Circulating Metabolites and Cardiovascular Disease in Asians with Chronic Kidney Disease

Chronic kidney disease (CKD) poses a significant global health problem, affecting more than 10% of the global population, corresponding to 843.6 million people worldwide as of 2017 [1]. Of the growing CKD burden, Asia is expected to contribute to a substantial proportion, with the greatest projected increase in use of renal replacement therapy relative to other regions worldwide [2].

Among patients with CKD, cardiovascular disease is the leading cause of mortality. Cardiovascular mortality accounts for at least 35–40% of all deaths in patients with CKD stages G3-5, compared to 26% in controls with normal kidney function [3]. In Asia, cardiovascular disease was the leading cause of death in 2019, accounting for 10.8 million deaths, which were 35% of total deaths in Asia. Nearly 39% of these CVD deaths were premature (defined as the death of a person age <70 years), which is remarkably higher than premature deaths in the USA (23%), Europe (22%), and globally (34%) [4]. This increased risk of cardiovascular disease and death in CKD is not fully accounted for by traditional risk factors, sparking interest in novel biomarkers that can be integrated into risk prediction models.

Kidney disease has complex interactions with metabolism, not only directly impacting systemic metabolism leading to peripheral insulin resistance and protein energy wasting, but also affecting the levels of circulating metabolites in the body. Second only to the heart in mitochondrial abundance, the kidney has significant intrinsic metabolic activity and performs the energy-intensive task of solute and water homeostasis [5]. Given the complex interaction between metabolism and kidney disease, metabolic profiling offers the opportunity to greatly understand the pathophysiology of cardiovascular disease in CKD, which involves the interplay between inflammation, oxidative stress, and endothelial dysfunction [3, 6].

Prior studies evaluating metabolic profile and cardiovascular risk were focused predominantly in healthy American [7] and European cohorts [8, 9]. Studies evaluating the metabolic profile of cardiovascular disease in the setting of CKD remain limited, other than two studies in an American cohort of patients (Clinical Phenotyping Resource and Biobank Core of the Michigan O'Brien Renal Center, C-PROBE) and one in a Danish cohort (Copenhagen Chronic Kidney Disease Cohort, CPG CKD) [10‒12], with scant information in the Asian population. However, ethnic differences in CKD and cardiovascular disease have been noted and possibly contributed by a combination of differential exposures and susceptibility, but not well understood. The relationship between metabolites and cardiovascular disease in CKD may not be extrapolated from predominantly White populations until inter-ethnic variations are better understood [13]. However, to the best of our knowledge, there are no prior studies evaluating metabolic profiling of cardiovascular disease in CKD in the Asian population.

In our study, we aimed to determine serum/plasma metabolites associated with prevalent cardiovascular disease in Asian patients with CKD and further elucidate their clinical implications. Of the metabolic profiling tools, mass spectrometry and nuclear magnetic resonance (NMR) spectroscopy, the latter enables untargeted analyses to explore molecular “fingerprints” and offers a cost-effective, reproducible, and high-throughput means of metabolite quantification that is useful in large epidemiological studies [14]. Hence, a NMR spectroscopy approach was used, allowing for comprehensive metabolic profiling of lipids, lipoproteins, apolipoproteins, fatty acids, ketone bodies, amino acids, and glycolysis-related metabolites [14], to identify useful biomarkers associated with increased cardiovascular risk.

Metabolic profiling was performed on blood samples from Chinese, Indian, Malay participants aged 40–80 years who participated in the baseline visit of three independent population-based cohort studies in Singapore, as part of the Singapore Epidemiology of Eye Disease (SEED) programme: the Singapore Malay Eye Study (SiMES, 2004–2006, n = 3,280), the Singapore Indian Eye Study (SINDI, 2007–2009, n = 3,400), and the Singapore Chinese Eye Study (SCES, 2009–2011, n = 3,353). Details of recruitment and methodology of these studies have been published previously [15].

All three studies had an identical methodology and were conducted in the same study clinic. All participants underwent a questionnaire interview, standardized clinical examination, and the collection of blood samples. The studies were conducted in accordance with the Declaration of Helsinki, with ethics approval obtained from the Singapore Eye Research Institute Institutional Review Board. Written informed consent was provided by all participants.

For the current analysis, only participants with estimated glomerular filtration rate (eGFR) of <60 mL/min/1.73 m² (CKD stage G3–5) were included. After exclusion of patient without eGFR data or cardiovascular disease status, 153 Chinese, 262 Indians, and 707 Malays were included for final analysis. Serum samples were collected from the Chinese and Indian participants, while plasma samples were collected for analysis from the Malay participants.

