Association of Circulating Metabolites With Risk of Coronary Heart Disease in a European Population: Results From the Biomarkers for Cardiovascular Risk Assessment in Europe (BiomarCaRE) Consortium | Cardiology | JAMA Cardiology | JAMA Network
[Skip to Navigation]
Sign In
Figure 1.  Metabolite Associations With Future Coronary Heart Disease
Metabolite Associations With Future Coronary Heart Disease

Five phosphatidylcholine metabolites were associated with future coronary heart disease after Bonferroni correction. Hazard ratios (HR) are per SD of log-transformed metabolite level and are adjusted for body mass index, systolic blood pressure, antihypertensive treatment, diabetes, total cholesterol, sex, daily smoker status, study center, and examination age. Error bars indicate 95% CIs. The log-metabolite levels were multiplied by −1 to obtain HRs greater or equal to 1. Thus, HRs are described as HRs per a 1-SD decrease. A logarithmic scale is used on the x-axis. PC indicates phosphatidylcholine.

Figure 2.  Comparison of Hazard Ratios (HRs) for Incident Coronary Heart Disease of Classic Risk Factors and Metabolites
Comparison of Hazard Ratios (HRs) for Incident Coronary Heart Disease of Classic Risk Factors and Metabolites

For each metabolite, a Cox model, adjusted for examination age, sex, systolic blood pressure, total cholesterol, body mass index, diabetes, daily smoker status, antihypertensive medication, and study center, was computed. The HRs for the risk factors were computed from a model that included only the aforementioned adjusting variables. No association could be estimated for age, since it was used as the time scale, nor for sex and daily smoker status, since they were used as strata in the Cox models. All HRs are per 1-SD increase. The log-metabolite levels were multiplied by −1 to obtain HRs greater or equal to 1. Thus, HRs are per a 1-SD decrease. A logarithmic scale is used on the x-axis. PC indicates phosphatidylcholine.

Figure 3.  Survival Curves According to Tertiles of the Metabolite Distributions in the Case-Cohort Set
Survival Curves According to Tertiles of the Metabolite Distributions in the Case-Cohort Set

Survival curves are shown separately for each of the metabolites with significant associations with coronary heart disease. The P values given in the respective survival curves are for the log-rank test. The number of persons at risk during the follow-up period is given per one-third in the table directly below the survival curves. Tertiles were used to categorize the metabolite distributions into thirds, and these values were 3.28μM (first tertile) and 4.54μM (second tertile) for acyl-alkyl-PC C40:6, 4.72μM (first tertile) and 6.61μM (second tertile) for acyl-alkyl-PC C38:6, 33.46μM (first tertile) and 46.37μM (second tertile) for diacyl-PC C38:5, and 46.98μM (first tertile) and 69.87μM (diacyl-PC C38:6) and 15.80μM (first tertile) and 23.58μM (second tertile) for diacyl-PC C40:6. PC indicates phosphatidylcholine.

Table 1.  Baseline Characteristics
Baseline Characteristics
Table 2.  Metabolite Associations With Future Coronary Heart Diseasea
Metabolite Associations With Future Coronary Heart Diseasea
Original Investigation
October 30, 2019

Association of Circulating Metabolites With Risk of Coronary Heart Disease in a European Population: Results From the Biomarkers for Cardiovascular Risk Assessment in Europe (BiomarCaRE) Consortium

Author Affiliations
  • 1Department of General and Interventional Cardiology, University Heart Center Hamburg, Hamburg, Germany
  • 2German Center for Cardiovascular Research, Partner Site Hamburg/Kiel/Lübeck, Hamburg, Germany
  • 3National Institute for Health and Welfare, Helsinki, Finland
  • 4Research Center in Epidemiology and Preventive Medicine, Department of Medicine and Surgery, University of Insubria, Varese, Italy
  • 5Center for Clinical Research and Prevention, the Capital Region of Denmark, Copenhagen, Denmark
  • 6Center for Clinical Research and Prevention, Bispebjerg and Frederiksberg Hospital, Copenhagen, Denmark
  • 7Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
  • 8Department of Public Health, Faculty of Health and Medical Science, University of Copenhagen, Denmark
  • 9Faculty of Medicine, Aalborg University, Aalborg, Denmark
  • 10Institute of Epidemiology, Helmholtz Zentrum München, German Research Center for Environmental Health GmbH, Neuherberg, Germany
  • 11Department of Epidemiology, University Center for Health Sciences at the Klinikum Augsburg (UNIKA-T), Ludwig-Maximilians-Universität München, Augsburg, Germany
  • 12Department of Epidemiology and Prevention, IRCCS Neuromed, Pozzilli, Italy
  • 13Cardiovascular Epidemiology Unit, Institute of Cardiovascular Research, University of Dundee, Dundee, United Kingdom
  • 14The George Institute for Global Health, University of Oxford, Oxford, United Kingdom
  • 15The George Institute for Global Health, University of New South Wales, Sydney, Australia
  • 16Institute of Epidemiology and Medical Biometry, University of Ulm, Ulm, Germany
  • 17Deutsches Herzzentrum München, Technische Universität München, Munich, Germany
  • 18German Center for Cardiovascular Research, Partner Site Munich Heart Alliance, Munich, Germany
JAMA Cardiol. 2019;4(12):1270-1279. doi:10.1001/jamacardio.2019.4130
Key Points

Question  Which circulating metabolites are associated with incident coronary heart disease, and what is their value for assessing the prospective incidence of coronary heart disease?

Findings  In the multinational and population-based Biomarkers for Cardiovascular Risk Assessment in Europe, the case cohort consisted of a weighted, random subcohort of an original cohort of more than 70 000 individuals, and 5 phosphatidylcholines were significantly associated with risk of incident coronary heart disease and showed comparable prognosticative values as individual classic risk factors.

Meaning  These findings underline the value of metabolomics for biomarker discovery and demonstrate the potential for use of phosphatidylcholines in coronary heart disease risk assessment.

Abstract

Importance  Risk stratification for coronary heart disease (CHD) remains challenging because of the complex causative mechanism of the disease. Metabolomic profiling offers the potential to detect new biomarkers and improve CHD risk assessment.

Objective  To evaluate the association between circulating metabolites and incident CHD in a large European cohort.

Design, Setting, and Participants  This population-based study used the Biomarkers for Cardiovascular Risk Assessment in Europe (BiomarCaRE) case-cohort to measure circulating metabolites using a targeted approach in serum samples from 10 741 individuals without prevalent CHD. The cohort consisted of a weighted, random subcohort of the original cohort of more than 70 000 individuals. The case-cohort design was applied to 6 European cohorts: FINRISK97 (Finland), Monitoring of Trends and Determinants in Cardiovascular Diseases/Cooperative Health Research in the Region of Augsburg (MONICA/KORA; Germany), MONICA-Brianza and Moli-sani (Italy), DanMONICA (Denmark), and the Scottish Heart Health Extended Cohort (United Kingdom).

