Distribution and prognostic value of high-sensitivity cardiac troponin T and I across glycemic status: a population-based study

Abstract

Background: Whether distributions and prognostic values of high-sensitivity cardiac troponin (hs-cTn) T and I are different across normoglycemic, prediabetic, and diabetic populations is unknown. Methods: 10127 adult participants from the National Health and Nutrition Examination Survey 1999–2004 with determined glycemic status and measurement of at least one of hs-cTn assays were included, from whom healthy participants and presumably healthy diabetic and prediabetic participants were selected to investigate pure impacts of glycemic status on distributions of hs-cTn. The nonparametric method and bootstrapping were used to derive the 99th upper reference limits of hs-cTn and 95% CI. Participants with available follow-up and hs-cTn concentrations of all 4 assays were included in prognostic analyses. Associations of hs-cTn with all-cause and cardiac-specific mortality were modeled by Cox proportional hazard regression under the complex survey design. The incremental value of hs-cTn to an established risk score in predicting cardiac-specific mortality was assessed by the 10-year area under time-dependent receiver operating characteristic curve (AUC) using the Fine-Grey competing risk model. Results: Among 9714 participants included in prognostic analyses, 5946 (61.2%) were normoglycemic, 2172 (22.4%) prediabetic, and 1596 (16.4%) diabetic. Hyperglycemic populations were older than the normoglycemic population but sex and race/ethnicity were similar. During the median follow-up of 16.8 years, hs-cTnT and hs-cTnI were independently associated with all-cause and cardiac-specific mortality across glycemic status. In the diabetic population, adjusted hazard ratios per 1-standard deviation increase of log-transformed hs-cTnT and hs-cTnI (Abbott) concentrations were 1.77 (95% CI 1.48–2.12; P < 0.001) or hs-cTnI (Abbott) (difference in AUC: 0.071; 95% CI 0.046–0.097; P < 0.001) would significantly increase the discriminative ability of the risk score; AUC of the score combined with hs-cTnT would be further improved by incorporating hs-cTnI (0.018; 95%CI 0.006–0.029; P = 0.002). The 99th percentile of hs-cTnT of the presumably healthy diabetic population was higher than the healthy population and had no overlap in 95% CIs, however, for hs-cTnI 99th percentiles of the two populations were very close and 95% CIs extensively overlapped. Conclusions: Hs-cTnT and hs-cTnI demonstrated consistent prognostic associations across glycemic status but incremental predictive values in hyperglycemic populations only. The susceptibility of hs-cTnT 99th percentiles to diabetes plus the additive value of hs-cTnI to hs-cTnT in diabetic cardiovascular risk stratification suggested hs-cTnI and hs-cTnT may be differentially associated with glycemic status, but further research is needed to illustrate the interaction between hyperglycemia and hs-cTn.