Association Between Trimethylamine N-oxide and Adverse Kidney Outcomes and Overall Mortality in Type 2 Diabetes Mellitus

In this prospective study, we recruited patients with T2D at Kaohsiung Medical University Hospital (KMUH) on a consecutive basis from October 2016 to June 2020. During this period, all eligible patients with T2D attending outpatient departments were systematically and continuously invited to participate. T2D was defined as a medical history, blood glucose levels in accordance with the American Diabetes Association's criteria or the use of antidiabetic medications. All participants received instructions on an appropriate diet for diabetes. Kidney function was assessed by the estimated glomerular filtration rate (eGFR), as determined using the 2009 Chronic Kidney Disease Epidemiology Collaboration creatinine formula (11). We excluded patients with an eGFR < 15 mL/min/1.73 m², those with acute illnesses, and those who had used antimicrobials or probiotics within the 1 month before enrollment to diminish the effect of baseline kidney function and gut microbiome on the relationship between TMAO and adverse kidney outcomes. The protocol for this study was approved by the institutional review board of KMUH (KMUHIRB-G(II)-20160021), informed consent was obtained in written form from all study subjects, and all clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki.

Demographic characteristics, a history of currently or ever having smoked cigarettes or drunk alcohol, and clinical data from interviews and medical records were collected at enrollment. Hypertension was based on history or the use of antihypertensive drugs, ischemic heart disease was defined as a history of acute or chronic ischemic heart disease, or myocardial infarction, whereas heart failure was defined according to the previously published Framingham criteria (12). Body mass index was calculated as weight divided by height squared. Information regarding the use of antidiabetic drugs, insulin, statins, angiotensin-converting enzyme inhibitors, or angiotensin II receptor blockers before and after enrollment was obtained from medical records.

The patients were instructed to abstain from food for a minimum of 12 hours before the collection of blood and single-point urine samples the following morning for biochemical analysis. Serum creatinine was measured according to the compensated Jaffé (kinetic alkaline picrate) method using an autoanalyzer (Roche/Integra 400, Roche Diagnostics) and a calibrator that could be used in isotope-dilution mass spectrometry (13). Protein in urine was expressed as urinary albumin/creatinine ratio (UACR).

Twelve-hour fasting blood samples were processed and immediately stored at −80 °C for future analysis. TMAO is stable when stored at −80 °C for extended periods and can withstand multiple freeze-thaw cycles. TMAO was measured using the stable isotope dilution method, combined with high-performance liquid chromatography and online tandem mass spectrometry (14). The laboratory assay's coefficient of variation for TMAO fluctuated across different batches, with a maximum value of 5.8%.

The participants were monitored until the occurrence of specified kidney outcomes, mortality, final contact, or the conclusion of the observation period in December 2021. The patients were regularly assessed at outpatient clinics every 3 months to monitor their clinical status and measure eGFR. eGFR data obtained during hospitalization and episodes of acute kidney injury were excluded from the analysis to minimize the impact of acute illness on the assessment of the decline in renal function. The primary kidney outcomes were the doubling of serum creatinine levels or progression to ESKD; the secondary kidney outcome was defined as a 30% decline in eGFR within the first 2 years (15). Changes in eGFR were calculated as the percentage change between the first and last measurements (16). Cardiovascular mortality was defined as death resulting from ischemic heart disease, myocardial infarction, congestive heart failure, arrhythmia, or cerebrovascular accident. Information on all-cause mortality was obtained by direct contact with the participants and their families and was further supplemented by reviewing medical records and the databank of the National Mortality File.