Main Outcomes and Measures  Associations with time to CHD onset were assessed individually by applying weighted and adjusted Cox proportional hazard models. The association of metabolites with CHD onset was examined by C indices.

Results  In 10 741 individuals (4157 women [38.7%]; median [interquartile range] age, 56.5 [49.2-62.2] years), 2166 incident CHD events (20.2%) occurred over a median (interquartile range) follow-up time of 9.2 (4.5-15.0) years. Among the 141 metabolites analyzed, 24 were significantly associated with incident CHD at a nominal P value of .05, including phosphatidylcholines (PCs), lysoPCs, amino acids, and sphingolipids. Five PCs remained significant after correction for multiple testing: acyl-alkyl-PC C40:6 (hazard ratio [HR], 1.13 [95% CI, 1.07-1.18]), diacyl-PC C40:6 (HR, 1.10 [95% CI, 1.04-1.15]), acyl-alkyl-PC C38:6 (HR, 1.11 [95% CI, 1.05-1.16]), diacyl-PC C38:6 (HR, 1.09 [95% CI, 1.04-1.14]) and diacyl-PC C38:5 (HR, 1.10 [95% CI, 1.05-1.16]). Lower levels of these metabolites were associated with increased risk of incident CHD. The strength of the associations competes with those of classic risk factors (C statistics: acyl-alkyl-PC C40:6, 0.756 [95% CI, 0.738-0.774], diacyl-PC C40:6, 0.754 [95% CI, 0.736-0.772], acyl-alkyl-PC C38:6, 0.755 [95% CI, 0.736-0.773], diacyl-PC C38:6, 0.754 [95% CI, 0.736-0.772]), diacyl-PC C38:5, 0.754 [95% CI, 0.736-0.772]). Adding metabolites to a base risk model including classic risk factors high-sensitivity C-reactive protein and high-sensitivity troponin I did not improve discrimination by C statistics.

Conclusions and Relevance  Five PCs were significantly associated with increased risk of incident CHD and showed comparable discrimination with individual classic risk factors. Although these metabolites do not improve CHD risk assessment beyond that of classic risk factors, these findings hold promise for an improved understanding of the pathophysiology of CHD.

Introduction

Coronary heart disease (CHD) constitutes an increasing public health burden worldwide.1 Improvement of cardiovascular risk stratification might advance preventive therapies. Current risk stratification is mainly based on classic risk factors, including the European Society of Cardiology or American Heart Association/American College of Cardiology recommended scores.2,3 However, these scores do not include the numerous environmental and genetic and epigenetic factors that reflect the complex and heterogenic causative mechanisms of CHD.4,5 Consequently, efforts have been undertaken to identify novel molecular markers of risk. Metabolomics, the systematic study of small molecules (<1.5 kDa), is a powerful approach to unravel metabolic disturbances linked to pathological conditions and may improve disease prognostication. The field of metabolomics and its applications to cardiovascular disease has evolved rapidly.6 Prior studies provided the first steps toward identification of metabolic disturbances in cardiovascular disease, suggesting a potential added value of metabolites as biomarkers for CHD.7-14

The aim of this study was to evaluate the association between circulating metabolites and CHD and assess the capability of metabolomics for the prognostication of incident CHD events in the general population. Using the harmonized database and biobank of the European Biomarkers for Cardiovascular Risk Assessment in Europe (BiomarCaRE) project and a large-scale approach, we evaluated the prognosticative values of 141 metabolites in more than 12 000 European individuals from the general population of Europe.

Methods
Study Cohorts and Definition of End Points

The study was conducted within the multinational BiomarCaRE project.15 Baseline variables, phenotypes, CHD events, and follow-up information from large population-based European cohorts have been harmonized in the Monica Risk, Genetics, Archiving and Monograph (MORGAM) database.16 All participating cohorts complied with the Declaration of Helsinki, and all local ethics review boards have approved the study protocol. The study protocol was approved by the local ethics review boards of all study centers, and all participants provided written informed consent.

In the current study, a case-cohort design was applied, including 6 European cohorts: FINRISK97 (Finland), Moli-sani (Italy), Monitoring of Trends and Determinants in Cardiovascular Diseases/Cooperative Health Research in the Region of Augsburg (MONICA/KORA; Germany), MONICA-Brianza (Italy), Danish-Multinational Monitoring of Trends and Determinants in Cardiovascular Disease (DanMONICA; Denmark), and Scottish Heart Health Extended Cohort (United Kingdom).15,17 Briefly, the present case-cohort set consisted of a weighted random subsample (subcohort) of those in the original cohort of more than 70 000 individuals who had no history of myocardial infarction, stroke, heart failure, and atrial fibrillation and who were selected independently of the definition of cases, plus all individuals of the cohort who developed CHD during the follow-up period. In SHHEC, the case-cohort selection was done from a random half of the full cohort.

Completeness of follow-up for the case cohort was 96.4% (12 462 of 12 928 individuals). Baseline serum samples from 12 928 individuals with 2166 incident CHD events were analyzed for the present study (eFigure 1 in the Supplement). Details of the individual cohorts are provided in eAppendix 1 in the Supplement. Baseline characteristics according to the individual cohorts are provided in eTable 1 and eTable 2 in the Supplement.

Incident CHD was defined as the first fatal or nonfatal coronary event (including definite or possible acute myocardial infarction, possible coronary death, or unstable angina pectoris) during the follow-up period. Details on the event classification in each cohort are provided in the description of MORGAM cohorts.18 Measurements of biomarkers high-sensitive C-reactive protein (hsCRP) and high-sensitive troponin I (hsTnI) have been described previously.19

Measurement of Metabolites and Data Processing

The AbsoluteIDQ p180 Kit (BIOCRATES Life Sciences AG) was used for sample preparation according to the manufacturer’s recommendation. A total of 184 metabolites from 5 analytic groups (amino acids, biogenic amines, acylcarnitines, glycerophospholipids, sphingolipids, and sugars) were quantified simultaneously using liquid chromatography and flow-injection analysis mass spectrometry. Measurements were performed on a TSQ Vantage (Thermo Fisher Scientific) connected to an UltiMate 3000 Pump/Autosampler (Thermo Fisher Scientific). For quantification, isotope-labeled internal standards and 7 calibrators of different concentration levels were used. In addition, 3 different levels of quality controls were included, from which quality control 2 was used for data normalization, containing metabolite levels most comparable with physiological levels. A detailed description of the sample preparation is provided in eAppendix 2 in the Supplement.

