Author + information
- Received August 2, 2010
- Revision received January 25, 2011
- Accepted January 31, 2011
- Published online August 9, 2011.
- Maria C. Ziadi, MD⁎,
- Robert A. deKemp, PhD⁎,
- Kathryn A. Williams, MS†,
- Ann Guo, MEng⁎,
- Benjamin J.W. Chow, MD⁎,
- Jennifer M. Renaud, MSc⁎,
- Terrence D. Ruddy, MD⁎,
- Niroshi Sarveswaran, BHSc⁎,
- Rebecca E. Tee, MSc⁎ and
- Rob S.B. Beanlands, MD⁎,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Rob S. B. Beanlands, National Cardiac PET Centre, University of Ottawa Heart Institute, 40 Ruskin Street, Ottawa, Ontario K1Y 4W7, Canada
Objectives We evaluated the prognostic value of myocardial flow reserve (MFR) using rubidium-82 (82Rb) positron emission tomography (PET) in patients assessed for ischemia.
Background The clinical value of MFR quantification using 82Rb PET beyond relative myocardial perfusion imaging remains uncertain.
Methods We prospectively enrolled 704 consecutive patients; 677 (96%) completed follow-up (median 387 days [interquartile range: 375 to 416 days]). Patients were divided into 4 groups: I, normal summed stress score (SSS) (<4) and normal myocardial flow reserve (MFR) (>2); II, normal SSS and MFR <2; III, SSS ≥4 and MFR ≥2; IV, SSS ≥4 and MFR <2.
Results For patients with a normal SSS and those with an abnormal SSS, there were significant differences in outcomes for hard events (cardiac death and myocardial infarction) between patients with MFR ≥2 and those with MFR <2 (I: 1.3% vs. II: 2% [p = 0.029]; III: 1.1% vs. IV: 11.4% [p = 0.05]) and for major adverse cardiac events (MACE) (p = 0.003 and p < 0.001, respectively). In the adjusted Cox model, MFR was an independent predictor of hard events (hazard ratio: 3.3; 95% confidence interval: 1.1 to 9.5; p = 0.029) and MACE (hazard ratio: 2.4, 95% confidence interval: 1.4 to 4.4, p = 0.003). The incremental prognostic value of the MFR over the SSS was demonstrated by comparing the adjusted SSS model with and without the MFR for hard events (p = 0.0197) and MACE (p = 0.002).
Conclusions MFR quantified using 82Rb PET predicts hard cardiac events and MACE independent of the SSS and other parameters. Routine assessment of 82Rb PET–quantified MFR could improve risk stratification for patients being investigated for ischemia.
The diagnostic and prognostic value of relative myocardial perfusion imaging (MPI) using single-photon emission tomography (1,2) and positron emission tomography (PET) (3–6) is well established. Relative MPI has limitations because it often uncovers only the territory supplied by the most severe coronary stenosis. This could underestimate the extent of coronary artery disease (CAD) (6–8). Also, relative MPI cannot define the presence of subclinical atherosclerosis.
In addition to relative MPI, PET imaging enables noninvasive quantification of myocardial perfusion. Various clinical applications for myocardial flow reserve (MFR) (stress myocardial blood flow [MBF]/rest MBF) have been proposed (8–15). To date, however, MFR measurement has not been integrated into clinical practice because studies demonstrating added clinical value have been limited.
Among available validated PET tracers for flow quantification, rubidium-82 (82Rb) has the most potential for broad clinical application. It is more widely available in North America than other cyclotron-based PET tracers. Still, there are no large studies that evaluated the prognostic value of flow quantification using 82Rb PET. If the added prognostic value of MFR quantification with 82Rb were demonstrated, this could have an important impact given the growing use of 82Rb PET worldwide.
We sought to assess the prognostic value of 82Rb PET–quantitated MFR in patients being investigated for ischemia. We hypothesized that patients with reduced MFR would have higher cardiac event rates than those with preserved MFR and 82Rb MFR would be an independent predictor of adverse outcomes.
We prospectively enrolled patients with known or suspected CAD, referred for dipyridamole 82Rb PET MPI for evaluation of ischemia at the University of Ottawa Heart Institute, Ottawa, Ontario, Canada. All patients provided written informed consent for inclusion in the study.
Patients were excluded if they did not have MBF data available because dynamic acquisition was not acquired or because of other technical factors (16). Patients who underwent dobutamine, exercise, and/or 13N-ammonia (13NH3) PET were also excluded. For those with more than 1 82Rb PET scan, only the first scan was used.
Patients refrained from caffeine ≥12 h and theophyllines for >48 h before the MPI study (17,18). Antianginal medications were withheld the morning of the test. After an overnight fast, patients were positioned in a 3-dimensional PET system (Discovery Rx/VCT, GE Healthcare, Milwaukee, Wisconsin) (19). A low-dose (∼0.5 mSv), fast helical (1.5 s) computed tomography scan (120 kpv with axial and angular milliampere modulation at a noise index of 50) was acquired for attenuation correction. Then, 10 MBq/kg of 82Rb was administered intravenously using a custom elution system to ensure dead-time losses remained <50% (20,21). A 17-frame, 10-min dynamic 82Rb scan was acquired with a parallel list-mode acquisition.
