A-031 High-Sensitivity Troponin in Routine Outpatient Testing: Laboratory Phenotypes and Explainable AI for Cardiovascular Risk Stratification

Abstract Background High-sensitivity cardiac troponins (hs-cTn) are cornerstone biomarkers in laboratory medicine. Beyond acute myocardial infarction, hs-cTn support risk stratification and detection of subclinical myocardial injury, with detectable baseline levels independently associated with adverse outcomes. However, laboratory phenotypes accompanying detectable or elevated hs-cTn in routine outpatient testing remain incompletely characterized. Integrative artificial intelligence (AI) enables multidimensional laboratory data analysis to identify biologically coherent patterns linked to cardiovascular vulnerability. This study aimed to characterize the distribution of hs-cTn in routine testing and evaluate its association with biochemical profiles using interpretable AI models to advance precision cardiovascular risk assessment. Methods This retrospective observational study included all hs-cTnI measurements performed between 2023 and 2025 at a private clinical laboratory in Brazil. Hs-cTnI was measured using a chemiluminescent microparticle immunoassay (Alinity i®, Abbott Diagnostics; limit of detection 0.7–1.6 pg/mL; limit of quantification 3.7–5.1 pg/mL; 10% CV). Troponin elevation was defined as concentrations above the sex-specific 99th percentile upper reference limit (=15.6 ng/L for women; =34.2 ng/L for men).Hs-cTnI results were evaluated alongside hematologic (hemoglobin, hematocrit), metabolic (fasting glucose, HbA1c), renal (urea, creatinine, estimated glomerular filtration rate), lipid (total cholesterol, HDL-C, LDL-C, triglycerides), and inflammatory (high-sensitivity C-reactive protein) biomarkers, all measured using standardized automated methods. Results A total of 7,518 hs-cTnI tests were included. Median age was 63.9 years (IQR 48.4–75.8), and 55.5% were female. Troponin concentrations were detectable in 97.6% of samples and quantifiable in 22%. The overall prevalence of hs-cTnI values above the sex-specific 99th percentile upper reference limit (URL) was 5.01% (5.63% in women vs. 4.24% in men), increasing progressively with age and reaching 12.27% among individuals >80 years. Hs-cTnI concentrations were significantly associated with markers of renal dysfunction, including higher urea (p<0.001), higher creatinine (p<0.001), and lower estimated glomerular filtration rate (p<0.001). Significant associations were also observed with inflammatory and metabolic markers, including elevated high-sensitivity C-reactive protein, higher fasting glucose and glycated hemoglobin, and lower HDL cholesterol (all p<0.001). In predictive models for hs-cTnI =URL, age, renal function markers, and high-sensitivity C-reactive protein emerged as the strongest predictors, achieving good discriminative performance (AUROC 0.80–0.84). AI-driven clustering identified a distinct renal–inflammatory phenotype predominantly in older individuals, exhibiting an almost threefold higher prevalence of hs-cTnI elevation (=URL: 14.6%) compared to the overall cohort. Conclusion In routine testing, hs-cTnI demonstrated a high prevalence of detectability and showed consistent associations with renal, inflammatory, and metabolic phenotypes, particularly in older individuals, suggesting a state of subclinical cardiovascular vulnerability. In predictive models, and high-sensitivity C-reactive protein emerged as the strongest predictors of hs-cTnI =URL (AUROC 0.80–0.84). The integration of multidimensional laboratory data through interpretable AI models enabled the identification of distinct laboratory patterns associated with elevated troponin levels, providing insight into hs-cTn behavior beyond acute clinical settings. These findings expand the clinical role of hs-cTnI as an integrated biomarker of cardiometabolic and inflammatory risk and support the application of AI-enhanced laboratory strategies for refined risk stratification and clinical prioritization.