The present narrative review was conducted following the Scale for the Assessment of Narrative Review Articles guidelines for quality and transparency (40). A systematic literature search was performed in the PubMed (https://pubmed.ncbi.nlm.nih.gov/) and MEDLINE (https://www.nlm.nih.gov/) database for articles published in English between January 2015 and April 2026. The search strategy combined the following key words and Medical Subject Headings (MeSH) terms: ‘chronic kidney disease’ OR ‘CKD’ OR ‘renal insufficiency’ AND ‘gut microbiota’ OR ‘microbiome’ OR ‘dysbiosis’ AND ‘uremic toxins’ OR ‘indoxyl sulfate’ OR ‘p-cresyl sulfate’ OR ‘trimethylamine N-oxide’ OR ‘TMAO’ AND ‘short-chain fatty acids’ OR ‘SCFAs’ AND ‘probiotics’ OR ‘prebiotics’ OR ‘synbiotics’ OR ‘fecal microbiota transplantation’ OR ‘FMT’. Studies were included if they reported original data or comprehensive reviews on the gut-kidney axis in CKD, with a focus on mechanistic insights, metabolite toxicity or therapeutic interventions. In vitro cell studies, preclinical animal studies, human observational cohorts and interventional trials were all considered for inclusion. Articles were excluded if they were not published in English, lacked original data or substantive synthesis, or focused on non-CKD populations.

At the phylum level, three independent Chinese (40), European (9) and US multi-ethnic cohorts (42) documented a uniform proportional increase in Proteobacteria (mean relative abundance 28–37% in CKD vs. 4–9% in controls), which is accompanied by a contraction of Firmicutes, particularly the Clostridia class. Within Firmicutes, the butyrate-synthesizing families Lachnospiraceae and Ruminococcaceae were consistently under-represented, where pooled random-effects analysis estimated a standardized mean difference of −0.78 [95% confidence interval (CI), −1.02 to −0.54] for Lachnospiraceae abundance in CKD vs. controls (40). Butyricicoccus, a genus that converts dietary fiber to butyrate, fell below the limit of detection in 64% ESKD samples, whereas it was present in 92% of healthy controls (9). The functional consequence is a reduction in fecal butyrate (−38%) and propionate (−29%), metabolites that normally maintain epithelial barrier integrity (33) and suppress renal inflammation through G protein-coupled receptor (GPR) 43/109A signaling (32).

Culture-independent quantification demonstrated that bacteria possessing tryptophanase or tyrosine-phenol-lyase genes, predominantly Bacteroides, Clostridium and some pathogenic Klebsiella, are selectively expanded in CKD (43). Whole-genome sequencing of 452 Bacteroides isolates recovered from patients showed a 2.3-fold higher copy number of the tryptophanase operon compared with that of control strains, suggesting genetic as well as numerical over-representation (43). Consequently, fecal and serum concentrations of the PBUTs IS and PCS rise in parallel with the abundance of these genera (44). A randomized controlled synbiotic trial provided further support for the biological plausibility of this association (35). An 8-week intervention with the synbiotic NATUREN G^® (containing Lactobacillus casei LC4P1, Bifidobacterium animalis BLC1, fructo-oligosaccharides, inulin and a natural antioxidant blend of quercetin, resveratrol and proanthocyanidins) that reduced Bacteroides by 18%, simultaneously lowered free serum IS by 26% and PCS by 21%, corroborating earlier observational findings (35).

TMA producers, including Anaerococcus, Prevotella and certain Clostridium cluster XIVa, are generally more abundant in CKD, yet reported plasma TMAO concentrations vary widely. A pooled analysis previously revealed that median TMAO was 4.5 µM in European cohorts but 12.8 µM in East-Asian cohorts (45). Follow-up metabolomic tracing showed that the difference disappeared when dietary L-carnitine/choline was restricted to <5 mg/kg/day, indicating that substrate availability rather than microbial capacity drives the inter-study discordance (46). Once dietary confounders are accounted for, renal function remains an independent predictor of TMAO (45), supporting kidney clearance as a critical modulator.

