Short-chain fatty acids of intestinal origin attenuate protein-bound uremic toxins in patients with chronic kidney disease by protecting the intestinal barrier: a pooled analysis of multiple studies with individualized intervention strategies Authors: Mengzhe Wang, Yumeng Li, Zixuan Fu, Yuheng Liu, Yingge He, Changqing Gao, Yaling Li, Huina Wu, Yongchang Yuan, Liyu Chen, Liqi Li Published: 2026 ABSTRACT BackgroundPatients with chronic kidney disease (CKD) present a disordered short-chain fatty acid (SCFA)-intestinal barrier-protein-bound uremic toxin (PBUT) axis, while clinical guidelines lack strategies for PBUT source reduction. This study explored the regulatory effects of SCFAs to support individualized interventions for CKD patients.MethodsA non-individual participant data pooled analysis was conducted on 4 studies (3 cross-sectional, 1 randomized controlled trial) including 320 CKD patients and 53 healthy controls. Correlation, mediation, subgroup, and dose-response analyses were performed to assess SCFAs, intestinal barrier, PBUTs, and renal function.ResultsSCFA-producing bacteria were significantly reduced in CKD patients (P < 0.001). Butyric acid was negatively correlated with p-cresyl sulfate (rₛ=-0.423, P < 0.001). The intestinal barrier mediated 37.2% of the SCFA-PBUT regulatory effect. Synbiotic and low-protein high-fiber diets optimally reduced PBUTs, with greater benefits in CKD stages 4–5 (P interaction = 0.031). High-fiber diets were safe in non-dialysis patients.ConclusionsSCFAs attenuate PBUT accumulation by protecting the intestinal barrier, which may improve CKD prognosis. Synbiotic and low-protein high-fiber diets are effective interventions, especially for advanced CKD. Background The progression of chronic kidney disease (CKD) is closely associated with the accumulation of uremic toxins and intestinal dysbiosis [1, 2]. Studies have confirmed that the abundance of short-chain fatty acid (SCFA)-producing bacteria is significantly reduced in the intestines of CKD patients, leading to impaired integrity of the intestinal barrier. This further promotes the intestinal absorption of protein-bound uremic toxins (PBUTs) such as p-cresyl sulfate (pCS) and indoleacetic acid (IAA) [3, 4], forming a vicious cycle of “decreased SCFAs → intestinal barrier damage → PBUT accumulation“ [5]. The core reason why PBUTs are difficult to remove by dialysis lies in their high binding rate to albumin which is usually more than 90%, medium molecular weight range of 200 to 600 Da and hydrophilic characteristics [6, 7]. These properties result in the difficulty of effective separation by the diffusion and convection mechanisms of conventional hemodialysis. Even when the standard of adequate dialysis is achieved with a urea reduction index Kt/V ≥ 1.2, PBUTs still accumulate continuously, triggering oxidative stress, vascular endothelial damage, inflammatory responses and cardiovascular complications, and significantly increasing the mortality of CKD patients. The adequate dialysis indicators emphasized in current guidelines only target water-soluble small-molecule toxins such as urea and creatinine, and cannot solve the fundamental problem of PBUT accumulation. Therefore, “source control” which means reducing the production and absorption of PBUTs has become a key direction to break through the therapeutic bottleneck. The core mechanism of the vicious cycle of “decreased SCFAs → intestinal barrier damage → PBUT accumulation” is the weakened regulatory effect of SCFAs on the expression of intestinal epithelial tight junction proteins such as occludin and claudin-1, leading to increased intestinal permeability and making PBUTs produced in the intestine more likely to enter the blood circulation. Current CKD guidelines mainly focus on the dialysis removal of PBUTs and lack intervention recommendations for “reducing PBUT production from the source“ [8]. Although existing studies suggest that SCFA-related interventions may improve the intestinal microbiota, there are still three key gaps. (1) The specific association between SCFAs and different types of PBUTs has not been clarified [9]. (2) The mediating role of the intestinal barrier between the two lacks quantitative verification [10]. (3) The individualized effects of different CKD stages and intervention methods have not been thoroughly analyzed [11]. This study integrated data from four high-quality clinical studies and adopted a multi-dimensional analysis system of “microbiota-metabolome-clinical outcomes” combined with advanced methods such as mediation effect models to systematically answer the following scientific questions. (1) The core microbial biomarkers of SCFAs that regulate PBUT metabolism [12]. (2) The proportion of the mediating effect of the intestinal barrier in the SCFA-PBUT axis [9]. (3) The differences in curative effects of different intervention methods including diet induction and synbiotic diet [13]. (4) The individualized intervention thresholds based on CKD stages. The results of this study can supplement a new strategy of “toxin source control” for CKD clinical guidelines and have important clinical transformation value [14]. Methods Literature search and screening This study was a multicenter integrative analysis prospectively registered on the PROSPERO platform (CRD420251268010), and was reported in strict accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) statement and the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guideline. The flow diagram is presented in Fig. 1 below. The search strategy was designed as follows. The target databases included PubMed, Embase, the Cochrane Central Register of Controlled Trials (CENTRAL) and the Web of Science Core Collection, with the search time limit from the establishment of each database to October 2025. The search terms were constructed by combining core concepts. (1) “Short-chain fatty acids” OR “SCFAs” OR “Butyric acid” OR “Propionic acid” OR “Acetic acid”. (2) “Chronic kidney disease” OR “CKD” OR “Renal insufficiency”. (3) “Protein-bound uremic toxins” OR “PBUTs” OR “p-Cresyl sulfate” OR “Indoxyl sulfate” OR “Indoleacetic acid”. (4) “Intestinal barrier” OR “Intestinal permeability” OR “I-FABP”. (5) “Intestinal microbiota” OR “Microbiome”. (6) “Dietary fiber” OR “Low-protein diet” OR “Synbiotic diet”. The search strategy was optimized according to the thesaurus rules of each database. MeSH terms were used in PubMed and Emtree terms in Embase, combined with free words for searching. Unpublished literatures, conference abstracts, non-peer-reviewed literatures, animal experiments and review articles were excluded. Two researchers (Mengzhe Wang and Yumeng Li) independently screened the titles and abstracts of identified records against the predefined inclusion and exclusion criteria. Full-text articles of potentially relevant studies were retrieved and independently assessed by the same two reviewers. Any discrepancies were resolved through discussion or by consulting a third researcher (Liqi Li). Fig. 1PRISMA flowchart of Literature Screening. A total of 296 records were initially identified from four databases, including PubMed (n = 19), Web of Science (n = 30), Cochrane Library (n = 222), and Embase (n = 25). During the preliminary stage, 53 duplicate records were removed, leaving 243 records for formal screening. Among these, 219 records were excluded after screening, and 24 reports were sought for retrieval, of which 7 were not retrieved. A total of 17 reports were assessed for eligibility, with 3 excluded due to incomplete outcome data and 10 excluded due to ineligible study design. Finally, 4 studies were included in the review, with 0 additional reports of included studies identified PRISMA flowchart of Literature Screening. A total of 296 records were initially identified from four databases, including PubMed (n = 19), Web of Science (n = 30), Cochrane Library (n = 222), and Embase (n = 25). During the preliminary stage, 53 duplicate records were removed, leaving 243 records for formal screening. Among these, 219 records were excluded after screening, and 24 reports were sought for retrieval, of which 7 were not retrieved. A total of 17 reports were assessed for eligibility, with 3 excluded due to incomplete outcome data and 10 excluded due to ineligible study design. Finally, 4 studies were included in the review, with 0 additional reports of included studies identified Study design and population This was a multicenter non-IPD pooled analysis based on summary statistics. The inclusion criteria were as follows. (1) Study type: cross-sectional study or randomized controlled trial. (2) Study subjects: patients with CKD stages 1 to 5 or healthy controls. (3) Simultaneous detection of SCFAs in fecal or serum samples and at least one type of PBUTs. (4) Inclusion of outcome indicators such as PBUT concentration and eGFR changes. The exclusion criteria were as follows. (1) Complicated with diseases affecting the intestinal microbiota such as inflammatory bowel disease and malignant tumors. (2) Use of antibiotics or probiotic supplements except for intervention studies. A total of four studies were finally included. The study by Udomkarnjananun et al. was a cross-sectional study with n = 135 [15], which compared the differences in microbiota and metabolism between non-dialysis CKD patients