Normalization of metabolite levels was performed across the entire BiomarCaRE case-cohort set to ensure harmonization. Normalizing each metabolite level was corrected by a factor determined from the ratio of the median of all quality control level 2 values across all plates and the median of the reference samples from the corresponding plates. Within each cohort, no strong batch effects were observed. After data normalization, metabolites were excluded from all subsequent analyses if more than 30% of measurements were missing or more than 50% of measurements were at less than the level of detection (LOD). Additionally, samples were excluded from the analyses if more than 50% of metabolites were missing. A global LOD value was defined for each metabolite of the flow-injection analysis part that corresponds to the highest LOD value across all plates being measured, while the liquid chromatography LOD values were taken from the Biocrates p180 Kit manual. As a consequence, all subsequent data analyses were performed on 141 metabolites and 10 741 samples in total (eFigure 1 in the Supplement). An equal distribution of risk factors and confounding variables were observed between the CHD case-cohort set before and after normalization and quality control (12 928 individuals and 10 741 individuals, respectively) (eTable 3 in the Supplement).

Statistical Analyses

To impute missing measurements, multiple imputation was performed using multivariate imputations by chained equations.20 A total of 20 imputed data sets were created. For the metabolites, a log-normal model was used (for missing values known to be less than the LOD, the imputed values were drawn from truncated log-normal distribution). For all other variables, predictive mean matching was used. The event indicator (CHD) and the Nelson-Aalen estimate of the cumulative survival function of the time to CHD were included in the imputation model. The imputation was done separately for each center and sex.

To calculate correlations of log-transformed metabolites with each other and with cardiovascular risk factors (sex, age at examination, systolic blood pressure, total cholesterol, high-density lipoprotein [HDL] cholesterol, body mass index [BMI; calculated as weight in kilograms divided by height in meters squared], diabetes, daily smoker status, and antihypertensive treatment), Pearson correlation coefficients were calculated. For this, only the subcohort was used, and the observations were weighted by the inverse of the inclusion probability.

Associations with time to CHD were assessed for each of the 141 log-transformed metabolites individually by applying weighted Cox proportional hazard models adjusted for BMI, systolic blood pressure, antihypertensive treatment, diabetes, total cholesterol, sex, daily smoker status, study center, and age at examination. Adjustment for age, sex, daily smoker status, and study center was performed without modeling their effects; age was used as the time scale, and the other 3 factors were used as stratification variables in the Cox models. For all Cox regression analyses, the case-cohort design was taken into account using methods described by Kulathinal et al.17 More specifically, in Cox regressions, individuals were weighted with inverse subcohort-sampling probability weights, and robust standard errors were used. Nominal significance was met at a P value of .05 or lower. After Bonferroni correction for multiple testing, significance was met if P values were less than 4.43 × 10−4 (.05 divided into 113). The number of tests used in the Bonferroni correction (n = 113) was estimated following the methods described in Gao et al.21 This method estimates the effective number of tests as the number of principal components of the log-metabolites that explain 99.5% of the variation. To further assess the effect of renal function and lipid metabolism, any significant findings after Bonferroni correction were additionally adjusted for estimated glomerular filtration rate and non-HDL cholesterol instead of total cholesterol. Hazard ratios (HRs) for the log-metabolites were reported per each 1-SD increase, in which the SD was calculated in the subcohort. For a direct comparison of HRs of metabolites and risk factors, log-metabolite levels were multiplied by −1. Sex-specific HRs for the log-metabolites were estimated by adding sex–log-metabolite interactions to the Cox models. Similarly, log-metabolite HRs for those using vs not using lipid-lowering medication were computed by including the interactions of lipid-lowering medication status and log-metabolites in the Cox models. This analysis was performed only on the Moli-sani and FINRISK97 cohorts because of the availability of information on lipid-lowering medications. Cox regressions in each study center were also performed, and the I2 statistic and prediction intervals for the HRs were calculated.22,23

Survival curves were estimated by the Kaplan-Meier method. Methods previously described17 were used to account for the case-cohort design. Age was used as the time scale. For each of the significant metabolites after Bonferroni correction, HRs according to tertiles of the metabolite levels were calculated using Cox models. The tertiles were defined using the subcohort, and analyses were weighted by the inverse of the inclusion probabilities. Two models of adjustment were used: (1) basic adjustment for age at examination, sex, and study center and (2) full adjustment using BMI, systolic blood pressure, antihypertensive treatment, diabetes, total cholesterol, sex, daily smoker status, study center, and examination age. As before, age was used as the time scale and sex, daily smoker status, and study center were used as the strata.

The prognosticative value of the metabolites was examined in terms of discrimination, as assessed by C indices. Probabilities of CHD within the next 10 years were estimated for each combination of sex, daily smoking status, and study center by fitting a Weibull model with age as the time scale and the linear predictor in the corresponding Cox model. In the estimation of these probabilities, 10-fold cross-validation was used. We computed the 95% CIs and P values for differences of the C indices using the methods described in Antolini et al.24 All statistical analyses were performed using R version 3.4.4 (R Development Core team).

Results

The analysis included 10 741 individuals (FINRISK, n = 1771; Brianza, n = 511; Moli-sani, n = 1364; MONICA/KORA, n = 1085; DanMONICA, n = 3432; and SHHEC, n = 2578) without CHD at baseline, including 4157 women (38.7%) and 6584 men, with a median (interquartile range) age of 56.5 (49.2-62.2) years. During a median (interquartile range) follow-up time of 9.2 (4.5-15.0) years, 2166 individuals (20.2%) experienced an incident CHD event, including 677 fatal CHD events (6.3% of participants) and 1489 nonfatal CHD events (13.9% of participants). The prevalence of hypertension, diabetes, and daily smoking at baseline was 48.1% (5169 participants), 5.2% (562 participants), and 31.8% (3416 participants), respectively. The baseline characteristics and CHD event numbers for the overall cohort, as well as individuals with and without incident CHD events, are summarized in Table 1.

The baseline characteristics, weighted to represent the full cohort according to the case-cohort design,17 as well as according to sex, are provided in eTable 4 and eTable 5 in the Supplement. Levels of 141 serum metabolites were analyzed, and mean metabolite levels are presented in eTable 6 in the Supplement.