Creatinine was measured from blood samples by spectrophotometry and estimated GFR was calculated based on the 2021 CKD-EPI equation. CKD was defined based on the Kidney Disease Improving Global Outcomes (KDIGO) 2012 Clinical Practice Guidelines, with the inclusion of participants with CKD stage G3–5, corresponding to an eGFR <60 mL/min/1.73 m². Prevalent cardiovascular disease was defined by the composite of participant-reported myocardial infarction, angina, or cerebrovascular disease at time of study.

Baseline information on age, sex, ethnicity, medical history, and medications were collected using standardized questionnaires. Blood pressure (BP) was measured using a digital automatic, oscillometric BP monitor after each participant was seated for at least 5 minutes and the mean of two measurements was recorded for the individual. Diabetes mellitus was defined as random serum glucose ≥11.1 mmol/L or glycosylated hemoglobin ≥6.5% or self-reported physician-diagnosed diabetes or use of glucose-lowering medication. Non-fasting venous blood was tested for serum lipids, glucose, and creatinine.

Blood samples were collected at baseline with storage at −80°C until time of metabolite analysis, which was between 8 and 15 years after collection. Mean specimen processing time from time of collection to time of storage was between 4 and 6 hours. The effect of storage conditions has been studied in multiple studies, with storage at −80°C and the avoidance of multiple freeze-thaw cycles found to be ideal for metabolomic and proteomics biomarkers [16].

A high-throughput NMR platform (Nightingale Health, Helsinke, Finland) was used to quantify 228 metabolites from baseline serum/plasma samples. Serum/plasma metabolic measures included lipids, 14 lipoprotein subclasses with lipid measures for each subclass, apolipoproteins, fatty acids, ketone bodies, amino acids, and glycolysis-related metabolites. The 14 lipoprotein subclasses included six subclasses of very low-density lipoprotein (VLDL: extremely large, very large, large, medium, small, very small), intermediate-density lipoprotein (IDL), three subclasses of low-density lipoprotein (LDL: large, medium, small), and four subclasses of high-density lipoprotein (HDL: very large, large, medium, small). Lipid concentration within each lipoprotein particle included triglycerides, free and total cholesterol, total lipids, phospholipids, esterified cholesterol. The NMR metabolomics quantification has been described in detail previously [14].

For analyses, metabolic concentrations below the detection limit were imputed by the minimum values detected for the corresponding metabolites. Three metabolites, namely, pyruvate, glycine, and glycerol, were excluded from the analyses as they were not measured in the Malay cohort. All included metabolites had less than 5% missing metabolic concentration values and missing metabolic concentrations were imputed by the mean values for the corresponding metabolites. All metabolite concentrations were log_e(metabolite) transformed to alleviate the skewness of distributions and then standardized to mean 0, variance 1 for each cohort separately to facilitate comparison of associations independent of the concentration ranges. Next, we used logistic regression model adjusting for age, gender, systolic BP, diabetes, total cholesterol, and HDL cholesterol as these were significantly associated with prevalent cardiovascular disease in univariate analysis. In each cohort separately, the odds ratio and 95% confidence interval for each metabolite were estimated using this model. Metabolites with p values <0.05 in at least one cohort and had associations of the same direction in all three ethnic cohorts (shown in online suppl. Fig. S1; for all online suppl. material, see https://doi.org/10.1159/000533741) were then included in a meta-analysis, where results from individual cohorts were combined using generic inverse variance-weighted random-effect model. Statistical significance of the meta-analyzed p values was pre-specified as 0.05/n using Bonferroni correction, where n was the number of serum/plasma metabolites passed to the meta-analysis accordingly. After Bonferroni correction, statistical significance for serum/plasma metabolites was considered at p value <0.003125 for the outcome. All statistical analyses were performed using R language (R 3.5.3, R Foundation for Statistical Computing 2019, Vienna, Austria). This study was reported in accordance to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement (online suppl. Table S1).

This study included 1,122 Asian participants with CKD after excluding participants with missing cardiovascular disease status and metabolic profiling (shown in Fig. 1). Of the participants included, 275 (24.5%) had cardiovascular disease. Among the ethnic groups, Indian participants had the highest prevalence of cardiovascular disease, followed by Malay and Chinese participants (shown in Fig. 2). Baseline characteristics presented in Table 1 showed that participants with or without cardiovascular disease were not statistically different in body mass index, smoking status, or alcohol use. Individuals with prevalent cardiovascular disease were older and more were male, had hypertension, hyperlipidemia, diabetes, and later stages of CKD. The participants with cardiovascular disease also had lower systolic and diastolic BP, total cholesterol, HDL cholesterol and LDL cholesterol, along with increased use of anti-hypertensive, anti-diabetic, and anti-cholesterol medication, but higher glycosylated hemoglobin.

