JACC: Cardiovascular Imaging
Comparison of Positron Emission Tomography Measurement of Adenosine-Stimulated Absolute Myocardial Blood Flow Versus Relative Myocardial Tracer Content for Physiological Assessment of Coronary Artery Stenosis Severity and Location
Author + information
- Received March 24, 2009
- Revision received April 6, 2009
- Accepted April 14, 2009
- Published online June 1, 2009.
Author Information
- Mohammad M. Hajjiri, MD⁎,
- Marcia B. Leavitt, BA⁎,
- Hui Zheng, PhD†,
- Amy E. Spooner, MD⁎,
- Alan J. Fischman, MD, PhD‡ and
- Henry Gewirtz, MD⁎,⁎ (hgewirtz{at}partners.org)
- ↵⁎
Reprint requests and correspondence:
Dr. Henry Gewirtz, Cardiac Unit/Yawkey 5E, Massachusetts General Hospital, 55 Fruit Street, Boston, Massachusetts 02114
Abstract
Objectives This study tests the hypothesis that absolute measurement of adenosine (Ado)-stimulated myocardial blood flow (MBFado) is superior to measurement of relative tracer uptake for identification of hemodynamically significant coronary artery disease (CAD).
Background Positron emission tomography measurement of absolute myocardial blood flow (MBF) (13N-ammonia) with Ado has the capability to more accurately assess hemodynamic severity of CAD than measurement of relative tracer content (TC) (nCi/ml) during Ado, which by definition depends on at least 1 normal zone to which others are compared.
Methods A total of 27 patients (20 male, 58 ± 11 years, mean ± SD) with known or suspected CAD and 21 normal subjects (13 male, 38 ± 10 years) were studied. Parametric (K1) MBF images and TC sum images were analyzed. A stenosis ≥70% defined significant CAD. The receiver-operator characteristic curve (ROC) analysis area under the curve (AUC) compared MBF and TC results. Cut-point analysis for sensitivity, specificity, and accuracy showed the best MBF criteria for CAD as MBFado <1.85 ml/min/g and the best TC as <70% maximum. The myocardial blood flow reserve ratio (MBFR) (optimal <2.0×) also was studied.
Results The ROC analysis of PET parameters showed that MBFado was superior to <70% maximum uptake for CAD detection (n = 144 vessels; AUC 0.900 vs. 0.690, respectively, p < 0.0001) and was marginally greater than MBFR (0.856; p = 0.10). For CAD cut-point analysis, MBFado accuracy exceeded TC (0.84 vs. 0.72, respectively, p = 0.005), as did sensitivity (0.81 vs. 0.48, respectively; p = 0.001). Specificity of MBFado for CAD classification (0.85) was comparable to TC (0.82; p = NS). Sensitivity, specificity, and predictive accuracy for MBFR were 0.62, 0.85, and 0.79, respectively. The difference in specificity was not significant versus MBFado. However, MBFado was more sensitive than MBFR (p = 0.01). The difference in predictive accuracy was borderline (p = 0.06) in favor of MBFado.
Conclusions Measurement of Ado-stimulated absolute MBF is superior to relative measurement of myocardial tracer retention for identification of CAD and can be accomplished with a single MBFado measurement.
Prior studies have shown that retention of 13N-ammonia in positron emission tomography (PET) myocardial perfusion scans may not accurately reflect the absolute level of myocardial blood flow (MBF) at the time of tracer injection (1,2). A widely accepted tracer kinetic model for obtaining K1 (1,3,4), a measure of MBF (ml/min/g), identifies back diffusion of the tracer into blood (k2) and incorporation of 13N-ammonia into glutamine (k3) as potential causes of disparity between initial myocardial uptake (i.e., K1) and equilibrium distribution of the tracer. Tracer retention (k3) has been shown by others to be an indicator of myocardial viability (2). Similar considerations apply to 82Rb, whose initial myocardial uptake may differ from retention at later time points depending on myocardial viability (5). Accordingly, the present study was designed to test the hypothesis that absolute measurement of MBF by the PET 13N-ammonia technique is superior to relative measurement of myocardial tracer retention for identification of hemodynamically significant coronary artery disease (CAD) (≥70% lumen diameter stenosis). We also tested the hypothesis that a single measurement of absolute MBF under conditions of adenosine (Ado) stress is adequate for this purpose and thereby eliminates the requirement for a 2-injection study comparing rest and stress MBF, or more commonly, relative tracer distribution.
