Towards a more robust estimation of the decision threshold for high-sensitivity cardiac troponins

Background The Universal Definition of Myocardial Infarction defines myocardial injury as an elevation in cardiac troponin (cTn) above the 99th percentile upper reference limit (URL) of a healthy reference population. Establishing precise URLs for cTn assays is therefore important, however recommended URLs across studies are heterogeneous and URLs for cTnI and cTnT are not bioequivalent. This study assesses the robustness of current methods for establishing cTn reference limits. Methods Generation Scotland participants (N=17,328) with high-sensitive cTnI and cTnT measurements and without self-reported cardiovascular disease were available for analysis, including 9,039 participants that fulfilled criteria of a healthy subcohort according to International Federation of Clinical Chemistry (IFCC) recommendations. Percentiles were determined using the non-parametric statistical method as recommended by the IFCC. To assess how the confidence interval around the 99th percentile varies with group size, we drew 50 independent samples for group sizes of 200, 400, 800, 1,600, and 6,000 individuals. Results Across key percentiles (0.5, 1, 2.5, 10, 25, 50, 75, 90, 95, 97.5) , cTnI values were lower than cTnT (median [IQR]: 1.8 [0.6–2.9] ng/L versus 3.0 [1.5–5.6] ng/L), yet this reversed at the 99th percentile with cTnI higher than cTnT (20.4 ng/L versus 17.6 ng/L), consistent with the presence of more uncommon high values in the upper tail of the cTnI distribution. At sample sizes of 400 male and female individuals recommended by the IFCC, the estimated 99th percentile is associated with a wide confidence interval, spanning approximately 10 ng/L from the lower to the upper bound for cTnT and 30 ng/L for cTnI. Increasing the size of reference cohorts (N>1,600) partially improved the robustness of the estimates of the 99th percentile, but did not resolve the fundamental fragility of relying on extreme tail 99th percentile values, even when applying stringent IFCC criteria for a healthy cohort. Sex-specific analyses showed similar results. Conclusion Our findings underscore fundamental limitations in the current definition of cardiac troponin URL. The 99th percentile is unstable at recommended minimum sample sizes and remains sensitive to extreme values even as sample sizes increases. Future approaches should consider more statistically robust thresholds, combined with greater individualization by sex, age, and comorbidity, to improve clinical relevance and precision.