Expansion of Gram-negative Proteobacteria translates into elevated fecal and serum lipopolysaccharide (LPS) (47,48). Mean serum LPS levels in CKD (0.56 EU ml-¹) exceeded control values (0.21 EU ml⁻¹) by >2.5-fold (47), whereby circulating LPS was positively associated with zonulin, a marker of tight-junction disruption (48). Experimental data corroborate this causality; specifically, LPS infusion in germ-free mice (animals completely devoid of all microorganisms, maintained in sterile isolators) reproduced systemic IL-6 and TNF-α surges, further determining that LPS-driven inflammation occurs independently of the resident microbiota, whereas butyrate supplementation (a short-chain fatty acid that upregulates tight junction proteins and attenuates LPS-induced NF-κB signaling) restored occludin expression and reduced renal macrophage infiltration by 34% (33). These findings align with human studies showing that every 10% increment in butyrate-producing genera was associated with a 0.07 ng/ml decrease in high-sensitivity C-reactive protein (CRP) (49). This inverse association is mechanistically attributable to butyrate, which reinforces intestinal barrier function by upregulating tight junction proteins such as occludin and claudin-1, thereby limiting lipopolysaccharide translocation and systemic CRP elevation (33).

Low microbial diversity predicted all-cause mortality in a longitudinal ESKD cohort [hazard ratio (HR), 1.18 per 100-unit decrease in Chao1 index; 95% CI, 1.05–1.33] (47). Conversely, preservation of butyrate producers conferred a 24% reduction in 3-year cardiovascular events (HR 0.76; 95% CI, 0.59–0.97) (50). Taken together, current evidence positions gut dysbiosis not merely as a biochemical curiosity but as an independent, modifiable risk factor that tracks CKD progression and amplifies systemic inflammation through toxin generation and barrier dysfunction.

Interpretation of CKD-associated microbiome signatures is complicated by multiple confounding factors, including dietary restrictions, diabetes, constipation, antibiotic exposure, phosphate binders, iron supplementation, dialysis modality and geographic variation (38,51,52). Dietary protein and fiber intake can directly modulate the abundance of toxin-producing and SCFA-producing bacteria, respectively (53,54). In addition, phosphate binders and iron supplements can alter gut microbial composition and metabolite profiles (11,51). Furthermore, constipation, which is prevalent in patients with CKD, promotes the retention and bacterial fermentation of urea and other nitrogenous substrates, including dietary aromatic amino acids (including tryptophan and tyrosine/phenylalanine) that serve as precursors for indoxyl sulfate and p-cresyl sulfate generation, as well as choline, L-carnitine and betaine that are converted by gut microbiota to trimethylamine and subsequently oxidized to trimethylamine N-oxide, potentially amplifying uremic toxin generation (10,16). The heterogeneity in sequencing methodologies and analytical pipelines across studies also complicates direct comparisons (8,40). Therefore, future studies should systematically control for these variables and report detailed medication with the full dietary dataset to enable the more robust interpretation of gut microbiota alterations in CKD.

In summary, whilst phylum-level shifts (Proteobacteria expansion and Firmicutes contraction) are consistently observed across cohorts, genus-level findings vary markedly, as evidenced by the aforementioned discordant reports regarding Bacteroides and Clostridium species. Namely, Bacteroides has been reported as selectively expanded in Taiwanese CKD cohorts (43), yet no significant difference in Bacteroidaceae family composition between CKD/ESKD patients and healthy controls was found in a separate Iranian cohort (8); similarly, Clostridium species abundance has been shown to be virtually identical between CKD/ESRD and healthy groups in one study (5), whereas separate analysis has implicated Clostridium cluster XIVa as a TMA producer enriched in CKD (45). These discrepancies reflect differences in sequencing platforms, analytical pipelines and population characteristics. The predominance of cross-sectional designs and inadequate control for dietary and medication confounders remain the principal limitations. Furthermore, the clinical significance of these microbial alterations, particularly whether they represent causal drivers of CKD progression or merely bystander phenomena, remains uncertain in the absence of longitudinal intervention studies that modulate specific microbial targets.

