Because of the novelty of the technique, current clinical evidence on dynamic myocardial perfusion CT is limited to single-center experiences [1
], which have shown the feasibility of the technique and its accuracy for identifying hemodynamically significant coronary artery stenosis with a variety of reference standards, such as cardiac catheterization with or without functional flow reserve measurements, cardiac MRI, and nuclear myocardial perfusion imaging (MPI). The available data suggest that CT has potential for evolving into a clinically useful imaging modality for the assessment of myocardial perfusion. The unique advantage of myocardial perfusion CT is the potential of combining it with CCTA so that coronary artery stenoses and their effects on myocardial perfusion can be assessed comprehensively and noninvasively with a single modality.
A large body of evidence has accumulated on the prognostic value of SPECT [15
] in the evaluation of myocardial perfusion, and results of a number of studies have established a similar role of PET [16
] and cardiac perfusion MRI [17
]. To our knowledge, however, no prognostic data are available on myocardial perfusion CT. The purpose of this investigation was to assess the prognostic value of stress dynamic myocardial perfusion CT for future major adverse cardiac events (MACE) in a multicenter population of patients at intermediate to high cardiovascular risk.
Materials and Methods
The population for this study was 242 patients enrolled in a multicenter registry. Baseline findings on a portion of these patients (146 and 137 patients in two studies [18
]) have been previously reported. However, neither outcome in general nor the predictive value of CCTA or myocardial perfusion CT for future MACE was assessed in those studies.
In the multicenter registry, we pooled data on patients from six centers in Asia, Europe, and North America who had undergone CCTA and dynamic myocardial perfusion CT between November 2009 and July 2011 as part of single-center studies. Patients were eligible if they had suspected or known CAD. Patients were not considered for study inclusion if they had contraindications to CT or to administration of iodinated contrast medium or adenosine. The respective research study protocols had been approved by the institutional review boards of all participating institutions, and written informed consent had been obtained from all research subjects before enrollment. The dynamic myocardial perfusion CT examinations were investigational. The results of these examinations were not reported or disclosed to clinicians and were not used to guide management decisions.
Collection of Clinical Risk Factor Baseline Data
At myocardial perfusion CT, the following demographic parameters and baseline clinical risk factors were collected for all patients: age; sex; history of diabetes, hypertension, or dyslipidemia; history of smoking (current or former smoker); personal history of CAD (previous myocardial infarction or previous angiographically documented clinically significant CAD); and family history of CAD. The total number (range, 0–6) of these cardiac risk factors was determined.
CT Image Acquisition
All image acquisitions were performed with a second-generation dual-source CT system (Somatom Definition Flash, Siemens Healthcare). Initially, CCTA was performed at rest after IV administration of 50–80 mL of iodinated contrast agent (concentration, 300–370 mg I/mL). Depending on the patient's heart rate and rhythm, CCTA was performed with retrospective ECG gating (patients with arrhythmia), prospectively ECG-triggered sequential acquisition (patients in sinus rhythm with a heart rate > 60 beats/min), or prospectively ECG-triggered high-pitch spiral acquisition (patients in sinus rhythm with a heart rate ≤ 60 beats/min).
After 3 minutes of continuous adenosine administration at an infusion rate of 140 μg/kg/min, myocardial perfusion CT acquisition was initiated. The scan delay was determined with a test bolus injection and set 4–6 seconds before arrival of the contrast agent in the aorta. Data acquisition was performed for 30 seconds with both x-ray tubes set at 100 kV, gantry rotation time of 0.28 seconds, and tube current of 300 mAs per rotation. Perfusion imaging was performed with an ECG-triggered shuttle mode in which the table shifted between two z-positions of the heart. Image acquisition was performed in systole 250 ms after the R wave. With a defined detector width of 38 mm, and 10% overlap between the two imaging positions, the coverage of the acquisition was 73 mm. A total of 14–15 image volumes during myocardial passage of the contrast bolus were acquired for each patient. Myocardial perfusion CT studies were contrast enhanced with 40–50 mL of iodinated contrast agent (concentration, 300–370 mg I/mL) administered at a flow rate of 4–7.5 mL/s to ensure an iodine delivery rate of 1.5–2.25 g I/s.
