Adults with chronic subdural hematoma, a common neurosurgical condition that mainly strikes people over 70, carry a measurably different community of gut bacteria than healthy adults of similar age, according to a case-control study published in Frontiers in Cellular and Infection Microbiology. Researchers at Beijing Tiantan Hospital, Capital Medical University, and collaborating institutions profiled the fecal bacterial communities of 40 adults with chronic subdural hematoma and 96 healthy controls using 16S rRNA gene sequencing, and found that the overall structure of the microbial community differed between the two groups even though the total diversity of species within each person’s gut did not. Thirteen bacterial genera showed statistically robust differences in abundance after the team adjusted for age, sex, hypertension, smoking, and alcohol use, and corrected for multiple comparisons.

Chronic subdural hematoma, or CSDH, was once viewed as a static collection of blood between the dura and the brain. It is now understood as a dynamic inflammatory and angiogenic disorder in which local inflammation, fragile new blood vessels, hyperfibrinolysis, and repeated microhemorrhage can drive hematoma enlargement and recurrence after surgery. Among people aged 70 or older, incidence estimates reach 58 cases per 100,000 person-years, and population-level data have linked rising subdural hematoma incidence with increasing antithrombotic drug use. As populations age, the clinical burden is expected to grow. What has remained poorly understood is whether systemic factors, including the trillions of microbes residing in the intestine, contribute to this inflammatory process.

The gut microbiota contributes to immune homeostasis and metabolism and communicates bidirectionally with the nervous system through immune, neural, endocrine, and metabolic pathways. Experimental studies indicate that perturbing intestinal microbes can modify immune responses and outcomes after cerebral ischemia, and altered microbial profiles have been reported in human neurological conditions including moyamoya disease, mild cognitive impairment, and migraine with irritable bowel syndrome. But findings vary across populations and analytical pipelines, and the gut microbiota of adults with CSDH had never been systematically characterized. The new study was designed to fill that gap while explicitly avoiding any claim of causation.

Between June 2023 and December 2024, the team enrolled adults with CSDH diagnosed by a neurosurgeon on the basis of compatible clinical findings and a hypodense or isodense subdural collection on head computed tomography or brain magnetic resonance imaging. Healthy controls were recruited from routine health examinations and had no evidence of CSDH on cranial computed tomography. All participants were older than 18, free of autoimmune or gastrointestinal disorders, and had not used systemic hormonal medication, Chinese herbal medicine, antibiotics, probiotics, prebiotics, or nonsteroidal anti-inflammatory drugs in the preceding three months. Fecal samples from patients were collected fasting on the morning after hospital admission, before burr-hole drainage, then stored and transported cold before DNA extraction and sequencing of the V3-V4 region of the 16S rRNA gene on an Illumina NovaSeq platform.

The two groups were closely matched on age, with a median of 69.5 years in the CSDH group and 68.0 years among controls, but differed in some baseline characteristics: the CSDH group had a higher proportion of men, 90.0 percent versus 75.0 percent, and more participants with diabetes, 20.0 percent versus 7.3 percent, while a smaller proportion were current smokers. To handle these imbalances, the researchers used linear models and covariate-adjusted centered log-ratio models that account for the compositional nature of sequencing data, applying Benjamini-Hochberg correction across all tested genera. After adjustment, there was no clear evidence of a difference in alpha diversity, measured by Chao1 richness, Shannon diversity, or Simpson diversity, between patients and controls.

Beta diversity, which captures differences in community composition between individuals, told a different story. Partial permutational multivariate analysis of variance detected a group difference across all three distance measures tested: Bray-Curtis, weighted UniFrac, and unweighted UniFrac, each with a p value of 0.001. The adjusted group term explained 10.6 percent of the total variation in weighted UniFrac distance, the largest of the three, compared with 3.6 percent for Bray-Curtis and 2.9 percent for unweighted UniFrac, meaning most of the variation in gut communities was not attributable to disease status. A complementary test found no evidence that the two groups differed in multivariate dispersion, which helps rule out the possibility that the apparent compositional shift was simply an artifact of unequal spread within groups.

