A retrospective study was conducted including 248 patients referred to the PET/CT Unit at the National Autonomous University of Mexico in Mexico City by their cardiologist for the evaluation of suspected myocardial ischemia. Demographic and clinical data were retrieved through the electronic patient archive system. Several clinically relevant confounders were included in this study: sex, age, body mass index (BMI), hypertension, dyslipidemia, type 2 diabetes mellitus (DM), smoking habit, and semiquantitative perfusion scores (see ahead). Ethical approval was obtained for the conduction of the study from the local institutional ethics committee.
PET images were acquired on a whole-body 64-slice PET/CT scanner (Biograph True Point; Siemens Medical Solutions). PET data were acquired in 3-D list mode. Patients were studied after an overnight fast, and all refrained from caffeine-containing beverages or theophylline-containing medications for 24 hours before the study. Myocardial perfusion was assessed at rest and during vasodilator stress with adenosine and Nitrogen-13 ammonia as the perfusion radiotracer. Two CT-based transmission scans (120 kVp; 20-30 mA; helical scan mode with a pitch of 1.35) were obtained before the rest perfusion studies and after the stress perfusion studies with normal breathing for correction of photon attenuation for PET. The registration of the CT attenuation map with the PET images was verified visually by an experienced technologist and alignment was corrected if necessary by manual 3-D translation. Regional myocardial perfusion was first assessed during rest using 740 MBq of Nitrogen-13 ammonia. Rest imaging extended for 10 minutes and began a few seconds before the injection. The radiotracer was administered as a single peripheral intravenous bolus (3-5 seconds) followed by a 10-mL saline flush. Thirty minutes later, a pharmacological stress test was performed, beginning with the injection of adenosine during a 6-minute period (140 mg/kg per minute). A second dose of 740 MBq of Nitrogen-13 ammonia was injected at the third minute of the adenosine infusion. Stress images acquisition was started a few seconds before the radiotracer injection. Sixteen dynamic frames were reconstructed (twelve 10-s, two 30-s, one 1-minute, and one 6-minute frames, for a total of 10 minutes). Standard reconstruction (2-D attenuation-weighted OSEM) was used with 3 iterations and 14 subsets and 3-D postfiltering with a 5-mm Gaussian kernel filter. Transverse data were reformatted to a 168×168×47 matrix with 2 mm pixels for each dynamic frame.
Perfusion Data Analysis
Semiquantitative Myocardial Perfusion
Images were interpreted semiquantitatively using the standard American Heart Association 5-point scoring system 17 and traditional metrics were documented: summed difference score (SDS), summed stress score (SSS), and summed rest score (SRS). SRS (in a fixed perfusion defect) was considered as a measure of the extent and severity of a previous MI.
Quantitative Myocardial Perfusion
Left ventricular contours and input function regions were obtained automatically allowing minimal observer intervention in QPS software package (Cedars-Sinai, Los Angeles, CA, USA). Dynamic myocardial samples were obtained from the polar map by analyzing all time frames within the fixed left ventricular contour boundaries. Quantitative rest and stress MBF values (mL/g/minute) were computed for each sample on the polar map as described previously18 using a previously described 2-tissue compartment pharmacokinetic model for Nitrogen-13 ammonia.19 MPR was calculated as the ratio between stress MBF and rest MBF (making it a unitless variable). Rest MBF was corrected accordingly for the rate-pressure product (RPP).20 The global MPR and stress MBF were calculated within the whole left ventricular region (as defined by the left ventricle long-axis plane) as parameters of interest for our analysis.
Ventricular Function Data Analysis
ECG-gated stress images were analyzed with the QGS software package (Cedars-Sinai, Los Angeles, CA, USA).21 Short-axis images were processed and ventricular edges and cavity volumes were calculated for each of the re-binned 8 dynamic frames that were reconstructed for the average cardiac cycle. The algorithm for determining edges and calculating volume has been described previously.22
Systolic function was evaluated through left ventricular ejection fraction (LVEF) (% of the end-diastolic volume ejected) using the average time-volume curves. Next, diastolic function was evaluated with the mean filling rate during the first third of diastole (MFR/3) as a surrogate marker. MFR/3 was obtained from the first derivative of the left ventricular time-volume curve and expressed in end-diastolic volumes per second (EDV/s) as it is adjusted to the ventricle dimensions on a patient-by-patient basis.23,24
Finally, ventricular synchrony was evaluated through Entropy, which is a measurement of uniformity of the onset and progression of wall motion throughout the cardiac cycle.25 Entropy was expressed as a percentage (ranging from 0 to 100) with greater percentages reflecting a lesser uniformity of ventricular fiber contraction.14 It was obtained by phase analysis (embedded in the QGS software package [Research Edition, PET Processing plugin, Cedars-Sinai]). Entropy was used because it constitutes an expression of (dys)synchrony that is not dependent on phase similarity25 as other described parameters (Bandwidth and Standard Deviation).
All continuous variables are described as mean ± standard deviation, and categorical variables are expressed as frequencies with percentages. Between-group comparisons were made using independent samples t tests.
The biserial correlations (using Pearson’s correlation coefficient) between LVEF, MFR/3, and Entropy were evaluated. Next, a general multivariate analysis of covariance (MANCOVA) was performed including sex, age, BMI, hypertension, dyslipidemia, type 2 DM, smoking habit, SRS, SSS, stress MBF, and MPR in the model as the independent (i.e. predictive) variables, and LVEF, MFR/3, and Entropy as the dependent (i.e. outcome) variables. The independent significance of the included predictors was tested using Pillai’s trace criterion with an approximate F statistic. Further, the multivariate analyses were repeated including the same independent and dependent variables for patients with and without imaging evidence of a previous MI. Additionally, the effect sizes for the predictors (η2) are reported and graphically depicted.
Finally, supplementary analyses were performed in both female and hypertensive patients without a previous MI.
All statistical analyses were performed with SPSS (Released 2013. IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY: IBM Corp., USA). A P-value of <0.05 was considered statistically significant.