New hypothesis: a gut-lipid-kidney axis in constipated CKD patients—insights from multi-omics triangulation
Yichen Liu, Jin Zhao, Jinguo Yuan, Zixian Yu, Jie Liu, Xiaoxuan Ning, Shiren Sun
TL;DR
Constipation in CKD patients increases mortality risk by disrupting a gut-lipid-kidney axis involving protective bacteria and a key fat molecule.
Contribution
Identification of a novel gut-lipid-kidney axis linking constipation, gut microbiota depletion, and CKD progression.
Findings
Functional constipation independently predicts all-cause and cardiovascular mortality in CKD patients.
Constipation causes depletion of the protective genus Herbidospora, which reduces PC (14:0_18:2) and worsens CKD.
PC (14:0_18:2) mediates 12.5% of the protective pathway from Herbidospora to CKD.
Abstract
Patients with chronic kidney disease (CKD) frequently experience constipation, often linked to gut microbiota disruptions. We hypothesized that constipation is not merely a comorbidity but a driver of adverse outcomes through a specific host-microbial metabolic axis. Employing a multi-cohort design for evidence triangulation, we first employed a large U.S. database and mortality data up to 31 December 2019 to investigate the association between constipation and overall mortality in patients with CKD using weighted Cox proportional hazards models. To explore potential microbial and metabolic mechanisms that could underlie such clinical observations, we then used genetic data from the independent FINRISK cohort to investigate causal relationships between specific gut microbiota, their metabolites, and CKD. In the observational cohort, we identified functional constipation as a potent,…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Fig 1
Fig 2
Fig 3
Fig 4
Fig 5
Fig 6
Fig 7| Characteristic | Unmatched cohort | PSM matched cohort (1:1) | |||||||
|---|---|---|---|---|---|---|---|---|---|
| No constipation | Constipation | SMD | No constipation | Constipation | SMD | ||||
|
| 2,291 | 278 | 278 | 278 | |||||
| Age (mean [SD]) | 63.52 (16.44) | 63.37 (16.67) | 0.881 | 0.009 | 63.04 (16.81) | 63.37 (16.67) | 0.818 | 0.020 | |
| Sex (%) | |||||||||
| Male | 1,093 (47.7) | 70 (25.2) | <0.001 | 0.481 | 69 (24.8) | 70 (25.2) | 1.000 | 0.008 | |
| Female | 1,198 (52.3) | 208 (74.8) | 209 (75.2) | 208 (74.8) | |||||
| Race (%) | |||||||||
| Mexican American | 332 (14.5) | 25 (9.0) | 0.022 | 0.220 | 25 (9.0) | 25 (9.0) | 0.901 | 0.087 | |
| Other Hispanic | 152 (6.6) | 24 (8.6) | 26 (9.4) | 24 (8.6) | |||||
| Non-Hispanic White | 1,291 (56.4) | 149 (53.6) | 157 (56.5) | 149 (53.6) | |||||
| Non-Hispanic Black | 458 (20.0) | 70 (25.2) | 60 (21.6) | 70 (25.2) | |||||
| Other race | 58 (2.5) | 10 (3.6) | 10 (3.6) | 10 (3.6) | |||||
| BMI (mean [SD]) | 30.13 (7.29) | 29.60 (7.34) | 0.256 | 0.072 | 29.34 (7.13) | 29.60 (7.34) | 0.671 | 0.036 | |
| HbA1c (mean [SD]) | 6.44 (1.58) | 6.41 (1.78) | 0.800 | 0.015 | 6.46 (1.57) | 6.41 (1.78) | 0.718 | 0.031 | |
| ALB (mean [SD]) | 41.23 (3.60) | 40.40 (3.99) | <0.001 | 0.221 | 40.26 (4.33) | 40.40 (3.99) | 0.707 | 0.032 | |
| BUN (mean [SD]) | 17.87 (9.47) | 17.92 (9.70) | 0.939 | 0.005 | 18.21 (10.74) | 17.92 (9.70) | 0.740 | 0.028 | |
| eGFR (mean [SD]) | 70.17 (28.54) | 67.18 (29.68) | 0.101 | 0.103 | 68.43 (30.57) | 67.18 (29.68) | 0.626 | 0.041 | |
| Marital status (%) | |||||||||
| Married | 1,182 (51.6) | 133 (47.8) | 0.892 | 0.082 | 137 (49.3) | 133 (47.8) | 0.929 | 0.099 | |
| Widowed | 463 (20.2) | 62 (22.3) | 55 (19.8) | 62 (22.3) | |||||
| Divorced | 277 (12.1) | 36 (12.9) | 43 (15.5) | 36 (12.9) | |||||
| Separated | 71 (3.1) | 8 (2.9) | 8 (2.9) | 8 (2.9) | |||||
| Never married | 200 (8.7) | 27 (9.7) | 25 (9.0) | 27 (9.7) | |||||
| Living with partner | 98 (4.3) | 12 (4.3) | 10 (3.6) | 12 (4.3) | |||||
| Education (%) | |||||||||
| <9th grade | 347 (15.1) | 54 (19.4) | 0.029 | 0.215 | 50 (18.0) | 54 (19.4) | 0.850 | 0.099 | |
| 9–11th grade | 412 (18.0) | 56 (20.1) | 60 (21.6) | 56 (20.1) | |||||
| High school | 577 (25.2) | 78 (28.1) | 70 (25.2) | 78 (28.1) | |||||
| Some college | 576 (25.1) | 60 (21.6) | 69 (24.8) | 60 (21.6) | |||||
| College grad | 379 (16.5) | 30 (10.8) | 29 (10.4) | 30 (10.8) | |||||