Analysis of Coronary CT Angiographic Studies
Analysis of all CCTA and myocardial perfusion CT data was performed in a central core laboratory at one site. All CCTA data were reconstructed with a section thickness of 0.75 mm and 0–5 mm increments with a vascular reconstruction kernel (B26F). Two experienced readers (3 and 6 years of experience in cardiac CT interpretation) independently evaluated all CCTA studies. For CAD evaluation, vessel-based analysis was performed to evaluate the presence of stenosis in the left anterior descending, left circumflex, and right coronary arteries. The left main coronary artery was included with the left anterior descending. Coronary artery dominance (right, left, or balanced) was recorded. The degree of stenosis was assessed with multiplanar reconstructions and curved multiplanar reconstructions along the vessel centerline (Circulation, Siemens Healthcare). Vessels were visually assessed as to whether they harbored stenosis with 50% or greater luminal narrowing. If the lumen of a given vessel was not evaluable because of the presence of heavy calcifications or motion artifacts, it was rated nondiagnostic. For the data analysis, these vessels were considered to have 50% or greater stenosis according to an intention-to-diagnose approach to avoid the risk of overlooking any stenoses possibly present in nondiagnostic vessels. Myocardial regions were then matched to the supplying vessels. Discordant findings were resolved in a final consensus reading with a third observer (more than 15 years of experience in cardiac CT). The results of this consensus interpretation were used for statistical analysis.
Qualitative Assessment of Myocardial Perfusion CT Studies
Qualitative interpretation of myocardial perfusion CT studies was performed independently by the two readers. Each vessel territory was assessed for the presence of perfusion defects. For this purpose, reconstructed perfusion datasets were viewed with dedicated software (VPCT body application, Syngo MMWP workstation, Siemens Healthcare), which analyzes and displays parameters of the myocardial blood supply, including myocardial blood flow and myocardial blood volume, as parametric maps. Visual analysis of myocardial perfusion CT data were performed in side-by-side correlation with CCTA studies to directly match perfusion defects to the supplying vessels. To ensure correct identification of vascular territories, angiographic visualization of vessel dominance was used to decide which vessel supplied the inferior and inferoseptal regions. A vessel territory was considered to harbor a perfusion defect if an area of decreased myocardial blood flow, volume, or both was present that was consistent with an ischemic cause and not suggestive of an artifact (whether or not ≥ 50% stenosis was detected in the supplying vessel at CCTA). Myocardial perfusion CT studies were viewed in axial, coronal, and sagittal orientation for reliable differentiation of artifacts from true perfusion defects. Any discrepancies between the two readers were resolved by consulting the third experienced reader. A myocardial perfusion CT stress test result was classified as positive if one or more territories had a perfusion deficit.
Prospective follow-up was conducted at each site by chart review and telephone interviews with the patient or a close relative 6, 12, and 18 months after imaging. Outcome data were collected from a standardized questionnaire. Reported clinical events were confirmed by contact with the patient's primary care physician or the admitting hospital. The occurrence of MACE was recorded at each time point. If an event was registered during follow-up but the exact time of the event could not be determined, it was assumed that the event occurred at the end of the 6-month time interval since the last follow-up examination because this would result in the most conservative risk estimates. MACE were defined as cardiac death, nonfatal myocardial infarction, unstable angina requiring hospitalization, or revascularization (percutaneous coronary intervention or coronary artery bypass grafting).
Continuous data (age) were tested for normal distribution by Kolmogorov-Smirnov test and found to be not normally distributed. Thus, age was presented as median and interquartile range (25th–75th percentile) and compared by nonparametric Mann-Whitney test. Categoric data were displayed as absolute frequencies and proportions. The chi-square test was used to compare the frequency distribution of binary data between groups. The chi-square test for trend was used to compare the distribution of the number of affected vessels or territories. Patients were divided into groups according to the number of vessels with 50% or greater stenosis and according to the number of vascular territories with perfusion defects.
