Gut microbiota and blood biomarkers as correlating factors in patients with postoperative delirium: analysis of three prospective observational studies.

Postoperative delirium is associated with both gut microbiota alterations and Tau phosphorylation; however, how these factors interact and jointly contribute to postoperative delirium remains poorly understood. This prospective observational cohort study screened 491 patients aged ≥65 years undergoing elective laminectomy or hip or knee replacement under general or spinal anesthesia at Massachusetts General Hospital (2016-2020). The study aimed to assess the correlation between plasma Tau protein levels, specific gut microbiota, and the gut microbiota-derived metabolite indole-3-propionic acid (IPA) in participants with and without postoperative delirium. Exclusion criteria included major neurological disease, smoking history, sensory impairment, and recent antibiotic use. Delirium was assessed 24-48 h postoperatively. Out of 491 screened participants, 139 had blood biomarker data, and 86 had gut microbiota data included in the final analysis. Ten percent of the participants experienced postoperative delirium. Co-occurrence network analysis demonstrated that microbial interactions differed between postoperative delirium and non-postoperative delirium groups. Significant associations among gut microbiota, IPA, and Tau biomarkers were observed only in postoperative delirium participants. Additionally, three machine learning classifiers distinguished postoperative delirium cases from non-postoperative delirium cases with Area Under the Curve values above 50%. The results suggest that plasma Tau may interact with gut microbiota, and this interaction is associated with postoperative delirium, suggesting the involvement of a gut-blood-brain axis in postoperative delirium vulnerability. These findings implicate a microbiota-Tau interaction that may represent both a potential pathogenic mechanism and a therapeutic target for postoperative delirium pending confirmative studies. Furthermore, machine learning classifiers utilizing microbiome-derived features demonstrate a potential in distinguishing patients with and without postoperative delirium.