In this analysis, 24 individual metabolites were nominally associated with CHD at a P value of .05 or less. These 24 metabolites belonged to 3 different metabolite classes (phosphatidylcholines [PC]), amino acids, and sphingolipids; Table 2). Pearson correlations of individual significant metabolites are shown in eFigure 2 in the Supplement. When controlling for multiple testing, 5 PCs remained significant (HRs: acyl-alkyl-PC C40:6, 1.13 [95% CI, 1.07-1.18]; P = 3.01 × 10−6; diacyl-PC C40:6, 1.10 [95% CI, 1.04-1.15]; P = 4.28 × 10−4; acyl-alkyl-PC C38:6, 1.11 [95% CI, 1.05-1.16]; P = 4.61 × 10−5; diacyl-PC C38:6, 1.09 [95% CI, 1.04-1.14]; P = 1.8 × 10−4; diacyl-PC C38:5, 1.10 [95% CI, 1.05-1.16]; P = 1.07 × 10−4; Figure 1; Table 2). All significant metabolites were negatively associated with CHD. The associations were consistent between men and women, with slightly higher HRs in women; all were nonsignificantly different, with the exception of acyl-alkyl-PC C40:6 (men: HR, 1.09 [95% CI, 1.02-1.15], P = .009; women: HR, 1.21 [95% CI, 1.11-1.31]; P = 1.3 × 10−5; interaction P = .04). Associations of the 5 significant PCs with CHD per study center are illustrated in eFigure 3 in the Supplement, including for acyl-alkyl-PC C40:6 (FINRISK: HR, 1.31 [95% CI, 1.15-1.50]; P = 4.77 × 10−5; SHHEC: HR, 1.29 [95% CI, 1.17-1.43]; P = 3.11 × 10−7), acyl-alkyl-PC C38:6 (FINRISK: HR, 1.26 [95% CI, 1.11-1.44]; P = 3.55 × 10−4; SHHEC: HR, 1.22 [95% CI, 1.13-1.32]; P = 1.42 × 10−7), diacyl-5 PC C38:5 (FINRISK: HR, 1.21 [95% CI, 1.06-1.37]; P = .004; SHHEC: HR, 1.23 [95% CI, 1.14-1.34]; P = 5.92×10−7), diacyl-PC C38:6 (FINRISK: HR, 1.30 [95% CI, 1.15-1.47]; P = 3.62×10−5; SHHEC: HR, 1.11 [95% CI, 1.04-1.19]; P = .003), and diacyl-PC C40:6 (FINRISK: HR, 1.25 [95% CI, 1.10-1.43]; P = 8.67×10−4; SHHEC: HR, 1.27 [95% CI, 1.15-1.40]; P = 1.32×10−6). Importantly, the associations of these 5 significant metabolites were not attenuated after additional adjustments for estimated glomerular filtration rate and non-HDL cholesterol instead of total cholesterol (eTable 7 in the Supplement).

Because medications, such as lipid-lowering drugs, might affect the levels of metabolites, an interaction analysis was performed in the subset of the case-cohort set in which data on lipid-lowering medications were available. As shown in eTable 8 in the Supplement, lipid-lowering medication use had no significant association with the prognosticative values of the PCs.

Comparing the association between the 5 significant CHD-associated PCs and classic cardiovascular risk factors (BMI, systolic blood pressure, diabetes, and total cholesterol), showed only a slightly weaker association for metabolites than some of the classic risk factors. For instance, for acyl-alkyl-PCs C40:6 (HR, 1.13 [95% CI, 1.07-1.18]; P = 3.01 × 10−6) and C38:6 (HR, 1.11 [95% CI, 1.05-1.16]; P = 4.61 × 10−5), the strength of association was similar to those of BMI (HR, 1.14 [95% CI, 1.07-1.20]; P = 3.10 × 10−5) and systolic blood pressure (HR, 1.17 [95% CI, 1.10-1.23]; P = 3.28 × 10−8) (Figure 2).

To assess the ability of the 5 PC metabolites to estimate CHD risk, Kaplan-Meier survival analyses according to tertiles of metabolite levels using age as the time scale were performed. Figure 3 shows unadjusted Kaplan-Meier curves of CHD event-free time for each of the 5 PC metabolites, while eFigure 4 in the Supplement shows HRs across metabolite tertiles. A decrease in CHD risk was observed across increasing tertiles. Individuals in the highest tertile of metabolite distributions had a decreased risk of developing future CHD compared with those in the lowest tertile (acyl-alkyl-PC C40:6: HR, 1.21 [95% CI, 1.06-1.37]; P = .004; acyl-alkyl-PC C38:6: HR, 1.29 [95% CI, 1.12-1.47], P = 2.29 × 10−4; diacyl-PC C38:5: HR, 1.14 [95% CI, 1.00-1.29], P = .05; diacyl-PC C38:6: HR, 1.21 [95% CI, 1.06-1.38], P = .004; diacyl-PC C40:6: HR, 1.09 [95% CI, 0.96-1.24], P = .17). These findings were similar after full adjustment.

To assess the discriminatory potential of the 5 significant metabolites, C indices were calculated for the 10-year prognostication of CHD. Acyl-alkyl-PC C40:6 performed best in the discrimination of CHD, with a C index of 0.756 (95% CI, 0.738-0.774), followed by acyl-alkyl-PC C38:6 (0.755 [95% CI, 0.736-0.773]). Diacyl-PC C38:5, C38:6, and C40:6 all performed with a C index of 0.754 (95% CI, 0.736-0.772). The C indices of the metabolites were comparable with those of classic cardiovascular risk factors and the 2 clinically established biomarkers, hsCRP, and hsTnI. However, the addition of the metabolites, either individually or as a combination to a base risk model (including sex, examination age, systolic blood pressure, total cholesterol, HDL cholesterol, BMI, diabetes, daily smoker status, antihypertensive treatment, hsCRP, hsTnI, and study center) did not improve discrimination by C statistics (C index base model, 0.828 [95% CI, 0.809-0.846]; C index plus metabolites, 0.828 [95% CI, 0.810-0.846]; eTable 9 in the Supplement).

Discussion

Using a comprehensive targeted metabolomics approach, we analyzed the association of 141 serum metabolites with risk of incident CHD in a population based case-cohort sample of more than 70 000 European individuals. We report 3 main findings: first, we identified 24 metabolites that were significantly and independently associated with incident CHD. These metabolites are mainly acyl-alkyl-PCs and diacyl-PCs. Five of these 24 metabolites remained significantly associated after correcting for multiple testing. Second, these 5 significant metabolites showed an inverse association with CHD risk, because lower levels were accompanied with higher risk. Third, the discriminatory potential of these metabolites for CHD events was comparable with those of classic risk factors and established biomarkers hsCRP and hsTnI. However, the metabolites did not relevantly improve the assessment of prospective incident CHD compared with existing conventional risk factors.