Cumulative event rates stratified by CCTA and myocardial perfusion CT features were analyzed with Kaplan-Meier survival curves fitted for MACE with patient data censored after the first event. The log-rank test was used to test for significant differences in cumulative event rates between groups. The univariate Cox proportional hazards regression model was used to analyze the predictive value of findings at CCTA and myocardial perfusion CT for MACE during follow-up. Adjusted Cox proportional hazards models were used to evaluate whether the predictive value of CCTA and myocardial perfusion CT findings was independent of age, sex, and clinical risk factors and whether findings at myocardial perfusion CT had predictive value independent of findings at CCTA. The first model was adjusted for sex and number of clinical risk factors (hypertension, hyperlipidemia, diabetes, smoking history, history of CAD, family history of CAD) with block entry of all variables. The second model additionally incorporated the number of vessels with 50% or greater stenosis at CCTA. All statistical analyses were performed with MedCalc Statistical Software (version 12.7.2, MedCalc Software). Two-sided p < 0.05 was considered to indicate statistical significance.
The emerging technique of stress myocardial perfusion CT has the potential to supplement the well-established clinical utility of CCTA with functional assessment of myocardial perfusion. Our data derived from a multicenter registry show that myocardial perfusion CT findings are predictive of adverse events and have incremental prognostic value over clinical risk factors and assessment of coronary artery stenosis with CCTA. Our primary findings are that the presence of at least one perfusion defect at myocardial perfusion CT is an independent prognostic marker for future MACE and that the risk of MACE increases with the number of affected myocardial territories. Our data show that a patient with any perfusion defect detected with myocardial perfusion CT is at approximately 2.5-fold increased risk of experiencing a future cardiac event compared with patients with normal myocardial perfusion CT findings. A positive myocardial perfusion CT study increases the risk of MACE twofold even after adjustment for conventional risk factors and CCTA findings. The risk increases approximately 1.4-, 3.4-, and 4.8-fold among patients with perfusion defects in one, two, and three territories.
The prognostic power of nuclear MPI is supported by a large body of evidence accumulated over 3 decades of experience with the technique [16
]. The presence and extent of ischemia at SPECT have been found to be highly predictive of cardiac events [20
]. Furthermore, the prognostic utility of nuclear MPI is incremental to clinical assessment, treadmill stress testing, and coronary catheter angiography [21
]. Cardiac PET is a more recently developed technique, and evidence on the prognostic value of PET is accumulating rapidly [24
]. Compared with SPECT, PET has a number of advantages, including superior spatial and temporal resolution and potential for quantifying absolute myocardial blood flow [16
]. In this regard, dynamic myocardial perfusion CT is similar to PET in enabling absolute quantification of myocardial blood flow. Absolute quantification of myocardial blood flow with PET allows calculation of myocardial flow reserve, defined as the ratio between myocardial blood flow at peak and flow at rest. Initial studies have shown the prognostic value of myocardial flow reserve at myocardial perfusion PET [30
]. In our study, we focused on visual rather than quantitative analysis of perfusion datasets, because this appears more suitable for clinical routine and reflects the most common approach of assessing myocardial perfusion in clinical practice with SPECT or cardiac MRI.
In line with the prognostic results of SPECT and PET, cardiac MRI, either with vasodilator or dobutamine stress, has been found to be an excellent prognostic test. Patients with known or suspected CAD and stress cardiac MRI findings positive for ischemia are at 5% annual risk of experiencing a future cardiac event, whereas patients with negative results are at less than 1% risk [34
]. The prognostic results of both nuclear and cardiac MRI functional testing are comparable to the prognostic value of CCTA findings reported in the Coronary CT Angiography Evaluation for Clinical Outcomes: An International Multicenter Registry (called the CONFIRM registry) [35
]. The presence of nonobstructive disease confers 1% annual risk, and high-risk CAD increases the risk to 5%. Moreover, CCTA findings of CAD severity have independent prognostic value over conventional clinical predictors [35
Combining CCTA and nuclear MPI has been found to improve risk stratification over that with either test alone [38
]. This combination of morphologic and functional testing is particularly valuable because findings can be directly correlated, as in hybrid imaging combining CT with SPECT or PET, which has been found to improve risk stratification of patients with known or suspected CAD [39
]. Patients with evidence of stenosis at CCTA and a matched reversible perfusion defect at SPECT are at highest risk of future MACE with an annual event rate of 6.0%. In unmatched patients this risk is lower at 2.8%, and further reduction to 1.3% is observed among patients without evidence of ischemia or obstructive CAD [39
A 2013 study [40
] showed that the prognostic value of myocardial perfusion SPECT strongly depends on the appropriate use of this test. When used within established appropriate use criteria, SPECT has high predictive value for future MACE, but the findings are not predictive of MACE if the technique is used inappropriately. Likewise, the clinical utility of myocardial perfusion CT as a prognostic tool will greatly depend on appropriate patient selection. Because myocardial perfusion CT is currently a research application and no accepted criteria for its clinical use exist, the development of such guidelines will be essential to translate the prognostic value of myocardial perfusion CT reported in our study into a clinically useful test with diagnostic and prognostic value.
