Tryptophan metabolites and stroke risk after acute myocardial infarction in patients with and without metabolic syndrome: insights from a MACCE-based cohort

Abstract Background Metabolic syndrome (MetS) obviously increases the risk of major adverse cardiac and cerebrovascular events (MACCE) in patients with acute myocardial infarction (AMI). However, the metabolic mechanisms underlying this heightened vulnerability remain unclear, and individualized predictive models are limited. Objective To elucidate the role of tryptophan metabolism in MetS-related MACCE risk after AMI and to establish a machine learning model (random forest) for MACCE prediction. Methods A total of 3223 AMI patients undergoing percutaneous coronary intervention were enrolled between 2017 and 2021. Through untargeted metabolomics analysis, potential MetS-related metabolites were screened and identified, followed by internal validation of tryptophan metabolites—indole-3-lactic acid (ILA), tryptophan (TRP), kynurenine (KYN), and indole-3-propionic acid (IPA)—in 3190 patients. Cox regression was performed separately for Mets and no-Mets participants to assess the associations of Tryptophan Levels and with a primary focus on stroke risk, and secondary analyses of other MACCE components. A random forest model was constructed to predict MACCE over a 72-month follow-up by integrating metabolic and clinical variables. Results Patients with MetS exhibited significant disturbances in tryptophan metabolism. Elevated levels of indole-3-lactic acid (ILA), tryptophan (TRP), and kynurenine (KYN) were independently associated with higher stroke risk in MetS patients (adjusted HR per twofold increase: ILA 1.34, TRP 1.46, KYN 1.47; all P < 0.05), but not in non-MetS individuals. The random forest model exhibited good prediction performance (AUC 0.715; 95% CI 0.683–0.747), identifying KYN, indole-3-lactic acid (ILA) as major predictive features. Conclusion Tryptophan metabolic dysregulation, particularly elevated KYN, was associated with a higher risk of stroke in AMI-MetS patients. By integrating untargeted discovery, targeted validation, and machine learning–based modeling, our study provides a novel framework for individualized risk stratification and supports further investigation into the translational potential of metabolic biomarkers in this high-risk cardiometabolic population.