Optimal prophylaxis for perioperative drug-induced delayed neurocognitive recovery: a Bayesian network meta-analysis of 18 interventions with mechanism-based stratification.

Background: Postoperative delayed neurocognitive recovery (dNCR) is a common and serious perioperative complication. Although there are various drugs available for prevention, there is a lack of efficacy comparison and significant controversy exists. Methods: Systematic searches were conducted in PubMed, Embase, Cochrane Library, Web of Science, World Health Organization International Clinical Trials Registry Platform (WHO ICTRP), and ClinicalTrials.gov databases from the establishment of the databases until July 2025. Randomized controlled trials (RCTs) evaluating drug interventions for preventing dNCR were included. According to the clinical mechanism of action, the intervention measures were divided into three independent networks: anesthetic maintenance drugs, adjunctive sedation and neuroprotective agents, and analgesics. Bayesian network meta-analysis was used for each of the three networks to integrate direct and indirect evidence for efficacy comparison and ranking. The effect size was reported as the odds ratio (OR) and its 95% credible interval (CrI). The quality of evidence was evaluated using the CINeMA framework. Results: A total of 69 randomized controlled trials were included, involving 9,636 patients, and 18 intervention measures (including the saline control group) were evaluated. The anesthesia maintenance drug network: Propofol (SUCRA = 0.908) was the most effective in preventing dNCR, outperforming sevoflurane (OR 0.60, 95% CrI 0.44-0.82) and midazolam (OR 0.21, 95% CrI 0.08-0.51). The adjunctive sedation and neuroprotective drug network: Dexmedetomidine (SUCRA = 0.887) and ulinastatin (SUCRA = 0.886) had similar efficacy rankings, both outperforming ketamine (SUCRA = 0.557), dexamethasone (SUCRA = 0.334), and minocycline (SUCRA = 0.080). The analgesic drug network: Ropivacaine (SUCRA = 0.775), lidocaine (SUCRA = 0.748), parecoxib (SUCRA = 0.691), and butorphanol (SUCRA = 0.679) showed a trend of reducing the incidence of dNCR, but the 95% CrI of each drug compared to saline all exceeded 1.0, and the differences were not statistically significant. The CINeMA evidence quality assessment showed that the majority of the comparative evidence in these three networks had a "low" or "very low" quality. Conclusion: In this meta-analysis, propofol ranked the highest among the anesthetic maintenance drugs, dexmedetomidine and ulinastatin ranked among the top for adjuvant drugs, and analgesic drugs did not show a definite protective effect. Due to the low quality of evidence, publication bias, and the lack of head-to-head comparisons, the above rankings are only preliminary references and are not sufficient to constitute clinical recommendations. Systematic Review Registration: https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42024607899, identifier CRD42024607899.