Impact of nutritional status and abnormal bone-muscle metabolism on chronic low back pain after lumbar decompression surgery: a multicenter predictive model study based on paraspinal muscle parameters.

Objective: This study aimed to investigate the associations of preoperative nutritional status, bone-muscle metabolic abnormalities, and paraspinal muscle degeneration with chronic low back pain (CLBP) after lumbar decompression surgery. A multicenter predictive model was developed to improve preoperative risk stratification, enable early identification of high-risk patients, and support individualized perioperative management. Methods: A total of 2,333 patients who underwent unilateral biportal endoscopic (UBE) decompression surgery at five centers were retrospectively enrolled. The cohort was divided into a training set, an internal validation set, and an external test set. Demographic characteristics, laboratory variables, and imaging parameters of the paraspinal muscles were collected. Multivariable logistic regression analysis was performed to identify independent factors associated with postoperative CLBP. Based on the selected variables, multiple machine learning models were developed. Model performance was evaluated using receiver operating characteristic (ROC) analysis, calibration curves, and decision curve analysis. In addition, model interpretability analyses were conducted to assess the contributions of key variables to the predictions. Results: Multivariable logistic regression analysis identified age, albumin (Alb), calcium (Ca), alkaline phosphatase (ALP), psoas muscle index (PMI), multifidus fat infiltration (MF FI), and erector spinae fat infiltration (ES FI) as independent factors associated with postoperative CLBP. Among the predictive models, machine learning models showed better discrimination than the traditional logistic regression model. This finding suggests that machine learning may be more effective in capturing complex nonlinear relationships associated with postoperative CLBP. The ExtraTrees model showed the best performance, with area under the curve (AUC) values of 0.834 in the internal validation set and 0.816 in the external test set. These values were higher than those of the logistic regression model. The model also showed good calibration, with Brier scores of 0.168 and 0.171 in the internal validation and external test sets, respectively, and provided stable clinical net benefit. Further analysis showed that age, PMI, and paraspinal muscle fat infiltration were the most important predictors of postoperative CLBP. A web-based calculator was subsequently developed to improve the clinical applicability of the model. Conclusion: Preoperative nutritional insufficiency, bone-muscle metabolic abnormalities, and paraspinal muscle degeneration were closely associated with CLBP after lumbar decompression surgery. A predictive model integrating these factors showed good discriminative ability for predicting postoperative pain risk. Preoperative assessment based on these variables may help identify high-risk patients at an early stage. It may also guide strategies for nutritional optimization, bone-muscle metabolism management, and perioperative rehabilitation, thereby supporting individualized perioperative care.