Novel Metabolomic Signature as a Biomarker to Detect Subclinical Atherosclerosis in Individuals Traditionally Classified as Low Risk
Yuxin Xue, Joshua Vescovi, Adam S. Chan, Stephen T. Vernon, Owen Tang, Dantong Zhu, Katharine A. Bate, Daniel Brieger, Michael P. Gray, Lake-Ee Quek, John F. O’Sullivan, Stuart M. Grieve, Ellis Patrick, Jean Y.H. Yang, Gemma A. FigtreeBACKGROUND:
Standard modifiable risk factors such as age, hypertension, diabetes, cholesterol, and smoking contribute to coronary artery disease risk. However, up to 27% of first-time heart attacks occur in individuals without standard modifiable cardiovascular risk factors. As medical management can stabilize coronary atherosclerosis, there is a need for biomarkers to identify disease before symptoms or a heart attack.
METHODS:
We performed targeted metabolomics, identifying 54 metabolites in 1212 adults undergoing coronary artery calcium score (CACS) from the BioHEART-CT Study. Individuals were classified as having clinically actionable CACS positive or not (clinically actionable CACS negative). Metabolite-to-metabolite ratios were then calculated to identify transferable biomarkers. Logistic regression models incorporating age, sex, and metabolite ratios were developed and classified as low risk by traditional scoring. The BioHEART Metabolite Risk Score (BioHEART-MRS) was derived from the best-performing model, trained in a Discovery cohort (n=635), adapted using transfer component analysis (n=266), and validated in an independent External cohort (n=311).
RESULT:
A model including age, sex, and 5 metabolite ratios: histidine-kynurenic acid, histidine-riboflavin, citrullin-proline, kynurenic acid–phosphocholine, and arginine-dimethylguanidino valerate, referred to as BioHEART-MRS, demonstrated the best performance in younger individuals for predicting CACS. In the External cohort, BioHEART-MRS outperformed the traditional risk score (area under the curve=0.67 versus 0.55). The traditional risk score underperformed in the under-60 cohort, misclassifying 62% of those with clinically actionable CACS as low risk. Within this group, BioHEART-MRS identified 46 additional cases among low-risk individuals (n=495). As a triage tool, BioHEART-MRS reduced the number needed to scan to detect one true positive case to 7, compared with scanning all low-risk individuals.
CONCLUSIONS:
In the External validation cohort, BioHEART-MRS showed greater discrimination than Multi-Ethnic Study of Atherosclerosis risk score among traditionally low-risk individuals and enriched for clinically actionable CACS positive among those selected for imaging. These findings support prospective evaluation of metabolite-based triage for CACS, particularly in younger and SMuRFless individuals.