| PIR cat (%) | |||||||||
| <1.3 | 706 (30.8) | 108 (38.8) | 0.004 | 0.215 | 111 (39.9) | 108 (38.8) | 0.566 | 0.091 | |
| 1.3–3.5 | 1,026 (44.8) | 123 (44.2) | 129 (46.4) | 123 (44.2) | |||||
| ≥ 3.5 | 559 (24.4) | 47 (16.9) | 38 (13.7) | 47 (16.9) | |||||
| Smoking (%) | |||||||||
| Never | 1,109 (48.4) | 158 (56.8) | 0.021 | 0.179 | 152 (54.7) | 158 (56.8) | 0.876 | 0.044 | |
| Previous | 776 (33.9) | 74 (26.6) | 78 (28.1) | 74 (26.6) | |||||
| Current | 406 (17.7) | 46 (16.5) | 48 (17.3) | 46 (16.5) | |||||
| Hypertension (%) | |||||||||
| No | 1,312 (57.3) | 169 (60.8) | 0.290 | 0.072 | 171 (61.5) | 169 (60.8) | 0.931 | 0.015 | |
| Yes | 979 (42.7) | 109 (39.2) | 107 (38.5) | 109 (39.2) | |||||
| Diabetes (%) | |||||||||
| No | 518 (22.6) | 61 (21.9) | 0.861 | 0.016 | 58 (20.9) | 61 (21.9) | 0.836 | 0.026 | |
| Yes | 1,773 (77.4) | 217 (78.1) | 220 (79.1) | 217 (78.1) | |||||
| Heart failure (%) | |||||||||
| No | 2,077 (90.7) | 251 (90.3) | 0.927 | 0.013 | 252 (90.6) | 251 (90.3) | 1.000 | 0.012 | |
| Yes | 214 (9.3) | 27 (9.7) | 26 (9.4) | 27 (9.7) | |||||
| Dialysis (%) | |||||||||
| No | 2,254 (98.4) | 269 (96.8) | 0.092 | 0.106 | 267 (96.0) | 269 (96.8) | 0.820 | 0.039 | |
| Yes | 37 (1.6) | 9 (3.2) | 11 (4.0) | 9 (3.2) | |||||
| Opioid use (%) | |||||||||
| No | 2,090 (91.2) | 256 (92.1) | 0.713 | 0.031 | 250 (89.9) | 256 (92.1) | 0.459 | 0.075 | |
| Yes | 201 (8.8) | 22 (7.9) | 28 (10.1) | 22 (7.9) | |||||
| Characteristic | HR | 95% CI | |
|---|---|---|---|
| Model 1: unadjusted | 1.21 | 1.02, 1.44 |
|
| Model 2: adjusted for demographics | 1.35 | 1.13, 1.60 |
|
| Model 3: fully adjusted | 1.33 | 1.11, 1.58 |
|
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsGut microbiota and health · Gastrointestinal motility and disorders · Chronic Kidney Disease and Diabetes
INTRODUCTION
Chronic kidney disease (CKD) is emerging as a global public health issue, affecting approximately 9.1% of the population, particularly as aging populations grow (1). It involves a gradual decline in kidney function, leading to impaired waste and water removal from the body. As the disease progresses, it can result in various complications that substantially affect the quality of life of patients. Constipation is one of the most common gastrointestinal issues in CKD. Renal dysfunction can cause toxin accumulation, thereby disturbing gastrointestinal function. On the other hand, the use of medications used to treat CKD and dietary fiber restriction can exacerbate constipation (2, 3). Beyond symptom discomfort, constipation also initiates a “gut‒kidney vicious cycle” by increasing intra-abdominal pressure and renal vascular resistance (4). Despite its clinical relevance, the significance of constipation as a CKD complication is often overlooked by clinicians and patients (5), which may stem from the limited research on the impact of constipation on CKD patient outcomes.
Gut microbiota dysbiosis may be a vital link between CKD and its associated constipation. In CKD, the progressive accumulation of uremic waste products compromises intestinal barrier integrity (6), facilitating the translocation of bacterial metabolites and pro-inflammatory molecules across the intestinal–kidney axis, thereby aggravating renal injury (7). Microbial metabolic disturbances impair intestinal neuromuscular function, slow transit time, and contribute directly to constipation (8). Among the altered metabolites, indole sulfate (IS) and trimethylamine N-oxide (TMAO) are particularly noteworthy: both accumulate due to reduced renal clearance, yet also exert pathogenic effects by promoting oxidative stress, endothelial dysfunction, and nephrotoxicity (9, 10). In particular, TMAO elevation is strongly linked to CKD progression, while cardiovascular disease remains the leading cause of mortality in CKD patients (11, 12). Emerging evidence further suggests that gut-derived metabolites can modulate enteric nervous system activity and gut motility (13), raising a critical research question: could specific microbiota-driven metabolites simultaneously regulate gastrointestinal function and renal pathology, thereby influencing the clinical trajectory of CKD with constipation?