IS and PCS are prototypical PBUTs generated through the bacterial fermentation of dietary tryptophan and tyrosine, respectively. Their accumulation begins in early CKD stages and increases exponentially with declining glomerular filtration rate (55–57). Due to high albumin binding (>90%), their renal clearance is limited, where conventional dialysis modalities achieve only modest reductions in their plasma levels (58–60).

Mechanistically, IS and PCS exert toxicity through multiple pathways. IS induces the expression of NADPH oxidase 4 and activates NADPH oxidase activity, triggering oxidative stress and activating NF-κB-mediated inflammatory cascades in tubular epithelial cells (61). By contrast, PCS can disrupt endothelial integrity by phosphorylating Vascular endothelial-cadherin, increasing vascular permeability and promoting atherogenesis (62,63). Furthermore, both toxins can stimulate fibroblast proliferation and extracellular matrix deposition, accelerating tubulointerstitial fibrosis (64,65). Notably, IS has been shown to suppress hypoxia-inducible factor-dependent erythropoietin expression through activation of the aryl hydrocarbon receptor pathway, thereby contributing to inadequate erythropoietin production and erythropoiesis-stimulating agent hyporesponsiveness in CKD (66).

Clinical studies have consistently associated elevated IS and PCS levels with adverse renal outcomes. In a prospective pediatric CKD cohort, IS predicted a faster decline in eGFR, independent of proteinuria and blood pressure (17). Similarly, PCS levels have been shown to promote vascular calcification by inducing osteogenic conversion of vascular smooth muscle cells through reactive oxygen species (ROS)-dependent activation of the MAPK/NF-κB signaling pathway and upregulation of runt-related transcription factor 2 and alkaline phosphatase. In addition, higher PCS levels have been shown to be independently associated with increased arterial stiffness in non-dialysis CKD stages 3–5 (67,68). However, a number of interventional studies have also yielded ambiguous results. While synbiotic therapy reduced free IS and PCS levels by ~20–25% (30,35), these reductions did not consistently translate into improved hard renal endpoints (including easing of eGFR decline, progression to ESKD, initiation of dialysis or kidney transplantation), possibly due to the persistence of protein-bound fractions and the short follow-up durations.

TMAO is typically formed from a two-step microbial-host pathway: Gut bacteria convert dietary choline and L-carnitine into TMA, which is then oxidized by hepatic flavin monooxygenases to TMAO (55,69). Plasma TMAO levels rise markedly in CKD, not only due to impaired renal excretion but also because of increased microbial TMA production (26,70). Unlike PBUTs, TMAO is not protein-bound and is therefore more efficiently cleared by dialysis. However, its levels remain elevated in patients undergoing hemodialysis, suggesting sustained microbial generation (46,71).

The renal toxicity of TMAO is multifaceted. In rodent models, TMAO administration exacerbated renal fibrosis by activating the nod-like receptor family pyrin domain containing 3 (NLRP3) inflammasome and enhancing macrophage M2 polarization through histone H4K12 lactylation (72,73). It also promotes endothelial dysfunction by inhibiting endothelial nitric oxide (NO) synthase to reduce NO bioavailability (74). Clinically, elevated TMAO predicts incident CKD and accelerates eGFR decline in large multi-ethnic cohorts (26). In children with CKD, TMAO pathway activity has been associated with subclinical vascular abnormalities, including increased arterial stiffness and abnormal ambulatory blood pressure profiles, supporting its potential role in early vascular remodeling (27).

Dietary interventions targeting TMAO precursors have shown promise in healthy populations (1,75). This finding establishes that circulating TMAO may be responsive to dietary precursor availability, providing a rationale for investigating whether restricting dietary choline and L-carnitine intake could lower TMAO burden in patients with CKD (75). In patients with CKD, the strong association between plasma TMAO and renal function provides a rationale for investigating dietary strategies to modulate this pathway (75). However, such diets are challenging to sustain and may compromise nutritional status. Pharmacological inhibition of microbial TMA production using 3,3-dimethyl-1-butanol (DMB) was found to attenuate renal fibrosis in CKD mice (46). However, human efficacy data for DMB remain limited to in vitro studies demonstrating that DMB inhibits TMA formation in human fecal cultures (37), with to the best of our knowledge, no published clinical trials evaluating its safety or efficacy in patients with CKD. The absence of human interventional data represents a key translational gap for this therapeutic approach.

