The Trouble with Troponin: Sensitivity and Patient Selection?

Troponin assays were developed to better diagnose acute myocardial infarction (AMI). Yet troponin elevation is frequently encountered in non-cardiac disease states. To investigate test performance in a large integrated healthcare system we reviewed all testing performed in a 14-hospital system over a 12-month period in 2014. Troponin-I results were judged to be either normal (=0.04mg/dl) based on laboratory reference ranges. We identified a total of 82,853 hospital encounters associated with testing, comprising 12% of all outpatient and 29% of all inpatient visits. The prevalence of AMI was low with only 3.5% of encounters associated with either a primary or secondary diagnosis of AMI. Both sensitivity and specificity were greater than 90%. However, positive predictive value (PPV) was low: 22% for a primary diagnosis of AMI and 29% for any AMI diagnosis. The majority (71%) of troponin elevations were not associated with a diagnosis of AMI. Sepsis, rather than AMI, was the most common diagnosis associated with troponin elevation. Given the low prevalence of AMI in our patient population, specificity rather than sensitivity would be expected to determine PPV. In fact, increased test sensitivity at the expense of specificity would further reduce PPV. Using our data set we calculated a specificity of 91%. Statistical modeling suggests that increasing specificity to 95% would increase PPV to almost 50% and that increasing specificity to 99% would raise PPV to greater than 75%. Our data suggests that the quest for higher troponin sensitivity may be misguided and that efforts to enhance test performance should focus on improving both specificity and patient selection. Reduced testing of patients with sepsis or at low risk for AMI would significantly improve PPV without requiring development of new “high specificity” assays.