Lipid metabolism disorders are common in patients with CKD and may interact with gut microbiota in complex ways. Lipid molecules, such as phosphatidylcholines (PCs), are metabolized by the gut microbiota (e.g., Clostridium spp.) to produce TMAO, which accelerates renal fibrosis and cardiovascular complications (14–16). However, the protective roles of specific PC subtypes and their underlying mechanism remain unclear. This scientific gap provides an opportunity to explore whether unique gut microbiota–derived lipid metabolites could mitigate CKD progression rather than exacerbate it.
To explore these unresolved questions, we employed a multi-cohort design to triangulate evidence from observational and genetic data. Our objectives were threefold: (i) to determine whether constipation independently predicts all-cause and cardiovascular mortality in a large CKD cohort ; (ii) to investigate the causal chain linking constipation, gut microbiota dysbiosis, and lipid metabolism using Mendelian randomization (MR), specifically focusing on whether constipation drives the depletion of renoprotective taxa and metabolites; and (iii) to provide biological plausibility for our epidemiological findings, we performed an exploratory computational analysis to generate a hypothesis regarding the potential molecular interactions between specific microbial metabolites and host lipid signaling. By integrating population-level evidence with genetic causality, our work aims to characterize a previously underexplored “gut-lipid-kidney” axis, offering a novel perspective on the non-renal drivers of CKD mortality.
MATERIALS AND METHODS
Overall study design
This study employed a multi-disciplinary design to investigate a potential gut-lipid-kidney axis.
Stage 1 was clinical association. We first assessed the clinical relevance of gut health in a large U.S. cohort. NHANES data—retrospectively accessed for this study—are overseen by the NCHS/CDC. Ethical clearance from the NCHS Ethics Review Committee and participants’ signed informed consent documents were obtained prior to data gathering. All procedures complied with institutional and federal research regulations (17, 18).
In stage 2, to overcome the inherent limitations of observational studies, such as residual confounding and reverse causation, we employed a two-sample MR design. We identified specific types of gut bacteria and lipid metabolites that have a causal relationship with CKD. Subsequently, we investigated whether constipation affects the abundance of these bacteria and quantified the mediating role of lipids in this microbial pathway. Valid MR analysis requires IVs to (i) strongly associate with the gut microbiota, (ii) show no pleiotropy (no confounding pathways), and (iii) affect outcomes solely via microbiota exposure (19).
Study participants
The participants were drawn from the NHANES database from 2005 to 2010 because data on constipation were recorded over three 2-year cycles (2005–2006, 2007–2008, and 2009–2010). The study included patients who satisfied the following criteria: (i) aged >20 years; (ii) had relevant defecation data; and (iii) had serum creatinine, urine albumin (ALB), and urine creatinine data. One of these conditions was excluded: (i) pregnant woman; (ii) self-reported celiac disease, colon cancer, or inflammatory bowel conditions; and (iii) variables with missing values that could not be interpolated (missing data >15%). The data analysis included 2,569 participants (Fig. 1A).
Study flowchart of participant selection and methodological framework for multi-omics causal inference. (A) Flowchart illustrating the inclusion and exclusion criteria for the clinical epidemiological study based on the National Health and Nutrition Examination Survey (NHANES) 2005–2010. Abbreviations: NHANES, National Health and Nutritional Examination Survey; CKD, chronic kidney disease. (B) Schematic overview of the Mendelian randomization (MR) design. (A) Illustration of the study design and workflow. (B) Mediated Mendelian randomization assessment detailing the effect of the gut microbiota on CKD via lipid composition. C, total effect; AB, indirect effect; C', direct effect. Abbreviations: SNPs, single nucleotide polymorphisms.*
CKD definition
The criteria for CKD were as follows: (i) estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m^2^; (ii) urine albumin-to-creatinine ratio (UACR) ≥ 30 mg/g; and (iii) self-reported CKD. A person is defined as having CKD if any one of these criteria is met. The eGFR is derived through the CKD-EPI equation (20).
Definition of constipation
Constipation was defined by the frequency of feces and fecal traits according to the NHANES database (21). Using the Bristol Stool Form Scale (BSFS), the rate of transport through the colon was assessed, and participants were asked to identify their typical stool type based on visual cues. Constipation was defined as Type 1 (separate hard lumps) or Type 2 (lumpy sausage) (22). The participants reported their bowel movement frequency: “How often do you have a bowel movement?” Those with fewer than three weekly occurrences were classified as constipated.
Data extraction
To minimize potential bias arising from subjective variable selection and to prevent model overfitting, we employed the least absolute shrinkage and selection operator (LASSO) regression algorithm to identify the most significant covariates for the final models (23). An extensive pool of candidate variables was initially screened, encompassing demographics (age, sex, education level, race, poverty income ratio, and marital status), lifestyle factors and medication history (specifically smoking status and opioid use), and major comorbidities (diabetes mellitus, hypertension, congestive heart failure, and history of dialysis). Additionally, clinical and laboratory parameters, including body mass index (BMI), alanine aminotransferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen (BUN), eGFR, glycated hemoglobin (HbA1c), and serum ALB, were evaluated. The penalty parameter (λ) was determined via 10-fold cross-validation, and only variables with non-zero coefficients at the optimal λ were retained for adjustment in the final multivariable Cox proportional hazards models and propensity score matching (PSM) analysis (selection process visualized in Fig. S2).
