Introduction: The Atherosclerosis Risk In Communities (ARIC) study conducted community surveillance of hospitalized myocardial infarction (MI) from 1987 to 2014 among four US communities (Jackson MS, Forsyth County NC, Washington County MD, and Minneapolis MN). Surveillance of MI during the troponin era (1996 - ) has been complicated by increasing sensitivity of troponin I assays. It is unclear to what extent increased assay sensitivity has affected the validity of event classification. We hypothesized that among events that would have been classified as a definite/probable MI or a suspect/no MI regardless of cardiac biomarkers, the sensitivity and specificity of troponin I assays to identify abnormal enzyme levels (ARIC community surveillance criterion: 2x upper limit of normal) has changed over time in hospitals participating in ARIC community surveillance.
Methods and results: From 33,995 community hospitalizations with suspicion of MI or coronary heart disease with a troponin I measurement that occurred between 1996 and 2014, 2,143 met ARIC criteria as MI cases (ARIC computer algorithm classification of definite or probable MI regardless of cardiac biomarker levels) and 9,909 we considered noncases (ARIC computer algorithm classification of no or suspect MI regardless of cardiac biomarker levels). In order to be an MI case, the event had to include an evolving diagnostic ECG. To be an noncase, the event had to include no chest pain and an equivocal or absent/uncodable ECG. We used survey-weighted logistic regression models to predict sensitivity and specificity from 1996 - 2014. The sensitivity of troponin I assay being abnormal was approximately 50% for most of the time period, with a potential increase to 70% after 2010. The specificity was approximately 90% in the late 1990s and early 2000s, but fell to 68% (95% confidence interval: 63% - 73%) by 2014. The median upper limit normal (ULN) for troponin I among all 33,995 hospitalizations dropped from 1.4 ng/mL in 1996 to 0.045 ng/mL in 2014.
Conclusions: For most of the troponin era among the four ARIC communities, hospitalizations that were classified as an MI regardless of cardiac biomarker levels were equally likely to have or not have abnormal troponin I. In recent years, the likelihood that hospitalizations that were classified as no MI regardless of cardiac biomarker levels had normal troponin I decreased, potentially affecting the usefulness of troponin I measurements to separate hospitalized MIs from non-MIs in the context of community surveillance, particularly among hospitalizations with few other signs of MI. These shifts in assay validity were likely due in part to a decrease in ULN for troponin I in the four ARIC communities.