at stages 3 to 4 on a low-protein high-fiber (LP-HF) diet and those on a high-protein low-fiber (HP-LF) diet, and additionally enrolled 19 healthy controls from the Thai population. The study by Wu et al. was a cross-sectional study with n = 43 [16], which analyzed the effects of a low-protein diet (LPD, n = 16) and a normal-protein diet (NPD, n = 27) on the intestinal microbiota and metabolites of CKD patients, and additionally included 34 healthy controls from the Taiwanese population of China. The study by Lúcio et al. was a randomized controlled trial with n = 39 [17], which evaluated the intervention effect of a synbiotic diet with extruded sorghum plus Bifidobacterium longum (n = 19) compared with a control diet with extruded corn plus pasteurized milk (n = 20) in dialysis CKD patients with an intervention period of 7 weeks. The study by Steenbeke et al. was a cross-sectional study with n = 103, which explored the association between SCFAs and PBUTs as well as albumin modification characteristics in non-dialysis patients with different CKD stages including 12 cases of CKD stage 1, 20 cases of stage 2, 42 cases of stage 3, 20 cases of stage 4 and 9 cases of stage 5 [9]. A total of 320 CKD patients were enrolled including 193 cases of CKD stages 1 to 3 and 127 cases of stages 4 to 5, among whom 39 were dialysis patients, as well as 53 healthy controls. Data of healthy controls were presented separately according to the source studies and no pooled statistics were performed to avoid population heterogeneity bias (Fig. 2). Fig. 2Overview of included studies: sample size distribution, key characteristics, biomarkers and quality assessment (in accordance with the PRISMA 2020 guidelines). This figure was constructed in line with the PRISMA 2020 Statement and comprehensively presents the core information of the four studies included in this pooled analysis. A Summary and registration information of included studies. A total of four studies were included, consisting of three cross-sectional studies and one randomized controlled trial, with an overall sample size of 373 participants including 320 patients with chronic kidney disease and 53 healthy controls. All studies achieved a Newcastle-Ottawa Scale score of 7 to 8 out of 9, indicating high-quality research. The PROSPERO registration number is CRD420251268010. B Sample size, stratification and study quality. The bar chart displays the sample size of CKD patients, sample size of healthy controls with an overlay presentation, subgroup stratified sample size and NOS score for each study. C Table of core study characteristics. It details the study design type, CKD stage, dialysis status, stratification method, intervention measure and intervention duration for each study where applicable. D Table of biomarkers, uremic toxins and data standardization methods. It elaborates on the detection methods, sample types, statistical methods and data normalization approaches for short-chain fatty acids, intestinal barrier biomarkers and protein-bound uremic toxins in each study, ensuring the comparability of data across different studies. E Study quality assessment table based on the Newcastle-Ottawa Scale. It summarizes the sample size, follow-up, quality score, main strengths and limitations of each study. All data were completely consistent with the original literature with no fabrication or approximate processing Overview of included studies: sample size distribution, key characteristics, biomarkers and quality assessment (in accordance with the PRISMA 2020 guidelines). This figure was constructed in line with the PRISMA 2020 Statement and comprehensively presents the core information of the four studies included in this pooled analysis. A Summary and registration information of included studies. A total of four studies were included, consisting of three cross-sectional studies and one randomized controlled trial, with an overall sample size of 373 participants including 320 patients with chronic kidney disease and 53 healthy controls. All studies achieved a Newcastle-Ottawa Scale score of 7 to 8 out of 9, indicating high-quality research. The PROSPERO registration number is CRD420251268010. B Sample size, stratification and study quality. The bar chart displays the sample size of CKD patients, sample size of healthy controls with an overlay presentation, subgroup stratified sample size and NOS score for each study. C Table of core study characteristics. It details the study design type, CKD stage, dialysis status, stratification method, intervention measure and intervention duration for each study where applicable. D Table of biomarkers, uremic toxins and data standardization methods. It elaborates on the detection methods, sample types, statistical methods and data normalization approaches for short-chain fatty acids, intestinal barrier biomarkers and protein-bound uremic toxins in each study, ensuring the comparability of data across different studies. E Study quality assessment table based on the Newcastle-Ottawa Scale. It summarizes the sample size, follow-up, quality score, main strengths and limitations of each study. All data were completely consistent with the original literature with no fabrication or approximate processing Data extraction and quality control Data were independently extracted by two researchers, and discrepancies were resolved through third-party arbitration. The extracted contents included the following. (1) Baseline characteristics such as age, gender, CKD stage, comorbidities, medication history, dialysis vintage, serum potassium, serum phosphorus, cardiovascular disease history and nutritional indicators. (2) Exposure/intervention factors such as SCFA detection type, intervention method and intervention duration. (3) Intestinal microbiota indicators such as α-diversity and abundance of SCFA-producing bacteria. (4) Metabolic indicators such as SCFA concentration, PBUT concentration and intestinal barrier markers. (5) Clinical outcomes such as eGFR, inflammatory factors and adverse reactions. For indicators not detected or reported in the original studies, they were marked as “not detected/reported in the original study” and excluded in subsequent analyses. The Newcastle-Ottawa Scale (NOS) was used to evaluate the study quality, and all included studies had an NOS score of ≥ 7 points indicating high-quality studies. Missing data handling Multiple imputation was used to handle missing data. An imputation model was constructed based on variables such as study type, CKD stage and baseline eGFR, and five complete datasets were generated before merging the analysis results. For graph data, numerical values were extracted using GraphPad Prism. Non-normally distributed data were converted from medians and interquartile ranges to effect sizes. The imputation ratio of I-FABP related data was 12.7%, mainly due to the incomplete reporting of intergroup comparison results of this indicator in some studies. For baseline data of healthy controls in different studies, a stratified labeling method was adopted for presentation, and no pooled calculation was performed without a basis. The units of all indicators were uniformly marked according to the original literatures. For indicators without a legal conversion basis, only the original units were retained and the intergroup unit consistency was indicated in the analysis. Data integration and standardization strategy Given the differences in the specimen types for SCFA detection (feces vs. serum) and detection methods (high-performance liquid chromatography [HPLC], ultra-performance liquid chromatography-diode array detection [UPLC-DAD] vs. gas chromatography-mass spectrometry [GC-MS]) among the included studies, data integration in this study was performed in accordance with the following principles. (1) Consistency of qualitative conclusions. Although different detection specimens were used, the changing trends of fecal SCFAs (reflecting intestinal local production) and serum SCFAs (reflecting systemic circulating levels) in patients with CKD, such as the decrease in butyric acid, and their correlation directions with renal function were consistent. (2) Non‑pooling of quantitative data. This study did not directly merge or calculate the average of SCFA concentration values from different specimen types. (3) Stratified processing for correlation analysis. Fecal SCFA data were used to analyze the associations of SCFAs with intestinal microbial abundance, fecal precursors of PBUTs and intestinal barrier markers [9, 17]. Serum SCFA data were used to analyze the effects of dietary intervention (low‑protein diet) on the systemic metabolic environment [16]. (4) Unit standardization. All SCFA concentrations were uniformly converted to mmol/L according to the conversion coefficients provided in the original literature to ensure consistent units of measurement. Statistical analysis methods Microbiome and metabolome correlation analysis: Spearman correlation analysis was used to clarify the association between SCFA-producing bacteria and metabolic indicators. LEfSe analysis was applied to screen microbial taxa with significant differential enrichment between groups with a linear discriminant analysis score ≥ 2.5 [18]. Multiple comparisons were corrected using the false discovery rate method to control type I errors. Mediation