Baseline characteristics were grouped by tertile TMAO level, with differences between continuous variables assessed using t-tests or Mann-Whitney U tests, and categorical variables using χ² tests. Continuous variables with skewed distribution underwent log10 transformation to approximate a normal distribution before further analysis. We used the cumulative incidence competing risk method to account for competing risks, where death could preclude potential subsequent primary kidney outcomes (17, 18). Additionally, the Kaplan-Meier method was used for the analysis of all-cause mortality. We compared cumulative incidence functions using the Gray test, a technique similar to the log-rank test for Kaplan-Meier estimators (17). Propensity scores for TMAO tertile groups were estimated by a multinomial logistic regression model, ensuring balanced baseline covariates and reducing confounding factors across the 3 groups (19). Cox proportional hazards analysis with inverse probability of treatment weighting was used to examine the associations between TMAO level and primary kidney outcomes and all-cause mortality (20). Logistic regression models with inverse probability of treatment weighting were used to assess the association between serum TMAO level and 30% decline in eGFR within the initial 2 years (20). To evaluate the model classification performance, time-dependent receiver operating characteristic (ROC) curves were used to account for dynamic changes in disease status, offering a more precise evaluation of biomarker predictivity over time (21). This method was applied to appraise the prediction accuracy of TMAO, UACR, and eGFR for primary kidney outcomes at 1-, 3-, and 5-year intervals (21). Additionally, the Hosmer-Lemeshow test was used to assess the calibrations of TMAO, eGFR, and UACR separately in predicting the primary kidney outcomes (22, 23). We further analyzed factors that might modify the association between serum TMAO levels and the risk of progression to ESKD or doubling of serum creatinine, including sex (male/female), age (≥60 or <60 years), T2D duration (≥10 or <10 years), baseline eGFR (≥60 or <60 mL/min/1.73 m²), glycated hemoglobin (HbA1c; >7% or ≤7%), and baseline UACR (≥30 mg/g or <30 mg/g). Statistical analyses were executed using SPSS (version 20, IBM Corporation, Armonk, NY) and R statistical analysis software, version 4.2.2, with the “survival,” “survminer,” “dplyr,” and “cmprsk” packages for time-to-event evaluations and competing risk analysis, “timeROC” for estimation of time-dependent ROC curve, “twang” for propensity score estimating and weighting, and “rms” and “ggplot2” were used for modeling and visualization purposes (Foundation for Statistical Computing, Vienna, Austria; available at: http://www.R-project.org↗). All P values were 2-sided, with a significance threshold established at <.05.

Table 1 presents comparisons of baseline clinical characteristics, medication records, and laboratory parameters among participants stratified by tertiles of serum TMAO levels cut at 0.40 µM and 0.88 µM. The study population consisted of 440 patients with T2D and a mean age of 62.6 ± 10.9 years, of whom 55.0% were male. The median T2D duration was 9.0 years, and the prevalence rates of hypertension, heart failure, ischemic heart disease, and hyperlipidemia were 60.9%, 24.1%, 9.5%, and 78.2%, respectively. The median eGFR and HbA1c level were 79.8 mL/min/1.73 m² and 7.0%, respectively. The patients in the highest tertile of serum TMAO levels were generally older, had a longer duration of T2D, and demonstrated higher usage of sulfonylurea and insulin compared with those in the lowest tertile. They also had higher baseline blood urea nitrogen and lower eGFR and hemoglobin levels than those in the lowest tertile.

Over a mean follow-up period of 4.0 years, 26 patients (5.9%) reached a doubling of serum creatinine level or progression to ESKD (the primary kidney outcomes), 24 patients (5.5%) reached a doubling of serum creatinine level, 10 patients (2.3%) developed ESKD, and 15 patients (3.4%) died (Table 2). The patients in the highest tertile of serum TMAO had a higher incidence rate of primary kidney outcomes, including either doubling of serum creatinine level or ESKD, with an incidence rate of 35.1 per 1000 person-years compared with 9.1 and 3.6 per 1000 person-years in the patients in the middle and lowest tertiles, respectively.

Univariate analysis was performed to ascertain individual risk factors linked to the primary kidney outcomes. The results revealed positive correlations between the primary kidney outcomes and duration of T2D, body mass index, smoking, hypertension, UACR, log-transformed triglyceride, serum cholesterol, and log-transformed TMAO levels. Cumulative probability analysis showed a significantly elevated risk of primary kidney outcomes in the highest tertile compared with the middle and lowest tertiles (Gray test, P < .001; Fig. 1A and Supplementary Table S1) (24). After propensity score weighting, serum log-transformed TMAO was positively associated with primary kidney outcomes including doubling of serum creatinine Level or progression to ESKD (hazard ratio [HR], 5.32; 95% CI, 1.99-14.25). The patients in the highest tertile had a 6.45-fold (95% CI, 1.42-29.25) higher risk of reaching a primary outcome compared with those in the lowest tertile (Table 3).