Methods
Patient/subject population
Subjects and patients were selected from a database (N = 157) studied according to the PET myocardial perfusion protocol used in our laboratory (1,3,6–11). The vast majority of patients have been reported previously in the context of other unrelated studies (1,3,6–11). We selected patients with known or suspected CAD who had had clinically indicated cardiac catheterization within 1 year of the PET study without interval change in their clinical condition (N = 27; 20 male, 58 ± 11 years, mean ± SD). Patients with coronary artery bypass graft or prior myocardial infarction were excluded. Normal volunteers (subjects, N = 21; 13 male, 38 ± 10 years) were in good health, asymptomatic, free of cardiac risk factors, and had not had cardiac catheterization. The Partners Institutional Review Board approved the study.
PET measurements of MBF and image analysis
The PET imaging was performed on a Scanditronix PC4096 (General Electric, Milwaukee, Wisconsin) whole-body tomography machine (1,3,6–11). At baseline, approximately 25 mCi of 13N-ammonia was administered intravenously over 30 s, with dynamic imaging begun just before injection.
Subsequently, radioactivity was allowed to decay for at least 30 min. Next, 2 min after starting an infusion of Ado (Adenoscan, Astellas Pharma US, Deerfield, Illinois) at 140 μg/kg/min × 6 min, dynamic data acquisition was begun, and several seconds later approximately 25 mCi of 13N-ammonia was administered.
Attenuation-corrected 13N-ammonia images were reconstructed with a filtered back projection algorithm and yielded an output resolution of 7.8 mm in the transverse plane. A region of interest placed over the left ventricular cavity was used to generate the arterial input function for the tracer kinetic model used to compute K1 (1). The K1 values were converted to MBF (ml/min/g) (3).
The K1 images were analyzed in a blinded fashion at base, mid, and distal left ventricular levels (Fig. 1). Regions of interest at each level were placed over myocardial segments of the left anterior descending artery (LAD) (septal, anteroseptal, anterior, anterolateral), left circumflex artery (LCX) (lateral, inferolateral), and right coronary artery (RCA) (inferior and inferoseptal) (3). The same regions of interest were analyzed for the summed images to measure tracer content (TC) (nCi/ml).
ROI Placement
A short-axis uptake image (base level) with ROIs drawn is shown for a study patient. The ROIs were reproduced at mid and distal LV levels. K1 images were analyzed similarly. LV = left ventricle; ROI = region of interest.
Data analysis
Definition of CAD
Coronary arteriograms were analyzed visually by an experienced invasive cardiologist blinded to PET data. Lumen diameter reduction ≥70% defined hemodynamically significant CAD.
Training and verification data sets
The list of patients and subjects was alphabetized, and every other one was assigned to a training group (n = 24) to develop optimal criteria for detection of CAD. The best criteria subsequently were applied to the verification set (n = 24 remaining patients/subjects). The PET criteria were tested for sensitivity and specificity to determine which performed best (greatest sensitivity–specificity product) for detection of CAD. Based on prior work (1,6), Ado-stimulated myocardial blood flow (MBFado) at cutoff points of <1.65 ml/min/g, <1.85 ml/min/g, and <2.00 ml/min/g in 2, 3, or 4 myocardial segments of a given coronary artery were tested (i.e., 9 combinations). The ratio of Ado-stimulated to rest MBF (myocardial blood flow reserve ratio [MBFR]) was tested at cut points of <2.0×, <2.5×, and <3.0× for 2, 3, or 4 myocardial segments of a given coronary artery (i.e., 9 combinations). Relative TC cutoff criteria of <70% maximum, <75% maximum, and <80% maximum for 2, 3, or 4 segments of a given coronary territory also were tested (i.e., 9 combinations).
Data reduction and receiver-operator characteristic curve (ROC) analysis
The PET scans had 24 segments (Fig. 1). Tracer uptake during Ado was normalized to the segment with highest concentration (nCi/ml). The PET data were averaged for each vascular territory (N = 48 × 3 = 144 vessels) to obtain a single value for each parameter. The vessels were treated as independent entities for statistical analyses. A subject-wise analysis also was performed for each parameter to determine which best classified all coronary vessels in a given individual.