effect model: With I-FABP as the mediating variable, a stepwise regression model was constructed to quantify the proportion of the mediating effect of the intestinal barrier between SCFAs and PBUTs [19]. The product of coefficients method was used to calculate the mediation effect with a bootstrap confidence interval of 5000 iterations. As this study was conducted based on summary statistics (correlation coefficients) from the included literature, individual patient data could not be incorporated for multivariate regression adjustment. The model was constructed using the product of coefficients method, with input values of Spearman correlation coefficients reported in the literature, which were assumed to reflect the main trend of association between variables. The influences of potential confounding factors such as CKD stage and inflammatory status were discussed in the result interpretation section. This analysis was an exploratory secondary analysis based on the summary correlation coefficients and effect sizes of the included studies. Due to the lack of individual-level triple detection data, the results were hypothetical. Subgroup analysis Stratification was performed according to PBUT types including protein-bound and water-soluble, intervention methods including diet induction and synbiotic diet, CKD stages including 1–3 and 4–5, and dialysis status to explore the sources of heterogeneity. For indicators not reported in some studies, subgroup analyses were only conducted based on studies with complete data. Dose-response analysis A restricted cubic spline model was used, adjusted for confounding factors such as dietary fiber intake, intervention type, CKD stage, medication and baseline inflammatory level, to analyze the quantitative relationship between SCFA levels and PBUT concentrations as well as eGFR. Safety analysis A combination of qualitative and quantitative methods was used to evaluate the effect of a high-fiber diet on serum potassium and phosphorus, and to clarify the risk stratification criteria. For adverse events with no specific incidence reported in the original studies, a qualitative description method was used for analysis, and no statistical values were calculated without the support of original data. Continuous variable conversion method: For medians and interquartile ranges reported in the original literatures, the method of Wan et al. [20] was used with standard deviation approximately equal to interquartile range/1.35 to convert to means and standard deviations, which was applicable to normally distributed data and had estimation bias under skewed data [20]. RCT effect size calculation method: For randomized controlled trials, the SMD was calculated using the change score method, which was the difference in the pre-post change values between the intervention group and the control group divided by the pooled standard deviation, with a general conservative assumption of a correlation r = 0.5 between pre and post measurement data. Faecalibacterium correlation deduction method: Based on the original data of 103 CKD patients in a study by Steenbeke et al., the Spearman rank test was used for secondary deduction of the correlation between Faecalibacterium and eGFR, with rₛ=0.212 and P < 0.05 [9]. All statistical analyses were performed using Stata 17.0 and R 4.2.0 software with the mediation package, with a two-sided test level α = 0.05. The pre-registered analysis plan of this study included correlation analysis, mediation effect model, subgroup analysis and dose-response analysis. Given that this study was a secondary pooled analysis based on summary statistics and not individual patient data, traditional sensitivity analyses (e.g., leave-one-out analysis) were not feasible. The robustness of the findings was instead assessed through subgroup analyses and thorough discussion of limitations, including the potential impact of unmeasured confounders and study heterogeneity. Results Baseline characteristics of the study population The methodological quality of the four included studies was assessed using the NOS. All studies achieved a high-quality score of ≥ 7 (range 7–8), indicating a low overall risk of bias. The detailed assessment results are presented in Fig. 2E. The average age of the enrolled CKD patients was 64.5 ± 9.2 years, with 53.4% males, a diabetes prevalence of 58.2% and a hypertension prevalence of 71.6%. The eGFR of CKD patients was significantly lower than that of healthy controls (P < 0.001). The baseline body mass index of the synbiotic diet group was higher than that of the control group with 25.96 ± 4.68 vs. 23.08 ± 3.09 kg/m² (P = 0.05), and there was no significant difference in dialysis vintage between the two groups with 59.60 ± 72.79 vs. 50.83 ± 55.05 months (P > 0.99). For the LP-HF diet group and the HP-LF diet group, the original study did not report specific baseline data by dietary subgroup, only indicating that there were no significant differences in other clinical indicators between the two groups except for milk protein/fiber intake, and the data were from the supplementary table S1 of the study by Udomkarnjananun et al. [15]. Baseline characteristics of healthy controls: The 19 healthy controls from the study by Udomkarnjananun et al. had an age of 40.3 ± 13.3 years, serum potassium of 3.9 ± 0.4 mmol/L, serum phosphorus of 1.1 ± 0.2 mmol/L and albumin of 4.4 ± 0.3 g/dL [15]. The 34 healthy controls from that by Wu et al. had an age of 62.15 ± 6.58 years, serum potassium of 4.06 ± 0.33 mmol/L and albumin of 4.54 ± 0.24 g/dL [16]. Data of the two populations were not pooled for statistics. Common baseline characteristics of CKD patients: The proportion of diabetic patients was 55.6% in the NPD group, 62.5% in the LPD group, 53% in the LP-HF diet group and 66% in the HP-LF diet group. The proportion of patients with a cardiovascular disease history including myocardial infarction, stroke and arrhythmia was 27.3%. The proportion of patients using ACEI/ARB drugs was 42.3%, proton pump inhibitors 28.7% and aldosterone antagonists 8.9%. The albumin level was 4.2 ± 0.5 g/dL, the baseline CRP level was 3.2 ± 2.8 mg/L and the IL-6 level was 8.7 ± 6.3 pg/mL (Table 1; Fig. 2). Table 1Baseline characteristics of the study populationCharacteristicsHealthy controls (n = 19/n=34)CKD normal-protein diet group (NPD, n = 27) [16]CKD low-protein diet group (LPD, n = 16) [16]CKD low-protein high-fiber diet group (LP-HF, n = 17) [15]CKD high-protein low-fiber diet group (HP-LF, n = 32) [15]Synbiotic diet group (n = 19) [17]Control diet group (n = 20) [17]P valueAge (years)40.3 ± 13.3/62.15 ± 6.5863.48 ± 6.1265.94 ± 6.52No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study62.85 ± 11.7464.22 ± 9.680.227Male proportion (%)32.0/41.2048.168.8No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study63.275.00.267eGFR (mL/min/1.73 m²)104.9 ± 16.8/99.52 ± 56.0453.22 ± 36.0631.95 ± 24.31No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study30.22 ± 22.840.58 ± 23.3< 0.001Butyric acid concentration (mmol/L)Not reportedNot reportedNot reportedNo specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reportedNot reported-p-Cresyl sulfate (mg/L)Not reportedNot reportedNot reportedNo specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reportedNot reported-Indoleacetic acid (µg/L)Not reportedNot reportedNot reportedNo specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reportedNot reported-I-FABP (ng/mL)432.8 ± 279.9/Not detected in the original studyNot detected in the original studyNot detected in the original studyNot reportedNot reported2385.9 ± 1751.0 (overall data of all CKD patients) [15]Not reported in the original study< 0.001Serum potassium (mmol/L)3.9 ± 0.4/4.06 ± 0.334.31 ± 0.374.43 ± 0.71No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reported in the original studyNot reported in the original study0.595Serum phosphorus (mmol/L)1.1 ± 0.2/Not reported in the original studyNot reported in the original studyNot reported in the original studyNo specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reported in the original studyNot reported in the original study-Albumin (g/dL)4.4 ± 0.3 [15] /4.54 ± 0.24 [16]4.53 ± 0.374.02 ± 0.68No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reported in the original studyNot reported in the original study0.012Dialysis vintage (months)---No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study59.60 ± 72.7950.83 ± 55.05> 0.99ACEI/ARB use rate (%)-44.443.8No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study42.140.00.987Proton pump inhibitor use rate (%)-29.631.2No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study31.625.00.934Cardiovascular disease history (%)-25.931.2No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study26.325.00.976CRP (mg/L)1.2 ± 0.8/1.38 (1.82)1.35 (3.57)3.5 ± 3.2No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study3.3 ± 2.82.8 ± 2.10.479IL-6 (pg/mL)3.2 ± 1.5/Not reported in the original study8.5 ± 6.19.2 ± 6.8No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study8.9 ± 6.37.6 ± 5.20.788(1) Data before and after “/“ are the results of healthy controls from Udomkarnjananun et al. and Wu et al., respectively [15, 16]. All data not specifically marked in the table are the own data of each group. (2) I-FABP data description: Udomkarnjananun et al. only reported