Time-dependent ROC analysis revealed that the area under the curve (AUC) for TMAO (0.953) was comparable to UACR (0.964) and eGFR (0.987) at 1 year, suggesting that serum TMAO had similar activity to predict kidney outcomes in T2D as UACR and eGFR at 1 year. At 3 and 5 years, the AUC of UACR was persistently higher than eGFR and serum TMAO, and the AUCs of serum TMAO and eGFR were similarly decreased over time (Fig. 2A–C). These findings indicated that TMAO's classification strength was not inferior with eGFR over time. The calibration of our prognostic models for serum TMAO, UACR, and eGFR was depicted graphically through scatter plots (Fig. 2D–2F), with predicted vs observed event numbers for primary kidney outcomes (Supplementary Tables S2A–C) (24). The TMAO-predicted events aligned closely with actual events, as indicated by the proximity of data points to the identity line (Fig. 2D). The nonsignificant Hosmer-Lemeshow test (P = .64) further supported that TMAO could adequately predict kidney outcome. Conversely, the calibration plots for UACR (Fig. 2E) and eGFR (Fig. 2F) identified a slight deviation from the expected number of primary kidney outcomes and failed Hosmer-Lemeshow test results (P = .018 for UACR and P = .015 for eGFR).

During the follow-up period, 22 (5.0%) patients had a decline in eGFR of ≥30% within the first 2 years. In addition, the patients in the highest tertile of serum TMAO had a higher incidence rate of kidney function decline (33.4 per 1000 person-years), compared with rates of 5.5 and 3.6 per 1000 person-years in the patients in the middle and lowest tertiles, respectively (Table 2). In univariate analysis, serum TMAO levels were positively correlated with a 30% decline in eGFR within the first 2 years. This association remained statistically significant after adjusting for confounders through propensity score weighting (Table 3). Serum log-transformed TMAO was positively associated with rapid decline in kidney function (HR, 9.59; 95% CI, 2.94-31.19). The patients in the highest tertile had a 5.86-fold (95% CI, 1.31-26.17) higher risk of reaching secondary outcome compared with those in the lowest tertile (Table 3).

We stratified the patients by sex (male/female), age (≥60 or <60 years), diabetes duration (≥10 or <10 years), baseline eGFR (≥60 or <60 mL/min/1.73 m²), HbA1c (> 7% or ≤7%), and UACR (≥30 mg/g or <30 mg/g) to examine the interaction effects of these factors on the relationship between circulating TMAO and primary kidney outcomes. The results revealed a significant interaction between TMAO level and sex (P = .018) in relation to the primary kidney outcomes (Supplementary Table S3) (24). There were no significant interaction effects between age, diabetes duration, baseline eGFR, UACR, or HbA1c on the association between circulating TMAO and primary kidney outcomes. These findings suggested a potential sex-specific relationship between TMAO and kidney health, with the male patients being more likely to reach a primary kidney outcome than the female patients.

The patients in the highest tertile of serum TMAO had a significantly higher all-cause mortality rate (5.9%, 13 deaths) compared with those in the middle (2.7%, 2 deaths) and lowest tertiles (0.7%, 1 death) (Table 2). The causes of mortality included cardiovascular events (4 cases), sepsis (5 cases), and cancer (6 cases). Besides, Kaplan-Meier analysis showed a greater cumulative probability of all-cause mortality in the highest TMAO tertile group (log-rank test, P < .001, Fig. 1B, and Supplementary Table S1) (24). After the propensity score weighting, serum log-transformed TMAO was found to be significantly correlated with all-cause mortality (HR, 4.58; 95% CI, 1.61-13.00). The patients in the highest tertile had an 9.04-fold increased risk of mortality (95% CI, 1.13-72.66) compared with their counterparts in the lowest tertile, indicating the potential link between elevated circulating TMAO level and increased risk of all-cause mortality.