Data from prospective studies are sparse, and previous smaller studies report conflicting results. Most recent data suggest that beyond the established association between low-density lipoprotein (LDL) cholesterol, and future cardiovascular events PCs might also contribute to cardiovascular disease.25,26 In cross-sectional analyses, PCs were identified in plasma lipoproteins and atherosclerotic plaques, and alterations in PC levels were described in patients with stable coronary artery disease, myocardial infarction, and atherosclerosis as compared with healthy participants.25,26 Other prospective analyses in primary and secondary cardiovascular prevention identified PCs as markers for cardiovascular mortality in patients with established coronary artery disease,27 in the population-based Bruneck study12 and as markers for risk of myocardial infarction or CHD in healthy adults in prospective cohorts.10,28 The group of PCs presents a heterogeneous group, and different subtypes have different associations and thereby confer different risks. While the PC subtypes identified by this study associated inversely with incident disease, Paytner et al28 identified different PC subtypes (C34:2 and C36:4) that were positively associated with increased risk of CHD.

The growing discipline of metabolomics and lipidomics provides new insights into the pathogenesis of cardiovascular disease. In this study, increasing levels of PCs were protective against incident CHD. This is in line with a previous interventional study29: using lipidomics, the Rosuvastatin and Atorvastatin in Different Dosages and Reverse Cholesterol Transport (RADAR) study analyzed in depth the lipid changes induced by rosuvastatin at different dosages. Liquid chromatography–mass spectrometry was applied to obtain the metabolite profiles of plasma samples, for which a randomly assigned treatment with rosuvastatin in increasing dosages was administered during an 18-week period. In result, rosuvastatin significantly increased the plasma levels of PCs after 6 and 18 weeks. In accordance, Paapstel et al25 observed inverse associations between PC and lysoPC profiles and inflammation as well as arterial function in patients with atherosclerosis using the same metabolomics approach as in this study. The authors investigated serum PC and lysoPC species in association with arterial stiffness, hemodynamics, and endothelial dysfunction in 32 patients with peripheral arterial disease, 52 patients with coronary artery disease, and in 40 healthy control participants. Contrarily, Floegel et al10 identified PC classes that were positively associated with risk of myocardial infarction in healthy adults. Furthermore, Tang et al30 demonstrated that higher levels of trimethylamine N-oxide, which is a product of PCs, are positively associated with cardiovascular event rates. Interestingly, trimethylamine N-oxide levels depend on gut flora metabolism, suggesting a possible therapy target.31 Notably, newly identified associations of PC classes with CHD need to be considered separately, because different pathophysiological pathways could underlie disease development.

The molecular bases of these findings still need to be elucidated. The 3 major phospholipid classes (sphingomyelins, PCs, and lysoPCs) are important structural components of plasma lipoproteins and cell membranes and are involved in regulation of cell function, membrane protein trafficking, and inflammation.32 In human plasma, PCs and lysoPCs make up about 60% to 70% and 10% to 20%, respectively, of circulating phospholipids. Production of lysoPCs is the result of partial hydrolysis of PCs, lecithin-cholesterol acyltransferase activity, or hepatic secretion.25 These lipid species are important components of oxidized LDL. Oxidized LDL has been hypothesized to induce foam-cell formation (an early yet critical step in the development of atherosclerosis); it downregulates endothelial nitric oxide synthase, increases the formation of metalloproteinases, and induces apoptosis in human coronary endothelial cells.33 One may hypothesize that the protective effects of increased PC levels are attributable to modification of oxidized LDL. However, this study is not capable of deciphering underlying molecular mechanisms, and potential effects on oxidized LDL and other molecules in this setting remain speculative.

Strengths

In these analyses, we used the harmonized database of the BiomarCaRE project, providing, to our knowledge, the best possible end point validation across Europe and including a large sample size and many incident CHD events. In addition to the use of a harmonized database, all metabolite measurements were performed in 1 central laboratory using a well-validated assay platform and stringent quality control was applied. Furthermore, we included study cohorts from different European regions. Most importantly, the identification of new and solid factors associated with CHD, which have not been reported yet in the context of CHD, might boost the way to new and unknown pathopysiologic pathways in cardiovascular disease.

The strength of the association of the 5 PCs with the risk of future CHD was comparable with that of classic risk factors, in particular BMI and systolic blood pressure. However, contrary to previous reports,12,34 the metabolites identified in this study did not appreciably improve the association of incident CHD compared with existing conventional risk factors. In contrast to this study, in which only PCs were identified and assessed in detail, the previous studies used combinations of several metabolite classes in their discrimination analyses. An explanation for the differences in the ability to improve risk assessment might be that the various metabolites classes contribute to different pathophysiological pathways based on their composition,25,35 thus leading to an improved risk assessment when combined.

Limitations

A limitation of our study is the use of a targeted metabolite approach and thereby the measurement of formally selected metabolites. The currently known and quantifiable metabolome consists of more than 100 000 metabolites.36,37 Thus, we cannot exclude the possibility that by using a nontargeted approach with a broad coverage of the entire metabolome, additional metabolite classes might have been identified. Furthermore, we were unable to control for dietary and lifestyle parameters in general, which have obviously changed during the last 3 decades of study time, or for medication, which may have affected our results of associated metabolites.3,38 Therefore, future studies with information on dietary intake and fasting status are needed to validate these data. Additionally, basic research is crucial to provide answers as to whether the identified metabolite markers are central to the pathogenesis of CHD.

Conclusions

In conclusion, by quantifying a large set of metabolites in 12 928 European individuals, we identified 5 PCs that were associated with increased risk of incident CHD. The strength of the association of metabolites was similar to that of individual classic cardiovascular risk factors, and C indices were comparable with those obtained from classic risk factors and established biomarkers hsCRP and hsTnI. However, these metabolites did not appreciably improve the prognostication of incident CHD compared with existing conventional risk factors. The present study demonstrates the value of metabolomics for biomarker discovery and holds promise for improved understanding of pathophysiology of CHD. Further studies to investigate the link between newly detected PCs and CHD development are needed.

Back to top
Article Information

Accepted for Publication: August 16, 2019.

Corresponding Author: Tanja Zeller, PhD, University Heart Center Hamburg, Clinic for General and Interventional Cardiology, Martinistraße 52, 20246 Hamburg, Germany (t.zeller@uke.de).

Published Online: October 30, 2019. doi:10.1001/jamacardio.2019.4130

Author Contributions: Drs Cavus and Karakas had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Cavus and Karakas contributed equally.

Concept and design: Cavus, Karakas, Jørgensen, Boernigen, Schnabel, Koenig, Kuulasmaa, Salomaa, Blankenberg, Zeller.

Acquisition, analysis, or interpretation of data: Cavus, Karakas, Ojeda, Kontto, Veronesi, Ferrario, Linneberg, Jørgensen, Meisinger, Thorand, Iacoviello, Boernigen, Woodward, Costanzo, Tunstall-Pedoe, Kuulasmaa, Salomaa, Zeller.