Statistics
Data are expressed as mean ± SD. The McNemar test compared the significance of differences between PET parameters for prediction of CAD. Paired and unpaired Student t test and chi-square test were used as indicated. ROC curves for detection of CAD were obtained for each PET parameter. The areas under the curve (AUC) were computed by fitting a logistic regression model and were compared using a nonparametric method (12), which takes into account the correlated nature of the data. The method avoids strong assumptions on the distribution of the data. A sample size of 144 arteries provides 80% power to detect a minimum difference of 0.20 AUC units between 2 ROC curves using 2-sided test with 5% type I error (13). Values of p < 0.05 were considered statistically significant.
Results
Patient/subject population—training versus verification data sets
Training and verification data sets did not differ significantly regarding age (51 ± 16 vs. 48 ± 17, respectively; p = 0.45), sex (male/female = 16/8 vs. 17/7, respectively; chi-square = 0.0, p = NS), and distribution of known or suspected CAD versus normal subjects (CAD/normal subjects = 14/10 vs. 13/11, respectively; chi-square = 0.0, p = NS) (Table 1). Accordingly, the data sets were combined. Patient/subject clinical and hemodynamic information are shown in Table 1. None of the normal subjects were taking medications.
Patient/Subject Clinical Characteristics and Hemodynamic Parameters
ROC analysis
The ROC curves for each PET parameter for all coronary vessels are shown in Figures 2A through 2C. The AUC for MBFado (0.900) exceeded that for <70% maximum tracer uptake (0.690, p < 0.0001). The MBFado AUC was marginally greater than MBFR (0.856; p = 0.10). The MBFR AUC exceeded that of <70% maximum tracer uptake (p < 0.005).
ROC Analysis, All Vessels (N = 144)
The ROC curves for PET parameters tested for all coronary vessels are shown. (A) MBFado; (B) MBFR (MBFado/rest); (C) normalized tracer uptake (nlUPTKado). The AUC for MBFado exceeded nlUPTKado (p < 0.0001). The MBFado did not differ from MBFR (p = 0.10). AUC = area under the curve; MBFado = adenosine-stimulated myocardial blood flow; MBFR = myocardial blood flow reserve ratio; PET = positron emission tomography; ROC = receiver-operator characteristic curve; SENS = sensitivity; SPEC = specificity.
The ROC analysis for the LAD alone is shown in Figures 3A and 3B, and for the LCX and RCA combined in Figures 4A and 4B. The AUC for MBFado for LAD (0.922) exceeded that for <70% maximum tracer uptake (0.709, p = 0.03). In contrast, the AUC for MBFR (0.863) did not differ from that of <70% maximum tracer uptake (p = 0.15).
ROC Analysis, LAD Only
The ROC curves are shown for PET parameters; LAD territory. (A) MBFado; (B) normalized tracer uptake (nlUPTKado). The AUC for MBFado exceeded that of nlUPTKado (p = 0.03). LAD = left anterior descending coronary artery; other abbreviations as in Figure 2.
ROC Analysis (LCX + RCA)
The ROC curves for PET parameters; LCX + RCA territories. (A) MBFado; (B) normalized tracer uptake (nlUPTKado). The AUC for MBFado exceeded that of nlUPTKado (p = 0.02). LCX = left circumflex artery; RCA = right coronary artery; other abbreviations as in Figure 2.
The AUC for MBFado for LCX + RCA (Figs. 4A and 4B) exceeded that for <70% maximum tracer uptake (0.897 vs. 0.738, respectively, p = 0.02). The AUC for MBFR (0.864) did not differ from that of <70% maximum tracer uptake (p = 0.09).
The ROC analysis of PET parameters in subjects on (n = 19) versus off (n = 29) beta-blockers was performed. Patients on beta-blockers were older (59 ± 12 years vs. 43 ± 14 years, respectively, p < 0.001), and were more likely to have CAD (chi-square = 12.3, p < 0.005) and hypertension (chi-square = 10.5, p < 0.005) versus individuals who were not, 72% of whom were normal subjects. The group not on beta-blockers had lower systolic arterial pressure (119 ± 13 mm Hg vs. 138 ± 17 mm Hg, respectively, p < 0.001) but identical resting heart rate (64 ± 11 beats/min vs. 64 ± 12 beats/min, respectively). Thus, the beta-blocker group had slightly higher rate–pressure product (mm Hg/min) than those not taking beta-blockers (8,817 ± 2,209 vs. 7,563 ± 1,436, respectively, p = 0.04), although rest MBF did not differ (0.86 ± 16 vs. 0.76 ± 06, respectively, p = NS).
