Derivation and validation of a high sensitivity troponin-T HEART pathway

Document Type


Publication Date


Publication Title

American heart journal


BACKGROUND: The HEART Pathway is widely used for chest pain risk stratification but has yet to be optimized for high sensitivity troponin T (hs-cTnT) assays.

METHODS: We conducted a secondary analysis of STOP-CP, a prospective cohort study enrolling adult ED patients with symptoms suggestive of acute coronary syndrome at 8 sites in the United States (US). Patients had a 0- and 1-hour hs-cTnT measured and a HEAR score completed. A derivation set consisting of 729 randomly selected participants was used to derive a hs-cTnT HEART Pathway with rule-out, observation, and rule-in groups for 30-day cardiac death or myocardial infarction (MI). Optimal baseline and 1-hour troponin cutoffs were selected using generalized cross validation to achieve a negative predictive value (NPV) >99% for rule out and positive predictive value (PPV) >60% or maximum Youden index for rule-in. Optimal 0-1-hour delta values were derived using generalized cross validation to maximize the NPV for the rule-out group and PPV for the rule-in group. The hs-cTnT HEART Pathway performance was validated in the remaining cohort (n = 723).

RESULTS: Among the 1452 patients, 30-day cardiac death or MI occurred in 12.7% (184/1452). Within the derivation cohort the optimal hs-cTnT HEART Pathway classified 36.5% (266/729) into the rule-out group, yielding a NPV of 99.2% (95% CI: 98.2-100) for 30-day cardiac death or MI. The rule-in group included 15.4% (112/729) with a PPV of 55.4% (95% CI: 46.2-64.6). In the validation cohort, the hs-cTnT HEART Pathway ruled-out 37.6% (272/723), of which 2 had 30-day cardiac death or MI, yielding a NPV of 99.3% (95% CI: 98.3-100). The rule-in group included 14.5% (105/723), yielding a PPV of 57.1% (95% CI: 47.7-66.6).

CONCLUSIONS: A novel hs-cTnT HEART Pathway with serial 0- and 1-hour hs-cTnT measures has high NPV and moderate PPV for 30-day cardiac death or MI.

PubMed ID



ePub ahead of print



First Page


Last Page