Drafting of the manuscript: Cavus, Karakas, Boernigen, Tunstall-Pedoe, Zeller.

Critical revision of the manuscript for important intellectual content: Cavus, Karakas, Ojeda, Kontto, Veronesi, Ferrario, Linneberg, Jørgensen, Meisinger, Thorand, Iacoviello, Woodward, Schnabel, Costanzo, Tunstall-Pedoe, Koenig, Kuulasmaa, Salomaa, Blankenberg.

Statistical analysis: Cavus, Ojeda, Kontto, Boernigen, Woodward.

Obtained funding: Thorand, Tunstall-Pedoe, Koenig, Zeller.

Administrative, technical, or material support: Cavus, Karakas, Linneberg, Costanzo, Kuulasmaa, Salomaa, Zeller.

Supervision: Cavus, Karakas, Veronesi, Ferrario, Jørgensen, Koenig, Salomaa, Blankenberg, Zeller.

Conflict of Interest Disclosures: Dr Karakas reports research grants from Vifor Pharma, Adrenomed, and Abbott Diagnostics and lecture fees and fees for serving on advisory boards from Vifor Pharma, Adrenomed, AstraZeneca, Bayer Pharma AG, Sanofi, and Amgen. Dr Blankenberg has received research funding from Abbott, Abbott Diagnostics, Bayer, Boehringer Ingelheim, Siemens, and ThermoFisher; honoraria for lectures from Abbott, Abbott Diagnostics, AstraZeneca, Bayer, Boehringer Ingelheim, Medtronic, Pfizer, Roche, Siemens Diagnostics, Siemens, and ThermoFisher; and honoraria for advisory board memberships and consulting for Boehringer Ingelheim, Bayer, Novartis, Roche, and ThermoFisher. Dr Koenig reports receiving fees for serving on advisory boards from Novartis, the Medicines Company, Amgen, AstraZeneca, Pfizer, Merck Sharp & Dohme, DalCor, and Kowa and lecture fees from Amgen, AstraZeneca, Novartis, Sanofi, Berlin-Chemie, and Merck Sharp & Dohme, as well as research grants from Roche, Beckmann, Singulex, and Abbott Diagnostics. Dr Zeller reports research grants from Vifor Pharma and Abbott Diagnostics and grants from European Union Framework FP7 during the conduct of the study. Dr Salomaa has participated in a conference trip sponsored by Novo Nordisk and received an honorarium from the same source for participating in an advisory board meeting; he also has ongoing research collaboration with Bayer Ltd and a grant from the Finnish Foundation for Cardiovascular Research paid to his institution. Dr Thorand reported grants from the European Union during the conduct of the study. Dr Woodward reported personal fees from Amgen and Kirin outside the submitted work. Dr Kuulasmaa reported grants from European Union and grants from University Clinic Hamburg–Eppendorf during the conduct of the study. No other disclosures were reported.

Funding/Support: The BiomarCaRE Project was funded by the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement HEALTH-F2-2011-278913. Within the BiomarCaRE Project, Thermo Fisher provided the TSQ Vantage mass-spectrometer for measurements. The activities of the Monica Risk, Genetics, Archiving and Monograph (MORGAM) data center have also been sustained by recent funding from the European Union FP7 project Consortium on Health and Ageing: Network of Cohorts in Europe and the United States (CHANCES [grant HEALTH-F3-2010-242244]). Funding of each study is provided in eAppendix 1 in the Supplement. Dr Zeller is supported by the German Center for Cardiovascular Research (grant 81Z1710101).

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Group Information: BiomarCaRE consortium: Ersin Cavus, MD, Department of General and Interventional Cardiology, University Heart Center Hamburg, Hamburg, Germany; Mahir Karakas, MD, Department of General and Interventional Cardiology, University Heart Center Hamburg, Hamburg, Germany; German Center for Cardiovascular Research, Partner Site Hamburg/Kiel/Lübeck, Hamburg, Germany; Francisco M. Ojeda, PhD, Department of General and Interventional Cardiology, University Heart Center Hamburg, Hamburg, Germany; Jukka Kontto, PhD, National Institute for Health and Welfare, Helsinki, Finland; Giovanni Veronesi, PhD, Research Center in Epidemiology and Preventive Medicine, Department of Medicine and Surgery, University of Insubria, Varese, Italy; Marco Mario Ferrario, PhD, Research Center in Epidemiology and Preventive Medicine, Department of Medicine and Surgery, University of Insubria, Varese, Italy; Allan Linneberg, MD, Center for Clinical Research and Prevention, the Capital Region of Denmark, Copenhagen, Denmark; Center for Clinical Research and Prevention, Bispebjerg and Frederiksberg Hospital, Copenhagen, Denmark; Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark; Torben Jørgensen, MD, Center for Clinical Research and Prevention, the Capital Region of Denmark, Copenhagen, Denmark; Department of Public Health, Faculty of Health and Medical Science, University of Copenhagen, Copenhagen, Denmark; Faculty of Medicine, Aalborg University, Aalborg, Denmark; Christa Meisinger, MD, Institute of Epidemiology, Helmholtz Zentrum München, German Research Center for Environmental Health GmbH, Neuherberg, Germany; University Center for Health Sciences at the Klinikum Augsburg (UNIKA-T), Ludwig-Maximilians-Universität München, Augsburg, Germany; Barbara Thorand, PhD, Institute of Epidemiology, Helmholtz Zentrum München, German Research Center for Environmental Health GmbH, Neuherberg, Germany; Licia Iacoviello, MD, Research Center in Epidemiology and Preventive Medicine, Department of Medicine and Surgery, University of Insubria, Varese, Italy; Department of Epidemiology and Prevention, IRCCS Neuromed, Pozzilli, Italy; Daniela Börnigen, PhD, Department of General and Interventional Cardiology, University Heart Center Hamburg, Hamburg, Germany; Mark Woodward, PhD, Cardiovascular Epidemiology Unit, Institute of Cardiovascular Research, University of Dundee, Dundee, United Kingdom; the George Institute for Global Health, University of Oxford, Oxford, United Kingdom; the George Institute for Global Health, University of New South Wales, Sydney, Australia; Renate Schnabel, MD, Department of General and Interventional Cardiology, University Heart Center Hamburg, Hamburg, Germany; German Center for Cardiovascular Research, Partner Site Hamburg/Kiel/Lübeck, Hamburg, Germany; Simona Costanzo, PhD, Department of Epidemiology and Prevention, IRCCS Neuromed, Pozzilli, Italy; Hugh Tunstall-Pedoe, MD, Cardiovascular Epidemiology Unit, Institute of Cardiovascular Research, University of Dundee, Dundee, United Kingdom; Wolfgang Koenig, MD, Institute of Epidemiology and Medical Biometry, University of Ulm, Ulm, Germany; Deutsches Herzzentrum München, Technische Universität München, Munich, Germany; German Center for Cardiovascular Research, Partner Site Munich Heart Alliance, Munich, Germany; Kari Kuulasmaa, PhD, National Institute for Health and Welfare, Helsinki, Finland; Veikko Salomaa, MD, National Institute for Health and Welfare, Helsinki, Finland; Stefan Blankenberg, MD, Department of General and Interventional Cardiology, University Heart Center Hamburg, Hamburg, Germany; German Center for Cardiovascular Research, Partner Site Hamburg/Kiel/Lübeck, Hamburg, Germany; Tanja Zeller, PhD, Department of General and Interventional Cardiology, University Heart Center Hamburg, Hamburg, Germany; German Center for Cardiovascular Research, Partner Site Hamburg/Kiel/Lübeck, Hamburg, Germany.

