1010-64 Left Ventricular Size Effects the Diagnostic Accuracy of SPECT Thallium-201 Perfusion Imaging in Both Men and Women

The diagnostic accuracy of SPECT TI-201 for coronary artery disease (CAD) is lower in women (F) than in men (M). Previous investigators have suggested a Bayesian explanation. More recently, it has been suggested that smaller left ventricular mass in women was the cause. The current study was undertaken to assess the effect of left ventricular size (LVsize) on the accuracy of SPECT TI-201. For this study patients who had either less than 5% pretest probability of CAD (nls) or had cardiac catheterization performed within 45 days of stress thallium testing without an intervening cardiac event (pts) were Identified. We excluded patients with pathologic Q waves, documented prior infarction, LBBB on EKG, prior CABG or non-ischemic cardiomyopathy. We Identified a total of 259 subjects (153 M, 106 F). Patients with<50% stenosis on cath were analyzed as nl. Twenty M and 20 F were taken from the nls and processed separately to create normal limits. Images were reconstructed with a Hanning filter. The fraction of the left ventricle less than 1, 2, 2.5, 3 and 4 standard deviations was calculated and used to calculate the area under the receiver operating characteristic (ROC) curves using the Dorfman maximum likelihood technique. A measure of left ventricular size was generated from short axis slice diameter and the number of slices. The number of stenosed vessels (2.0±0.8 vs. 1.9±0.9, P=NS) and worst stenosis (91±12% vs. 89±15%, p=NS) were similar for M and F. M exercised to higher workloads (estimated METS 6.7±3.1 vs. 4.3±2.3, p<0.0001) but similar peak heart rates (HR) and percent of maximal predicted HR (133±19 vs. 130±22 and 78±11% vs. 79±13%, both p=NS). The LVsize was greater in men (105 vs. 79 p<0.001). The area under the ROC curves for M was marginally greater than F (0.92 vs. 0.82 P=0.10). LV size significantly affected accuracy; the ROC area for M and F with LVsize>75 was much greater than for those<75 (0.89 vs. 0.64 P=0.007). There was no significant difference in accuracy between M and F with LVsize>75 (0.91vs. 0.85, P=0.47). We conclude that a small LVsize significantly reduces the accuracy of SPECT TI-201 in M and F, but since F have, on average, smaller LVsize, they are more likely to be affected