We chose a qualitative visual analysis for the detection of perfusion defects in myocardial perfusion CT in this study. This approach is comparable to how the established techniques of SPECT and myocardial perfusion MRI are commonly analyzed and reported. Other authors have used quantitative measurements of myocardial blood flow to define myocardial segments with perfusion defects. However, there is currently no consensus on how ROIs are best placed and which cutoff is most suitable for discriminating normal from ischemic myocardium. Studies [19
] have shown that relative measurements of myocardial blood flow are more suitable than absolute measurements, which introduces the additional difficulty of defining remote, presumably normal myocardium.
We defined perfusion defects as areas of decreased myocardial blood flow or myocardial blood volume that are consistent with an ischemic cause and not suggestive of an artifact. In theory, one could expect cases of compensated stenosis causing a decrease in myocardial blood flow with no change in myocardial blood volume. In practice, we found that perfusion defects on myocardial blood flow maps almost always represented corresponding abnormalities on myocardial blood volume maps, although the precise extent of the perfusion defect could differ slightly. This finding is consistent with animal data showing that both myocardial blood flow and myocardial blood volume are decreased in both infarcted and ischemic myocardium [42
]. Nevertheless, the relative changes in myocardial blood flow and myocardial blood volume in relation to the degree of coronary artery stenosis merit further investigation.
Our data relied on dynamic CT measurements of myocardial perfusion. Other authors have investigated single-shot approaches to myocardial perfusion CT, which capture a snapshot of iodine distribution within the myocardium during the first-pass inflow of contrast medium. The multicenter CORE320 study [4
] showed that adding single-shot myocardial perfusion CT to CCTA significantly improved accuracy for identifying flow-limiting coronary stenosis. CT-derived fractional flow reserve has been investigated as an alternative to determining the hemodynamic relevance of coronary stenosis. This approach uses computational fluid dynamics to calculate the fractional flow reserve based on the CCTA dataset. Several studies of CT-derived fractional flow reserve have shown high diagnostic accuracy [6
Several limitations of our study merit consideration. Our data relied on the acquisition of dynamic perfusion datasets and should not be extrapolated to static approaches to assessment of myocardial blood supply [5
]. Furthermore, all examinations were performed with dual-source CT and are thus derived from a single CT vendor. The results may not apply to other vendors with different technical implementations of dynamic myocardial perfusion CT. We used qualitative evaluation of myocardial perfusion CT datasets and per-territory analysis. Further studies may clarify whether a quantitative approach and per-segment analysis further improve risk stratification. In addition, most of the MACE in our study were comparatively soft events (unstable angina, revascularization), which introduces the potential for bias. It is important to consider, however, that the results of myocardial perfusion CT were purely investigational and not disclosed to clinicians. If present, any bias in the outcomes would likely favor CCTA over myocardial perfusion CT and underestimate the incremental value of myocardial perfusion CT. Nevertheless, our results must be verified in large-scale prospective trials.
Despite the limitations our study provides initial evidence that in a population at intermediate to high cardiovascular risk, myocardial perfusion CT has incremental predictive value for future MACE beyond clinical risk factors and assessment of coronary artery stenosis with CCTA.