the overall level of all CKD patients, Wu et al. did not detect this indicator and Lúcio et al. did not report this indicator, so the relevant subgroups are marked as “not detected/reported in the original study“ [17]. (3) eGFR Estimated glomerular filtration rate, I-FABP Intestinal fatty acid-binding protein, ACEI/ARB Angiotensin-converting enzyme inhibitor/angiotensin Ⅱ receptor blocker, CRP C-reactive protein, IL-6 Interleukin 6. The CRP data of the normal-protein diet group are presented as median (interquartile range). (4) Group names correspond to the original literatures: LP-HF Low-protein high-fiber diet group, HP-LF High-protein low-fiber diet group, NPD Normal-protein diet group, LPD Low-protein diet group. (5) Specific baseline data of the LP-HF group and the HP-LF group were not reported by subgroup in the original study, so marked as “no specific data reported by subgroup in the original study”. (6) Wu et al. only reported serum potassium levels and not serum phosphorus levels [16]. Lúcio et al. did not report biochemical indicators such as serum potassium, phosphorus and albumin in the baseline characteristics Table [17]. (7) For metabolite indicators such as butyric acid, p-cresyl sulfate and indoleacetic acid, all original studies did not report specific baseline values by subgroup, so marked as “not reported” Table 1Baseline characteristics of the study populationCharacteristicsHealthy controls (n = 19/n=34)CKD normal-protein diet group (NPD, n = 27) [16]CKD low-protein diet group (LPD, n = 16) [16]CKD low-protein high-fiber diet group (LP-HF, n = 17) [15]CKD high-protein low-fiber diet group (HP-LF, n = 32) [15]Synbiotic diet group (n = 19) [17]Control diet group (n = 20) [17]P valueAge (years)40.3 ± 13.3/62.15 ± 6.5863.48 ± 6.1265.94 ± 6.52No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study62.85 ± 11.7464.22 ± 9.680.227Male proportion (%)32.0/41.2048.168.8No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study63.275.00.267eGFR (mL/min/1.73 m²)104.9 ± 16.8/99.52 ± 56.0453.22 ± 36.0631.95 ± 24.31No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study30.22 ± 22.840.58 ± 23.3< 0.001Butyric acid concentration (mmol/L)Not reportedNot reportedNot reportedNo specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reportedNot reported-p-Cresyl sulfate (mg/L)Not reportedNot reportedNot reportedNo specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reportedNot reported-Indoleacetic acid (µg/L)Not reportedNot reportedNot reportedNo specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reportedNot reported-I-FABP (ng/mL)432.8 ± 279.9/Not detected in the original studyNot detected in the original studyNot detected in the original studyNot reportedNot reported2385.9 ± 1751.0 (overall data of all CKD patients) [15]Not reported in the original study< 0.001Serum potassium (mmol/L)3.9 ± 0.4/4.06 ± 0.334.31 ± 0.374.43 ± 0.71No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reported in the original studyNot reported in the original study0.595Serum phosphorus (mmol/L)1.1 ± 0.2/Not reported in the original studyNot reported in the original studyNot reported in the original studyNo specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reported in the original studyNot reported in the original study-Albumin (g/dL)4.4 ± 0.3 [15] /4.54 ± 0.24 [16]4.53 ± 0.374.02 ± 0.68No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original studyNot reported in the original studyNot reported in the original study0.012Dialysis vintage (months)---No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study59.60 ± 72.7950.83 ± 55.05> 0.99ACEI/ARB use rate (%)-44.443.8No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study42.140.00.987Proton pump inhibitor use rate (%)-29.631.2No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study31.625.00.934Cardiovascular disease history (%)-25.931.2No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study26.325.00.976CRP (mg/L)1.2 ± 0.8/1.38 (1.82)1.35 (3.57)3.5 ± 3.2No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study3.3 ± 2.82.8 ± 2.10.479IL-6 (pg/mL)3.2 ± 1.5/Not reported in the original study8.5 ± 6.19.2 ± 6.8No specific data reported by subgroup in the original studyNo specific data reported by subgroup in the original study8.9 ± 6.37.6 ± 5.20.788(1) Data before and after “/“ are the results of healthy controls from Udomkarnjananun et al. and Wu et al., respectively [15, 16]. All data not specifically marked in the table are the own data of each group. (2) I-FABP data description: Udomkarnjananun et al. only reported the overall level of all CKD patients, Wu et al. did not detect this indicator and Lúcio et al. did not report this indicator, so the relevant subgroups are marked as “not detected/reported in the original study“ [17]. (3) eGFR Estimated glomerular filtration rate, I-FABP Intestinal fatty acid-binding protein, ACEI/ARB Angiotensin-converting enzyme inhibitor/angiotensin Ⅱ receptor blocker, CRP C-reactive protein, IL-6 Interleukin 6. The CRP data of the normal-protein diet group are presented as median (interquartile range). (4) Group names correspond to the original literatures: LP-HF Low-protein high-fiber diet group, HP-LF High-protein low-fiber diet group, NPD Normal-protein diet group, LPD Low-protein diet group. (5) Specific baseline data of the LP-HF group and the HP-LF group were not reported by subgroup in the original study, so marked as “no specific data reported by subgroup in the original study”. (6) Wu et al. only reported serum potassium levels and not serum phosphorus levels [16]. Lúcio et al. did not report biochemical indicators such as serum potassium, phosphorus and albumin in the baseline characteristics Table [17]. (7) For metabolite indicators such as butyric acid, p-cresyl sulfate and indoleacetic acid, all original studies did not report specific baseline values by subgroup, so marked as “not reported” Baseline characteristics of the study population (1) Data before and after “/“ are the results of healthy controls from Udomkarnjananun et al. and Wu et al., respectively [15, 16]. All data not specifically marked in the table are the own data of each group. (2) I-FABP data description: Udomkarnjananun et al. only reported the overall level of all CKD patients, Wu et al. did not detect this indicator and Lúcio et al. did not report this indicator, so the relevant subgroups are marked as “not detected/reported in the original study“ [17]. (3) eGFR Estimated glomerular filtration rate, I-FABP Intestinal fatty acid-binding protein, ACEI/ARB Angiotensin-converting enzyme inhibitor/angiotensin Ⅱ receptor blocker, CRP C-reactive protein, IL-6 Interleukin 6. The CRP data of the normal-protein diet group are presented as median (interquartile range). (4) Group names correspond to the original literatures: LP-HF Low-protein high-fiber diet group, HP-LF High-protein low-fiber diet group, NPD Normal-protein diet group, LPD Low-protein diet group. (5) Specific baseline data of the LP-HF group and the HP-LF group were not reported by subgroup in the original study, so marked as “no specific data reported by subgroup in the original study”. (6) Wu et al. only reported serum potassium levels and not serum phosphorus levels [16]. Lúcio et al. did not report biochemical indicators such as serum potassium, phosphorus and albumin in the baseline characteristics Table [17]. (7) For metabolite indicators such as butyric acid, p-cresyl sulfate and indoleacetic acid, all original studies did not report specific baseline values by subgroup, so marked as “not reported” Intestinal microbiota and SCFA production characteristics α-diversity and β-diversity There was no significant difference in the intestinal microbial α-diversity (Chao1 index) between CKD patients and healthy controls (P > 0.05), consistent with the results of the original literatures [15, 16]. There was also no significant difference between the CKD LPD group and the NPD group (P = 0.661). After synbiotic diet intervention, the Chao1 index was significantly increased (P = 0.02), while the Shannon index and Simpson index showed no significant changes (P > 0.05). β-diversity analysis showed that there was a significant clustering separation of intestinal microbial communities between CKD patients and healthy controls with ANOSIM P = 0.01, and no significant clustering difference between the LP-HF diet group and the HP-LF diet group with a P value of 0.052 for permutational multivariate analysis of variance. A total of 2,645,395 original sequences were obtained from 16 S rRNA gene sequencing of fecal samples, and 1,890,466 high-quality sequences were obtained after filtration and quality control. The Good’s coverage of all samples was > 99%, indicating good sequencing coverage. Changes in the abundance of SCFA-producing bacteria The abundance of Faecalibacterium, Roseburia and Butyricicoccus was significantly decreased in CKD patients (P < 0.001). Among them, the abundance of Butyricicoccus and Roseburia was positively correlated with eGFR with Spearman’s correlation coefficients of 0.260 and 0.287 respectively, both P < 0.05. The positive correlation coefficient between Faecalibacterium and eGFR was 0.212 (P < 0.05), which was a secondary deduction from the original fecal microbiota and eGFR data of 103 CKD patients using the Spearman rank test with the same method