In subjects on beta-blockers, the AUC for MBFado (0.839) exceeded that for MBFR (0.720, p < 0.05) as well as <70% maximum tracer uptake (0.645, p = 0.03). In individuals off beta-blockers, the AUC for MBFado (0.925) did not differ from that of MBFR (0.932) but exceeded that for <70% maximum tracer uptake (0.699, p < 0.005).
Optimal MBF and tracer retention cut-point analysis
The best MBFado cut-point was <1.85 ml/min/g. The best MBFR was <2.0×. Thus, sensitivity, specificity, and predictive accuracy for MBFado were 0.81, 0.85, and 0.84, respectively versus 0.62, 0.85, and 0.79 for MBFR. The difference in specificity was not significant. However, MBFado was more sensitive than MBFR (p = 0.01), whereas the difference in overall predictive accuracy was borderline (p = 0.06).
The best 13N-ammonia retention criterion was <70% maximum. Thus, sensitivity, specificity, and predictive accuracy for 13N-ammonia retention criteria were 0.48, 0.82, and 0.72, respectively.
Comparison of optimal MBF versus 13N-ammonia retention criteria for CAD detection
In a subject-wise analysis requiring all 3 coronary vessels to be correctly classified, MBFado exceeded tracer uptake <70% maximum (10 of 10 discordant pairs correct by MBFado (p < 0.005). The MBFado also classed 2.52 ± 0.71 coronary arteries per individual correctly vs. 2.17 ± 0.86 by tracer uptake (p < 0.005).
Detection of CAD was better by MBFado versus tracer uptake <70% maximum (25 of 33 discordant pairs, 76%, p = 0.005). Accordingly, the predictive accuracy (true positive + true negative/all tests) of MBFado (0.84) for detection of CAD exceeded that of tracer uptake <70% maximum (0.72) and the McNemar test result, by definition, identical to that for detection of CAD. The specificity of MBFado (0.85) did not differ significantly versus tracer uptake <70% maximum (0.82). However, sensitivity of MBFado (0.81) exceeded that of tracer uptake <70% maximum (0.48; 15 of 16 discordant pairs, 94% correct by MBFado, p = 0.001).
Detection of LAD stenoses by MBFado was more sensitive than that of tracer uptake <70% maximum (8 of 8 discordant pairs, p < 0.02). Specificity and accuracy did not differ. Because the LCX and RCA territories overlap and clinically carry a better prognosis in terms of ischemia or infarct vis-à-vis the LAD, they were combined to compare MBFado and tracer uptake criteria for detection of CAD. The MBFado specificity did not differ significantly from that of <70% maximum tracer uptake. The MBFado sensitivity was marginally greater than that for <70% maximum tracer uptake (7 of 8 discordant pairs, 78%, p = 0.08). Accuracy for MBFado exceeded that of <70% maximum tracer uptake (17 of 23 discordant pairs, 74%, p < 0.04).
Discussion
This study tests the hypotheses that: 1) measurement of Ado-stimulated absolute MBF is superior to that of quantitative measurement of relative myocardial tracer retention for identification of CAD; and 2) a single measurement of MBFado would prove adequate for this purpose. The data obtained support both hypotheses.
The ROC analysis showed that MBFado exceeded <70% maximum uptake with Ado stress (p < 0.0001) (Figs. 2A to 2C); MBFado also showed marginally greater AUC versus MBFR (p = 0.10) (Figs. 2A to 2C). The ROC analysis for CAD in the LAD alone and in the combined LCX + RCA distributions showed that MBFado exceeded <70% maximum uptake in both the LAD distribution (p = 0.03) (Figs. 3A to 3B) and in the combined LCX + RCA distributions (p = 0.02) (Figs. 4A and 4B).