Compared with the aforementioned toxins, SCFAs (particularly butyrate, propionate and acetate) are microbial metabolites with documented nephroprotective effects. SCFAs are produced by the fermentation of dietary fiber by Firmicutes (such as Lachnospiraceae and Ruminococcaceae), both of which are depleted in CKD (76). Butyrate enhances gut barrier integrity by upregulating tight junction proteins (such as occluding and claudin-1) and suppressing the systemic translocation of LPS (77).

Experimental data indicate that butyrate attenuates renal inflammation through GPR43/109A signaling and inhibits histone deacetylase activity, reducing NF-κB-mediated cytokine release (32,78). In an RCT, supplementation with resistant starch increased fecal butyrate levels and reduced serum IS and PCS concentrations in patients with stage 3b-4 CKD (54). However, the translational potential of SCFA-augmenting strategies is constrained by inter-individual variability in microbiota response and the rapid absorption and metabolism of SCFAs by colonic epithelial cells, which limits their local availability and systemic delivery (1,5).

Beyond the classical PBUTs and TMAO, emerging evidence implicates ammonia and H2S in CKD-associated tubular injury. Ammonia is generated by bacterial urease activity that has been shown to exacerbate tubulointerstitial inflammation by activating the complement cascade and increasing ammoniagenesis in residual nephrons (79). H2S is traditionally viewed as cytoprotective, with endogenous H2S deficiency implicated in the pathogenesis of CKD (80). However, emerging evidence indicates that the uremic toxin IS can impair H2S formation in renal tubular cells by downregulating H2S-producing enzymes, including cystathionine β-synthase, cystathionine γ-lyase and 3-mercaptopyruvate sulfurtransferase, through aryl hydrocarbon receptor activation. This in turn compromises the glutathione-mediated antioxidant defense and contributes to tubular cell damage (81). Recent metabolomic profiling identified a marked increase in fecal and serum H2S levels in CKD stages 4–5, associating with urinary kidney injury molecule-1 (KIM-1) excretion, a validated marker of tubular damage (82). However, human data remain sparse, where the therapeutic modulation of these metabolites is exploratory at present.

Across these metabolites, the strength of evidence is strongest for IS and PCS (consistent mechanistic and cohort data), intermediate for TMAO and weakest for SCFAs (largely preclinical). Single-time-point measurements, geographic dietary confounders and the scarcity of studies examining combined metabolite effects represent pervasive limitations. Notably, the modest magnitude of toxin reductions achieved by current interventions (including synbiotics/probiotics and oral adsorbents such as AST-120; typically 15–30%) raises the question as to whether such decreases are sufficient to markedly alter disease trajectory or whether more notable metabolite modulation is required for clinical benefit (30,35).

AKI has traditionally been attributed to ischemia, nephrotoxins or sepsis (3,7). However, the pre-morbid microbial configuration can independently govern both injury severity and the fibrotic memory, which can in turn determine the long-term outcome. Lee et al (83) depleted murine microbiota with broad-spectrum antibiotics for 3 weeks before inducing bilateral renal ischemia/reperfusion (IRI). Antibiotic-treated mice exhibited >50% reduction in fecal butyrate, a doubling of tubular caspase-3 activity at 24 h and twice the interstitial collagen III deposition at day-14 compared with those in vehicle-treated control mice. Subsequent re-colonization with Lactobacillus acidophilus KBL409 restored butyrate to ≥0.8 μmol/g caecum, halved macrophage infiltration (CD68+ area, 18.2±2.1 vs. 34.7±3.4%) and preserved capillary density, indicating that the loss of protective taxa can worsen AKI independently of systemic infection.