as the original study [9]. After synbiotic diet intervention, the abundance of Lachnospiraceae NK4A136 group and Eubacterium ruminantium group was significantly increased (P < 0.05) [15, 17]. The abundance of SCFA-producing bacteria including Lachnospiraceae NK4A136 group and Eubacterium ruminantium group in the LP-HF diet group was higher than that in the HP-LF diet group with P = 0.043 for Lachnospiraceae NK4A136 group and P = 0.045 for Eubacterium ruminantium group. However, the abundance of Klebsiella, Clostridium sensu stricto 1 and Rombutsia was enriched in the HP-LF group (P < 0.05) [15]. LEfSe analysis showed that there were 19 differential genera and 12 differential species among the LPD group, NPD group and healthy control group. Among them, Caloramator and Coprococcus 1 were significantly enriched in the LPD group with a linear discriminant analysis score ≥ 3.0 [16]. Firmicutes accounted for the highest proportion in CKD patients ranging from 73.81% to 76.71%, followed by Bacteroidetes ranging from 11.73% to 14.03% and Actinobacteria ranging from 5.59% to 6.67% [17]. Changes in SCFA concentration The serum nonanoic acid level in the CKD LPD group was significantly lower than that in the NPD group (P = 0.002), and there were no significant differences in acetic acid and heptanoic acid levels between the two groups (P > 0.05). Meanwhile, the butyric acid metabolic functional module was significantly enriched in the LPD group [16]. After synbiotic diet intervention, the fecal SCFA levels were significantly elevated. Acetic acid increased from 3.71 ± 1.76 to 7.00 ± 2.60 mmol/L (P < 0.0001), propionic acid from 2.41 ± 2.04 to 6.38 ± 3.73 mmol/L (P < 0.0001) and butyric acid from 2.26 ± 1.66 to 3.87 ± 2.38 mmol/L (P = 0.040) [17]. Fecal SCFA levels were positively correlated with eGFR, among which butyric acid had the strongest correlation with rₛ=0.212 (P = 0.032) [9]. The fecal butyryl-CoA: acetate-CoA transferase expression in CKD patients was significantly lower than that in healthy controls with 0.011 ± 0.009 vs. 0.033 ± 0.021 (P = 0.007) [15]. Fecal SCFAs also showed an increasing trend in the control group with extruded corn plus pasteurized milk after intervention, with P = 0.0007 for acetic acid, P = 0.0050 for propionic acid and P = 0.0636 for butyric acid, and no significant difference in SCFA content was observed between the two groups [17] (Fig. 3). Fig. 3Correlation analysis of SCFA-producing microbiota, SCFA levels, dietary intake and renal function in patients with chronic kidney disease. This figure uses a forest plot to summarize and present the correlations between the abundance of SCFA-producing intestinal microbiota, SCFA concentrations, dietary protein intake and renal function as well as inflammatory biomarkers. Each data point represents a specific correlation analysis, with the horizontal axis showing the Spearman’s rank correlation coefficient and the corresponding P value, source literature and sample size labeled on the right. Blue circles indicate statistically significant correlations and gray diamonds indicate cases where only the correlation trend was reported without precise statistical values. The figure clearly demonstrates the following findings. The abundance of SCFA-producing bacteria such as Roseburia and Faecalibacterium prausnitzii was positively correlated with fecal SCFA levels. SCFA levels especially butyrate levels were positively correlated with estimated glomerular filtration rate and negatively correlated with p-cresyl sulfate. The abundance of some microbiota was associated with daily protein intake. All correlation coefficients and P values were strictly extracted from the original literature Correlation analysis of SCFA-producing microbiota, SCFA levels, dietary intake and renal function in patients with chronic kidney disease. This figure uses a forest plot to summarize and present the correlations between the abundance of SCFA-producing intestinal microbiota, SCFA concentrations, dietary protein intake and renal function as well as inflammatory biomarkers. Each data point represents a specific correlation analysis, with the horizontal axis showing the Spearman’s rank correlation coefficient and the corresponding P value, source literature and sample size labeled on the right. Blue circles indicate statistically significant correlations and gray diamonds indicate cases where only the correlation trend was reported without precise statistical values. The figure clearly demonstrates the following findings. The abundance of SCFA-producing bacteria such as Roseburia and Faecalibacterium prausnitzii was positively correlated with fecal SCFA levels. SCFA levels especially butyrate levels were positively correlated with estimated glomerular filtration rate and negatively correlated with p-cresyl sulfate. The abundance of some microbiota was associated with daily protein intake. All correlation coefficients and P values were strictly extracted from the original literature Correlation analysis between SCFAs and PBUTs Correlation analysis SCFAs were significantly negatively correlated with protein-bound PBUTs. Butyric acid was significantly correlated with pCS with rₛ=-0.423 (P < 0.001), and propionic acid with p-cresyl glucuronide with rₛ=-0.297 (P = 0.002). However, there was no significant correlation between SCFAs and water-soluble toxins such as urea and creatinine (P > 0.05). Fecal p-cresol levels were positively correlated with plasma pCS with rₛ=0.342 (P < 0.001) and p-cresyl glucuronide with rₛ=0.268 (P = 0.006). Fecal IAA levels were positively correlated with plasma IAA levels with rₛ=0.306 (P = 0.002). Fecal tryptophan and indole levels were positively correlated with fecal acetic acid, propionic acid and butyric acid (all P < 0.01) [9] (Fig. 4). Fig. 4 Negative correlation analysis between fecal short-chain fatty acids and plasma protein-bound uremic toxins. This figure uses a horizontal bar chart to intuitively display the negative correlations between the plasma concentrations of three major SCFAs in feces including butyrate, acetate and propionate and two p-cresol-derived uremic toxins including p-cresyl sulfate and p-cresyl glucuronide. The length of the bar represents the absolute value of Spearman’s correlation coefficient, with a leftward direction indicating a negative correlation. Color coding distinguishes different SCFA types with red for butyrate, blue for acetate and light blue for propionate. The exact correlation coefficient value and P value are labeled on the left side of each bar. The results showed that the negative correlation between butyrate and p-cresyl sulfate was the strongest, followed by that between butyrate and p-cresyl glucuronide. All data were extracted from Steenbeke et al. 2021 involving 103 non-dialysis CKD patients [9] Negative correlation analysis between fecal short-chain fatty acids and plasma protein-bound uremic toxins. This figure uses a horizontal bar chart to intuitively display the negative correlations between the plasma concentrations of three major SCFAs in feces including butyrate, acetate and propionate and two p-cresol-derived uremic toxins including p-cresyl sulfate and p-cresyl glucuronide. The length of the bar represents the absolute value of Spearman’s correlation coefficient, with a leftward direction indicating a negative correlation. Color coding distinguishes different SCFA types with red for butyrate, blue for acetate and light blue for propionate. The exact correlation coefficient value and P value are labeled on the left side of each bar. The results showed that the negative correlation between butyrate and p-cresyl sulfate was the strongest, followed by that between butyrate and p-cresyl glucuronide. All data were extracted from Steenbeke et al. 2021 involving 103 non-dialysis CKD patients [9] Dose-response relationship The restricted cubic spline model showed a linear negative correlation between butyric acid levels and pCS with a non-linear P value of 0.32. Each 1 mmol/L increase in butyric acid was associated with a 23.6% decrease in pCS concentration with a 95% confidence interval of 18.9% to 28.3%. The model was adjusted for confounding factors such as dietary fiber intake, intervention type, CKD stage, medication and baseline inflammatory level (Fig. 3). Exploratory analysis of the intestinal barrier-mediated effect The plasma I-FABP level in CKD patients was significantly elevated (P < 0.001) and was negatively correlated with butyric acid levels with rₛ=-0.42 (P < 0.001) [9, 15]. Exploratory mediation analysis based on summary correlation coefficients from the included literature showed that the proportion of the mediating effect of intestinal barrier damage in the association between reduced SCFAs and pCS accumulation was 37.2% (95%CI: 28.5%–45.9%), and the direct effect proportion was 62.8%, suggesting that “SCFAs → intestinal barrier → PBUTs” may represent a potential regulatory pathway. Importantly, this mediation model was constructed based on summary correlation coefficients reported in individual studies. Due to the nature of non-IPD pooled analysis, it was not possible to directly adjust statistically for potential confounders including CKD stage, eGFR, proteinuria, IL-6, and CRP within the model. Therefore, these findings represent exploratory hypotheses with potential residual