Preoperative and early postoperative predictors of nausea and vomiting after laparoscopic abdominal surgery in an enhanced recovery programme

Postoperative nausea and vomiting (PONV) is a frequent, distressing complication of laparoscopic abdominal surgery, yet widely used risk scores such as the Apfel score predate enhanced recovery (ERAS) practice and do not capture perioperative management. In a prospective cohort study at a single tertiary centre, we characterised PONV, temporally tested a modified Apfel score, and quantified potentially modifiable postoperative exposures. A development cohort ( n = 315; August–December 2023) and a temporally independent validation cohort ( n = 64; February–March 2025) were assessed for PONV within 24 h, which occurred in 17.5% (55/315) and 23.4% (15/64), respectively. Because 24-h tramadol and early oral intake are postoperative exposures that may be influenced by the outcome, the prediction model was fixed at a landmark (post-anaesthesia care unit discharge) using only information available at that point: history of motion sickness, smoking, postoperative patient-controlled analgesia (PCA) and surgical site. Motion sickness was an independent risk factor (odds ratio 4.90) and current smoking was protective (0.40); PCA and surgical site were not independently significant. Discrimination was moderate (area under the curve 0.789 development, 0.832 validation; optimism-corrected C-statistic 0.776), and calibration point estimates suggested under-prediction in the higher-incidence validation cohort (observed-to-expected ratio 1.31, 95% CI 0.79 to 2.18). The model discriminated better than a modified Apfel score in development (ΔAUC + 0.039, p = 0.003) but not in validation (ΔAUC − 0.014, p = 0.49). In a separate exploratory analysis, the apparent protective effect of early oral intake weakened substantially under a time-ordered landmark analysis, consistent with reverse causation. These findings identify modifiable perioperative exposures as hypothesis-generating targets rather than establishing a superior prediction tool, and require multicentre external validation before clinical use.