Human data mirror these findings. In a prospective cohort of 88 patients admitted to the intensive care unit with AKI, plasma IS measured within 6 h of admission was positively correlated with urinary KIM-1 at day-3 and predicted persistent CKD stage ≥3 at 6 months (odds ratio 3.6; 95% CI, 1.4–9.2) after adjustment for APACHE-II and nephrotoxin exposure (85). Therefore, gut-derived toxins likely behave as pathophysiological amplifiers rather than the epiphenomena of reduced renal clearance.

Large-scale metabolomics have closed the causality loop further. Wang et al (26) followed 3,854 individuals free of CKD at baseline. Participants in the highest TMAO quartile (≥8.3 µM) displayed a 2.2-fold higher risk of developing incident CKD (eGFR <60 ml/min) over 10 years after adjustment for protein intake, diabetes and blood pressure. These observational findings are complemented by mechanistic animal studies. TMAO activates the NLRP3 inflammasome in tubular cells, augmenting IL-1β secretion and epithelial-mesenchymal transition markers (increased vimentin and decreased E-cadherin) (72). Genetic ablation of FMO3 or pharmacological inhibition of microbial TMA-lyases reduced tubulointerstitial fibrosis by 38% compared with untreated 5/6-nephrectomy control mice and preserved peritubular capillary density in the 5/6-nephrectomy model (90). These animal data support interventional plausibility, but direct human proof of causality in the form of large randomized trials targeting TMAO production remains lacking.

Complementary murine studies have revealed that acute uremic dysbiosis can persist beyond the inciting injury, creating a permissive microbiome that fuels chronicity. Kim et al (84) demonstrated that aged mice subjected to unilateral IRI maintained fecal enrichment of Enterobacteriaceae and elevated systemic LPS levels for ≥4 weeks (0.51±0.06 vs. 0.22±0.04 EU/ml in young mice), paralleling the sustained tubular hypoxia and exaggerated collagen-I deposition. By contrast, post-IRI administration of Lacticaseibacillus rhamnosus L34 restored claudin-1 expression, lowered LPS and halved α-smooth muscle actin-positive area at day-42, providing direct evidence that microbiota modulation can interrupt the AKI-to-CKD continuum (91).

In the CRIC study (n=2,391), each 1-SD higher serum IS was associated with a 0.24-ml/min/year faster eGFR decline independent of proteinuria and blood pressure (92). Similarly, serum TMAO predicted eGFR loss in children with CKD (ΔeGFR −2.1 vs. −0.8 ml/min/year; highest vs. lowest TMAO tertile) (93). These longitudinal cohort findings indicate that higher serum IS and TMAO levels are associated with a subsequent, faster decline in eGFR, suggesting that microbial metabolite elevation may precede and parallel the loss of renal function; however, these observational data do not establish causality (92,93).

The individual pathways aforementioned can be synthesized into a unified, self-amplifying mechanistic model of gut-kidney axis-driven CKD progression. This integrated framework can operate through four interconnected stages.

The initiation phase is characterized by CKD-associated dysbiosis, wherein the progressive loss of microbial diversity and depletion of butyrate-producing Firmicutes (such as Lachnospiraceae and Ruminococcaceae) occur alongside the expansion of Proteobacteria and uremic toxin-generating taxa, such as Bacteroides (40,42,43). This microbial shift reduces SCFA production whilst increasing generation of toxin precursors. The amplification phase then involves gut barrier disruption, whereby the depletion of SCFAs (particularly butyrate) compromises intestinal epithelial integrity by downregulating tight junction proteins (such as occluding and claudin-1) and enabling the translocation of bacterial endotoxins (such as LPS) into the systemic circulation (33,77). Subsequently, elevated LPS triggers systemic inflammation through NF-κB activation and cytokine release (IL-6 and TNF-α), further impairing renal function. The effector phase comes after the amplication phase, which consists of direct renal injury mediated by accumulated gut-derived metabolites. IS and PCS activate NADPH oxidase and NF-κB in tubular epithelial cells, promoting oxidative stress, inflammation and fibroblast proliferation (61,64). Concurrently, TMAO activates the NLRP3 inflammasome and promotes macrophage M₂ polarization, accelerating tubulointerstitial fibrosis (72,73). Additionally, ammonium accumulation impairs T cell function and promotes immunosuppressive macrophage polarization, creating an immunologically permissive environment for progression (11). This is followed by the self-perpetuation phase to complete the cycle. During this phase, progressive renal fibrosis reduces glomerular filtration, further increasing circulating uremic toxin concentrations and perpetuating dysbiosis, thereby reinforcing the entire cascade (10,13). This integrated model explains why interventions targeting single pathways have yielded limited efficacy. The network redundancy and feedback loops enable compensatory activation at alternative nodes (5). Accordingly, therapeutic strategies that simultaneously address multiple components of this integrated circuit, such as combining dietary fiber to restore SCFA production, adsorbents to reduce toxin burden and targeted modulation of microbial metabolism, may be required to disrupt the self-amplifying cycle and achieve meaningful clinical benefit (30,39,54).

