A Risk Prediction Model for Redislocation Following First Time Lateral Patellar Dislocation (LPD)

Abstract

Objectives:There is not a formula that is regularly used to help predict the risk of redislocation for an individual patient following first time lateral patellar dislocation (LPD). Our objective was to develop a model to help predict the risk of redislocation that can be easily applied in a clinic visit following first time LPD.Methods:Between 2008-2012, patients were prospectively identified at 2 musculoskeletal outpatient clinics. Patients were included if they had a history, physical exam, & magnetic resonance imaging (MRI) consistent with LPD without other significant ligamentous injury. Multiple anatomic & injury variables were obtained from the MRI including tibial tubercle - trochlear groove distance, patellar tilt, trochlear depth, trochlear facet asymmetry, trochlear condyle asymmetry, lateral trochlear inclination angle, trochlear sulcus angle (SA), Insall-Salvati ratio (ISR), Caton-Deschamps index, patellotrochlear index, location & severity of chondral injury, location & severity of MPFL injury, location of bone bruising, & whether the patient had open or closed physes (GP). Demographic patient information was also collected to include age, sex, & whether the injury was contact or non-contact. Patients were contacted at a minimum of 2 years post injury to determine if they had experienced a redislocation injury. Demographic and MRI variables were compared between patients with and without patellar redislocation using two group t-tests for continuous variables and Fisher’s exact tests for categorical variables; the statistical analysis was completed by an independent professional statistician at our institution. Stepwise logistic regression models were used to identify potential predictors of redislocation. Initially, MRI variables that had continuous data were treated as continuous variables and then a second stepwise logistic regression model was run using binary indicator variables. A cutpoint for each variable was determined using an outcome-oriented approach; the cutpoint yielding the most significant association with the redislocation outcome in a simple logistic regression model was chosen. P-values less than 0.05 were considered statistically significant. A prediction model was then created using the statistically significant variables, and a receiver operating characteristic (ROC) curve was applied to the model to evaluate for diagnostic performance.Results:Inclusion criteria were met by 108 patients. Of all of the variables, the statistically significant variables were sex, GP, SA, and ISR. The cutpoints were SA ≥ 154 degrees and ISR ≥ 1.3. The predicted probability of redislocation (percent) was 6.4 for males and 2.3 for females if the patient had closed growth plates and normal SA and ISR. The predicted probability of redislocation (percent) was 89.9 for males and 75.7 for females if the patient had open growth plates and abnormal SA and ISR. If two of three variables (GP, SA, ISR) were positive, the risk was M70.4/F45.4 (GP, SA), M46.5/F23.3 (GP, ISR), and M72.5/F47.9 (SA, ISR). The area under the curve for the ROC curve was 0.81 indicating good diagnostic performance.Conclusion:Our model demonstrates high risk of redislocation with open growth plates and SA & ISR values above the cutpoints, whereas there are low rates of redislocation with the inverse findings. This model may serve as a clinically applicable means of predicting redislocation following first time LPD.