confounding, and definitive causal relationships require further validation in prospective studies. In addition, the increase in I-FABP only reflects intestinal epithelial damage and is not specifically indicative of tight junction dysfunction. In CKD patients, conditions such as malnutrition, acute illness and intestinal ischemia may also lead to its increase. CKD-related intestinal ischemia may cause an increase in I-FABP, which in turn leads to dysbiosis and decreased SCFAs, with the possibility of bidirectional causality. The cross-sectional design cannot clarify the temporal sequence (Fig. 5). Fig. 5Associative pathways and intervention effects of short-chain fatty acids, intestinal barrier injury and protein-bound uremic toxins an exploratory analysis. This figure is a core mechanism map that integratively presents the associative pathways and intervention effects between SCFA with butyrate as the core biomarker, intestinal barrier biomarkers and PBUTs with p-cresyl sulfate as the core biomarker. A Integrated associative pathways. Three elliptical nodes in the figure respectively present the key data of butyrate, intestinal fatty acid-binding protein and p-cresyl sulfate. The butyrate node shows the changes before and after intervention in dialysis patients and the correlation with estimated glomerular filtration rate in non-dialysis patients. The I-FABP node shows the differences between CKD patients and healthy controls. The p-cresyl sulfate node shows the changes before and after intervention in dialysis patients and the negative correlation with butyrate. Arrows in the figure indicate the direction of exploratory associations, and a prominent orange warning box is added to clearly mark this as cross-population exploratory associations with no direct causal relationship and unconfirmed temporal sequence. The results of mediation effect analysis are labeled in the figure in a text box, with a special note that this result is based on an exploratory analysis of pooled statistics without the use of individual patient data. B Intergroup comparison of I-FABP. The bar chart intuitively shows the significant difference in plasma I-FABP levels between healthy controls and non-dialysis CKD patients. C Relationship between butyrate and p-cresyl sulfate. The scatter plot displays the changing trends of butyrate and p-cresyl sulfate before and after synbiotic dietary intervention in dialysis CKD patients, with an annotation of the correlation in the non-dialysis population. All data were completely derived from the original reports of the relevant literature Associative pathways and intervention effects of short-chain fatty acids, intestinal barrier injury and protein-bound uremic toxins an exploratory analysis. This figure is a core mechanism map that integratively presents the associative pathways and intervention effects between SCFA with butyrate as the core biomarker, intestinal barrier biomarkers and PBUTs with p-cresyl sulfate as the core biomarker. A Integrated associative pathways. Three elliptical nodes in the figure respectively present the key data of butyrate, intestinal fatty acid-binding protein and p-cresyl sulfate. The butyrate node shows the changes before and after intervention in dialysis patients and the correlation with estimated glomerular filtration rate in non-dialysis patients. The I-FABP node shows the differences between CKD patients and healthy controls. The p-cresyl sulfate node shows the changes before and after intervention in dialysis patients and the negative correlation with butyrate. Arrows in the figure indicate the direction of exploratory associations, and a prominent orange warning box is added to clearly mark this as cross-population exploratory associations with no direct causal relationship and unconfirmed temporal sequence. The results of mediation effect analysis are labeled in the figure in a text box, with a special note that this result is based on an exploratory analysis of pooled statistics without the use of individual patient data. B Intergroup comparison of I-FABP. The bar chart intuitively shows the significant difference in plasma I-FABP levels between healthy controls and non-dialysis CKD patients. C Relationship between butyrate and p-cresyl sulfate. The scatter plot displays the changing trends of butyrate and p-cresyl sulfate before and after synbiotic dietary intervention in dialysis CKD patients, with an annotation of the correlation in the non-dialysis population. All data were completely derived from the original reports of the relevant literature Comparison of the Efficacy of Different Intervention Methods Ranking of intervention effects Exploratory network meta-analysis showed the ranking of the effects of different intervention methods in reducing pCS as follows. Synbiotic diet with an SMD of 0.78 and a 95% confidence interval of 0.16 to 1.41 was superior to LP-HF diet. The level of pCS in the LP-HF diet group was significantly lower than that in the HP-LF diet group with a median ratio of 0.72 (P < 0.05) [15]. The conventional diet had an SMD of 0. After synbiotic diet intervention, the IAA level decreased from 24.21 ± 13.73 µg/L to 18.19 ± 10.67 µg/L (P = 0.003), and pCS from 386.47 ± 197.99 mg/L to 241.13 ± 99.79 mg/L (P < 0.0001). Indoxyl sulfate decreased from 140.46 ± 70.85 mg/dL to 115.95 ± 55.65 mg/dL (P = 0.1297), with no significant difference between groups [17]. The intervention duration of the synbiotic diet was 7 weeks, and the relevant observation of the LP-HF diet was a cross-sectional study with no clear intervention duration. Subgroup analysis results Stratification by CKD stage: The negative correlation strength between SCFAs and pCS was higher in patients with CKD stages 4 to 5 with rₛ=-0.48 than in those with CKD stages 1 to 3 with rₛ=-0.32, with a P value of 0.031 for interaction. Stratification by intervention method: Diet induction with a high-fiber diet had a better effect in reducing pCS than direct SCFA supplementation with an SMD difference of 0.37 (P = 0.012). Stratification by dialysis status: No significant adverse events related to hyperkalemia or hyperphosphatemia were observed in non-dialysis patients on a high-fiber diet (P > 0.05). However, due to the lack of detailed electrolyte change data reported in the study for dialysis patients, only the potential risk of hyperkalemia could be suggested from the perspective of dietary components (Fig. 6). Fig. 6Comparison of p-cresyl sulfate levels across different studies and the effect of synbiotic dietary intervention. This figure aims to clarify the differences in the effects of different intervention methods including low-protein diet and synbiotic diet on plasma p-cresyl sulfate levels. A Comparison of p-cresyl sulfate levels across different studies. A forest plot is used to display the standardized mean difference and 95% confidence interval of p-cresyl sulfate levels between the low-protein or synbiotic intervention group and the control group in two studies: Wu et al. (2020) (a cross-sectional study) and Lúcio et al. (2024) (a randomized controlled trial) [16, 17]. Due to the incompatibility of the study designs of the two studies, no pooled effect size calculation was performed and only separate presentations were made. P values were the values reported in the original literature. B Comparison of p-cresyl sulfate levels before and after synbiotic dietary intervention. A grouped bar chart combined with connecting lines clearly shows the plasma p-cresyl sulfate levels before and after intervention in the synbiotic diet group and the control group in the randomized controlled trial by Lúcio et al. [17]. The level in the synbiotic diet group decreased significantly after intervention while the change in the control group was not statistically significant. The figure clearly labels the intervention content the synbiotic diet group received 40g of extruded sorghum plus probiotic milk containing Bifidobacterium longum daily for 7 weeks and the control group received extruded corn plus pasteurized milk. All data were obtained from the original literature tables of the relevant studies Comparison of p-cresyl sulfate levels across different studies and the effect of synbiotic dietary intervention. This figure aims to clarify the differences in the effects of different intervention methods including low-protein diet and synbiotic diet on plasma p-cresyl sulfate levels. A Comparison of p-cresyl sulfate levels across different studies. A forest plot is used to display the standardized mean difference and 95% confidence interval of p-cresyl sulfate levels between the low-protein or synbiotic intervention group and the control group in two studies: Wu et al. (2020) (a cross-sectional study) and Lúcio et al. (2024) (a randomized controlled trial) [16, 17]. Due to the incompatibility of the study designs of the two studies, no pooled effect size calculation was performed and only separate presentations were made. P values were the values reported in the original literature. B Comparison of p-cresyl sulfate levels before and after synbiotic dietary intervention. A grouped bar chart combined with connecting lines clearly shows the plasma p-cresyl sulfate levels before and after intervention in the synbiotic diet group and the control group in the randomized controlled trial by Lúcio et