Perioperative neurofilament light chain and glial fibrillary acidic protein as biomarkers of postoperative delirium: a systematic review and meta-analysis.

BACKGROUND

Postoperative delirium (POD) is a frequent and serious neurocognitive complication reflecting acute brain dysfunction and predicting long-term cognitive decline. Neurofilament light chain (NfL) and glial fibrillary acidic protein (GFAP) are biomarkers of neuronal and astroglial injury, but their relevance to POD remains uncertain.

METHODS

This systematic review and meta-analysis aimed to determine whether there is an association between perioperative NfL or GFAP levels and POD. PubMed, Embase, and Web of Science were searched until January 2026. Odds ratios (ORs), standardized mean differences (SMDs), and diagnostic performance (AUC) were calculated using random-effects models.

RESULTS

Fourteen studies involving 2,604 participants were included, of whom approximately 30% developed POD. In categorical analyses, higher NfL levels were associated with increased POD risk across nine studies (pooled OR = 1.27, 95% CI 1.08-1.50), consistent across plasma (OR = 1.27) and cerebrospinal fluid (CSF) (OR = 1.32) samples. Continuous analyses also showed higher NfL concentrations in delirious patients (Plasma SMD = 0.48; CSF SMD = 0.40). GFAP showed a significant but weaker and less consistent association (overall OR = 1.68, 95% CI 1.15-2.43), mainly driven by plasma data, while continuous analyses were nonsignificant. Diagnostic meta-analysis showed fair discriminative performance for NfL (AUC = 0.68) and moderate accuracy for GFAP (AUC = 0.63).

CONCLUSIONS

This meta-analysis demonstrates that perioperative elevations of NfL and GFAP are associated with POD. However, their diagnostic performance remains modest, and current evidence does not support their use as stand-alone diagnostic biomarkers. NfL showed a more stable and reproducible relationship, suggesting its potential role as a marker of perioperative neural stress and as a component of future multimodal delirium risk models.