Central versus local adjudication of myocardial infarction in a cardiac biomarker trial

Objectives

The impact of regulatory requirements, which require central adjudication for the diagnosis of acute myocardial infarction (AMI) in cardiac biomarker studies, is unclear. We determined the impact of local (at the site of subject enrollment) versus central adjudication of AMI on final diagnosis.

Methods

This is a retrospective analysis of data from the Myeloperoxidase in the Diagnosis of Acute Coronary Syndromes Study, an 18-center prospective study of patients with suspected acute coronary syndromes, with enrollment from December 19, 2006, to September 20, 2007. Local adjudication of AMI was performed by a single site investigator at each center following the protocol-specified definition and according to the year 2000 definition of AMI, which based cardiac troponin (cTn) elevation on local cut points for each of the 13 different assays. After completion of the Myeloperoxidase in the Diagnosis of Acute Coronary Syndromes Study primary analysis and to evaluate a new troponin assay, a Food and Drug Administration–mandated central adjudication was performed by 3 investigators at different institutions. This adjudication used the 2007 Universal Definition of AMI, which differs by use of the manufacturer's 99th percentile cTn cut point. We describe the outcome of this process and compare it with the local adjudication. Central adjudicators were not blinded to local adjudications. For central adjudication, discrepant diagnoses were resolved by consensus. Local versus central cTn cut points differed for 6 assays. Both definitions required a rise and/or fall of cTn. Discrepant cases were reviewed by the lead author. Difficult cases were defined as having a difference between local and central adjudication, an elevated cTn with a temporal rise and fall, and a negative or absent risk stratification test. Statistics were by χ², κ, and logistic regression.

Results

Of 1,107 patients enrolled, 11 had indeterminate central adjudication, leaving 1,096 for analysis. In spite of high agreement across central versus local adjudicators, κ = 0.79 (95% CI [0.73, 0.85]), AMI was diagnosed more often by central adjudication, 134 (12.2%) versus 104 (9.5%), with 44 local diagnoses (4%) changed from non-AMI to AMI (n = 37) or AMI to non-AMI (n = 7) (P < .001). These 44 represented 34% (95% CI 26%-42%) of 141 cases in which either central or local adjudication was AMI. Of diagnoses changed to AMI, 3 reasons contributed approximately one-third each: the local use of a non-99th percentile cTn cutoff (32%), the possibility of human error (34%), and difficult cases (34%).

Conclusion

Despite an acceptable κ, over a third of patients with a diagnosis of AMI were not assigned that diagnosis by both sets of adjudicators. This supports the importance of 1 standard method for diagnosis of AMI.