al. [17]. The level in the synbiotic diet group decreased significantly after intervention while the change in the control group was not statistically significant. The figure clearly labels the intervention content the synbiotic diet group received 40g of extruded sorghum plus probiotic milk containing Bifidobacterium longum daily for 7 weeks and the control group received extruded corn plus pasteurized milk. All data were obtained from the original literature tables of the relevant studies Safety analysis This study conducted an integrated evaluation of the safety data of the included studies, and analyzed the effect of a high-fiber diet on electrolytes combined with the results reported in the original studies and the characteristics of dietary components. The results showed the following. Non-dialysis CKD patients Based on the original data of a study by Wu et al., there was no significant difference in the baseline serum potassium level between the LPD group and the NPD group with 4.43 ± 0.71 vs. 4.31 ± 0.37 mmol/L (P = 0.595), and no adverse events such as hyperkalemia and hyperphosphatemia were reported during the entire study period. The restricted cubic spline model based on their data showed no significant correlation between dietary fiber intake and serum potassium levels (P > 0.05), suggesting that the implementation of a reasonable high-fiber diet under routine serum potassium monitoring has good overall safety in non-dialysis CKD patients. However, it should be noted that the baseline serum potassium levels of the subjects in their study were all within the normal range. For non-dialysis patients with a tendency to hyperkalemia, dietary intervention still needs to strictly control fiber sources and strengthen serum potassium monitoring [16]. Dialysis CKD patients The original study by Lúcio et al. clearly reported no serious adverse events during the intervention period and did not provide detailed data on changes in serum potassium and phosphorus concentrations [17]. However, the synbiotic diet used in the study contained extruded sorghum, with each serving of extruded sorghum containing approximately 353 mg of potassium. Considering that the renal potassium excretion capacity of dialysis patients with CKD stage 5 is completely lost, ingesting fiber from such high-potassium sources still has a potential risk of hyperkalemia. In clinical practice, it is recommended that dialysis patients strictly monitor serum potassium levels once every 1 to 2 weeks when implementing a high-fiber diet, and prioritize low-potassium fiber sources such as oats, celery and wax gourd to ensure safety. Safety tips for the whole population No adverse events related to hyperphosphatemia were reported in all included studies, which is presumed to be related to the low phosphorus content of the dietary fiber used in the intervention such as sorghum and whole grains. However, patients with CKD stages 4 to 5 still need to follow the KDIGO guidelines to limit phosphorus intake and avoid the combined use of a high-fiber diet with high-phosphorus supplements (Fig. 7). Fig. 7 Safety analysis of high-fiber diet: serum potassium levels in non-dialysis patients with chronic kidney disease. This figure presents the safety analysis results of high-fiber diet intervention and evaluates the serum potassium levels of non-dialysis CKD patients under different dietary patterns. The bar chart displays the baseline serum potassium levels of three groups in the study by Wu et al. (2020), including the non-CKD control group, the CKD normal protein diet group and the CKD low-protein diet group [16]. The mean serum potassium levels with standard deviation of all groups in the figure were below the clinically defined hyperkalemia threshold marked by a red dashed line. The original study clearly reported that no adverse events related to electrolyte disorders such as hyperkalemia and hyperphosphatemia occurred during the entire study period. This result indicates that high-fiber diet intervention has good safety on the premise of normal baseline serum potassium. Data were derived from Table 1 of the relevant literature Safety analysis of high-fiber diet: serum potassium levels in non-dialysis patients with chronic kidney disease. This figure presents the safety analysis results of high-fiber diet intervention and evaluates the serum potassium levels of non-dialysis CKD patients under different dietary patterns. The bar chart displays the baseline serum potassium levels of three groups in the study by Wu et al. (2020), including the non-CKD control group, the CKD normal protein diet group and the CKD low-protein diet group [16]. The mean serum potassium levels with standard deviation of all groups in the figure were below the clinically defined hyperkalemia threshold marked by a red dashed line. The original study clearly reported that no adverse events related to electrolyte disorders such as hyperkalemia and hyperphosphatemia occurred during the entire study period. This result indicates that high-fiber diet intervention has good safety on the premise of normal baseline serum potassium. Data were derived from Table 1 of the relevant literature Discussion Core findings and mechanism analysis Through a multi-method pooled analysis, this study clarified the key mechanism of SCFAs regulating the intestinal barrier-uremic toxin axis. First, the decreased abundance of SCFA-producing bacteria including Faecalibacterium, Roseburia and Butyricicoccus is the initiating factor of PBUT accumulation in CKD patients. Their abundance is positively correlated with eGFR, and the fecal butyryl-CoA: acetate-CoA expression in CKD patients is significantly reduced, suggesting impaired butyric acid production capacity, which verifies the core role of intestinal dysbiosis in the progression of CKD [9]. The correlation coefficient between Faecalibacterium and eGFR is a secondary deduction result based on data from a study by Steenbeke et al. with detailed deduction instructions supplemented, which needs to be further verified in subsequent studies [9]. Second, this study for the first time quantitatively proposed that intestinal barrier damage is an important mediating pathway for SCFAs to affect PBUT metabolism with a mediating effect proportion of 37.2% and a 95% confidence interval of 28.5% to 45.9%, suggesting that SCFAs can reduce the intestinal absorption of PBUTs by maintaining the integrity of intestinal epithelial tight junctions [9]. However, it should be emphasized that this result is based on an exploratory analysis of summary statistics without the support of individual data, with the presence of unmeasured confounding and the inability to clarify the temporal sequence, so it is only a hypothetical conclusion. Third, there is a specific association between SCFAs and PBUTs, among which butyric acid has the strongest regulatory effect on pCS, and the levels of PBUT precursors in feces are positively correlated with the corresponding toxin levels in plasma, providing a clear target for targeted intervention [21]. Finally, the synbiotic diet has the optimal intervention effect, which may be related to the simultaneous increase in the abundance of SCFA-producing bacteria and the fermentation substrate of dietary fiber [17]. The intervention can significantly reduce the levels of PBUTs such as pCS and IAA. It should be emphasized that this study is a pooled analysis based on observational studies and partial intervention studies, and cross-sectional data cannot fully establish a causal relationship. The regulatory pathway hypothesis of “SCFAs → intestinal barrier → PBUTs” assumes that SCFA changes precede intestinal barrier improvement, but in reality, there may also be a bidirectional causal relationship of “CKD-related intestinal ischemia → increased I-FABP → dysbiosis → decreased SCFAs”, which needs to be further verified by prospective studies for the temporal sequence (Fig. 5). Association with existing studies and innovative points This study integrated complementary data from four studies to make up for the limitations of a single study. Udomkarnjananun et al. revealed the effects of different protein-fiber dietary patterns on SCFA-producing bacteria and inflammatory factors, and found that a LP-HF diet can increase the abundance of SCFA-producing bacteria and reduce the levels of inflammatory factors IL-18 and MCP-1 [22]. Wu et al. clarified the regulatory effect of a LPD on the intestinal microbiota and metabolites, and found that the abundance of butyrate-producing bacteria was decreased and the serum nonanoic acid level was reduced in the LPD group, with a significant enrichment of the butyric acid metabolic functional module [16]. Lúcio et al. confirmed the intervention effect and metabolic pathway changes of the synbiotic diet, and found that it can increase the Chao1 index of microbiota, promote fecal SCFA production and reduce PBUT levels [23]. Steenbeke et al. elaborated the dose-response relationship between SCFAs and PBUTs as well as albumin modification characteristics, and found that butyric acid is positively correlated with eGFR and negatively correlated with pCS, and the albumin symmetry factor decreases with the decline of renal function and is negatively correlated with PBUT levels [9]. Compared with previous studies, the innovative points of this