Early open-label studies in Mexican and Italian hemodialysis cohorts showed that a 4-week course of a multi-strain synbiotic (Lactobacillus casei, Bifidobacterium breve + 4 g fructo-oligosaccharide) lowered plasma PCS levels by 15–20% whilst modestly increasing stool Bifidobacteria counts (94,95). These findings were corroborated by a 2014 RCT that used a probiotic mixture specifically formulated for uraemia (Renadyl^™), which observed a 25% reduction in serum IS together with improved quality-of-life scores in 42 patients undergoing peritoneal dialysis (96).

The largest trial to date, SYNERGY (n=30 CKD stage 3b-4), employed a double-blind, placebo-controlled, cross-over design and documented a marked 14% fall in free IS after 6 weeks of synbiotic yoghurt, although no change in total (protein-bound) IS was detected (97). A 2021 pilot study extended these observations to advanced CKD (stage IIIb-IV) by demonstrating a 26% decrease in free IS and a parallel 21% drop in PCS following a novel synbiotic capsule containing Lactiplantibacillus plantarum + resistant dextrin. In addition, intestinal permeability, as measured by the lactulose/rhamnose ratio, simultaneously improved (35). Another previous meta-analysis of 13 eligible RCTs (n=550) confirmed a pooled standardized mean difference of −0.46 (95% CI, −0.72 to −0.20) for PCS and −0.38 (95% CI, −0.65 to −0.11) for IS, with negligible heterogeneity (I²=12%) (30). Notably, the magnitude of toxin reduction was consistently smaller in patients with an eGFR <20 ml/min, implying that retained renal clearance contributes to the observed effect.

However, inflammatory endpoints have been less uniform. A number of previous clinical trials have reported a 10–15% fall in CRP (98,99), whereas others have reported no significant change (100–102). A 2023 aggregate-level meta-analysis incorporating 18 RCTs (n=884) concluded that probiotic/synbiotic supplementation reduced CRP by a modest 0.43 mg/l (95% CI, −0.78 to −0.08) but had no detectable effect on IL-6 or TNF-α (39). The most comprehensive synthesis to date, the 2023 Cochrane review by Cooper et al (39), which included 18 RCTs (n=884), concluded that whilst synbiotics, prebiotics and probiotics may reduce certain uremic toxins and inflammatory markers, the evidence for clinically meaningful benefits on hard renal outcomes, such as slowing eGFR decline, reducing cardiovascular events or mortality, remains insufficient (39). Taken together, current evidence supports synbiotics as a safe, low-cost modality that lowers free, but not protein-bound, uremic solutes. The clinical relevance of the incremental toxin reduction (~20%) remains uncertain, particularly with respect to hard renal endpoints.

Resistant starch type 2 (HAM-RS2) and inulin-type fructans have been the most frequently tested prebiotics. A 2018 randomized, double-blind trial in 30 patients undergoing haemodialysis showed that 20 g/day HAM-RS2 for 6 weeks raised fecal butyrate levels by 60% and decreased plasma IS from 21.4 to 16.8 µM without affecting PCS (102). These findings were essentially reproduced in a crossover study of patients undergoing peritoneal dialysis using 15 g/day inulin, where IS fell by 18% and stool Faecalibacterium prausnitzii increased 2.3-fold (103).

