Clinical correlates, threshold sensitivity, and parsimonious modeling of high-sensitivity cardiac troponin I in health examination cohorts: a retrospective cross-sectional study with same-center temporal validation

Background High-sensitivity cardiac troponin I (hs-cTnI) interpretation depends on assay-specific reference limits. We examined correlates of hs-cTnI above a study-specific cutoff, compared two parsimonious logistic models, and assessed temporal transportability and threshold sensitivity. Methods The 2025 development cohort comprised 1,758 adults undergoing a standardized health-examination package; 86 (4.9%) had hs-cTnI >19.8 ng/L. Unweighted Age + Sex and Age + Sex + Creatinine models were fitted using all participants, with 1,000-bootstrap internal validation. Fixed equations were evaluated in a same-center cohort examined during January–June 2026: 2,181 participants for Age + Sex and 2,179 for Age + Sex + Creatinine because two participants lacked creatinine. Threshold sensitivity analyses were feasible only in 2026, using report-based sex-specific limits (15.6 ng/L for women; 34.2 ng/L for men), FDA-reported sex-specific limits (17/35 ng/L), and an overall 28 ng/L cutoff. A left-censored analysis evaluated continuous hs-cTnI. Results Development Age + Sex ROC-AUC was 0.838, with an optimism-corrected AUC of 0.836. Adding creatinine changed AUC by +0.002 (95% CI −0.003 to +0.006). Temporal ROC-AUC was 0.716 (95% CI 0.649–0.774) for Age + Sex and 0.705 (0.640–0.762) for Age + Sex + Creatinine; both models overpredicted the probability of hs-cTnI >19.8 ng/L. In 2026, the age-adjusted male-sex OR was 2.39 under the common cutoff, 0.64 (95% CI 0.33–1.24) with report-based sex-specific limits, and 0.72 (0.36–1.43) with FDA-reported sex-specific limits. Both sex-specific intervals included 1; these threshold-dependent estimates do not establish an inverse male-sex association. Age remained positively associated across threshold definitions. Conclusions Age was the most consistent correlate among available variables. Male-sex estimates varied by cutoff definition in the 2026 cohort. Adding creatinine provided no evidence of improved discrimination beyond Age + Sex. Calibration deteriorated in the temporal cohort, but its cause remains uncertain. These descriptive findings do not support clinical risk categories, electronic alerts, or replacement of direct hs-cTnI measurement.