In the analysis of the SEED CKD population, 6 serum/plasma metabolites (tyrosine, isoleucine, leucine, valine, cholesteryl esters in small HDL; phospholipids in small LDL) and 10 serum/plasma metabolite ratios (IDL: cholesterol, cholesteryl esters, and triglycerides to total lipid ratio; large HDL: cholesterol, cholesteryl esters, and phospholipids to total lipid ratio; large LDL: cholesterol, cholesteryl esters, and triglycerides to total lipid ratio; medium HDL: cholesteryl esters to total lipid ratio) were found to be significantly associated with prevalent cardiovascular disease in at least one ethnic cohort, after adjusting for age, gender, systolic BP, diabetes, total cholesterol, and HDL cholesterol (online suppl. Fig. S1). Meta-analysis combining the Chinese, Indian, and Malay cohorts with Bonferroni correction for multiple comparisons found that lower tyrosine, leucine, and valine concentrations and lower cholesteryl esters to total lipid ratio in IDL were associated with cardiovascular disease (shown in Fig. 3).

In this study, Asian participants with cardiovascular disease were more likely to be male, older, and had hypertension, hyperlipidemia, and diabetes, consistent with known cardiovascular risk factors. The lower BP, total cholesterol, and LDL cholesterol in individuals with cardiovascular disease may be due to greater use of anti-hypertensive, anti-diabetic, and anti-hyperlipidemia medications for more intensive cardiovascular risk factor control. Across Chinese, Indian, and Malay ethnicities, three metabolites (tyrosine, leucine, valine) and 1 metabolite ratio (cholesteryl esters to total lipid ratio in IDL) were significantly associated with prevalent cardiovascular disease among Asian participants with CKD.

Branched-chain amino acids (BCAAs), such as leucine and valine, and aromatic amino acids, such as tyrosine, are essential amino acids, contributing to metabolic homeostasis, and have been associated with metabolic disorders including obesity, diabetes, insulin resistance [17, 18]. In the context of cardiovascular disease, BCAAs have been implicated in heart failure, arrhythmias, atherosclerosis, and coronary artery disease [19]. However, the pathogenesis still remains unclear, with studies looking into tissue-specific BCAA metabolism and inter-organ BCAA metabolic interactions.

Our study shows that lower levels of leucine and valine were associated with prevalent cardiovascular disease in the Asian participants, concurring with findings among African American cohorts, but converse to studies in American and European cohorts [20‒22]. Prior studies in American and European population noted that BCAAs were positively associated with adverse cardiovascular outcomes, but in the Diabetes Heart Study that comprised a racially diverse cohort, increased BCAA levels were associated with subclinical coronary atherosclerosis (assessed by coronary artery calcium) in European American participants, but not in African American participants [23]. In another study in the African American population, there was a statistically significant inverse relationship between leucine and incident cardiovascular disease, with a similar trend for other BCAAs (isoleucine, valine) that were not statistically significant [24]. These results suggest that there may be differences in the interactions between BCAAs and cardiovascular disease in different ethnic groups and lend support to previous observations from a population-based study with European, South Asian, and African Caribbean participants that found that inter-ethnic group differences in cardiovascular risk persisted despite adjustment for traditional risk factors, suggesting that other factors may contribute to cardiovascular risk [25].

However, these studies did not specifically examine individuals with CKD since the baseline kidney function was not mentioned, or the study cohorts had normal kidney function [23, 26]. Yet the kidneys play the crucial role of amino acid homeostasis, contributing to between 5 and 9% of BCAA metabolism [19]. In individuals with CKD, abnormalities in BCAA metabolism may develop due to a loss of renal contribution to amino acid metabolism, while the effects of metabolic acidosis on peripheral and hepatosplanchnic nitrogen metabolism can lead to BCAA depletion [27]. Moreover, CKD is a known cardiovascular risk enhancer that confers additional risks beyond that of traditional cardiovascular risk factors via chronic inflammation and endothelial dysfunction [28]. Limited data from a Danish study in individuals with CKD revealed that increased BCAAs are associated with prevalent and also subclinical cardiovascular disease. Our study findings instead show an inverse relationship between BCAAs and prevalent cardiovascular disease in an Asian cohort of CKD patients, suggesting ethnic differences in cardiovascular disease risk and BCAA metabolism. Ethnic differences have been described in metabolic profile of other diseases [29], and it will be important to gain a greater understanding of inter-ethnic variations to guide the delivery of care to various ethnic groups and bridge any disparity in healthcare delivery.

