Our study examined the values of various gated SPECT parameters in the detection of CAD in patients with normal perfusion images using ROC AUC analysis. The AUC categorized the stress ESV as an excellent classifier and categorized all the stress EDV, the stress EF, the rest ESV, and the rest EF as good classifiers. Together with a closer inspection of the confident intervals of each parameters, the stress ESV was likely the most reliable parameter. In contrast, both the ΔEF and the TID ratio were not useful in detection of CAD in patients with normal perfusion images in our cohort.
Our result emphasized that the calculated left ventricular volumes, particularly the stress ESV, and the LVEF are valuable for detection of CAD in patients with normal perfusion images (Fig. 4). This was concordant with the finding of Peix, who found that the EDV of patients with ischemic were approximately 50% higher than those of patients without ischemic and the ESV of patients with ischemic were almost twice as high as those without ischemic [9]. We further strengthened those findings by the inclusion of ICA as a reference standard and proposed the possible cutoffs for those values. We postulated that the larger volumes of the left ventricle in patients with CAD might represent the impaired contractile function of the ischemic myocardium and the contractility dysfunction might be exacerbated by the stunning of myocardium after stress [9]. It was also possible that the subendocardial ischemic contributed to the apparently larger stress ESV in patients with CAD [10].
TID ratio and ΔEF were long believed to be crucial ancillary findings that could indicate the presence of CAD but our result indicated otherwise. TID is a phenomenon in which the left ventricular cavity after stress appears larger than the cavity at rest. It was thought to be a result of subendocardial ischemic or myocardial stunning. Increased TID ratio in patients with perfusion defects was proven to be a good indicator of severe CAD. However, there has been conflicting evidence on the utility of TID ratio in patients with normal perfusion images. Some studies suggest that increased TID ratio was a strong predictor for CAD in patients without apparent perfusion defect [11, 12]. Others found that high TID ratio in the otherwise normal perfusion images did not correlate with CAD [13, 14]. Many researchers speculated that the inconsistency resulted from the differences in stress protocols, radiotracers, and imaging techniques. Because a meta-analysis found that diagnostic values of TID ratio was remarkably poorer in vasodilator stress compared to those in exercise stress (AUC = 0.78. vs AUC = 0.92) [10], we suspected that the exclusive use of vasodilator stress in our cohort could be one culprit for the poor diagnostic performance of TID ratio in our study. The interpretation of ΔEF was similarly controversial and several cutoffs have been proposed to indicate abnormal EF response to stress. A publication form the International Atomic Energy Agency (IAEA) stated that the post stress LVEF that was at least 10 EF units lower than rest LVEF indicated abnormal EF response [8] but a more recent study proposed a smaller threshold at 5 EF units [9]. Because the EF response gradually faded after stress and it was found that the normalization of post stress LVEF could occur as early as 13 min [15]. The explanation for the poor diagnostic performance of ΔEF in our study was most likely the 1-h-long interval before the image acquisition.
Several limitations of this study were noticeable. First, it should be reminded that perfusion defect on perfusion images is still the most crucial findings for diagnosis of CAD from the MPI [16]. Although up to 7.2% of patients with normal perfusion images had CAD [12], correctable causes of false negative perfusion images must be excluded before the perfusion be considered normal. These causes include, but not limited to, unrecognized ingestion of caffeine-containing products, insufficient coronary vasodilatation, motion artifacts, and attenuation artifacts [2].
Secondly, our sample size was rather small, which could hamper the precision of our result and reduce our ability to detect the differences in diagnostic values among the examined parameters. The limitations resulted from the stringent need for the ICA, which was not frequently done in patient with normal perfusion images. Inclusion of alternative reference standard, such as clinical information after sufficient follow up may increase the number of eligible subjects and improve the precision of future study.
Thirdly, due to the small sample size, we could not provide gender specific cutoffs for the gated SPECT parameters. There has been plenty of evidences showing that the reference ranges for gated SPECT parameters could differ significantly between sexes and ethnicities. A study from Germany established normal ranges of 41–47 mL, 102–111 mL, and 51–55 EF units for ESV, EDV and LVEF in women and 68–89 mL, 150–170 mL, and 45–50 EF units for the respective values in men [17]. In contrast, the Japanese Nuclear Society working group published normal ranges of 1–30 mL, 32–86 mL, and 60–90 EF units for ESV, EDV and LVEF in women and 8–47 mL, 42–120 mL and 54–80 EF units in men [18]. Both the German and the Japanese studies demonstrated significant different gated SPECT parameters between women and men, which was consistent with many other studies [19]. With larger sample size in future studies, we hope to establish sex specific reliable cutoffs of the gated SPECT parameters, which will further improve the accuracy of MPI in the detection of CAD.