At the genus level, the pattern was striking. Of 344 detected genera, 71 met prevalence and abundance thresholds, and 13 remained significant after covariate adjustment and multiple-testing correction. Nine were more abundant in patients with CSDH, including Dorea, Collinsella, Erysipelotrichaceae UCG-003, Blautia, the Eubacterium hallii group, the Ruminococcus torques group, Romboutsia, Clostridium sensu stricto 1, and the Christensenellaceae R-7 group. Four were more abundant in healthy controls: Bacteroides, Flavonifractor, Lachnoclostridium, and Lachnospira. Collinsella carried the largest positive adjusted coefficient, at 3.134, followed by Dorea at 2.251, while Bacteroides showed the largest negative coefficient, at minus 1.988. At the phylum level, Firmicutes, Actinobacteriota, and Desulfobacteriota were relatively enriched in patients, whereas Bacteroidota was markedly enriched in controls, averaging 35.83 percent of sequences versus 14.27 percent in the CSDH group.

The associations proved resilient to sensitivity testing. All 13 primary genus associations retained their direction and significance after additional adjustment for diabetes, and again after further adjustment for statin, anticoagulant, and antiplatelet exposure, with the Dorea coefficient easing from 2.251 to 2.054 and then 2.029. Analyses of non-rarefied counts retained all 13 associations and additionally identified Olsenella. An exploratory bootstrap analysis of 1,000 resamples found that percentile intervals for all 13 genera excluded zero across the models. The team also built a random forest classifier using genus-level profiles with repeated nested cross-validation to limit feature-selection and tuning bias. Across 10 repeated outer cross-validation partitions, the mean out-of-fold area under the receiver-operating-characteristic curve was 0.938, plus or minus 0.010, and averaging each participant’s out-of-fold probabilities yielded an aggregated AUC of 0.951. Bacteroides, Lachnospira, and Collinsella were selected in all 50 outer models.

Two correlations with clinical characteristics survived multiple-testing correction across 130 genus-characteristic pairs. Dorea abundance was positively correlated with diabetes, with a partial Spearman coefficient of 0.334 and a corrected q value of 0.011, and Bacteroides abundance was negatively correlated with cerebral infarction, with a coefficient of minus 0.292 and a q value of 0.041. Notably, these correlations were estimated across the full cohort rather than specifically among CSDH cases, and the authors caution that shared exposures, reverse causation, and differences in medical care remain alternative explanations. Persistence of the Dorea association after diabetes adjustment, they note, does not exclude residual metabolic or treatment-related confounding.

The authors are explicit that these are exploratory associations in a selected case-control sample. The cross-sectional, single-center design precludes causal inference, and the classifier has not established calibration, external validity, or clinical utility; an AUC quantifies ranking discrimination, not diagnostic usefulness, and the case-control sampling fraction does not reflect clinical prevalence. Recruitment setting was inseparable from disease group, since cases were hospitalized while controls attended health examinations, and hospitalization, acute illness, altered intake, and medication changes could all shape the microbiome. Dietary intake and nutritional status were not systematically measured. The 16S rRNA approach does not measure microbial functional genes, metabolites, intestinal permeability, or circulating immune mediators, and relative abundance does not reflect absolute bacterial load, so the data cannot link the observed genera to inflammation or angiogenesis in hematoma membranes. Clustering sequences at 97 percent similarity may also merge distinct strains that newer methods can resolve. The authors call for multicenter longitudinal studies with detailed dietary, nutritional, and medication assessment, direct functional measurements, and independent classifier validation before any temporal or clinical relevance can be established.

Subject of Research: Gut microbiota composition in adults with chronic subdural hematoma

Article Title: Altered gut microbiota composition in adults with chronic subdural hematoma: a case-control study

Article References: Li, Y., Yu, X., Lin, C., Gao, G., Liu, W., & Wu, L. (2026). Altered gut microbiota composition in adults with chronic subdural hematoma: a case-control study. Frontiers in Cellular and Infection Microbiology, 16, Article 1962090. https://doi.org/10.3389/fcimb.2026.1962090

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1962090

Keywords: chronic subdural hematoma, gut microbiota, 16S rRNA sequencing, case-control study, microbiome, beta diversity, random forest classifier, Bacteroides, Collinsella, Dorea, neurosurgery, inflammation