Coronary artery disease (CAD) is the leading cause of death in many countries. Stress electrocardiogram (ECG) is able to detect myocardial ischemia and also has prognostic value, which may be impaired in presence of electrical baseline abnormalities. Stress myocardial single photon emission tomography (SPECT) has recognized utility yield in assessment of CAD, requiring independent interpretation of ECG and myocardial images. Purpose: To analyze stress ECG interpretation reproducibility among observers with different training level, including pre- and post-graduate medical students, compared to an experienced cardiologist and also with SPECT. Method: We studied 95 patients under CAD evaluation, mean age 61±9.3 years (range: 42-85), 56% male. Interobserver correlation kappa (k) between perfusion gated 99mTc-Sestamibi SPECT and exercise ECG were calculated as normal/abnormal, presence of necrosis, ischemia or mixed pattern. Interobserver kappa (k) analysis was made. Results: 49.5% stress ECG and 45% SPECT studies were abnormal at the initial report with 62.1% concordance (k: 0.24) for normal/abnormal and 58.9% for ischemia (k: 0.14). Agreement between stress ECG initial report and independent cardiologist was 89.5% (k: 0.78). The correlation between independent observers and the initial report ranged between 62.1% and 48.4%, for baseline ECG between 41.1% and 90.5% considering normal/abnormal, and between 80% and 93.7% (k: 0.59-0.87) for the presence of ischemia. Conclusion: Stress ECG interpretation presented adequate interobserver reproducibility with greater agreement in the most experienced observers, confirming the importance of training. © 2009 Elsevier Espaa, S.L. y SEMN.