The best cut-point for CAD detection byMBFado exceeded <70% maximum uptake for classification of all 3 major coronary arteries in a subject-wise analysis (p < 0.005), correct classification of all coronary vessels (p = 0.005), and thus predictive accuracy. Although sensitivity for detection of CAD in the LAD distribution was superior by MBFado vs. <70% maximum uptake (p < 0.02), specificity and overall accuracy did not differ. However, accuracy for detection of CAD in the combined LCX + RCA territories by MBFado exceeded that of <70% maximum uptake (p < 0.04). Sensitivity of MBFado was marginally better than <70% maximum uptake (p = 0.08). Specificity did not differ between the two.
A single measurement of MBFado was more sensitive, equally specific, and modestly although not significantly better for accuracy in detection of CAD versus MBFR. Although MBFR with a cutoff of <2.0-fold increase has been used to detect CAD (14), such measurements are quite sensitive to the level of rest MBF. Thus, ratios >2 may result when rest MBF is relatively reduced (Fig. 5). Accordingly, MBFado likely is superior to MBFR for CAD detection.
Patient Example (K1 vs. Uptake Images)
Shown are short-axis, K1 (left), and uptake (right) images obtained during adenosine (Patient A.B.-3VD, on beta-blocker). The MBFado was true positive for all coronary vessels (RCA = 0.57, LCX = 0.99, LAD = 1.13 ml/min/g). Relative uptake was false negative for the LAD (uptake 75% maximum). Moreover, visual interpretation of the uptake image indicates LAD territory is best, and so by definition normal. The MBFR for the LAD (2.5×; MBFrest = 0.45) also is false negative. Note in each image that brighter colors indicate, respectively, higher values of K1 (left) or tracer content (right). LAD = left anterior descending coronary artery; other abbreviations as in Figure 4.
Measurement of MBFado may be thought of as the noninvasive equivalent of the invasive fractional flow reserve measurement (15). A fractional flow reserve <0.75 is a reliable indicator of flow-limiting CAD (15), and because of compensatory microvascular dilation distal to the stenosis at rest (16–18), reflects failure of MBF to augment sufficiently to prevent a decline in distal coronary pressure with intracoronary Ado. The measurement also is not influenced by beta-blockers (19), another similarity to MBFado (see the following text). Accordingly, measurement of absolute MBF with PET, especially with 18F-fluorine labeled tracers (20), offers the prospect of improved detection of CAD with a substantially simplified clinical protocol.
Both MBFado and MBFR may be abnormal in the presence of diffuse microvascular disease absent CAD and will be reduced in proportion to the severity of it (7,21,22). However, MBFR will be of no more value than MBFado in making the distinction and, as noted previously, may be misleading. Thus, determination of optimal MBFado criteria for detection of microvascular disease without CAD, observed by ourselves (7) and others (23–25), will require additional quantitative PET studies of MBFado because it was not investigated in this report.
Beta-blockers played only a limited role in the outcome of the present study. As anticipated, MBFado was not impacted by the presence or absence of beta-blocker (AUCs did not differ significantly), a result consistent with microsphere data showing no effect of beta-blockade on the hyperemic response to a specific A2A receptor agonist (26). Further, for subjects on beta-blockers, the AUC for MBFado (0.839) exceeded that for MBFR (0.720, p < 0.05). Thus, when rest MBF is at usual levels (approximately 0.75 to 1.00 ml/min/g), MBFR is driven by MBFado, adds little additional information, and may at times be misleading (Fig. 5). Finally, the seemingly paradoxical finding that patients taking beta-blockers had a somewhat higher rate–pressure product and MBF than individuals who were not is accounted for by differences in age and health status, especially hypertension, between the groups. Nonetheless, differences in rest MBF were modest (12%) and not statistically significant, with absolute levels in the expected range for hemodynamic conditions of the study.
The applicability of these data to PET 15O-water and 82Rb myocardial perfusion imaging (MPI) is uncertain. The 82Rb retention in the myocardium varies with time (5). Regions with scar lose 82Rb more rapidly versus viable myocardium (5). Because early (15 to 110 s) and late (120 to 360 s) distributions may differ (5), results of relative analysis will vary with data acquisition timing. Accordingly, results of the present study concerning relative uptake analysis likely are applicable to clinical PET 82Rb MPI. Absolute measurement of MBF with 82Rb PET (27), however, could provide quantitative results similar to those reported herein. PET 15O-water measurement of absolute MBF is comparable to that of 13N-ammonia (28). Thus, MBFado with 15O-water should be comparable to 13N-ammonia for CAD detection.