Missing value treatment
For variables with < 5% missing data, we applied the mice package in R to perform multiple imputation using chained equations (MICE) (Fig. S1) (24).
MR data sources
Constipation statistics were obtained from the FinnGen consortium via the OpenGWAS database (data set ID: finn-b-K11_CONSTIPATION). This data set includes 17,246 cases and 201,546 controls of European ancestry (total N = 218,792). The phenotype was defined based on electronic health records, with associations adjusted for age, sex, genotyping batch, and principal components. Gut microbiota summary statistics were obtained from the FINRISK 2002 cohort (FR02), comprising 5,959 participants. Notably, the microbiome was profiled using shallow shotgun metagenomic sequencing rather than 16S rRNA methods, allowing for species-level resolution. In the source GWAS, microbial abundances underwent centered log-ratio (CLR) transformation, and associations were adjusted for age, sex, genotyping batch, and the first 10 genomic principal components (25). For the lipid mediator analysis, we utilized data from the GeneRISK cohort (N = 7,174), representing the Finnish population. Circulating plasma lipid levels were quantified via high-throughput mass spectrometry-based shotgun lipidomics. The source study adjusted for potential confounders, including age, sex, collection site, lipid medication status, and ancestry, followed by inverse-normal transformation of residuals (26). CKD outcomes were sourced from the Genome-Wide Association Study (GWAS) repository (ID: ebi-a-GCST90018602; 2,117 cases and 174,345 controls).
We initially identified single nucleotide polymorphisms (SNPs) through GWAS with a threshold of P < 1 × 10^−5^, a genetic span of 10,000 kb, and a linkage disequilibrium (LD) coefficient r^2^ < 0.001. To reduce bias for weak IVs, we set F < 10 to be excluded (27). For complete details, please refer to Fig. 1B. The remaining relevant documents are listed in the Supplementary Materials (Tables S6 to S10).
Statistical analysis
Considering NHANES’s intricate multistage sampling strategy, the analyses were adjusted using weights (1/3 × WTSAF2YR). Continuous variables are presented as the means ± SD, whereas categorical variables are shown as frequencies (%). Comparisons between groups were performed via t-tests or ANOVA for continuous data and χ² tests for categorical data. Baseline characteristics were compared using weighted Student’s t-tests for continuous variables and weighted χ tests for categorical variables.
To robustly evaluate the association between constipation and mortality, we constructed three progressive multivariable Cox proportional hazards models. Model 1 was an unadjusted model. Model 2 adjusted for demographic and socioeconomic factors, including age, sex, race, BMI, education level, family income to poverty ratio (PIR), and smoking status. Model 3 (the fully adjusted model) further controlled for dietary factors (dietary fiber and water intake), clinical comorbidities (hypertension, diabetes, history of CVD, dialysis status), medication history (opioid use), and laboratory biomarkers (glycohemoglobin [HbA1c], ALB, BUN, and eGFR). The proportional hazards assumption was verified using Schoenfeld residuals.
To further strengthen causal inference in this observational setting, three sensitivity analyses were performed: (i) PSM was employed to balance baseline characteristics between constipated and non-constipated groups (1:1 matching, caliper 0.05); (ii) a Fine-Gray competing risk model was used to estimate sub-distribution hazard ratios (SHRs) for cardiovascular mortality, accounting for non-CVD death as a competing event; and (iii) a restricted cubic spline (RCS) analysis was conducted to visualize the dose-response relationship between stool frequency/consistency and mortality risk.
Mendelian randomization and mediation analysis
Two-sample MR was conducted to investigate the causal pathway from gut microbiota to CKD outcomes, mediated by lipid traits. Instrumental variables (IVs) for gut microbiota and lipid metabolites were selected from large-scale GWAS summary statistics. The primary causal estimate was derived using the IVW method. To ensure robustness, we performed extensive sensitivity analyses, including MR-Egger regression (to detect pleiotropy), weighted median, and MR-PRESSO (to correct for outliers). Heterogeneity was assessed using Cochran’s Q statistic. Finally, mediation analysis was performed to quantify the proportion of the total effect of gut dysbiosis on CKD risk that is mediated by specific lipid metabolites. All analyses were conducted using R software (version 4.4.2). Two-sided P-values <0.05 were considered statistically significant (28, 29).
RESULTS
Baseline characteristics of participants with CKD
A total of 2,569 participants with CKD were included in the final analysis, among whom 278 (10.8%) reported constipation. In the unmatched cohort, patients with constipation were significantly more likely to be female (P < 0.001) and from lower socioeconomic backgrounds, as evidenced by lower educational attainment (P = 0.029) and a lower PIR (P = 0.004). Clinically, the constipation group presented with significantly lower serum ALB levels compared to the non-constipated group (P < 0.001). Significant differences were also observed in racial distribution and smoking status (P < 0.05). Notably, no significant differences were found in age, BMI, eGFR, BUN, or the prevalence of comorbidities—including hypertension, diabetes, heart failure, dialysis status, and opioid Use—between the two groups (all P > 0.05). Following PSM (1:1), 278 pairs were generated. All baseline characteristics were well-balanced between the constipated and non-constipated groups, with all standardized mean differences (SMD) < 0.10 and P > 0.05 (Table 1).
