CT-based 3D bone shape modes are associated with clinically meaningful functional improvement after total knee arthroplasty

Objective: To evaluate whether preoperative CT-derived three-dimensional bone shape modes are associated with failure to achieve the Knee Society Score Function subscore (KSS-Function) minimal clinically important difference (MCID) after total knee arthroplasty (TKA). Design: This retrospective study included patients with knee osteoarthritis who underwent primary robotic-assisted knee arthroplasty between January 2019 and January 2024. Bone shape modes were extracted from preoperative CT-based segmentations of the distal femur, proximal tibia, and patella using statistical shape modeling. Clinical outcome analysis was restricted to TKA knees with complete 1-year KSS-Function follow-up. MCID non-achievement was defined as failure to achieve a 10-point improvement in KSS-Function. Associations between shape modes and MCID non-achievement were assessed using logistic regression, with adjustment for clinical covariates. Internal discrimination was evaluated using repeated patient-level grouped cross-validation. Results: Among 151 TKA knees, 15.9% did not achieve the KSS-Function MCID. Female sex and higher BMI were associated with greater odds of MCID non-achievement. After adjustment for sex and BMI, five shape modes were associated with MCID non-achievement. These modes reflected less pronounced patellofemoral bony remodeling, patellar median ridge curvature, and localized contour variation near the central tibial plateau. The clinical-only, shape-only, and combined models achieved AUCs of 0.680, 0.696, and 0.754, respectively. The combined model had a 0.074 higher AUC than the clinical-only model, but paired bootstrap analysis was not significant (p = 0.130). Conclusions: CT-derived bone shape modes were associated with MCID non-achievement after TKA, suggesting bone morphology may offer insights, but their incremental value requires confirmation in larger, adequately powered cohorts.