The study was conducted in Hangzhou, China, between January 2022 and January 2023. Patients were at least 65 years old and scheduled for hip fracture surgery under general anesthesia. Those with pre-existing cognitive impairment, psychiatric or neurological disorders, inflammatory bowel disease, recent antibiotic or probiotic use, or unwillingness to provide consent were excluded. Cognitive screening used the Chinese version of the Mini-Mental State Examination, and patients scoring 24 or below were excluded. Delirium was assessed every day from postoperative day one through day five with the Chinese version of the 4AT assessment tool; a score of four or more was considered diagnostic. The investigators recorded the onset and duration of delirium and collected fecal samples within 24 to 48 hours after surgery, while control samples were collected during routine physical examinations.

The three groups were reasonably similar in measured demographic and clinical characteristics. Mean ages were 78.30 years in the control group, 80.10 in the non-POD group and 81.40 in the POD group, with no significant difference among them. Sex, education, body mass index, hypertension, diabetes and cerebrovascular disease history also did not differ significantly. The two surgical groups underwent similar procedures: most received total hip replacement or intramedullary nailing, 60% in each group received a blood transfusion, and 90% in each group received postoperative antibiotics. Delirium began within one to three days after surgery in seven patients and within three to five days in three others, lasting an average of 2.23 days. These similarities helped reduce measured differences between the delirium and non-delirium surgical groups, although the sample was too small to eliminate all potential confounding.

Sequencing targeted the hypervariable V3-V4 region of the bacterial 16S rRNA gene. Amplicons were generated with primers 341F and 806R and sequenced on an Illumina MiSeq platform. Raw reads were processed in QIIME2, quality filtered, denoised and merged into amplicon sequence variants with DADA2, while taxonomy was assigned against the SILVA database, version 138. The investigators calculated Chao1 and Shannon indices to assess within-sample richness and evenness, and weighted UniFrac distances to assess differences in community composition between samples. Principal coordinates analysis was used to visualize beta diversity, while LEfSe and non-parametric tests were used to explore differential abundance. Because many taxa were examined in only 30 participants, the researchers applied Benjamini-Hochberg false discovery rate correction.

Neither the Chao1 nor the Shannon index differed significantly among the three groups. Although controls had higher mean values for both measures, the difference was not statistically significant. By contrast, weighted UniFrac beta diversity showed a significant difference across all three groups, with PERMANOVA yielding a P value of 0.001. Importantly, the POD and non-POD groups did not differ significantly from one another. The overall separation therefore appeared to be driven mainly by the contrast between healthy controls and the two surgical groups rather than by delirium status itself. The authors said this pattern suggests surgery, anesthesia or hospitalization may have exerted a stronger influence on overall gut community structure than delirium, although the observational design could not distinguish among those factors.

Firmicutes and Bacteroidota were the dominant phyla in all three groups. At lower taxonomic levels, healthy controls had significantly higher relative abundances of Lactobacillaceae, Lactobacillus and Pseudobutyrivibrio than both surgical groups, each reported with a nominal P value of 0.001. When the researchers compared the two surgical groups directly, the POD group showed higher nominal relative abundance of Erysipelotrichales, Erysipelotrichaceae, Lachnospiraceae and Atopobium. The non-POD group showed higher levels of Verrucomicrobiaceae and Lactobacillaceae. These between-group differences did not remain significant after false discovery rate correction, meaning that adjusted P values were above 0.05. The paper accordingly treats the patterns as exploratory rather than statistically established POD-specific effects.

LEfSe analysis identified additional candidate taxa that discriminated among the groups. Lactobacillus, Vibrionaceae and Shigella were associated with the non-POD group, while Lachnospiraceae, Erysipelotrichaceae, Collinsella and Helicobacter were associated with the POD group. The authors described these organisms as candidate discriminative features, not biomarkers. No independent cohort, cross-validation, receiver operating characteristic analysis, sensitivity or specificity assessment, or experimental validation was conducted. The taxonomic findings are also limited by what 16S sequencing can reveal: the method identifies bacterial sequences and estimates their relative abundance, but it does not directly measure microbial functions, metabolites, inflammatory activity or effects on the nervous system.

The pattern nevertheless provides a hypothesis for further investigation. The POD group appeared to have lower relative abundance of organisms commonly treated as commensals, including Lactobacillus, together with higher relative abundance of families and genera previously associated in other research with inflammation, including Erysipelotrichaceae and Collinsella. One possible gut-brain-axis mechanism would involve microbial disruption affecting intestinal barrier integrity, allowing bacterial products to enter the circulation and contributing to systemic or neuroinflammation. However, this study measured none of those proposed links. It did not quantify inflammatory cytokines, intestinal permeability, bacterial metabolites, circulating microbial products, blood-brain barrier integrity or neuroinflammatory pathways. Descriptions of taxa as beneficial or inflammation-associated therefore reflect prior associations rather than measured functions in these participants.

Timing and selection of controls further limit interpretation. Samples were taken after surgery, while delirium was assessed over the following five days, so the study cannot show whether the observed microbial patterns preceded delirium or arose in response to it. Surgery, anesthesia, hospitalization, diet, bowel function, analgesics, transfusion and antibiotics could all have influenced the postoperative samples; postoperative antibiotics were used by 90% of both surgical groups. The healthy controls were also fundamentally different from the surgical patients because they had neither fractures nor the associated hospitalization. Their distinct microbiota may therefore reflect those circumstances rather than a direct effect of delirium. For those reasons, the POD-versus-non-POD comparison is the more relevant contrast for generating hypotheses, but it was based on only 20 surgical patients and remains underpowered.

Overall, the study supports the existence of preliminary taxonomic associations between postoperative delirium and gut microbial composition in elderly hip fracture patients, but it does not demonstrate dysbiosis as a cause, consequence or diagnostic marker of delirium. The clearest statistical separation was between healthy controls and surgical patients, while the overall community structures of the POD and non-POD groups were not significantly different. The named candidate taxa require confirmation in larger, independent cohorts with preoperative sampling and, ideally, metagenomic, metabolomic and immune measurements. Until such validation, the results are best treated as a starting point for hypothesis-driven research rather than a basis for diagnosis, prognosis or intervention. The study data are available through the NCBI Sequence Read Archive under accession PRJNA154139.

Subject of Research: Associations between gut microbiota composition and postoperative delirium in elderly hip fracture patients

Article Title: A case-control study of gut microbiota structure in elderly patients with postoperative delirium after hip fracture based on 16S rRNA sequencing

Article References: Zhuang, X., Ge, G., Fan, Y., Liu, D., Zhu, C., Qi, W., & Shao, R. (2026). A case-control study of gut microbiota structure in elderly patients with postoperative delirium after hip fracture based on 16S rRNA sequencing. Frontiers in Cellular and Infection Microbiology, 16, 1905029. https://doi.org/10.3389/fcimb.2026.1905029

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1905029

Keywords: gut microbiota, postoperative delirium, hip fracture, 16S rRNA sequencing, elderly patients, case-control study, alpha diversity, beta diversity, gut-brain axis, microbiome dysbiosis, case-control, microbiota