Association of clinically determined extended perioperative antibiotic therapy with postoperative neurocognitive outcomes in older adults: a prospective cohort study with microbiome and biomarker analyses

Postoperative neurocognitive decline is common in older surgical patients and has multiple determinants, including patient vulnerability, operative stress, inflammation, infection, and perioperative complications. The gut microbiota may contribute to systemic and central inflammatory signalling through the microbiota-gut-brain axis. Human data linking clinically determined perioperative antibiotic exposure with postoperative cognition remain limited. This prospective cohort study examined whether patients receiving extended therapy had different neurocognitive, microbiome, short-chain fatty acid, and inflammatory profiles than patients receiving standard prophylaxis. This prospective observational cohort study enrolled 156 patients aged ≥ 60 years undergoing elective non-cardiac surgery at Loudi Central Hospital from January 2022 to January 2024. The study was approved by the Research Ethics Committee of Loudi Central Hospital (2021CH003274), and all participants provided written informed consent. Participants were classified by clinically determined antibiotic duration as standard prophylaxis (SP; ≤24 h) or extended therapy (ET; >24 h). Neuropsychological testing was performed at baseline and on postoperative days (PODs) 3, 7, and 30; the primary outcome was test-defined postoperative neurocognitive decline at POD 7. Multivariable logistic regression and propensity-score matching adjusted for measured baseline and perioperative covariates. Fecal microbiota, fecal short-chain fatty acids (SCFAs), and serum inflammatory and exploratory perioperative stress markers were also assessed. Test-defined neurocognitive decline at POD 7 occurred in 33/78 patients (42.3%) in the ET group and 19/78 (24.4%) in the SP group (relative risk 1.73, 95% CI 1.12–2.68; P = 0.013). The association remained in the reported multivariable model (adjusted odds ratio 2.34, 95% CI 1.18–4.64; P = 0.015) and in the propensity-matched sensitivity analysis (odds ratio 2.08, 95% CI 1.02–4.26; P = 0.044). Compared with SP, patients receiving ET also had lower Shannon diversity, larger reductions in selected SCFA-associated genera, lower fecal butyrate, and higher systemic inflammatory markers. Lower butyrate and higher lipopolysaccharide-binding protein were associated with greater cognitive decline. Because ET was prescribed for contamination or postoperative infection and was associated with greater clinical complexity, these findings do not isolate an effect of antibiotic duration itself. Patients who received clinically determined extended perioperative antibiotic therapy had a higher frequency of postoperative neurocognitive decline and a different microbiome, SCFA, and inflammatory profile than patients receiving standard prophylaxis. These observations are biologically coherent but do not establish that antibiotic duration caused the cognitive or biological differences because treatment indication, surgical complexity, antibiotic spectrum, and other perioperative factors were incompletely separable. Clinically indicated antibiotics should not be withheld on the basis of these results. Future multicenter studies should evaluate infectious and neurocognitive outcomes together and, where clinical equipoise exists, compare acceptable antibiotic-duration strategies while measuring the factors that determine treatment duration.