Associations among serum trimethylamine-N-oxide (TMAO) levels, kidney function and infarcted coronary artery number in patients undergoing cardiovascular surgery: a cross-sectional study

This cross-sectional study was performed at the intensive care unit of Jichi Medical University Saitama Medical Center. Data were monitored at the Division of Molecular Epidemiology, Jikei University School of Medicine. Planned analyses proceeded in full compliance with the principles of the Declaration of Helsinki. The study protocol was reviewed and approved by the ethics committees of Jichi Medical University School of Medicine (approval #09-23) and Jikei University School of Medicine (approval #21-184 6062).

Clinical information such as baseline characteristics, comorbidity factors, various clinical scores, and medical history was extracted from medical records and pre-anesthesia interview forms. Age and sex were checked in the clinical chart. Body mass index (BMI) (kg/m²) was calculated as weight divided by the square of height. Laboratory data for serum creatinine (Cr) and low-density lipoprotein-cholesterol (LDL-C) were collected in the outpatient clinic before surgery in the fasting state. Estimated glomerular filtration rates (eGFR) (mL/min/1.73 m²) were calculated separately for men (eGFR = 194 × Cr^−1.094 × age^−0.287) and women (eGFR = 194 × Cr^−1.094 × age^−0.287 × 0.739) [5]. CKD stages were defined based on eGFR levels as follows: Stage 1 CKD, eGFR ≥90 ml/min/1.73 m²; Stage 2 CKD, eGFR ≥60 to <90 ml/min/1.73 m²; Stage 3 CKD, eGFR ≥30 to <60 ml/min/1.73 m²; Stage 4 CKD, eGFR ≥15 to <30 ml/min/1.73 m²; Stage 5 CKD, eGFR <15 ml/min/1.73 m²; and Stage 5D, patients treated with hemodialysis. Pre-operative use of β-blockers, angiotensin-converting enzyme inhibitors (ACEIs), angiotensin II receptor blockers (ARBs), statins, and insulin was checked. The following definitions were used: diabetes, fasting plasma glucose ≥126 mg/dL, HbA1c ≥6.5 % based on the Japanese Diabetes Society, or medication with insulin or hypoglycemic agents; hypertension, systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or medication with antihypertensive agents [6]; dyslipidemia, fasting LDL-C ≥140 mg/dL, triglycerides ≥220 mg/dL, or medication with lipid-altering agents [7]; and hemodialysis, maintenance hemodialysis before cardiovascular surgery. The pre-operative clinical and physical status of each patient was evaluated using the New York Heart Association (NYHA) guidelines [8].

The number of infarcted coronary arteries was determined using coronary angiography by cardiac surgeons before the cardiovascular surgery, which was used as the outcome measure in this study.

Serum TMAO levels were measured in a post hoc analysis of our previous study [4]. Serum samples for TMAO measurements were taken in the fasting state, and were collected in the operating room before the induction of anesthesia to avoid the dilution effect of intravenous fluid infusion and stored at −80 °C. TMAO levels were quantified using stable-isotope dilution high-performance liquid chromatography-atmospheric pressure chemical ionization tandem mass-spectrometry (HPLC-APCI-MS/MS) in positive multiple-reaction monitoring (MRM) mode using deuterated internal standards on a QTRAP triple-quadruple mass spectrometer (AB SCIEX, Framingham, MA) and trimethylamine-d9 N-oxide (d9-TMAO) (CAS number #1161070-49-0, Sigma-Aldrich Corp., St. Louis, MO) as the internal standard.

To 100 μL of human serum samples at room temperature were added 429 μL of isopropanol, 286 μL of acetonitrile, and 286 μL of Milli-Q water (Millipore, Bedford, MA), followed by the internal standard d9-TMAO. Mixtures were separated by centrifugation for 10 min at 13,360×g (Microcentrifuge 5415R, Eppendorf Inc., Enfield, CT), and supernatant samples (500 μL) were transferred to 1.5-mL Eppendorf tubes. The supernatants were evaporated to dryness using a centrifugal concentrator CC-105 (Tomy Tech USA, Fremont, CA) for 10 h, resuspended in 40 μL of ultrapure water (Wako Pure Chemical Industries, Tokyo, Japan), and vortex-mixed for 10 s. Supernatant (10 μL) was directly injected into the HPLC-APCI-MS/MS system.