Additional Contributions: We thank Tarja Palosaari, PhD, National Institute for Health and Welfare, Teemu Niiranen, MD, PhD, National Institute for Health and Welfare and Department of Internal Medicine, Turku University Hospital and University of Turku, and Laura Paalanen, PhD, National Institute for Health and Welfare, for their help with data harmonization; Ari Hauijärvi, National Institute for Health and Welfare, and Satya Bhowmik, PhD, Department of General and Interventional Cardiology, University Heart Center–Hamburg, for their substantial contribution in data management; Caroline Röthemeier and Tim Hartmann, Department of General and Interventional Cardiology, University Heart Center–Hamburg, and Judith Wahrheit and Manuel Kratzke, PhD, BIOCRATES Life Sciences AG, for providing their expertise in the metabolite measurements. None of these individuals were compensated by a study funder/sponsor for their contributions.

References
1.
Townsend  N, Wilson  L, Bhatnagar  P, Wickramasinghe  K, Rayner  M, Nichols  M.  Cardiovascular disease in Europe: epidemiological update 2016.  Eur Heart J. 2016;37(42):3232-3245. doi:10.1093/eurheartj/ehw334PubMedGoogle ScholarCrossref
2.
Piepoli  MF, Hoes  AW, Agewall  S,  et al; ESC Scientific Document Group.  2016 European Guidelines on cardiovascular disease prevention in clinical practice: the Sixth Joint Task Force of the European Society of Cardiology and Other Societies on Cardiovascular Disease Prevention in Clinical Practice (constituted by representatives of 10 societies and by invited experts), developed with the special contribution of the European Association for Cardiovascular Prevention & Rehabilitation (EACPR).  Eur Heart J. 2016;37(29):2315-2381. doi:10.1093/eurheartj/ehw106PubMedGoogle ScholarCrossref
3.
Goff  DC  Jr, Lloyd-Jones  DM, Bennett  G,  et al; American College of Cardiology/American Heart Association Task Force on Practice Guidelines.  2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines.  Circulation. 2014;129(25)(suppl 2):S49-S73. doi:10.1161/01.cir.0000437741.48606.98PubMedGoogle ScholarCrossref
4.
Khera  AV, Emdin  CA, Drake  I,  et al.  Genetic risk, adherence to a healthy lifestyle, and coronary disease.  N Engl J Med. 2016;375(24):2349-2358. doi:10.1056/NEJMoa1605086PubMedGoogle ScholarCrossref
5.
Khera  AV, Kathiresan  S.  Genetics of coronary artery disease: discovery, biology and clinical translation.  Nat Rev Genet. 2017;18(6):331-344. doi:10.1038/nrg.2016.160PubMedGoogle ScholarCrossref
6.
McGarrah  RW, Crown  SB, Zhang  GF, Shah  SH, Newgard  CB.  Cardiovascular metabolomics.  Circ Res. 2018;122(9):1238-1258. doi:10.1161/CIRCRESAHA.117.311002PubMedGoogle ScholarCrossref
7.
Brindle  JT, Antti  H, Holmes  E,  et al.  Rapid and noninvasive diagnosis of the presence and severity of coronary heart disease using 1H-NMR-based metabonomics.  Nat Med. 2002;8(12):1439-1444. doi:10.1038/nm1202-802PubMedGoogle ScholarCrossref
8.
Wang  TJ, Larson  MG, Vasan  RS,  et al.  Metabolite profiles and the risk of developing diabetes.  Nat Med. 2011;17(4):448-453. doi:10.1038/nm.2307PubMedGoogle ScholarCrossref
9.
Rhee  EP, Gerszten  RE.  Metabolomics and cardiovascular biomarker discovery.  Clin Chem. 2012;58(1):139-147. doi:10.1373/clinchem.2011.169573PubMedGoogle ScholarCrossref
10.
Floegel  A, Kühn  T, Sookthai  D,  et al.  Serum metabolites and risk of myocardial infarction and ischemic stroke: a targeted metabolomic approach in two German prospective cohorts.  Eur J Epidemiol. 2018;33(1):55-66. doi:10.1007/s10654-017-0333-0PubMedGoogle ScholarCrossref
11.
Würtz  P, Havulinna  AS, Soininen  P,  et al.  Metabolite profiling and cardiovascular event risk: a prospective study of 3 population-based cohorts.  Circulation. 2015;131(9):774-785. doi:10.1161/CIRCULATIONAHA.114.013116PubMedGoogle ScholarCrossref
12.
Stegemann  C, Pechlaner  R, Willeit  P,  et al.  Lipidomics profiling and risk of cardiovascular disease in the prospective population-based Bruneck study.  Circulation. 2014;129(18):1821-1831. doi:10.1161/CIRCULATIONAHA.113.002500PubMedGoogle ScholarCrossref
13.
Ward-Caviness  CK, Xu  T, Aspelund  T,  et al.  Improvement of myocardial infarction risk prediction via inflammation-associated metabolite biomarkers.  Heart. 2017;103(16):1278-1285. doi:10.1136/heartjnl-2016-310789PubMedGoogle ScholarCrossref
14.
Li  Y, Zhang  D, He  Y,  et al.  Investigation of novel metabolites potentially involved in the pathogenesis of coronary heart disease using a UHPLC-QTOF/MS-based metabolomics approach.  Sci Rep. 2017;7(1):15357. doi:10.1038/s41598-017-15737-3PubMedGoogle ScholarCrossref
15.
Zeller  T, Hughes  M, Tuovinen  T,  et al.  BiomarCaRE: rationale and design of the European BiomarCaRE project including 300,000 participants from 13 European countries.  Eur J Epidemiol. 2014;29(10):777-790. doi:10.1007/s10654-014-9952-xPubMedGoogle ScholarCrossref
16.
The MORGAM Project. MORGAM manual. http://www.thl.fi/publications/morgam/manual/contents.htm. Published 2001. Accessed September 18, 2019.
17.