Survival analysis
The death data were backdated to December 31, 2019, and after a median follow-up of 114 months, there were 1,742 deaths. Kaplan-Meier survival analysis demonstrated that CKD patients with constipation had a significantly lower cumulative survival probability compared to those without constipation (log-rank P = 0.027, Fig. 2A).
Impact of constipation on survival outcomes in patients with CKD. (A) Kaplan-Meier survival curves for all-cause mortality stratified by constipation status. The red line represents the constipation group, and the blue line represents the non-constipation group. Shaded areas indicate 95% confidence intervals. (B) Cumulative incidence function (CIF) curves for cardiovascular disease (CVD) mortality estimated using the Fine-Gray competing risk model. The analysis adjusted for potential confounders, including opioid usage, with non-CVD death treated as a competing event. The red line indicates the constipation group, showing a trend toward higher cumulative incidence of CVD death compared to the non-constipation group (blue line).
To determine whether constipation is an independent risk factor for mortality or merely a marker of poor health, we constructed three sequential multivariate Cox proportional hazards models (Table 2). In the unadjusted model (Model 1), constipation was associated with a 21% increased risk of all-cause mortality (HR: 1.21, 95% CI: 1.02–1.44, P = 0.027). Notably, after adjusting for demographic and socioeconomic factors (Model 2), the association strengthened (HR: 1.35, 95% CI: 1.13–1.60, P < 0.001), suggesting that demographic variables might partially mask the true impact of constipation in the crude analysis. In the fully adjusted model (Model 3), which extensively controlled for dietary habits (intake of water and fiber), comorbidities (hypertension, diabetes, CVD), kidney function markers (eGFR, BUN, UACR, ALB), and crucially, opioid use, the association remained robust and significant. Patients with constipation faced a 33% higher risk of mortality compared to those without (HR: 1.33, 95% CI: 1.11–1.58, P = 0.002).
Cause-specific mortality and dose-response relationships
Given the high burden of cardiovascular disease in the CKD population, we further performed a competing risk analysis using the Fine-Gray model to specifically assess CVD mortality (Fig. 2B). The results revealed a notable trend toward increased CVD mortality risk in patients with constipation. Although the association did not reach the conventional threshold for statistical significance (SHR: 1.31, 95% CI: 0.98–1.75, P = 0.066), the cumulative incidence function (CIF) curves demonstrated a clear visual separation, with the constipation group consistently exhibiting a higher incidence of CVD death throughout the follow-up period compared to the non-constipated group. This suggests that while statistical power may be limited by the number of cause-specific events, the direction of the association is consistent with the all-cause mortality findings.
To clarify the biological gradient of the associations, we utilized RCS analyses, which revealed distinct non-linear patterns (Fig. 3). Regarding stool consistency (Fig. 3A), a U-shaped relationship was observed. Using Type 4 (normal consistency) as the reference, the mortality risk increased significantly at both extremes of the BSFS. Specifically, harder stools (Types 1 and 2, indicating severe constipation) were associated with a steep increase in hazard ratios. Interestingly, loose or liquid stools (Types 6 and 7) also showed an upward trend in risk, suggesting that any deviation from normal bowel consistency—whether constipation or diarrhea—is detrimental in CKD patients. Regarding stool frequency (Fig. 3B), the analysis demonstrated a sharp inverse association. The risk of mortality rose precipitously as the weekly stool frequency dropped below seven times per week. In contrast, for frequencies exceeding seven times per week, the risk curve flattened and remained relatively stable, indicating that low bowel movement frequency is a specific and potent driver of excess mortality in this population.
Dose-response relationship between bowel habits and all-cause mortality in CKD patients. Restricted cubic spline (RCS) models were used to visualize the non-linear association between stool characteristics and mortality hazard ratios (HRs). The solid red line represents the estimated HR, and the shaded red area indicates the 95% CI. The horizontal dashed line represents the reference hazard ratio of 1.0. (A) Stool consistency: analysis of the Bristol Stool Form Scale (BSFS) scores reveals a U-shaped association. (B) Stool frequency: analysis of stool frequency (times/week) demonstrates an L-shaped association.
Subgroup analyses and interaction assessment
Stratified analyses (Fig. 4) confirmed that the mortality risk associated with constipation was particularly pronounced in females (HR: 1.38, 95% CI: 1.11–1.71), participants aged ≥60 (HR: 1.28, 95% CI: 1.07–1.54), and non-Hispanic Whites (HR: 1.29). The direction of the associations remained consistent across other subgroups.
Stratified analyses of the association between constipation and all-cause mortality across key clinical subgroups. A forest plot displaying the multivariable-adjusted hazard ratios (HRs) for all-cause mortality associated with constipation. The red squares represent the point estimates of the HRs, and the horizontal blue lines indicate the 95% confidence intervals (CIs). The vertical gray line at HR = 1.0 serves as the reference for no effect.
