Blinded evaluation of planar technetium-99m-sestamibi myocardial perfusion studies.

Sestamibi planar myocardial perfusion studies were performed at Hotel-Dieu de Montreal on 28 patients with documented coronary artery disease and 16 normal subjects. Stress and rest studies were performed on separate days. These studies were sent to Virginia for interpretation while blinded as to age, sex, and other clinical information. Studies were quantitated independently by two operators (using a computer program modified for Sestamibi), and interpreted independently by two experienced interpreters. Computer quantitation of 2816 segments gave an average interoperator deviation of 2.2%. Pure quantitative criteria were applied for computer interpretation. By varying the detection threshold, we produced the entire ROC curve relating sensitivity and specificity as a function of detection threshold. Using only computer criteria for normal or abnormal, interoperator agreement by patient was 98% and 93% by view. The computer could achieve equal positive and negative predictive accuracy of 87%. Interpreters, allowed both quantitative and subjective judgment, agreed on 91% of 44 patients, 90% of 132 views, and 92% of 660 segments. Interpreters averaged 94% positive and 86% negative predictive accuracy.