Use of clinically based troponin underestimates the cardiac injury in non-cardiac surgery: a single-centre cohort study in 51,701 consecutive patients

Postoperative myocardial infarction causes hundreds of thousands of deaths annually, and failure to rescue is a leading cause of hospital mortality. Strategies to recognize cardiac injury are important to reduce the burden of cardiac-related morbidity. For these reasons, we chose to assess the association between postoperative troponin I elevations and 30-day in-hospital mortality and, secondarily, to compare the predictive value of regularly scheduled troponin estimates with troponin ordered in response to clinical indications. We carried out a retrospective cohort analysis of 51,701 consecutive patients throughout 2003 to 2009. All patients were from a single university referral hospital and included all non-cardiac non-transplant surgery patients requiring overnight admission. Logistic regression was used to assess the risk-adjusted associations between troponin I and 30-day in-hospital mortality. The multivariable predictive model for death improved after troponin I was included. The receiver operating characteristic was 0.902 before troponin I vs 0.934 after troponin I (P < 0.0001). The likelihood ratio for troponin was 3.0 (95% confidence interval 2.8 to 3.2) and evident in each surgical service. Increasing troponin I showed a dose-response associated with increased mortality, and compared with clinically based measurements, a regularly scheduled postoperative troponin protocol showed a threefold increase in the probability of detecting myocardial injury. However, troponin I was not found to improve the risk prediction model in the lowest risk patients (n = 18,953; probability of death < 0.02%) with one cardiac death. Postoperatively elevated troponin I is associated with 30-day in-hospital mortality in a dose-dependent manner. A postoperative measurement protocol provides a threefold increase in the ability to detect myocardial injury. Conversely, in patients with a low mortality risk, cardiac injury is low; there is minimal improvement in the ability to detect cardiac injury, and the rescue rates from cardiac injury are excellent. These findings suggest that a surveillance protocol of troponin I would be optimal when limited to moderate to high-risk patients.