Kulathinal  S, Karvanen  J, Saarela  O, Kuulasmaa  K.  Case-cohort design in practice—experiences from the MORGAM Project.  Epidemiol Perspect Innov. 2007;4:15. doi:10.1186/1742-5573-4-15PubMedGoogle ScholarCrossref
18.
Kulathinal  S, Niemelä  M, Niiranen  T,  et al. Description of MORGAM cohorts. https://www.thl.fi/publications/morgam/cohorts/index.html. 2005.
19.
Blankenberg  S, Salomaa  V, Makarova  N,  et al; BiomarCaRE Investigators.  Troponin I and cardiovascular risk prediction in the general population: the BiomarCaRE consortium.  Eur Heart J. 2016;37(30):2428-2437. doi:10.1093/eurheartj/ehw172PubMedGoogle ScholarCrossref
20.
Buuren  S, Groothuis-Oudshoorn  K.  Mice: multivariate imputation by chained equations in R.  J Stat Softw. 2011;45(3). doi:10.18637/jss.v045.i03Google Scholar
21.
Gao  X, Starmer  J, Martin  ER.  A multiple testing correction method for genetic association studies using correlated single nucleotide polymorphisms.  Genet Epidemiol. 2008;32(4):361-369. doi:10.1002/gepi.20310PubMedGoogle ScholarCrossref
22.
Borenstein  M, Higgins  JP, Hedges  LV, Rothstein  HR.  Basics of meta-analysis: I2 is not an absolute measure of heterogeneity.  Res Synth Methods. 2017;8(1):5-18. doi:10.1002/jrsm.1230PubMedGoogle ScholarCrossref
23.
Higgins  JP, Thompson  SG.  Quantifying heterogeneity in a meta-analysis.  Stat Med. 2002;21(11):1539-1558. doi:10.1002/sim.1186PubMedGoogle ScholarCrossref
24.
Antolini  L, Nam  B-H, D’Agostino  RB.  Inference on correlated discrimination measures in survival analysis: a nonparametric approach.  Commun Stat Theory Methods. 2004;33(9):2117-2135. doi:10.1081/STA-200026579Google ScholarCrossref
25.
Paapstel  K, Kals  J, Eha  J,  et al.  Inverse relations of serum phosphatidylcholines and lysophosphatidylcholines with vascular damage and heart rate in patients with atherosclerosis.  Nutr Metab Cardiovasc Dis. 2018;28(1):44-52. doi:10.1016/j.numecd.2017.07.011PubMedGoogle ScholarCrossref
26.
Moxon  JV, Jones  RE, Wong  G,  et al.  Baseline serum phosphatidylcholine plasmalogen concentrations are inversely associated with incident myocardial infarction in patients with mixed peripheral artery disease presentations.  Atherosclerosis. 2017;263:301-308. doi:10.1016/j.atherosclerosis.2017.06.925PubMedGoogle ScholarCrossref
27.
Sigruener  A, Kleber  ME, Heimerl  S, Liebisch  G, Schmitz  G, Maerz  W.  Glycerophospholipid and sphingolipid species and mortality: the Ludwigshafen Risk and Cardiovascular Health (LURIC) study.  PLoS One. 2014;9(1):e85724. doi:10.1371/journal.pone.0085724PubMedGoogle Scholar
28.
Paynter  NP, Balasubramanian  R, Giulianini  F,  et al.  Metabolic predictors of incident coronary heart disease in women.  Circulation. 2018;137(8):841-853. doi:10.1161/CIRCULATIONAHA.117.029468PubMedGoogle ScholarCrossref
29.
Bergheanu  SC, Reijmers  T, Zwinderman  AH,  et al.  Lipidomic approach to evaluate rosuvastatin and atorvastatin at various dosages: investigating differential effects among statins.  Curr Med Res Opin. 2008;24(9):2477-2487. doi:10.1185/03007990802321709PubMedGoogle ScholarCrossref
30.
Tang  WH, Wang  Z, Levison  BS,  et al.  Intestinal microbial metabolism of phosphatidylcholine and cardiovascular risk.  N Engl J Med. 2013;368(17):1575-1584. doi:10.1056/NEJMoa1109400PubMedGoogle ScholarCrossref
31.
Wang  Z, Klipfell  E, Bennett  BJ,  et al.  Gut flora metabolism of phosphatidylcholine promotes cardiovascular disease.  Nature. 2011;472(7341):57-63. doi:10.1038/nature09922PubMedGoogle ScholarCrossref
32.
Gault  CR, Obeid  LM, Hannun  YA.  An overview of sphingolipid metabolism: from synthesis to breakdown.  Adv Exp Med Biol. 2010;688:1-23. doi:10.1007/978-1-4419-6741-1_1PubMedGoogle ScholarCrossref
33.
Koenig  W, Karakas  M, Zierer  A,  et al.  Oxidized LDL and the risk of coronary heart disease: results from the MONICA/KORA Augsburg Study.  Clin Chem. 2011;57(8):1196-1200. doi:10.1373/clinchem.2011.165134PubMedGoogle ScholarCrossref
34.
Ganna  A, Salihovic  S, Sundström  J,  et al.  Large-scale metabolomic profiling identifies novel biomarkers for incident coronary heart disease.  PLoS Genet. 2014;10(12):e1004801. doi:10.1371/journal.pgen.1004801PubMedGoogle Scholar
35.
van der Veen  JN, Kennelly  JP, Wan  S, Vance  JE, Vance  DE, Jacobs  RL.  The critical role of phosphatidylcholine and phosphatidylethanolamine metabolism in health and disease.  Biochim Biophys Acta Biomembr. 2017;1859(9 pt B):1558-1572. doi:10.1016/j.bbamem.2017.04.006PubMedGoogle ScholarCrossref
36.
Noronha  A, Modamio  J, Jarosz  Y,  et al.  The virtual metabolic human database: integrating human and gut microbiome metabolism with nutrition and disease.  Nucleic Acids Res. 2018;47(1):614-624.PubMedGoogle Scholar
37.
Wishart  DS, Feunang  YD, Marcu  A,  et al.  HMDB 4.0: the human metabolome database for 2018.  Nucleic Acids Res. 2018;46(D1):D608-D617. doi:10.1093/nar/gkx1089PubMedGoogle ScholarCrossref
38.
Geidenstam  N, Magnusson  M, Danielsson  APH,  et al.  Amino acid signatures to evaluate the beneficial effects of weight loss.  Int J Endocrinol. 2017;2017:6490473. doi:10.1155/2017/6490473PubMedGoogle Scholar
×