Interhospital observer agreement in interpretation of exercise myocardial Tc-99m tetrofosmin SPECT studies

We sought to determine the degree of interhospital agreement in the interpretation of exercise myocardial technetium-99m tetrofosmin single photon emission computed tomography (SPECT).

Five experienced hospital laboratories were asked to submit 2 sets of myocardial Tc-99m tetrofosmin SPECT images obtained in 150 patients undergoing coronary angiography: group A used a uniform color scale for all hospitals, and group B used the individual color scale in place at each hospital (uniform color scale, nonuniform color scale, and black-and-white scale). Thus a total of 300 images were interpreted by each center without knowledge of any other patient data. Angiographically significant coronary artery disease (< or =50% diameter stenosis) was present in 90 patients (60%). By a majority decision (3 or more centers), the sensitivity was found to be similar for groups A and B (82% and 84%, respectively), but the specificity was significantly higher for group A (87% vs 73%; P =.021). Four or all 5 of the centers agreed on abnormal or normal results of SPECT images in 87% of patients in group A (kappa 0.626) and in 78% of patients in group B (kappa 0.528). The kappa value of 0.617 was obtained for the uniform color scale, 0.467 for the uniform black-and-white scale, and 0.444 for the nonuniform color scale. Agreement on the left anterior descending artery territory (81% for group A and 78% for group B) was similar to that of the right coronary artery territory (79% for A and 75% for B) and to that of the left circumflex artery territory (91% for A and 85% for B). Agreement was similar in patients with 1-, 2-, and 3-vessel coronary artery disease (91%, 88%, and 86% for group A and 81%, 82%, and 82% for group B, respectively).

In the interpretation of myocardial Tc-99m tetrofosmin SPECT images, good interinstitutional observer agreement was found, mainly when the uniform display method was adopted.