic limbs were obtained from fresh canine cadavers. Lateral humeral condylar fractures were created and reduced, with or without an articular step-off (0-, 1-, 2-, 3-mm in a caudal or distal direction), and stabilised using a transcondylar screw and a supracondylar pin. Mediolateral and craniocaudal radiographs of each limb were taken, followed by CT imaging in a sagittal, dorsal, and transverse plane. All images were assessed for malreduction by 11 board-certified specialists. The accuracy of assessment was modelled using multivariable mixed logistic regression, and consistency between assessors was assessed with Gwet's first order agreement coefficient (AC1). The regression indicated that the estimated mean probability of an accurate assessment of anatomical reduction was 0.83 (95% CI = 0.56-0.96) and 0.81 (95% Cl = 0.51-0.97) for radiographs and CT scans, respectively. There was no evidence that the size of the articular step influenced the probability of correctly identifying a malreduction (p = 0.18), but the correct identification of malreduction depended on its direction and the type of imaging modality (p < 0.001). Distal malreductions were reliably diagnosed for both radiographs and CT, with estimated mean probability of incorrect diagnosis of 0.005 (95% Cl = 0.001-0.037) and 0.029 (95% Cl = 0.003-0.2), respectively. However, for caudally malreduced fractures, the model estimated the mean probability of misdiagnosis by radiographs was 0.58 (95% CI = 0.5-0.76) compared to 0.012 (95% CI = 0.001-0.104) for CT scans. Overall inter-reviewer agreement was moderate to substantial (Gwet's AC1 = 0.64; 95% CI = 0.55-0.74). Agreement for distal malreductions was significantly higher than for caudal malreductions and slightly higher for CT scans than for radiographs. Highest agreement was seen in reviewers using CT scans to correctly identify malreductions or radiographs for identifying distal malreductions. Radiographs and CT were accurate for detecting distal malreductions and identifying anatomically reduced fractures. Radiographs incorrectly assessed caudal malreduction in more than half the cases. Overall, there was good inter-observer agreement between the two types of imaging, except when using radiography to identify caudal malreductions. CT was found to accurately assess articular step-offs in both directions (caudal, distal), showing that this modality is more reliable than radiographs for assessing multidirectional articular misalignments. GLMM: Generalised linear mixed model; Gwet's AC1: Gwet's first order autocorrelation coefficient; MCP: Medial coronoid process; PTOA: Post-traumatic osteoarthritis.
Lukaszewicz et al. (Sun,) studied this question.