The present study compares the on-site interpretation of an admission electrocardiogram (ECG) with core laboratory results in a large, multicenter trial of 516 patients diagnosed with unstable angina pectoris or non-Q-wave myocardial infarction. The local investigators evaluated the admission ECG regarding ST-T changes before the ECGs were sent to the core laboratory for blinded interpretation. The strength of agreement between the observations was described by κ statistics. There was a poor agreement regarding identification of ST-segment elevation, with 17 patients identified by the local investigator versus 92 by the core laboratory (κ = 0.05). There was a fair agreement on ST-segment depression with 158 patients diagnosed on-site versus 64 by the core laboratory (κ = 0.38). Identification of T-wave inversion demonstrated good agreement with 306 patients diagnosed on-site versus 280 by the core lab- oratory (κ = 0.63). A moderate agreement regarding identification of a normal ECG was found with 101 patients on-site versus 135 in the core laboratory (κ = 0.42). Independent variables, including peak creatine kinase-MB and 30-day outcome, were more closely related to core laboratory results than the local investigator's interpretation of the admission ECG. Thus, in the present study, considerable differences were demonstrated between the on-site interpretation of the admission ECG and the blinded evaluation performed in the core laboratory regarding relatively simple electrocardiographic variables. The results suggest that more widespread use of independent evaluation of clinical data should be incorporated in future clinical trials.