Gut dysbiosis and mortality in hemodialysis patients

Patients with end-stage kidney disease (ESKD) on hemodialysis have substantially reduced life expectancy, with both cardiovascular (CV) and non-CV mortality risks elevated. Traditional risk factors (age, diabetes, CVD) and nontraditional factors (malnutrition, inflammation, uremic toxins) contribute to poor outcomes. Gut dysbiosis—loss of diversity and compositional imbalance—occurs in uremia and may promote inflammation and cardiovascular pathology. While lower microbial diversity has been linked to worse outcomes in other diseases, its association with mortality in ESKD is unclear. This study aimed to determine whether gut microbial diversity predicts all-cause mortality in hemodialysis patients and to characterize microbiome differences between nonsurvivors and survivors.

Prior research indicates the gut microbiota is central to immune and metabolic homeostasis, with high alpha diversity often reflecting better health. Uremia disrupts gut microbiota (dysbiosis), contributing to inflammation and uremic toxins implicated in CVD. Lower microbial diversity has been associated with worse survival in allogeneic hematopoietic-cell transplantation and COPD cohorts. In ESKD, malnutrition and inflammation correlate with decreased diversity. Microbiome-related metabolites (e.g., TMAO, secondary bile acids, SCFAs) have been linked to CVD. Some studies paradoxically note positive correlations between certain uremic toxins and microbial diversity in healthy cohorts, underscoring complex host–microbiome interactions and the need for integrative omics in CKD contexts.

Design: Prospective cohort study in an outpatient dialysis unit (Taipei Tzu Chi Hospital, Taiwan), with a nested matched case–control analysis for microbiome composition comparisons by mortality status. Population: 250 ESKD patients on thrice-weekly hemodialysis screened (Nov 2017–Feb 2018). Exclusions: active malignancy, liver cirrhosis, antibiotic use within 3 months before enrollment. A total of 109 patients provided fecal samples and were analyzed. Demographics and comorbidities were ascertained from interviews and medical records. Clinical measures: Anthropometrics and body composition (BCM bioimpedance) measured 1 h post mid-week dialysis. BMI, lean tissue index, fat tissue index calculated. Diet via modified short-form FFQ; nutrition via 7-point SGA; physical activity via PASE; labs included albumin (bromocresol green), glucose, lipids, electrolytes; inflammation markers IL-6 and TNF-α via ELISA (R&D Systems). Adequacy measures included Kt/V and nPNA. Outcomes: All-cause mortality (causes from official death certificates); secondary events: CV events (nonfatal MI, stroke, HF hospitalization, CV death) and infection-related hospitalizations. Censoring at transfer, transplantation, or end of follow-up (Feb 2020). Fecal sampling and sequencing: Home stool collection, refrigerated transport (4 °C) within 24 h, aliquoted; 200 mg in InhibitEx buffer. DNA extracted (Qiagen DNA Mini Kit). V3–V4 16S rRNA gene amplified with barcoded primers (341F/805R). Libraries (insert 465 bp) prepared with TruSeq Nano; sequencing on Illumina MiSeq v3 kit. All samples sequenced in one batch to minimize batch effects. Bioinformatics: Paired-end read merging with USEARCH (min 8 bp overlap), quality trimming with Mothur (min Q 27; length 400–550 bp); chimera removal with USEARCH (reference mode, 3% divergence). OTU clustering at 97% identity via UPARSE; taxonomy assigned against the Greengenes reference database. Alpha diversity (Simpson, Shannon) computed with phyloseq; beta diversity (Bray–Curtis) visualized by PCoA using ade4; significance tested by Monte Carlo permutations (1000). No rarefaction prior to diversity calculations. Differential taxa assessed using LEfSe (Kruskal–Wallis/Wilcoxon, LDA score >2, alpha <0.05). Enterotypes inferred using a classifier trained on MetaHIT samples. Statistical analysis: Patients dichotomized by median Simpson index into lower vs higher diversity groups. Survival compared with Kaplan–Meier and log-rank tests. Cox proportional hazards models estimated hazard ratios (HRs) for all-cause mortality and secondary outcomes; adjusted models included age, sex, and Charlson comorbidity index to avoid overfitting given low event rates. Correlations between diversity and clinical variables assessed by Spearman and partial Spearman (adjusted for age, sex, Charlson index). Case–control matching for microbiome composition: 14 nonsurvivors matched 1:4 to 56 survivors by age (±5 years) and sex.

