Circulating trimethylamine N-oxide is correlated with high coronary artery atherosclerotic burden in individuals with newly diagnosed coronary heart disease

This was a prospective observational single-center study in Peking University First Hospital, Beijing, China. We consecutively included adults with suspected CHD undergoing coronary angiography. Electronic medical records were utilized to collect demographics, lifestyle, medical history, and disease status information. We enrolled individuals aged ≥ 18 years with suspected CHD underwent coronary angiography. The exclusion criteria were: (1) Previous diagnosed CHD by coronary angiography; (2) previous percutaneous coronary intervention (PCI) or coronary artery bypass surgery (CABG); (3) without informed consent for sample collection. The definition of CHD was in accordance with the 2019 ESC guideline of CHD [18]. All subjects signed an informed consent form, and the study complied with medical ethics standards.

From January 1, 2019, to December 31, 2019, 1061 patients were included in the participants screening. According to the inclusion and exclusion criteria, 875 were eligible for this study. 18 subjects without blood sample were excluded. Further, 428 participants without CHD (the greatest coronary stenosis degree < 50%) were excluded. As a result, 429 newly diagnosed CHD cases we finally included in the statistical analysis, including 317 chronic coronary syndrome (CCS) cases and 112 acute coronary syndrome (ACS) cases (89 unstable angina pectoris (UA) cases, 19 NSTEMI cases, and 4 STEMI cases).

Blood samples were collected from radial or femoral access before heparinization during angiography using vacutainer tubes containing EDTA. Then the blood sample was centrifuged and stored at -80 °C. API 3200 triple quadrupole mass spectrometer (AB SCIEX, USA) was used to measure plasma TMAO according to the manufacturer’s instructions [19].

The SYNTAX score was assessed by trained physicians who were blinded to the TMAO levels. The SYNTAX score was computed using the online SYNTAX score calculator version 2.28 (www.syntaxscore.com/calculator/syntaxscore/frameset.htm↗). Patients were divided into three groups according to the SYNTAX score: low SYNTAX score group (SYNTAX score < 23), intermediate SYNTAX score group (23 ≤ SYNTAX score < 33), and high SYNTAX score group (SYNTAX ≥ 33). During coronary angiography, the stenosis degree of each involved vessel was recorded, and the vessel displaying the most severe stenosis was included in the final statistical analysis as the severest degree of stenosis.

Hypertension was regarded as meeting one or more of the criteria: SBP ≥ 140 mmHg or DBP ≥ 90 mmHg or physician-diagnosed hypertension or using antihypertensive drugs. Diabetes mellitus was regarded as meeting one or more criteria: physician-diagnosed diabetes, taking antidiabetic drugs, using insulin, glycated hemoglobin level ≥ 6.5%, FBG level ≥ 7.0 mmol/L, or a 2 h glucose level ≥ 11.1 mmol/L after an oral glucose tolerance test. Hyperlipidemia was regarded as physician-diagnosed hyperlipidemia or using lipid-lowering drugs.

Continuous variables were presented as mean ± standard deviation (SD) or median (IQR). Categorical variables were expressed as proportions. Due to the distribution of values for TMAO being strongly skewed, TMAO underwent log₂ transformed. The correlation between TMAO and SYNTAX score was investigated using linear regression model as a continuous variable per 1 µmol/L increase in log₂TMAO. Odds ratios (ORs) and 95% confidence intervals (95% CIs) for SYNTAX in relation to TMAO were calculated using logistic regression models. Two-tailed p < 0.05 was considered as statistically significant. R software (version: 4.2.2; http://www.R-project.org↗) was adopted for statistical analysis.

A total 429 newly identified CHD cases met the inclusion criteria were included in the final analysis (Fig. 1). Patients were grouped into low, intermediate, and high atherosclerosis burden groups based on their SYNTAX scores. The mean age (SD) was 63.89 (10.48) years; with 46.9% being male; 72.7% were hypertensive patients; 48.1% had a diagnosis of diabetes mellitus; 71.1% of subjects were diagnosed with dyslipidemia; 46.5% subjects had a smoking history. The average syntax score was 17.44 (13.41) points and the median TMAO level was 5.90 µmol/L [IQR: 4.90 to 8.30]. Individuals with greater atherosclerotic burden are more likely to be male (44.3% vs. 67.5% in SYNTAX < 23 and SYNTAX ≥ 33 groups, respectively; p = 0.021), with higher SBP (132.28 mmHg vs. 137.13 mmHg in SYNTAX < 23 and SYNTAX ≥ 33 groups, respectively; p = 0.019), had a larger number of involved arteries (1.84 vs. 2.85 in SYNTAX < 23 and SYNTAX ≥ 33 groups; p < 0.001), and had a more severe coronary stenosis degree (87.65% vs. 96.47% in SYNTAX < 23 and SYNTAX ≥ 33 groups, respectively; p < 0.001) (Table 1). As the distribution of TMAO was skewed (Supplementary Fig. 1A), we therefore conducted log₂ transformation for TMAO to enhance its suitability for further analysis (Supplementary Fig. 1B).

Plasma log₂TMAO levels were significantly correlated with the SYNTAX score in newly diagnosed CHD patients (Pearson’s correlation coefficient: r = 0.439, p < 0.001). Based on the linear regression model, in model 1, we treated SYNTAX score as dependent variable and _log2TMAO as independent variable. The results showed a positive correlation between log₂TMAO and SYNTAX score. For each 1 µmol/L increase in log₂TMAO there was associated with a 0.287 SD increase in the SYNTAX score (p = 0.001). In model 2, we used SYNTAX score as the dependent variable and including log₂TMAO, age, BMI, SBP, DBP, smoking status, drinking status, self-reported hypertension, self-reported diabetes, self-reported hyperlipidemia, family history of CHD, LDL-C and Hcy as independent variables. Significant correlation was also identified in this multivariable adjusted model (β = 179, p = 0.006). Additionally, when log₂TMAO was evaluated in quantiles, significantly higher SYNTAX score was identified in the third (Model 1: β = 0.241, p < 0.001; Model 2: β = 0.229, p < 0.001) and the fourth (Model 1: β = 0.433, p < 0.001; Model 2: β = 0.412, p < 0.001) quantiles in both unadjusted and adjusted models. See Table 2 for detailed information.

Given the significant positive correlation between TMAO and SYNTAX, stratified analysis was conducted to explore whether the correlation remains significant in different subgroups. Individuals with SYNTAX ≥ 23 were classified as subjects with intermediate to high coronary atherosclerotic burden, while individuals with SYNTAX < 23 were regarded as subjects with low burden. The results indicated that most subgroups exhibited positive correlations between TMAO and SYNTAX score. However, subjects who were non-smoker, and having diabetes mellitus failed to show significant correlation. None of these variables were observed to alter the correlation between TMAO and atherosclerotic burden (all p for interaction > 0.05) (Table 3; Fig. 2).

The discrimination abilities of a model incorporating traditional cardiovascular risk factors, such as BMI, age, SBP, DBP, smoking, drinking, hypertension history, diabetes mellitus history, CHD history, dyslipidemia history, LDL-C, and Hcy, were initially estimated using ROC curves and AUC. The results showed that the AUC of the combination of traditional risk factors was 0.7312. Subsequently, we added the log₂TMAO to the traditional model to construct a new model. The AUC of the new model increased to 0.7502 (Fig. 3). Model comparison revealed the discrimination ability of the new model outperformed the traditional model. The difference between models reaching statistical significance (p = 0.0027).

Below is the link to the electronic supplementary material.

Supplementary Material 1