Novel CT-based simulation reveals anterolateral surgical approach avoids acromial collision in antegrade humeral nailing

Sixty-eight anonymized shoulder CTs were segmented to generate scapular and humeral surface models. A straight 8-mm nail was virtually advanced along the medullary axis under two conditions: (I) transdeltoid with 0° glenohumeral extension and (II) anterolateral with 30° extension. Two trained investigators independently repeated all simulations and recorded nail-acromion collisions. Agreement was assessed with Cohen's κ and intraclass correlation coefficients (ICC). Collision rates were compared with paired McNemar tests. The predictive value of AI was analyzed using logistic regression and receiver-operating-characteristic (ROC) analysis. RESULTS: Implant-acromion collision occurred in 32/68 shoulders (47.1%) with the transdeltoid approach, versus 2/68 (2.9%) with the anterolateral approach (absolute risk reduction 44%; number-needed-to-treat 2.3; p< 10⁻⁶). AI independently predicted transdeltoid failure (adjusted odds ratio 2.81 per +0.10; p = 0.011); a cut-off ≥ 0.69 yielded an AUC of 0.72 (63% sensitivity, 72% specificity). Inter-observer reliability was substantial for collision (κ 0.86) and excellent for morphometrics (ICC ≥ 0.90). CONCLUSION: CT-based simulation demonstrates that the anterolateral approach markedly reduces the risk of implant-acromion collision in SAHN and should therefore be considered the preferred option. However, if the transdeltoid approach is selected, preoperative AI screening with a threshold of ≥0.69 is essential, as larger acromia increase collision risk and must be factored into surgical planning. LEVEL OF EVIDENCE: Level IV - paired in-silico simulation study.