• Cohort: 109 hemodialysis patients; mean age 68.4±10.4 years; 57 men; 49.5% with diabetes; 45.9% with CVD; dialysis vintage 8.0 (4.6–11.0) years. 11,968,852 high-quality merged 16S reads analyzed.
• Diversity correlations: Simpson index correlated with BMI (rs=0.252, P=0.008), SGA score (rs=0.263, P=0.008), PASE score (rs=0.309, P=0.004), and inversely with IL-6 (rs=−0.293, P=0.002) and TNF-α (rs=−0.264, P=0.011); associations remained significant after partial adjustment.
• Mortality: Over median 2.1 years, 15 deaths (13.8%): 11 non-CV, 4 CV (infections n=5, malignancies n=2 most common non-CV causes). Kaplan–Meier showed higher mortality in lower-diversity group (P=0.015). Estimated overall survival: 94.5% (higher diversity) vs 77.8% (lower diversity).
• Cox models (Simpson above vs below median): Unadjusted HR 0.24 (95% CI 0.07–0.84; P=0.026); adjusted for age/sex HR 0.24 (0.07–0.87; P=0.030); adjusted for age, sex, Charlson index HR 0.26 (0.07–0.95; P=0.041) indicating 74% lower mortality risk with higher diversity.
• Secondary outcomes: Higher diversity associated with lower risk of CV events (adjusted HR 0.36; 95% CI 0.15–0.88; P=0.026). No significant association with infection-related hospitalizations (adjusted HR 0.71; 95% CI 0.33–1.52; P=0.374).
• Baseline characteristics by mortality: Nonsurvivors were older; had lower total cholesterol and LDL; higher IL-6 and TNF-α; lower use of arteriovenous fistula; other factors largely similar.
• Matched case–control microbiome comparison (14 dead, 56 alive): Alpha diversity lower in nonsurvivors (Simpson P=0.007; Shannon P=0.028). Beta diversity differences were not significant (P=0.482). Phylum-level composition similar (Bacteroidetes ~60–62%, Firmicutes ~23–26%, Proteobacteria ~7%). Enterotypes predominately Bacteroides in both groups.
• Differential taxa (LEfSe): Survivors enriched in Parabacteroides, Succinivibrio, and Anaerostipes (SCFA producers); nonsurvivors enriched in Oscillospira, Achromobacter, Agrobacterium, PSB_M_3, Lactobacillus, vadinCA02, Alloscardovia, Anoxybacillus, Devosia, and Yersinia.

Lower gut microbial alpha diversity was associated with higher all-cause mortality in maintenance hemodialysis patients, independent of age, sex, and comorbidity burden. Diversity also inversely related to inflammatory cytokines (IL-6, TNF-α) and positively to nutritional and activity measures, aligning with the concept of diversity as a marker of microbiome health. Higher diversity correlated with fewer CV events, suggesting potential cardiometabolic pathways linking microbiota to outcomes in ESKD. The enrichment of SCFA-producing genera (Succinivibrio, Anaerostipes, Parabacteroides) in survivors supports a role for SCFAs in anti-inflammatory signaling and gut barrier integrity. Although phylum-level composition and enterotypes were similar and beta diversity did not differ significantly after matching, specific genera distinguished survivors from nonsurvivors, indicating that particular taxonomic configurations rather than broad compositional shifts may relate to prognosis. The absence of an association with infection-related hospitalizations may reflect low bloodstream infection incidence in this cohort. Overall, findings suggest inflammation may mediate the link between dysbiosis and adverse outcomes in ESKD.

Gut microbial diversity and specific compositional features are strongly associated with all-cause mortality in ESKD patients on hemodialysis. Higher alpha diversity confers substantially lower mortality and fewer cardiovascular events, while nonsurvivors exhibit lower diversity, higher inflammatory cytokines, and depletion of SCFA-producing genera. These results underscore the potential of microbiome-derived markers to identify higher-risk patients and point to inflammation as a plausible mediator. Future research should validate these findings in larger, more diverse cohorts with longer follow-up, incorporate multi-omics to elucidate functional microbiome pathways, and test whether interventions targeting specific microbial taxa or metabolites can improve outcomes in hemodialysis.

• Observational design precludes causal inference.
• Single baseline fecal sampling; temporal changes in diversity near events could bias associations toward the null.
• Prior antibiotic exposures beyond the 3-month exclusion window may have lingering effects on microbiota.
• Study focused on taxonomic associations without functional microbiome profiling; functional characteristics may be equally or more important.
• Small sample size and low number of deaths limited covariate adjustment and precluded subgroup analyses by cause-specific mortality.
• Cohort characteristics (older age, longer dialysis vintage, higher prevalence of diabetes and CVD) may limit generalizability to broader hemodialysis populations.