Study limitations
The study sample size and composition are reasonable (21 patients with CAD, 5 with atypical chest pain and normal coronary arteries, 1 with dilated cardiomyopathy, and 21 normal subjects). Thus, the incidence of CAD was 44%, an excellent one with intermediate pre-test probability of CAD (29) despite small sample size. However, because the study was retrospective, it suffers from all of the limitations of retrospective analysis. In addition, inclusion of subjects known to be normal may cause positive bias for estimates of sensitivity, specificity, and accuracy for PET parameters, a limitation however, that applies to each and does not change the major conclusions of the study.
Conclusions
The results of this study show that absolute measurement of MBFado by PET 13N-ammonia is superior to that of relative measurement of myocardial TC for identification of CAD. Accordingly, quantitative MBFado measurement has the potential to greatly simplify clinical PET MPI for evaluation of patients with known or suspected CAD.
Acknowledgments
Mr. Steve Wise and Sandy Barrow, RN, provided invaluable assistance in the performance of the PET studies. Mr. Jim Kean (formerly Astellas Pharma US, Inc.) supplied Ado.
Footnotes
Supported in part by unrestricted grants from Astellas Pharma US to Dr. Gewirtz.
- Abbreviations and Acronyms
- Ado
- adenosine
- AUC
- area under the curve
- CAD
- coronary artery disease
- LAD
- left anterior descending coronary artery
- LCX
- left circumflex coronary artery
- MBF
- myocardial blood flow
- MBFado
- adenosine-stimulated myocardial blood flow
- MBFR
- myocardial blood flow reserve ratio
- MPI
- myocardial perfusion imaging
- PET
- positron emission tomography
- RCA
- right coronary artery
- ROC
- receiver-operator characteristic curve
- TC
- tracer content
- Received March 24, 2009.
- Revision received April 6, 2009.
- Accepted April 14, 2009.
- American College of Cardiology Foundation
References
- ↵
- ↵
- Kitsiou A.N.,
- Bacharach S.L.,
- Bartlett M.L.,
- et al.
- ↵
- Gewirtz H.,
- Fischman A.J.,
- Abraham S.A.,
- et al.
- Hutchins G.D.,
- Schwaiger M.,
- Rosenspire K.C.,
- et al.
- ↵
- Gould K.L.,
- Yoshida K.,
- Hess M.J.,
- et al.
- Holmvang G.,
- Fry S.,
- Skopicki H.A.,
- et al.
- ↵
- Huggins G.S.,
- Pasternak R.C.,
- Alpert N.M.,
- et al.
- Skopicki H.A.,
- Abraham S.A.,
- Picard M.H.,
- et al.
- Skopicki H.A.,
- Abraham S.A.,
- Weissman N.J.,
- et al.
- Tawakol A.,
- Forgione M.A.,
- Stuehlinger M.,
- et al.
- Tawakol A.,
- Migrino R.Q.,
- Aziz K.S.,
- et al.
- ↵
- ↵
- ↵
- Kern M.J.,
- Dupouy P.,
- Drury J.H.,
- et al.
- ↵
- Pijls N.H.J.,
- Van Gelder B.,
- Van der Voort P.,
- et al.
- ↵
- Gould K.L.
- Gould K.L.
- Sun Y.,
- Gewirtz H.
- ↵
- ↵
- Huisman M.C.,
- Higuchi T.,
- Reder S.,
- et al.
- Shannon R.P.,
- Komamura K.,
- Shen Y.T.,
- et al.
- Neglia D.,
- Parodi O.,
- Gallopin M.,
- et al.
- ↵
- Dayanikli F.,
- Grambow D.,
- Muzik O.,
- et al.
- Pitkanen O.P.,
- Nuutila P.,
- Raitakari O.T.,
- et al.
- Yokoyama I.,
- Murakami T.,
- Ohtake T.,
- et al.
- ↵
- Riou L.M.,
- Ruiz M.,
- Rieger J.M.,
- et al.
- ↵
- El Fakhri G.,
- Sitek A.,
- Guerin B.,
- et al.
- ↵
- Bol A.,
- Melin J.A.,
- Vanoverschelde J.L.,
- et al.
- ↵
- Reyes E.,
- Underwood S.R.