全膝関節置換術後の長期オピオイド使用に関連する因子の予測モデル：人口ベースの後ろ向き研究

Knee osteoarthritis (OA) affects approximately 10% of the general population. Of these, nearly 20% of patients undergo total knee arthroplasty (TKA). A proportion subsequently develops long-term opioid use, yet comprehensive evaluations of perioperative risk factors remain limited.

A total of 15,051 patients with knee OA who underwent TKA between 2002 and 2018 were included. Our primary analysis used a Least Absolute Shrinkage and Selection Operator (LASSO)-Elastic Net logistic regression model with a 70/30 training-validation split to identify predictors of long-term opioid use and to estimate model performance using the area under the receiver operating characteristic curve (AUC) and calibration. As supplementary analyses, hierarchical logistic regression models progressively incorporated demographic, preoperative, in-hospital, and early post-discharge variables.

Overall, 1029 patients were identified as long-term opioid users. The LASSO-Elastic Net model achieved an AUC of 0.8213 (95% CI: 0.8080-0.8347). The AUCs of the hierarchical logistic models were 0.7523 (95% CI: 0.7364-0.7681) for demographic and preoperative factors (Model 1), 0.7682 (95% CI: 0.7534-0.7829) with added hospitalisation factors (Model 2), and 0.8251 (95% CI: 0.8119-0.8383) after further including post-discharge variables (Model 3).

This study systematically identifies risk factors for long-term opioid use across different phases of TKA care. A predictive model based on LASSO-Elastic analysis demonstrated an AUC of 0.8213, highlighting its potential for early identification of high-risk patients.

Using a population-based analysis, this study identifies risk factors for prolonged opioid use following TKA, aiding clinicians in early recognition and targeted preventive strategies.