Patellar shift ratio (PSR) is the optimal measurement for characterising lateral patellar shift and a reliable predictor of recurrent patellar dislocation

Purpose To propose a new measurement, the patellar shift ratio (PSR), for characterising lateral patellar shift; to determine its diagnostic accuracy in predicting recurrent patellar dislocation (RPD); and to identify the effect of patellar tilt and morphology on shift measurements. Methods Variables including the PSR, bisect offset (BSO), BSO adjusted by trigonometric analysis, linear patellar displacement (LPD) distance and congruence angle were measured on axial computed tomography (CT). Furthermore, PSR was classified into grades determined by the threshold generated by receiver operating characteristic (ROC) curves. Youden's index was used to identify the ideal threshold on the curve. The area under the ROC curve (AUC) values and likelihood ratios were calculated to assess diagnostic performance in predicting RPD. Correlation analysis was conducted to identify the effects of lateral tilt and the Wiberg index on lateral shift. Intra- and interobserver reliability were evaluated using the intraclass correlation coefficient. Results The study included 54 RPD patients and 54 controls matched by sex and age. The threshold for PSR was 24.3% (sensitivity, 96.30%; specificity, 85.19%). AUCs were categorised as excellent for all protocols (P < 0.001). The positive likelihood ratio of the PSR for predicting RPD was 6.50. PSR was categorised in terms of lateral trochlear quadrants consisting of 25% increments. Lateral patellar tilt showed a strong positive correlation with BSO (P < 0.001); moderate correlations with PSR, LPD distance and adjusted BSO (P <= 0.001); and a weak correlation with congruence angle (P = 0.034). The Wiberg index presented a positive moderate correlation with BSO (P < 0.001), while no correlation was found with PSR, LPD distance, or congruence angle (n.s.). All parameters showed good repeatability (intraobserver) and reproducibility (interobserver). Conclusions The PSR showed excellent diagnostic performance as indicated by likelihood ratios, measurement accuracy (being relatively resistant to the effects of tilt and the Wiberg index), and intra- and interobserver reliability. The optimal PSR threshold for predicting RPD risk was 24.3%, and the grade of PSR (normal, 1-4) was also a reliable and easily calculated predictor of RPD. This quantifying method to characterise lateral patellar shift is an alternative useful method for clinical assessments of RPD patients and for research on patellofemoral congruence.