Author(s): VOLMER, J., ABERMANN, E., HOSER, C., FINK, C., FEDEROLF, P., RASCHNER, C., HOLLAUS, B., Institution: MANAGEMENT CENTER INNSBRUCK, Country: AUSTRIA, Abstract-ID: 871

INTRODUCTION:
The decision to perform a lateral extra-articular tenodesis (LET) alongside anterior cruciate ligament reconstruction (ACLR) is an important factor in reducing ACL re-rupture rates[1]. Although effective in reducing re-rupture rates, the LET procedure comes at the cost of increased surgery time, increased pain, a possible decreased range of motion of the knee and could increase the risk of osteoarthritis in the lateral compartment[2]. Hence, it is important to identify the cases that would benefit from a LET. To this end, we developed a machine learning based LET score, identified the key patient characteristics that influence this score and showed the potential of the LET score to identify re-rupture cases within a cohort that received ACLR without LET.
METHODS:
A retrospective analysis was conducted on 547 patients who underwent ACLR between January 2021 and March 2024. Demographic, clinical and radiological data, including age at surgery, pivot shift test results and posterior tibial slope (PTS), were collected. Five machine learning models, including Random Forest and XGBoost, were trained using a 10-fold cross-validation approach. SHAP (Shapley Additive Explanations) values were computed to quantify the contribution of individual features to the LET score.
RESULTS:
The Random Forest and XGBoost models achieved high predictive performance, with an AUC of 0.90 ± 0.02, demonstrating their reliability in predicting LET decisions. The key factors influencing LET predictions included high-grade pivot shift tests, younger age at surgery, and revision procedures, with these characteristics contributing up to 30% to the LET score. Patients who experienced re-ruptures within two years had significantly higher predicted LET scores compared to those without re-ruptures (mean score: 0.49 vs. 0.23 respectively, p=0.011 ± 0.01). These results suggest that the machine learning models potentially captured latent risk factors associated with re-rupture, beyond the explicit decision labels provided by surgeons.
CONCLUSION:
The machine learning models demonstrated high performance in predicting LET decisions and identified key characteristics of patients that influence surgical outcomes. These findings provide a quantitative framework for optimising LET decisions, potentially reducing re-rupture rates. Future studies should validate these models across diverse datasets to improve generalisability and clinical applicability.
References
1. Getgood, A. M. J. et al. Lateral Extra-articular Tenodesis Reduces Failure of Hamstring Tendon Autograft Anterior Cruciate Ligament Reconstruction: 2-Year Outcomes From the STABILITY Study Randomized Clinical Trial. Am. J. Sports Med. 48, 285–297 (2020).
2. Castoldi, M. et al. A Randomized Controlled Trial of Bone–Patellar Tendon–Bone Anterior Cruciate Ligament Reconstruction With and Without Lateral Extra-articular Tenodesis: 19-Year Clinical and Radiological Follow-up. Am. J. Sports Med. 48, 1665–1672 (2020).