When Should I Be Associating a Lateral Extra-Articular Procedure to My Anterior Cruciate Ligament Reconstruction? AI vs. Surgeon Decision-Making.

Abstract

Residual rotational instability following anterior cruciate ligament reconstruction (ACLR) remains a clinical challenge, leading to renewed interest in adjunctive lateral extra-articular procedures (LEAP). This study aimed to compare clinical decision-making between experienced orthopaedic surgeons and an artificial intelligence (AI) model regarding indications for LEAP in ACLR, to assess concordance, and to explore the potential role of AI in surgical planning. A cross-sectional comparative study was conducted using 40 hypothetical ACLR case profiles, reflecting a range of patient demographics, injury characteristics, and activity levels. An AI model trained on literature-based criteria and expert input generated binary recommendations ("perform LEAP" or "do not perform LEAP") for each case. Twenty-two high-volume knee surgeons independently reviewed all cases, blinded to AI recommendations, and indicated whether they would recommend a LEAP. Agreement between surgeon decisions and AI recommendations was calculated, and factors influencing concordance were analysed using chi-square tests, <i>t</i>-tests, and Pearson correlations (<i>p</i> < 0.05). Overall, surgeon agreement with AI recommendations was high but varied by clinical factors. A positive pivot shift test was the strongest predictor of concordance (93.9% ± 4.8 vs. 71.8% ± 25.7; <i>p</i> = 0.0004, Cohen's <i>d</i> = 1.23). Surgeons agreed more often when the AI recommended LEAP (92.7% ± 6.5) than when it advised against it (70.9% ± 27.0; <i>p</i> = 0.0006). Male patient cases yielded higher agreement (91.1% ± 7.4) compared with female cases (75.7% ± 27.0; <i>p</i> = 0.018). Ligamentous laxity (Beighton score) showed a moderate positive correlation with agreement (<i>r</i> = 0.39; <i>p</i> = 0.013), while age, revision status, associated lesions, and time from injury to surgery were not significant predictors. Surgeons demonstrated strong alignment with AI recommendations in clear-cut scenarios, particularly when traditional clinical signs such as a positive pivot shift were present. Discordance emerged in borderline cases, notably when AI recommended against LEAP or in female patients. These findings suggest AI could support orthopaedic decision-making by standardising criteria for LEAP, enhancing consistency in ambiguous cases, and prompting the development of evidence-based scoring systems to refine indications. IRB cleared, no need for trial registration as all cases hypothetical and no patient data included in the study.