Risk Factors and Efficacy of Combined Neurometabolic–Digital Cognitive Training for Perioperative Neurocognitive Disorder in School-Age Children After General Anesthesia: A Prospective Cohort Study

Abstract

Aims: To identify independent risk factors for perioperative neurocognitive disorder (PND) in school-age children after general anaesthesia, develop a predictive risk model, and evaluate the clinical efficacy of combined neurometabolic pharmacotherapy and digital cognitive training in children with confirmed PND. Study Design: A prospective cohort study with a nested, non-randomised, two-arm treatment comparison. Place and Duration of Study: The study was conducted in the Departments of Pediatric Surgery, Anesthesiology-Reanimatology, and Dermatology at the Clinic of Tashkent Pediatric Medical Institute, Tashkent, Uzbekistan, from March 2024 to January 2025. Methodology: A total of 206 school-age children aged 7–17 years were studied: 181 underwent elective non-cardiac surgery under general anaesthesia, while 25 healthy children served as normative controls. PND was defined on postoperative day 7 using the ISPOCD composite Z-score method (|Z| ≥ 1.96 on at least two of seven instruments). Independent predictors were modelled using multivariable logistic regression. Among 56 children with confirmed PND, 30 received one month of combined therapy (hopantenic acid [Pantogam] plus the NeyroCog.uz digital cognitive-training platform), while 26 received standard follow-up. The groups were compared using analysis of covariance (ANCOVA) adjusted for baseline scores, and effect sizes were expressed as Cohen’s d. Results: PND occurred in 56 of 181 children (30.9%). Independent predictors were the number of anaesthetic agents used (OR=3.81, P<.001), premorbid neurological background (OR=3.62, P=.02), anaesthesia duration per minute (OR=1.02, P<.001), and age, which was associated with lower odds of PND (OR=0.67, P<.001). The composite model showed good discrimination (AUC=0.78; Nagelkerke R²=0.45). Combined therapy produced significantly greater covariate-adjusted gains than standard follow-up in selective attention (Δadj=15.13, P<.001), verbal memory (Δadj=16.03, P<.001), psychomotor speed, working memory, and auditory gnosis (all P<.001; Cohen’s d=1.0–2.9). No significant between-group differences were observed in clinical-neurological status or EEG (P=.37 and P=.77, respectively). Conclusion: The four-factor model showed potentially useful discrimination for PND risk. Combined neurometabolic–digital cognitive rehabilitation was associated with large, domain-selective gains compared with standard observation. These findings support external validation of the risk model and further randomised evaluation of the combined rehabilitation protocol before routine clinical adoption.