Joint effect analyses visualized via heatmaps (Fig. 5) revealed distinct interaction patterns. First, a strong synergistic effect was observed between constipation and opioid use (Fig. 5A); patients with both conditions faced a threefold increase in mortality risk (HR: 3.06, 95% CI: 2.09–4.48) compared to controls. Second, regarding lipid status (Fig. 5B), constipation significantly elevated mortality risk regardless of hyperlipidemia. Notably, the constipation group with normal lipid levels exhibited the highest risk (HR: 1.44, P < 0.05), suggesting that constipation exerts a detrimental effect independent of traditional lipid metabolic disorders.
Heatmaps illustrating the joint effects of constipation and comorbidities/medications on all-cause mortality. (A) Joint effect of constipation and opioid use. The heatmap displays the hazard ratios (HRs) for combinations of constipation status (y-axis) and opioid usage (x-axis). The group with “Non-Constipation” and “Non-User” serves as the reference (HR = 1.00). The intensity of the red color corresponds to the magnitude of the mortality risk. (B) Joint effect of constipation and hyperlipidemia. This panel illustrates the interplay between constipation and hyperlipidemia status.
Sensitivity and robustness checks
To strictly rule out potential biases and validate the reliability of our findings, we performed a comprehensive set of sensitivity analyses (Fig. S4). In the “no opioids“ sensitivity model, the association between constipation and mortality remained statistically significant (HR: 1.32, 95% CI: 1.11–1.60), confirming that our results were not driven by opioid-induced constipation. Second, to mitigate reverse causality (where severe illness leads to both constipation and death), we excluded participants who died within the first two years of follow-up. The results remained consistent (HR: 1.36, 95% CI: 1.15–1.70), demonstrating that the association is not an artifact of terminal illness.
MR mediation analysis
Absence of direct causal effect of constipation on CKD
To delineate the causal architecture underlying the associations observed in the NHANES cohort, two-sample MR analyses were initially performed to test for a direct causal link. The inverse variance weighted (IVW) method revealed no statistically significant direct causal effect of genetically predicted constipation on CKD susceptibility (P > 0.05; Fig. S5). This null result suggests that the clinical correlation between constipation and mortality is likely mediated by indirect ecological or metabolic pathways rather than direct pathogenic mechanisms.
Influence of the gut microbiota on CKD
Unless otherwise specified, the odds ratios (ORs) reported in this section represent the risk of CKD per 1-standard deviation (SD) increase in the genetically predicted relative abundance of the respective bacterial taxa. Multivariate regression analysis identified four bacterial groups associated with an increased risk of CKD, among which the Parabacillus genus showed the strongest effect (OR = 2.477, 95% CI = 1.473–4.165, P = 0.001), followed by Clostridium spp., Eubacterium CAG-274 spp., and Blautia spp.
Conversely, eight taxonomic groups were identified as potential protective factors. These include Herbidospora (OR = 0.614, 95% CI: 0.386–0.978, P = 0.040), as well as other bacteria with protective effects, including some well-known groups, such as Bifidobacterium and Prevotella, as well as Parabacteroides johnsonii spp., Fenollaria spp., and other bacterial species detailed in Fig. 6A and Fig. S7.
Mendelian randomization (MR) evidence elucidating the causal components of the gut-lipid-kidney axis. (A) Forest plot displaying the causal effects of specific gut microbiota taxa on CKD risk. Estimates are derived from the inverse variance weighted (IVW) method and sensitivity analyses (MR-Egger, weighted median). The x-axis represents the odds ratio (OR). Taxa with an OR < 1 are identified as renoprotective, while those with an OR > 1 are associated with increased risk. (B) Causal associations between genetically predicted plasma lipid metabolites and CKD. (C) Scatter plot demonstrating the upstream causal impact of constipation on Herbidospora abundance. The x-axis represents the SNP effects on constipation (exposure), and the y-axis represents the SNP effects on Herbidospora abundance (outcome).
Impact of lipids on CKD
MR analysis revealed causal correlations linking lipid-related variables to CKD development. Specifically, four factors, namely, PC (O-18:0_16:1) levels (OR = 1.209, 95% CI: 1.004–1.457, P = 0.045), PC (O-18:2_20:4) levels (OR = 1.243, 95% CI: 1.025–1.506, P = 0.027), sphingomyelin (d32:1) levels (OR = 1.164, 95% CI: 1.028–1.318, P = 0.017), and triacylglycerol (54:6) levels (OR = 1.205, 95% CI: 1.003–1.448, P = 0.046), were positively correlated, suggesting that they may increase the risk of developing CKD. PC (14:0_18:2) levels (OR = 0.836, 95% CI: 0.704–0.994, P = 0.043) and PC (18:2_20:1) levels (OR = 0.855, 95% CI: 0.742–0.985, P = 0.03) were negatively associated with CKD risk, suggesting a potential protective effect against CKD (Fig. 6B; Fig. S8). A sensitivity analysis was also conducted to validate these findings.
Mediation analysis: the Herbidospora-PC axis
To understand the mechanism by which Herbidospora (a subgenus of the phylum Actinomycetes) confers protection, we investigated its metabolic output. Mediation analysis revealed that PC (14:0_18:2) mediates a significant proportion of the causal relationship between Herbidospora and CKD. The total protective effect of Herbidospora (beta = −0.487) was partially mediated by PC (14:0_18:2), accounting for a mediation proportion of 12.5%. This establishes a functional Herbidospora-PC (14:0_18:2) axis that synergistically protects against CKD.
