Demography-based prediction of implant size in total knee arthroplasty: development and internal validation

Abstract

Abstract Background Accurate estimation of component size may facilitate preoperative planning in total knee arthroplasty (TKA). Demographic characteristics offer a simple approach that requires no additional imaging measurements, but prediction models require appropriate internal validation and comparison with simpler benchmarks. This study aimed to develop and internally validate a demographic model for predicting implanted femoral and tibial component sizes in a Taiwanese TKA cohort. Methods We retrospectively analyzed 633 primary TKA procedures performed in 575 patients between January 2022 and December 2024. Sex, height, weight, and age were used to develop linear regression models for femoral and tibial component size. Internal validation was performed using 1,000 patient-clustered bootstrap resamples, and patient-grouped 10-fold cross-validation was used to assess out-of-sample performance and calibration. The full models were compared with a reduced sex-plus-height model and a sex-specific modal-size benchmark. Performance was assessed using R², mean absolute error, root mean squared error, exact-size accuracy, and accuracy within one component size. Results The femoral and tibial models had apparent R² values of 0.661 and 0.687, with optimism-corrected values of 0.656 and 0.682, respectively. Optimism-corrected exact-size accuracy was 69.0% for femoral and 51.9% for tibial components, while 99.7% and 95.5%, respectively, were predicted within one size. Patient-grouped out-of-sample exact-size accuracy was 68.6% and 52.3%, and ± 1-size accuracy was 99.7% and 95.7%, respectively; both components were within one size in 95.6% of procedures. Calibration slopes were 0.994 for femoral and 0.996 for tibial prediction. Sex and height accounted for most of the predictive information, and the reduced sex-plus-height model performed close to the full model. Prediction was less accurate for uncommon component-size extremes. Conclusions A demographic model using routinely available variables estimated implanted TKA component sizes with limited optimism after patient-level internal validation. Its principal value was narrowing the likely component requirement to a three-size planning range rather than exact-size determination. Demographic prediction may provide a low-complexity adjunct for preoperative planning, but it should not replace intraoperative sizing or established imaging-based methods. External validation is required before broader clinical or inventory-management application. Clinical trial number Not applicable.