Cholesterol in blood is present as either unesterified (free cholesterol) or esterified (cholesteryl esters), which are both constituents of circulating lipoproteins (chylomicrons, VLDL, LDL, IDL, HDL). In our study, lower ratio of cholesteryl esters to total lipid in IDL associated with prevalent cardiovascular disease, consistent with prior lipidomic analysis done by Gerl et al. [30], which suggests inefficient conversion of cholesterol to cholesteryl esters in patients with cardiovascular disease. Though LDL has been generally accepted as a major CVRF in the progression of atherosclerosis, the traditional means of LDL determination involves the indirect method of calculating LDL from total cholesterol, triglycerides, and HDL via formula, thus including IDL and lipoprotein(a) in its measurement [31]. In a study by Hodis et al. [32] on atherosclerosis in general population, progression of cIMT as a surrogate of atherosclerosis was associated with IDL, but not LDL and VLDL, suggesting atherosclerotic risk attributable to LDL may be the result of lipoproteins in the IDL fraction of indirect LDL measurements. When IDL is assessed separately, it has been associated with presence, severity, and progression of atherosclerosis and cardiovascular disease [33]. Patients with CKD appeared to have a specific lipoprotein pattern termed “uremic dyslipidemia,” characterized by normal LDL, low HDL, and high triglyceride levels with accumulation of IDL [34]. The atherogenicity of IDL has similarly been suggested in studies in hemodialysis patients, which found IDL as the lipoprotein fraction most closely associated with atherosclerosis (with aortic pulse wave velocity as surrogate) [35]. Thus, our study supports the significance of IDL in cardiovascular risk among individuals with CKD, beyond alterations in HDL proteome reported in previous studies that focused on individuals with CKD (online suppl. Table S2) [10, 12].

As this is a cross-sectional study, it is not possible to determine causal relationship between the metabolites or metabolite ratios and cardiovascular disease. The outcome of cardiovascular disease may be underestimated since cardiovascular symptoms such as angina may be under-recognized [36]. Three metabolites, namely, pyruvate, glycine and glycerol, were not evaluated in the Malay cohort and hence excluded from the study. Glycine has been reported to have a protective effect in cardiovascular and metabolic diseases, while glycerol has been described to be associated with subclinical atherosclerosis and cardiovascular disease [37].

In Asia, the management of cardiovascular disease and CKD have been largely guided by guidelines developed in predominantly North American and European populations, but yet the epidemiological features of cardiovascular disease and CKD in Asia differ significantly [4, 38]. This is the first study of utilizing metabolic profiling in identifying metabolites in CKD associated with cardiovascular disease risk in a population-based Asian cohort, adding to currently sparse literature on the metabolite profile of cardiovascular disease in CKD [12].

Our study utilized a validated metabolomics tool to evaluate extensive array of potential metabolites, of which few had missing data, and these were dealt with by imputation. Moreover, the use of a standardized NMR platform will also allow for reproducibility and comparisons with other population-based cohorts. Metabolic profiles have been shown to provide added predictive value over clinical risk factors/variables in common diseases including cardiovascular disease, kidney disease, diabetes, and dementia [39]. With progressive maturation of NMR platforms, these tools may become increasingly cost-effective and accessible, allowing for assessment of a comprehensive array of metabolites for risk stratification and early identification of high-risk individuals.

In our study with participants with CKD across Chinese, Indian, and Malay ethnicities, lower tyrosine, leucine, valine, and cholesteryl esters to total lipid ratios in IDL were associated with prevalent cardiovascular disease. Altered amino acid and lipoprotein metabolism may play a role in increased cardiovascular risk, and the presence of CKD and ethnic differences may affect interactions between metabolites in health and disease. Comprehensive metabolic profiling of population-based cohorts may reveal inter-ethnic variations that allow us to better risk stratify patients and also individualize care with consideration of ethnic disparities.

This study was conducted in accordance with the Declaration of Helsinki, and the SEED programme was approved by the SingHealth Centralized Institutional Review Board (2018/2717, 2018/2921, 2012/487/A, 2015/2279, 2018/2006, 2018/2594, and 2018/2570). Informed consent was obtained from all participants.

Cynthia Ciwei Lim has received honoraria from AstraZeneca and Sebia for speaker fees.

This study was supported by the National Medical Research Council, NMRC/STaR/016/2013, NMRC/CIRG/1371/2013, and NMRC/CIRG/1417/2015.

Jiashen Cai, Cynthia Ciwei Lim, and Charumathi Sabanayagam conceived and designed the study. Jiashen Cai, Crystal Chun Yuen Chong, and Cynthia Ciwei Lim analyzed and interpreted the data. Jiashen Cai drafted the manuscript, with critical input and review by Ching Yu Cheng, Cynthia Ciwei Lim, and Charumathi Sabanayagam. All authors reviewed and approved the final version of the manuscript.

The data are available from the corresponding author on reasonable request.