Serum TMAO was quantified by HPLC using an Agilent 1100 series (Life Technologies, Santa Clara,CA) and column (CAPCELL CORE ADME S2.7, 2.1 × 75 mm, Shiseido, Tokyo, Japan) in positive mode. Separation proceeded using a linear gradient of 1 % formic acid (A) and 100 % methanol (B) as follows: 1 % formic acid for 0.5 min, then 100:0 (A:B) to 40:60 (A:B) for 6 min and 40:60 (A:B) to 100:0 (A:B) for 6.1 min at a flow rate of 0.3 mL/min at 20 °C. The HPLC instrument was coupled to a QTRAP triple-quadruple mass spectrometer (Life Technologies) with atmospheric pressure chemical ionization. The precursor-production ion pairs used in MRM mode were: m/z 76 → 59 for TMAO and m/z 85 → 68 for the internal standard. Other mass spectrometer parameters included ion spray voltage 5000 V and vaporizer temperature 400 °C. Data acquisition was controlled using Analyst 1.6.1 software (AB SCIEX). The calibration standard was 150 μL of TMAO dissolved in ultrapure water. The concentrations of the calibration standards were 2–50 μM TMAO. A calibration curve was constructed by plotting the peak area ratio of TMAO to the internal standard against its concentration. Accuracy is expressed as a ratio (%) of the nominal concentration, and precision is expressed as relative standard deviation. Fresh calibration standards were prepared every week, and all remained stable during the analysis.

The covariates were defined a priori as: (1) age; (2) sex; (3) smoking status divided into three categories (never, previous, and current smoker); (4) BMI; (5) CKD stage; (6) diabetes; (7) HbA_1c; (8) insulin use; (9) hypertension; (10) ACEI or ARB use; (11) dyslipidemia; (12) LDL-C; (13) statin use; (14) congestive heart failure; (15) NYHA; (16) β-blocker use; and (17) previous history of cerebrovascular disease. In multiple regression analysis, multicollinearity was resolved by selecting covariates that reflects the status of the same disease; in the category of diabetes, covariates were (6) diabetes, (7) HbA_1c and (8) insulin use; in the category of hypertension, covariates were (9) hypertension and (10) ACEI or ARB use; in the category of dyslipidemia, covariates were (11) dyslipidemia, (12) LDL-C and (13) statin use; and in the category of heart failure, covariates were (14) congestive heart failure, (15) NYHA and (16) β-blocker use. From each category, the one covariate with the lowest p value and the highest odds ratio in the single regression analysis was selected. Finally, the absence of multi-collinearity among these 10 covariates was confirmed by variance inflation factors. Patient characteristics were stratified by quartiles of TMAO levels in serum and compared with a priori-defined 17 covariates, of which significant deviations were evaluated by linear regression analyses or Chi-square tests. First, the association between the number of infarcted coronary arteries and the quartiles of TMAO or each covariate was assessed with single ordered logistic regression models. Second, a multiple ordered logistic regression model was used to compute the association between the number of infarcted coronary arteries and quartiles of TMAO with adjustment by 10 covariates. Third, a multiple linear regression model instead of the multiple ordered logistic regression model was used to confirm the same trend with a different statistical model. To ensure robust main results in the sensitivity analyses, quintiles of TMAO levels, instead of quartiles of TMAO levels, and natural logarithm-transformed values of TMAO were used to confirm the main results. The risks were evaluated by odds ratios (ORs) and coefficients such as β with 95 % confidence intervals (95 % CIs). Values of p < 0.05 were considered significant. All data were statistically analyzed using STATA version 14.0 (STATA Corp., College Station, TX).

On multivariate adjustment using these 10 covariates, the association between quartiles of TMAO and the outcome measure was re-evaluated using a multiple ordered logistic regression model. A significantly increased number of infarcted coronary arteries was identified in the highest quartile of TMAO versus the lowest quartile (OR 11.9; 95 % CI 3.88–36.7; p < 0.001), independent of diabetes, dyslipidemia, and other covariates. Third, the association was reanalyzed with a multiple linear regression model, and similar results were obtained; a significantly increased number of infarcted coronary arteries was seen in the highest quartile of TMAO versus the lowest quartile (β 1.33; 95 % CI 0.83–1.83; p < 0.001), independent of other covariates.