Complementary prognostic value of stress perfusion imaging and global coronary flow reserve derived from cardiovascular magnetic resonance: a long-term cohort study

A total of 1366 consecutive patients ≥ 45 years old with suspected or known CAD who underwent comprehensive CMR between January 2009 and December 2015 at our hospital were considered for participation in this study. Known CAD was defined as having a history of coronary revascularization or myocardial infarction (MI). Exclusion criteria were: (1) non-ischemic cardiomyopathy; (2) valvular disease; (3) congenital heart disease; (4) non-analyzable images in phase-contrast cine CMR; or (5) early revascularization within 90 days after CMR. The institutional review board approved this retrospective study and waived the need to obtain individual consent.

All CMR images were acquired using a 1.5T or 3T scanner (Achieva 1.5T/Achieva 3T; Philips Healthcare, the Netherlands). A 32 element cardiac-surface coil wase used for radiofrequency reception. Each CMR examination consisted of cine CMR, stress and rest perfusion CMR, stress and rest phase-contrast cine CMR of the CS, and late gadolinium enhancement (LGE) CMR. To assess left ventricular (LV) volumes, systolic function and LV mass, breath-holding cine CMR was acquired with a segmented balanced steady-state-free precession (bSSFP) sequence in the short-axis planes covering the entire LV. Scan parameters for the 1.5T scanner were: slice thickness, 10 mm; cardiac phases, 20; repetition time (TR), 3.2 ms; echo time (TE), 1.6 ms; flip angle (FA), 55°; field of view (FOV), 35 × 35 cm; acquisition matrix, 192 × 192; and reconstruction matrix, 256 × 256. Scan parameters for the 3T scanner were: TR, 2.8 ms; TE, 1.4 ms; FA, 55°; FOV, 35 × 35 cm; acquisition matrix, 176 × 308; and reconstruction matrix, 352 × 352.

Stress perfusion CMR consists of 4 or 3 short-axis slices acquired every 2 cardiac cycles or every cardiac cycle with a saturation-recovery balanced bSSFP or turbo field echo (TFE) sequence for 1.5T scanner (TR, 3.0 ms; TE, 1.5 ms; FA, 40°; FOV, 36 × 32 cm; acquisition matrix, 192 × 153; SENSE factor, 2; time between saturation preparation pulse and center of k-space acquisition, 200 ms; duration of image data acquisition, 211 ms; slice thickness section, 8 mm) or 3T scanner (TR, 2.9 ms; TE, 1.4 ms; FA, 20°; FOV, 32 × 29 cm; acquisition matrix, 224 × 131; SENSE factor, 2.5; time between saturation preparation pulse and center of k-space acquisition, 110 ms; duration of image data acquisition, 141 ms; slice thickness section, 10 mm). For stress perfusion CMR, dynamic imaging was initiated 3 min after starting adenosine triphosphate infusion (160 mg/kg/min) and continued for 1 min. Gadoterate meglumine (Gd-DOTA; Guerbet Japan, Tokyo, Japan) was injected at a dose of 0.03 mmol/kg and a flow rate of 4 mL/s, followed by a 20-mL saline flush for stress perfusion CMR. Rest perfusion CMR was performed with an identical set-up at 10 min following stress perfusion CMR.

CS blood flow during stress and at rest were measured by breath-hold phase-contrast cine CMR for 1.5T scanner (TR, 8.6 ms; TE, 5.6 ms; FA, 15°; FOV, 36 × 26 cm; acquisition matrix, 192 × 112; 16 phases per cardiac cycle; and velocity encoding, ± 70 cm/s) or 3T scanner (TR, 4.8 ms; TE, 3.1 ms; FA, 10°; FOV, 25 × 21 cm; acquisition matrix, 176 × 97; 25 phases per cardiac cycle; and velocity encoding, ± 80 cm/s) as shown in Fig. 1A. Stress and rest phase-contrast cine CMR images were acquired during breath-holding with shallow inspiration (15 s) on an imaging plane perpendicular to the CS immediately after stress and rest perfusion CMR [7]. The imaging position was carefully chosen so that blood flow in the CS was measured on the slice as close as possible to the orifice to the right atrium and the CS was visible throughout the cardiac cycle.

LGE CMR was acquired in the same LV short-axis planes as used for cine CMR using a three-dimensional inversion recovery TFE sequence 5–10 min after intravenous administration of Gd-DOTA with a cumulative dose of 0.15 mmol/kg for 1.5T scanner (TR, 3.8 ms; TE, 1.2 ms; FA, 15°; FOV, 40 × 36 × 5 cm; acquisition matrix, 224 × 156 × 5; reconstructed matrix, 256 × 256 × 10; SENSE factor, 2; TFE factor, 24) or 3T scanner (TR, 4.6 ms; TE, 2.2 ms; FA, 15°; FOV, 38 × 34 × 5 cm; acquisition matrix, 240 × 192 × 5; reconstructed matrix, 384 × 384 × 10; SENSE factor, 3; TFE factor, 33).

