Somatic and psychological predictors of chronic postsurgical pain in cancer patients: a machine learning approach in a longitudinal two-centre study.

BACKGROUND: Chronic postsurgical pain is a significant medical concern, particularly in cancer patients. However, most previous studies overlooked the psychological mechanisms contributing to this risk. The present study aimed to identify baseline predictive factors for chronic postsurgical pain in order to build and validate a predictive algorithm based on somatic and psychological predictors in a highly phenotyped longitudinal cohort of adult patients with breast or lung cancer from two European centres. METHODS: Comprehensive preoperative data, including patient characteristics and clinical, psychological, and social variables, were collected between 2017 and 2018. Four distinct machine learning models were trained and validated. Patients were examined at baseline and re-evaluated in face-to-face interviews 1 yr later. RESULTS: The sample included 255 patients (mean age 62.3 [range, 28-83] yr, 89.4% female). Chronic postsurgical pain was present in 83 patients (32.5%), of whom 72 (28% of the total sample) had neuropathic pain. We developed a predictive algorithm based on three independent variables: younger age (odds ratio [OR], 0.47; 95% confidence interval [CI], 0.32-0.69), preoperative pain outside of the surgical area (OR, 2.45; 95% CI, 1.46-4.10), and a specific anxiety symptom (overwhelming worries) (OR, 1.81; 95% CI, 1.05-3.13). CONCLUSIONS: This simple algorithm, requiring only three easily accessible inputs, offers a practical tool in routine preoperative settings, supporting timely, targeted interventions to improve pain management in cancer patients at risk of chronic postsurgical pain. CLINICAL TRIAL REGISTRATION: NCT02368275, NCT03124511, NCT02960971.