In this mediated MR study, causal associations between 12 microbial taxa, including Herbidospora, and CKD risk were identified. The mediation analysis revealed that PC accounted for 12.5% of the total effect of Herbidospora on CKD. These findings suggest that Herbidospora-derived PC (14:0_18:2) may exert renoprotective effects by targeting key regulatory pathways.
Constipation drives the depletion of the Herbidospora axis
Having established the renoprotective role of the Herbidospora-PC axis, we sought to determine if constipation disrupts this specific pathway. MR analysis confirmed a significant inverse causal association between genetic liability to constipation and the abundance of Herbidospora (beta = −0.36, P = 0.02). This finding completes the causal chain: constipation causally drives the depletion of Herbidospora, thereby impairing the downstream production of protective PC (14:0_18:2) and compromising renal health (Fig. 6C).
Assessing the robustness of the results
Sensitivity analyses confirmed the robustness of our MR findings. No significant heterogeneity or horizontal pleiotropy was detected via MR-Egger and MR-PRESSO tests (Tables S1 to S4; Fig. S9 and S10). Significance thresholds were adjusted using the Bonferroni correction based on the number of taxa tested at each level (30); associations with P < 0.05 but exceeding these thresholds were considered suggestive. Finally, leave-one-out analyses demonstrated that the results were not driven by any single SNP (Fig. S11 and S12).
Reverse Mendelian randomization
Inverse MR analysis revealed causal associations between Bifidobacterium spp. (OR = 0.955, 95% CI = 0.916−0.996; P = 0.035) and Paenibacillus J*.* spp. *(*OR = 0.98, 95% CI = 0.960−0.999; P = 0.049) and the prevalence of CKD. These results suggest that the potential role of Bifidobacterium spp. and Paenibacillus J spp. should be considered in future studies on the impact of CKD on the gut microbiota (Table S5).
Hypothesis generation: from molecular recognition to environmental transmission
To explore molecular feasibility, we performed molecular docking and 100 ns MD simulations. Results showed that PC (14:0_18:2) stably occupies the PPARγ hydrophobic pocket, maintained by van der Waals interactions with minimal conformational fluctuations (Fig. S6). These data suggest PC (14:0_18:2) as a putative PPARγ ligand (details in Note S1).
Beyond the molecular interaction, we further integrated the epidemiological associations with clinical dietary realities to propose a macroscopic transmission route. As illustrated in Fig. 7, we hypothesize that the depletion of Herbidospora in CKD patients is driven by the disruption of the “soil-plant-gut” transmission chain, primarily due to the strict low-potassium dietary management. This schematic diagram summarizes the proposed multi-level mechanism: from environmental intake interruption to the loss of renoprotective lipid metabolites.
Schematic representation of the environment-diet-gut-kidney axis involving Herbidospora. (Top) Homeostatic pathway. Consumption of unrefined, potassium-rich plant foods permits the ingestion of soil-associated Herbidospora. This taxon functions as a metabolic source of phosphatidylcholine (14:0_18:2) (represented by blue hexagons). These lipid metabolites enter the systemic circulation and support the maintenance of renal tissue integrity. (Bottom) Disrupted pathway in CKD. The therapeutic implementation of a low-potassium diet necessitates food processing methods (e.g., soaking, thorough cooking) that eliminate the exogenous source of Herbidospora. This exclusion, compounded by the physiological stress of constipation, results in a marked depletion of Herbidospora abundance. The consequent deficiency in protective phospholipids diminishes the metabolic defense against renal injury, potentially facilitating fibrosis.
DISCUSSION
Our study integrates clinical epidemiology with multi-omics causal inference to elucidate the prognostic role of intestinal motility in CKD and to uncover its potential microbial underpinnings. Our findings provide robust evidence that constipation is an independent risk factor for all-cause mortality in the CKD population, distinct from the confounding effects of opioid usage or metabolic comorbidities. Beyond this clinical association, our MR analysis reveals a novel, protective role of the environmental genus Herbidospora, suggesting that the gut-kidney axis extends beyond resident flora to include environmental microbial exposures.
Clinical data from large cohorts emphasize that constipation in CKD is not only a benign functional disorder but also a key driver of adverse outcomes. Unlike the general population, CKD patients have long been in a uremic environment characterized by impaired intestinal barrier integrity. We believe that constipation plays a double blow in this situation: the prolonged colonic transport time to some extent increases the exposure of the epithelium to uremic toxins, promoting their systemic translocation. This mechanism explains the independent risk of death observed in our study, even when traditional cardiovascular and kidney-specific risk factors were strictly adjusted through LASSO regression.
We acknowledge that the two cohorts differ in geographical and genetic backgrounds. Despite these differences, the consistent direction of our findings across distinct populations suggests that the gut-kidney axis may be a generalizable phenomenon, though direct causal inference from our integrated approach requires further validation. Our identification of constipation-associated mortality risk in the NHANES cohort may correlate with the protective deficiency observed in the FINRISK cohort. Constipation potentially disrupts the gut microbiota balance, increasing the production of harmful substances. The accumulation of these toxins may induce oxidative stress, increase intestinal permeability, and hinder protective microbiota colonization. Consequently, this reduces beneficial metabolite generation, ultimately compromising renal health (30, 31).