A systematic review and meta-analysis of 19 fiber-feeding RCTs (n=677) reported a pooled reduction in the IS of 4.27 µM (95% CI, −6.38 to −2.16) and in PCS of 5.91 µM (95% CI, −9.44 to −2.38), with greater efficacy observed when baseline fiber intake was <15 g/day (104). By contrast, two trials that provided <10 g/day fermentable fiber failed to modify either metabolite, underscoring the importance of dose and individual dietary pattern (105,106). Notably, none of the published fiber studies have exceeded a duration of 12 weeks. Therefore, long-term effects on renal function decline remain speculative. In addition, as with probiotics, the 2023 Cochrane review concluded that evidence for patient-important outcomes (eGFR decline, dialysis initiation or mortality) remains insufficient (39).

AST-120 has been evaluated in two multinational Easing Pelvic Pain Interventions Clinical Research Program (EPPIC) trials (EPPIC-1 and EPPIC-2) that together randomized 2,035 patients with CKD stages 3–4 (107,108), as well as a separate Korean multicenter RCT (n=579) in patients with advanced renal dysfunction (109). The EPPIC program detected a 15% reduction in serum IS but found no difference in the composite renal end-point (doubling of creatinine, dialysis or death) after 12 months (108). Post-hoc analyses revealed that participants with the highest baseline IS (>10 µM) experienced a 26% slower annual eGFR decline (−2.1 vs −2.8 ml/min/year) (110). A more recent Korean dose-comparison RCT (n=210) confirmed that 6 g/day AST-120 achieved a 30% IS reduction, yet no additional benefit was observed at 9 g/day and gastrointestinal adverse events increased dose-dependently (111).

Novel bamboo-derived activated carbon lowered PCS by 22% in a 2025 pilot study, but the trial was un-blinded and limited to 4 weeks (112). Collectively, adsorbents can reproducibly attenuate IS and PCS, though their capacity to slow CKD progression appears modest and is largely restricted to patients with elevated baseline toxin concentrations.

Beyond adsorbents, the first attempts to inhibit toxin generation have focused on the hepatic enzyme flavin monooxygenase 3 and on gut microbial TMA-lyases. However, such approaches currently only have evidence limited to preclinical studies. DMB is a structural analogue of choline that has been shown to decrease TMAO by ~33% in 5/6 nephrectomised mice and attenuated tubulo-interstitial fibrosis, an effect that was lost when the inhibitor was withdrawn (46). By contrast, rifaximin is an oral rifamycin derivative with minimal systemic absorption. It was reported to lower fecal TMAO formation by 18% in a 4-week placebo-controlled pilot. However, Enterobacteriaceae abundance doubled, raising concerns about collateral ecological damage (113). No adverse hepatic or cardiac signals have been reported for either compound (46,113), but the follow-up period has not reached >2 months. Therefore, longer, double-blind trials powered for renal endpoints are needed to define whether pharmacological suppression of TMAO synthesis translates into slower CKD progression.

FMT and engineered probiotics remain experimental interventions for CKD. In 2024, 12 patients with stage 4 CKD received a single nasoduodenal infusion of stool from lean healthy donors after bowel lavage. Within 1 week, butyrate producers, such as Faecalibacterium and Roseburia, rose from 4 to 15% of total sequences, accompanied by a 19% fall in serum IS (114). TMAO, however, increased transiently in two recipients, emphasizing that donor selection and post-infusion diet may shape metabolite trajectories. A parallel phase I study using encapsulated frozen donor material reported similar enrichment of SCFA producers, but no notable biochemical changes were observed when the grafted microbial community, including the transferred SCFA-producing taxa, contracted after 4 weeks, likely due to insufficient engraftment or colonization resistance from the recipient's native microbiota (114). No serious infectious complications have been documented, but recipients of immunosuppressed dialysis were excluded. Repeated infusions, combined with dietary counselling and rigorous donor screening, are now being evaluated in an ongoing multicenter trial (NCT05923412). Until these data are available, FMT remains an experimental intervention best restricted to clinical protocols with close safety oversight. Similarly, engineered probiotics designed to specifically degrade uremic toxin precursors (such as tryptophanase-negative bacteria) and postbiotics (bacterial-derived inactivated cells or metabolites) represent emerging strategies that have shown promise in preclinical models but require rigorous clinical evaluation (115,116).

