Predicting syndesmotic injuries in ankle fractures: a new system based on the medial malleolar focus

The early establishment of the diagnosis of a syndesmotic injury is essential for treatment selection. However, such injuries may not be apparent radiographically. Previous studies have attempted to describe correlations between medial malleolar fracture geometry and syndesmotic disruption. The main objective of this study was to create predictive models for assessing syndesmotic injuries based on an originally described angle, i.e., the medial crural-focal angle (MCFA). This study included 138 ankle fractures involving the medial malleolus. Any measure from the plain radiograph that could potentially lead to the suspicion of a syndesmotic disruption was recorded, and the newly described MCFA (formed by the main line of the medial malleolus fracture and a line perpendicular to the bearing surface of the tibial plafond) was also recorded. The inter- and intraobserver reliabilities were obtained using Krippendorff's alpha coefficients. To examine the predictive abilities of every parameter, several statistical methods were applied including logistic regression, an ad hoc clinical rule, and discriminant analysis. After variable selection, we obtained the best possible logistic model. The variables that were found to be statistically significant were the MCFA, the tibiofibular clear space (TFCS) and the type of injury in the Lauge-Hansen (L-H) classification. This model was tested by cross validation, which revealed a mean percentage of correctly classified patients of 88%. A simpler and more intuitive alternative model was sought that was based solely on the influences of the MCFA and the TFCS. Our study revealed that an absence of syndesmotic disruptions when the MCFA was under 60^°, and there were no uninjured patients with tibiofibular clear space values over 6mm. Cross-validation revealed that the mean percentage of patients who were correctly classified with this model was 86%. The application of discriminant analysis to this combination of variables resulted in a function was able to correctly classify a mean of 84% of patients. In conclusion, three models that can predict syndesmotic injury using parameters from preoperative plain radiographs were obtained and validated. The MCFA measurement was in these models and found to be a reliable technique.