Report / Paper metadata
Optimal prophylaxis for perioperative drug-induced delayed neurocognitive recovery: a Bayesian network meta-analysis of 18 interventions with mechanism-based stratification
This review compared drugs used around surgery in three separate evidence networks. Its rankings are exploratory: most comparisons had low or very low confidence, cognitive outcomes differed across trials, and several reporting inconsistencies limit how precisely the estimates can be interpreted.
What question was reviewed?
The authors searched for randomized trials of adult surgical patients reporting delayed neurocognitive recovery within 30 days or during hospital care. Eligible studies used validated cognitive tests, but could label the outcome as postoperative cognitive dysfunction, perioperative neurocognitive disorder or delayed neurocognitive recovery. Assessment times ranged from one hour to 30 days. These varied early assessments do not establish prevention of long-term dementia or prove that any cognitive change was caused by a drug.
Why are there three networks?
Maintenance anesthetics, adjunctive sedatives or neuroprotective agents, and analgesics were analyzed separately. The paper reports 69 trials and 9,636 patients overall, with 18 interventions including saline, not 18 active drugs. It excluded combination regimens and certain pre-existing neurological disorders, and included only English full texts. A ranking within one network is not a comparison across all drug categories or evidence for an optimal multi-drug anesthetic plan.
What patterns were reported?
Propofol ranked highest within the maintenance-anesthetic network. Dexmedetomidine and ulinastatin had nearly identical top rankings in the adjunctive network. In the analgesic network, all drug-versus-saline credible intervals crossed the no-effect value despite different rankings. This means those data did not establish a clear preventive effect for an analgesic; a higher rank alone does not demonstrate superiority or quantify a patient's absolute benefit.
How much confidence is justified?
Most comparisons were rated low or very low confidence. The authors report local inconsistency for propofol versus sevoflurane and evidence of publication bias in the adjunctive network. Studies differed in surgery, cognitive tests, assessment timing and administration regimens. The authors describe the analgesic network as relying on indirect comparisons without adequate head-to-head support. Neither a high ranking score nor a stable model ranking removes those limits.
What reporting differences need care?
Some statements do not reconcile cleanly. The network-specific trial and participant totals differ from the reported overall totals without a clear explanation. Methods specify a random-effects model, while a sensitivity-analysis passage describes replacing a fixed-effects model. The description of an exclusively indirect analgesic network also sits alongside a certainty section naming direct comparisons. The publication-bias correction paragraph contains interval values that do not match its description of proximity to one. This account does not silently repair those details or treat them as settled estimates.
Can the ranking guide a patient's medication?
The abstract explicitly says the rankings are insufficient for clinical recommendations, although the closing conclusion uses stronger language about drug selection. The detailed low-certainty evidence and reporting gaps support the more cautious interpretation. This is an original account of the full review, not a prescribing guide, a safety comparison or a recommendation to change an anesthetic regimen. Stronger claims would need consistent reporting, better direct comparisons and relevant patient-level benefits and harms.
- Kind
- Paper metadata
- Identifier
- PMC13642292