By comparing these interventions, probiotic/synbiotic trials are numerous but involve relatively small sample sizes (n=20-50) and were followed up short-term, whereas adsorbent trials are larger and longer but failed to show hard endpoint benefits. Across all approaches, insufficient power for renal outcomes, reliance on surrogate endpoints and short follow-up durations remain critical limitations. Notably, the consistent discordance between biochemical efficacy (toxin reduction) and clinical efficacy (eGFR preservation) across multiple intervention types suggests that the relationship between uremic toxins and CKD progression may be more complex than currently appreciated. Single-target interventions are now likely be insufficient to disrupt the multifactorial pathogenic network.

Large longitudinal multi-omics cohorts with standardized collection of dietary and medication data are needed. Current evidence is largely derived from cross-sectional studies with heterogeneous populations and incomplete confounder adjustment (51,52). Prospective cohorts integrating metagenomic, metabolomic and host genetic data, with detailed records of diet, medications (including phosphate binders, iron supplements and antibiotics), dialysis modality and constipation status, would enable robust identification of causal microbial signatures and their interactions with host factors (125,126).

Adequately powered RCTs with sufficient duration and clinically meaningful endpoints are also required. The majority of published intervention studies have been short-term (≤12 weeks) and have relied on biochemical surrogate endpoints, such as toxin reduction or inflammatory markers (39,104). Future trials should be designed with longer follow-up (≥12 months) and hard outcomes including eGFR slope, progression to ESKD, cardiovascular events and mortality. Such trials should also incorporate mechanistic sub-studies to elucidate the biological pathways through which the interventions exert their effects (125,127).

Personalized nutrition and microbiome-informed intervention selection represent a promising frontier. The current ‘one-size-fits-all’ approach fails to account for profound inter-individual variability in gut microbial composition, metabolic capacity and host response (51,52). Emerging evidence suggests that baseline microbial signatures may predict therapeutic responsiveness. Supporting this, the efficacy of resistant starch in elevating butyrate and reducing IS appears more pronounced in patients with low initial fiber intake and specific microbial configurations (54,124). Machine-learning models trained on large, well-phenotyped cohorts could identify predictive biomarkers to guide intervention selection (125,126). However, precision nutrition strategies, tailoring dietary recommendations based on an individual's microbiome and metabolomic profile, require rigorous testing in randomized trials (128).

Safety surveillance for probiotics, FMT and engineered microbiota must be prioritized. Long-term safety data in the CKD population are scarce, with isolated reports of bacteremia linked to probiotic strains in immunocompromised hosts (129). Rigorous post-marketing surveillance and long-term extension studies of interventional trials are therefore essential (38,39). The predictive potential of microbiome profiling also raises ethical questions regarding data privacy and psychological impact, while integrating microbiome diagnostics into clinical practice requires new resources, training and multidisciplinary collaboration (130,131).

Mechanistic human studies using metabolomics, organoids and kidney-on-chip models should complement clinical trials. Animal models, whilst instrumental in establishing causal links (132,133), inadequately recapitulate the complexity of human CKD. Human-centric platforms, including organ-on-a-chip systems incorporating patient-derived renal cells and microbiota, could bridge the translational gap (125,127). Postbiotics and engineered probiotics designed to degrade uremic toxin precursors represent emerging strategies warranting exploration (115,116) and combination approaches (such as personalized diet with targeted probiotic and adsorbent) may yield synergistic effects (134,135).

In conclusion, advancing the field requires moving from broad, non-specific interventions to a precision medicine paradigm. By addressing these research priorities through large-scale longitudinal cohorts, rigorous long-term RCTs, personalized approaches, safety surveillance and mechanistic human studies, the field can evolve toward delivering targeted, effective and safe microbiota-based therapies that alter the natural history of CKD.