The accuracy of technetium-99m (Tc-99m) sestamibi single-photon emission computed tomography (SPECT) for the assessment of myocardial viability in patients with coronary artery disease and left ventricular (LV) dysfunction is not defined completely. This study determines whether the performance of Tc-99m sestamibi SPECT for viability detection differs between patients with mild-to-moderate coronary artery disease. Patients with regional and/or global LV dysfunction were separated into 2 groups on the basis of LV ejection fraction (EF) at rest: group 1 (LVEF > 25%, mean 36 +/- 6%, n = 9), and group 2 (LVEF < or = 25%, mean 17 +/- 5%, n = 11). All patients underwent semiquantitative Tc-99m sestamibi SPECT and positron emission tomography (PET) at rest with N-13 ammonia and F-18 fluorodeoxyglucose. The overall regional concordance of SPECT and PET for viability detection was 89% in group 1 and 78% in group 2 (p = 0.002). Discordance in group 2 was almost exclusively due to PET viable and/or SPECT nonviable regions. In regions with hypoperfusion at rest by PET, concordance was 78% in group 1 and only 64% in group 2 (p = 0.0015). In regions with reduced perfusion and relatively increased metabolic activity ("flow: metabolism mismatch"), Tc-99m sestamibi SPECT identified 88% of regions in group 1 as viable, but only 42% of regions in group 2 (p = 0.002). Thus, while Tc-99m sestamibi semiquantitative SPECT at rest shows a good concordance with PET for the detection of myocardial viability in patients with coronary artery disease with mild-to-moderate LV dysfunction, it may underestimate myocardial viability in patients with severe LV dysfunction, particularly in those patients with hypoperfusion at rest as assessed by PET.