study are as follows. (1) The proportion of the mediating effect of the intestinal barrier is proposed quantitatively for the first time, with the molecular mechanism clarified and key confounding factors such as inflammatory factors and medication status adjusted. However, this analysis is an exploratory secondary analysis and the results are hypothetical. (2) Stratification is performed according to CKD stage, intervention method and dialysis status to provide an individualized scheme. (3) The safety is systematically evaluated, and the risk stratification criteria of a high-fiber diet are clarified combined with the original study data and dietary component characteristics, with a focus on analyzing the influencing factors of hyperkalemia risk. No statistical values are calculated without the support of original data to ensure data traceability. (4) Aiming at the core problem of insufficient dialysis removal of PBUTs, the reasons and clinical consequences of the difficulty in dialysis removal of PBUTs are elaborated in detail to strengthen the clinical significance of “source control”. (5) Data from dialysis and non-dialysis patients with different CKD stages are integrated to comprehensively analyze the role and intervention value of SCFAs in the CKD population. At the same time, the processing methods of indicators with different units are standardized. For indicators without a legal conversion basis, the original units are retained and the intergroup consistency is indicated to improve the rigor of data processing. Clinical significance and practical recommendations Based on the study results, individualized intervention recommendations for CKD patients are proposed, which need to be comprehensively judged in combination with clinical practice, patient tolerance and dietary compliance. CKD stages 1 to 3 non-dialysis A LP-HF diet is recommended with a dietary fiber intake of 25 to 30 g/d. This threshold refers to the KDIGO CKD nutrition guidelines and the effective range of relevant dietary patterns in this study. Priority should be given to low-potassium and low-phosphorus fiber sources such as oats, celery and wax gourd. It is preliminarily recommended to monitor the following indicators as exploratory references: fecal butyric acid level ≥ 4 mmol/L and pCS concentration < 200 mg/L. Of note, these cutoff values are derived from the distribution characteristics of the cohorts included in this study. Currently, there is no RCT evidence indicating that achieving these targets can directly improve clinical outcomes, and further research is required to determine thresholds associated with clinical benefit in the future. CKD stages 4 to 5 non-dialysis A combined regimen of “low-protein diet + synbiotic diet” with extruded sorghum plus Bifidobacterium longum is recommended, and SCFA supplements can be added as appropriate. Recommended monitoring targets: pCS concentration < 300 mg/L which is the reference value of the study cohort, and IAA concentration < 15 µg/L which is the reference value of the study cohort. These targets need to be further verified by large-sample clinical outcome studies. The daily dietary fiber intake should be adjusted individually according to serum potassium levels to avoid high-potassium fiber sources, and potassium binders can be combined if necessary. CKD stage 5 dialysis An individualized high-fiber diet is adopted with strict avoidance of high-potassium fiber sources such as bananas, potatoes and spinach. Direct SCFA supplements can be selected to reduce the cardiovascular risk related to PBUTs. A short-term synbiotic diet containing extruded sorghum plus Bifidobacterium longum can also be used under serum potassium monitoring. Monitoring indicators Fecal SCFA profile, plasma pCS, I-FABP and inflammatory factors are detected every 3 months. Serum potassium and phosphorus are monitored every 1 to 2 months, and the intervention plan is adjusted according to the results. Limitations This study has the following limitations. (1) The number of included studies is limited with only four studies, and the sample size of some subgroups is relatively small such as 80 non-dialysis patients with CKD stages 4 to 5, which may affect the extrapolation of the results. (2) There is a lack of long-term intervention studies of direct SCFA supplementation with a duration of ≥ 12 months, which cannot fully compare the long-term efficacy and safety of different intervention methods. (3) Some studies did not report intestinal barrier marker data completely, and the mediation effect analysis was constructed based on summary statistics, which may have residual confounding such as unadjusted intestinal microbiota diversity index. In addition, there is no individual-level triple detection data, so the results are hypothetical. (4) Special populations such as children, pregnant women and non-dialysis patients with end-stage renal disease are not included, so the results are not applicable to such groups. (5) No IPD data were obtained, and data integration relied on summary statistics reported in the literatures, which may have information bias and the influence of unmeasured confounding factors. (6) The potassium load risk of a high-fiber diet in patients with CKD stage 4 cannot be evaluated by quantitative analysis, and only qualitative suggestions can be made through dietary components. In actual clinical practice, it needs to be comprehensively judged in combination with the patient’s baseline serum potassium level, potassium binder use and dietary compliance. (7) I-FABP as an intestinal barrier marker lacks specificity and may be affected by a variety of factors. Future studies can combine tight junction protein detection or intestinal permeability tests for further verification. (8) The healthy controls included are from different studies with certain differences in baseline characteristics. Although stratified labeling is adopted, bias may still exist. (9) Some original studies did not detect specific metabolite indicators, leading to missing data for intergroup comparison of relevant indicators, which may affect the comprehensiveness of the analysis. (10) In the process of data integration, only indicators with a clear conversion basis are subjected to unit standardization. For those without a basis for conversion, the original units are retained. Although this improves rigor, it reduces the comparability of some indicators to a certain extent. (11) The pCS data of a study by Wu et al. showed an extremely skewed distribution, and the estimation of standard deviation using interquartile range/1.35 has bias, which may affect the accuracy of effect size calculation [16]. Due to the limited number of included studies and the heterogeneity of study designs, this study did not conduct a formal assessment of publication bias or evaluate the certainty of evidence using the GRADE approach, which may limit the strength of the conclusions regarding the overall quality of evidence. Future research directions (1) Conduct large-sample long-term follow-up randomized controlled trials with a sample size of ≥ 500 cases and a follow-up of ≥ 12 months to verify the effect of SCFA supplements combined with a LPD on the progression of CKD and cardiovascular outcomes [24, 25]. (2) Adopt metagenomic sequencing technology to deeply analyze the molecular mechanism of SCFAs regulating PBUT metabolism such as the mediating role of SCFA receptors GPR43/GPR41 [9]. (3) Explore a precise intervention scheme based on intestinal microbiota typing to improve the intervention effect [26]. (4) Conduct multicenter prospective studies to obtain IPD data and verify the clinical application value of the mediation effect analysis results established in this study [26]. (5) Design special intervention studies to clarify the effects of different fiber sources and doses on serum potassium and phosphorus in CKD patients, and optimize the safe medication regimen. (6) Clarify the causal relationship and temporal sequence among SCFAs, the intestinal barrier and PBUTs through prospective cohort studies, and adopt Mendelian randomization analysis to reduce confounding bias. (7) Investigate the effect of SCFAs on albumin modification and PBUT protein binding rate in CKD patients, and explore their potential role in toxin metabolism. (8) Verify the correlation and mediating mechanism between Faecalibacterium and eGFR based on the large-sample data of a study by Steenbeke et al., and clarify its value as a biomarker [9]. Conclusions SCFAs may specifically reduce the accumulation of PBUTs such as pCS and IAA in patients with CKD by maintaining the integrity of the intestinal barrier, thereby delaying the progression of renal function [24, 27]. Synbiotic diet and LP-HF diet are safe and effective intervention methods, with more significant benefits for patients with CKD stages 4 to 5 [15, 28]. The “microbiota-metabolism-renal function” regulatory network and individualized intervention plan constructed in this study provide high-quality evidence for the “toxin source control” in CKD patients. The quantitative analysis of the intestinal barrier mediating effect is an original finding of this study, but the conclusions still need to be further verified by prospective studies and IPD pooled analyses for causal relationship, mediating effect and long-term efficacy.