We discovered that Herbidospora mongoliensis plays a primary role within the Actinomycetota phylum. This bacterium predominantly inhabits soil or grassland environments and has previously been extremely rare in the human gut. We propose an ecological hypothesis consistent with these findings and amenable to verification. The “soil-plant-gut” constitutes a continuous microbial transmission chain. Environmental microorganisms entering the human gut via the alimentary pathway may temporarily influence host metabolism or establish functional colonization. Specifically, CKD patients require a strict low-potassium diet and often need to peel, soak, and thoroughly cook root vegetables, which may unintentionally cut off the main contact route of soil-related flora (32, 33). Notably, Herbidospora mongoliensis is classified as a phospholipid-producing PIV strain. This isolate possesses the unique capability to synthesize mixed-acyl PC, consistent with its role in CKD protection (34). Our findings broaden therapeutic strategies and advocate for intervention measures targeting specific functional metabolites to compensate for the inevitable loss of environmental microbial exposure due to necessary dietary restrictions.
This study has several limitations. First, due to the cross-sectional design of the NHANES baseline survey, our survival analysis relied on single-point measurements of covariates. Consequently, we could not account for time-varying changes or cumulative effects of lifestyle factors (e.g., changes in smoking or alcohol habits) or medication adjustments during the follow-up period. Second, data on severe constipation and dietary variables were primarily self-reported, inevitably introducing subjective bias and recall errors. Third, differences in geographical distribution and population composition between the U.S. (NHANES) and European (GWAS) cohorts may limit direct comparability. However, the identification of a plausible gut-kidney axis signal across these diverse settings strengthens the possibility that we have uncovered a fundamental and conserved biological pathway. Fourth, while MR minimizes confounding, it cannot prove causality like randomized trials, and molecular docking only predicts structural interactions without confirming biological activation. Therefore, our findings should be interpreted as supportive mechanistic hypotheses requiring experimental validation.
Conclusion
Evidence from this multi-cohort study outlines a novel gut-lipid-kidney axis, linking constipation-associated mortality in CKD to gut dysbiosis. The trajectory extends from genetic identification of Herbidospora mongoliensis and its metabolite PC (14:0_18:2) as protective factors to a structural mechanism via stable binding to PPARγ. Consequently, targeting this axis—through next-generation probiotics or metabolite supplementation—represents a translatable strategy for slowing CKD progression.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Bikbov B, Purcell CA, Levey AS, Smith M, Abdoli A, Abebe M, Adebayo OM, Afarideh M, Agarwal SK, Agudelo-Botero M, et al.. 2020. Global, regional, and national burden of chronic kidney disease, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. The Lancet 395:709–733. doi:10.1016/S 0140-6736(20)30045-3PMC 704990532061315 · doi ↗ · pubmed ↗
- 2Kim K, Kim J-E, Kim JH, Ahn SH, Jung CY, Hwang SD, Lee SW, Song JH. 2024. Real-world evidence of constipation and laxative use in the Korean population with chronic kidney disease from a common data model. Sci Rep 14:6610. doi:10.1038/s 41598-024-57382-738503885 PMC 10951406 · doi ↗ · pubmed ↗
- 3Stevens PE, Levin A, Kidney Disease: Improving Global Outcomes Chronic Kidney Disease Guideline Development Work Group Members. 2013. Evaluation and management of chronic kidney disease: synopsis of the kidney disease: improving global outcomes 2012 clinical practice guideline. Ann Intern Med 158:825–830. doi:10.7326/0003-4819-158-11-201306040-0000723732715 · doi ↗ · pubmed ↗
- 4Platell C, Hall J, Dobb G. 1990. Impaired renal function due to raised intraabdominal pressure. Intensive Care Med 16:328–329. doi:10.1007/BF 017063592212259 · doi ↗ · pubmed ↗
- 5Coluzzi F, Caputi FF, Billeci D, Pastore AL, Candeletti S, Rocco M, Romualdi P. 2020. Safe use of opioids in chronic kidney disease and hemodialysis patients: tips and tricks for non-pain specialists. Ther Clin Risk Manag 16:821–837. doi:10.2147/TCRM.S 26284332982255 PMC 7490082 · doi ↗ · pubmed ↗
- 6Vaziri ND, Goshtasbi N, Yuan J, Jellbauer S, Moradi H, Raffatellu M, Kalantar-Zadeh K. 2012. Uremic plasma impairs barrier function and depletes the tight junction protein constituents of intestinal epithelium. Am J Nephrol 36:438–443. doi:10.1159/00034388623128155 PMC 3725306 · doi ↗ · pubmed ↗
- 7Vaziri N.D., Zhao Y-Y, Pahl MV. 2016. Altered intestinal microbial flora and impaired epithelial barrier structure and function in CKD: the nature, mechanisms, consequences and potential treatment. Nephrol Dial Transplant 31:737–746. doi:10.1093/ndt/gfv 09525883197 · doi ↗ · pubmed ↗
- 8Obata Y, Pachnis V. 2016. The effect of microbiota and the immune system on the development and organization of the enteric nervous system. Gastroenterology 151:836–844. doi:10.1053/j.gastro.2016.07.04427521479 PMC 5102499 · doi ↗ · pubmed ↗