Abstract
Patients with stable coronary artery disease (CAD) are at an increased risk of acute myocardial infarction (AMI), particularly among older individuals. Developing a reliable model to predict AMI occurrence in these patients holds the potential to expedite early diagnosis and intervention. This study is aimed at establishing a circulating amino acid-assisted model, incorporating amino acid profiles alongside clinical variables, to predict AMI risk. A cohort of 874 CAD patients from two independent centers was analyzed. Plasma amino acid levels were quantified using liquid chromatography tandem mass spectrometry (LC-MS/MS) employing a targeted metabolomics approach. This methodology incorporated (13)C isotope-labeled internal standards for precise quantification of 27 amino acids. Univariate logistic regression was applied to identify differentially expressed amino acids that distinguished between stable CAD and AMI patients. To assess prediction performance, receiver operating characteristic (ROC) curve and nomogram analyses were utilized. Five amino acids-lysine, methionine, tryptophan, tyrosine, and N6-trimethyllysine-emerged as potential biomarkers (p < 0.05), exhibiting significant differences in their expression levels across the two centers when comparing stable CAD with AMI patients. For AMI risk prediction, the base model, utilizing 12 clinical variables, achieved areas under the curve (AUC) of 0.7387 in the discovery phase (n = 623) and 0.8205 in the external validation set (n = 251). Notably, the integration of these five amino acids into the prediction model significantly enhanced its performance, increasing the AUC to 0.7651 in the discovery phase (Delong's test, p = 1.43e-02) and to 0.8958 in the validation set (Delong's test, p = 8.91e-03). In conclusion, the circulating amino acid-assisted model effectively enhances the prediction of AMI risk among CAD patients, indicating its potential clinical utility in facilitating early detection and intervention.