Development and validation of a predictive model for chronic pain after thoracoscopic pulmonary resection.

BACKGROUND

Chronic postsurgical pain (CPSP) is a major complication following video-assisted thoracoscopic surgery (VATS) for lung resection. Despite its clinical importance, the reported incidence of CPSP varies considerably across studies, and a standardized tool for risk prediction remains lacking. This study aimed to identify predictors associated with CPSP after VATS lung resection within a single-center prospective observational cohort and to develop predictive models to guide personalized analgesic strategies.

METHODS

Clinical data were analyzed from patients who underwent thoracoscopic pulmonary resection at Tangdu Hospital between August 2024 and July 2025. Eligible patients were randomly allocated into a training cohort (80%) for model development and an internal validation cohort (20%) for model validation. Acute pain was assessed using the highest Numerical Rating Scale (NRS) score recorded within 72 h postoperatively, while CPSP was evaluated at 3 months after surgery. Patients were dichotomized based on the presence of CPSP into the CPSP group and the non-CPSP group. Variables with a P-value <0.1 in univariate analysis were entered into multivariate logistic regression to establish a prediction model for CPSP using R software.

RESULTS

A total of 744 patients were included in this study, with 595 assigned to the training set and 149 to the validation set. Among all patients, 284 (38.1%) developed CPSP. The incidence rates in the training and validation cohorts were 37.8 and 36.9%, respectively. Multivariate logistic regression identified gender, postoperative acute pain, postoperative pneumonia, duration of chest tube drainage, and opioid rescue dose (in morphine milligram equivalents) as independent predictors associated with CPSP. The nomogram model constructed based on these predictors demonstrated good performance, with an area under the curve (AUC) of 0.880 (95% CI: 0.850-0.910). Calibration curve analysis confirmed the high predictive accuracy of the model (Hosmer-Lemeshow test, P = 0.737), and decision curve analysis indicated a significant clinical net benefit. Furthermore, in the validation cohort, the model maintained good discriminative ability (AUC = 0.829) and calibration (Hosmer-Lemeshow test, P = 0.765), enabling reliable individualized risk prediction.

CONCLUSIONS

CPSP remains highly prevalent after thoracoscopic pulmonary resection.