CMR images were analyzed using dedicated software (cvi⁴²; Circle Cardiovascular Imaging Inc., Calgary, Canada). At end-diastole and end-systole, endocardial LV borders were manually traced in contiguous short-axis cine CMR images covering the apex to mitral valve plane to calculate LV end-diastolic volume (LVEDV) and LV end-systolic volume (LVESV) and ejection fraction (LVEF). After tracing epicardial LV borders at end-diastole, LV mass was calculated as the sum of the myocardial volume at end-diastole multiplied by the specific gravity (1.05 g/mL) of myocardial tissue.

Two independent observers (one with 20 years and one with 8 years of experience in CMR), blinded to clinical information and other imaging results, evaluated stress perfusion CMR and LGE CMR (Fig. 1B). Disagreement between the two observers was settled by a consensus reading. Perfusion defects were defined as hypo-enhanced regions that persisted for at least three phases after peak contrast enhancement and followed a coronary distribution. Inducible ischemia was defined as a stress perfusion defect in a segment without LGE or larger than LGE. Myocardial ischemia was defined when inducible ischemia was present in > 10% of LV myocardium, namely in ≥ 3 of 32 subsegments (endocardial and epicardial sectors for each of the 16 segments) [10]. On LGE images, presence or absence of LGE was visually assessed in 16 segments.

CS blood flow (mL/min) was calculated by integrating the product of cross-sectional area and mean velocity in the CS from the 16 images across the cardiac cycle. Global CFR was defined as CS blood flow during stress divided by CS blood flow at rest. Contours of the CS were manually traced on magnitude images at each cine frame. The traced region of interest was applied on corresponding phase-difference image, and cross-sectional area and mean velocity were recorded. Impairment of global CFR was defined as showing a lower tertile in global CFR.

Follow-up information was collected through a review of hospital records or telephone interviews with the interviewer blinded to CMR results between 2019 and 2020. If the same patient had more than one CMR between 2009 and 2015, the first CMR was used for this cohort study and the patient was followed up from that time. Major adverse cardiac events (MACE) comprised cardiovascular death, non-fatal MI, unstable angina, hospitalization for heart failure, stroke, ventricular arrhythmia and late (> 90 days after CMR) revascularization. Severe events were defined as MACE other than late revascularization. Hard events were defined as cardiovascular death and non-fatal MI. Cardiovascular death was defined as death caused by acute MI, ventricular arrhythmias, heart failure or stroke. Non-fatal MI was defined as prolonged angina accompanied by new electrocardiographic (ECG) abnormalities and increased cardiac biomarkers. Unstable angina was defined as new-onset, worsening, or rest angina requiring hospital admission.

Continuous variables are presented as the mean ± standard deviation and categorical variables are expressed as frequency (percentage). Univariate Cox proportional hazards regression analysis was performed to identify potential predictors of MACE among coronary risk factors and CMR imaging parameters and multivariate Cox proportional hazards regression analysis was performed using the “enter” method for variables from univariate analysis showing values of p < 0.05 to determine independent predictors of MACE These Cox analyses were performed in all patients and subsets of patients with known/suspected CAD. The results are reported as hazard ratios (HRs) with 95% confidence intervals (CIs). The incremental value of a predictor over another predictor was evaluated by calculating global chi-square values. Net reclassification improvement (NRI) with the addition of global CFR to coronary risk factors plus ischemia was calculated, and continuous NRI and integrated discrimination improvement were estimated. Kaplan–Meier curves were used to estimate event-free rates for MACE, severe events or hard events in all patients and subsets of patients with known/suspected CAD or with 1.5T or 3T CMR and differences between time-to-event curves were compared using the log-rank test. Annualized event rates were calculated by dividing 5-year Kaplan–Meier event rates by 5. Interobserver agreement was evaluated using Bland–Altman analysis, intraclass correlation coefficient (ICC), and quadratic-weighted kappa analysis by two observers (one with 8 years and one with 7 years of experience in CMR) in a subset of 40 randomly selected patients. Two-sided p-values < 0.05 were considered statistically significant. All analyses were performed using the SPSS (version 23.0; Statistical Package for the Social Sciences, International Business Machines, Inc., Armonk, New York, USA) and the R statistical package (version 3.4.4, R Foundation for Statistical Computing, Vienna, Austria).

The study has several limitations. First, this was a single-center, retrospective study. A prospective multi-center study is warranted to confirm the results of this study. Second, this study used both 1.5- and 3T CMR scanners, which differed in several respects. It is of importance to note that the CMR parameters between 1.5T and 3T were not consistent, but when the analyses were performed for 1.5T and 3T separately, the results were not significantly different between 1.5T and 3T and were similar to those of the overall analysis. Third, the present study did not use quantitative perfusion mapping by CMR, which also allows assessment of both local and global CFR. Quantitative stress perfusion CMR might be able to show results similar to those of this study in a single test. However, stress perfusion CMR during the study period did not include quantitative method. The prognostic value of quantitative stress perfusion CMR may be a subject for future research.