Pharmacological stress and imaging
After rest PET MPI, a dipyridamole stress test was performed (0.14 mg/kg/min over 5 min). Then 10 MBq/kg of 82Rb was infused 3 min after completion of the vasodilator infusion. Stress images were acquired per rest MPI. A repeat low-dose computed tomography scan was acquired after stress images for attenuation correction.
Images were reconstructed using Fourier rebinning and filtered back-projection with a 12-mm 3-dimensional Hann window of the ramp filter. Automatic reorientation of the images, automatic extraction of mean myocardial and cavity time-activity curves (21,22), and generation of polar maps of absolute MBF and MFR were performed using our FlowQuant software (Ottawa Heart Institute Research Corporation, Ottawa, Ontario, Canada) (16).
The list-mode data from 2.5 to 10 min were replayed to reconstruct electrocardiographic-gated images. Left ventricular ejection fractions (LVEFs) were determined using 4DM software (INVIA, Ann Arbor, Michigan).
82Rb PET analysis
Static Image Interpretation
Images were interpreted using a 17-segment model (23) and a 5-point scoring system blinded to clinical, imaging, and flow data by an experienced blinded observer and then independently compared with the clinical imaging report. Any discrepancies were then reviewed independently by an additional experienced blinded observer. Any remaining discrepancies were settled by consensus. Summed stress score (SSS), summed rest score, and summed difference score (summed difference score = SSS − summed rest score) were calculated. An SSS ≥4 was considered abnormal (2,14). The LVEF during rest and stress and LVEF reserve (stress-rest LVEF) (6) were determined. The presence or absence of transient ischemic dilation was noted.
82Rb Flow Quantification
MBF was quantified using a 1-tissue compartment model with a flow-dependent extraction correction (0). The washout rate was expressed as k2 = K1/DV; DV, the distribution volume of 82Rb in tissue, was set to a constant value (25) for each scan by fitting the model to the region of normal relative uptake (75% to 100% of maximum). MFR <2.0 was considered abnormal (9,14).
Stress electrocardiograms were reviewed and interpreted by blinded observers using recommended practice guidelines (26).
Cardiac event definitions and follow-up
The primary outcome was the prevalence of hard cardiac events: myocardial infarction (MI) and cardiac death. A secondary outcome was the prevalence of major adverse cardiac events (MACE): cardiac death, MI, late revascularization (percutaneous coronary intervention or coronary artery bypass graft) and cardiac hospitalization (e.g., acute coronary syndrome and heart failure). Coronary artery bypass graft or percutaneous coronary intervention within 90 days after the PET scan was considered to be triggered by the relative MPI results and therefore censored from analysis (4). Definitions of each variable were described previously (4,27).
Elective admissions for procedures (e.g., implantable cardioverter-defibrillator) were not counted as events. For patients with ≥2 events, the first event date was considered for analysis.
Follow-up information was acquired for 677 of 704 patients (96%), the majority via telephone interview (successful in 609 of 677 patients; 90%). Events are based on the best available data as of March 2010. When telephone contact was unsuccessful, a record search (n = 12; 2%) was used. Telephone contact with a close family member of the patient was obtained for some patients (n = 56; 8%). Additional data were gathered from medical charts and/or referring physicians. Verification of events was obtained from patient charts, hospital records, death certificates, and/or contact with referring physicians.
Multivariable Cox proportional hazards models were used to assess the independent prognostic value of the MFR. Individual predictor's Wald chi-square statistics are provided as an indicator of the relative importance of the predictor. For the MFR and SSS, accepted cutoff values were used to create the variables of primary interest. Adjusted Kaplan-Meier curves illustrate the incremental value of the MFR over the SSS. Cox model contrasts were used to test for all MFR and SSS group differences. Interactions between the MFR and SSS were tested in the models and were not statistically significant. Therefore, the SSS and MFR could be assessed as independent predictors without the interaction.
To prevent overfitting of the multivariable Cox proportional hazards models, only baseline characteristics included in Tables 2 and 3 with p values <0.05 with MFR and SSS were considered in the full models. For the hard events model, only 3 significant baseline characteristics (previous MI, stress LVEF, and peripheral vascular disease) were considered in the model with the MFR and SSS. Stepwise selection was used to create the adjusted model controlling for confounding. PVD was not significant in the final hard events model.
To show the incremental value of the MFR, the adjusted model with the SSS + MFR was compared with the model with the SSS alone using a likelihood ratio chi-square test. We demonstrated the added value of the MFR as both a continuous variable and a binary variable. We used the dichotomous variable because it represents a more straightforward way to display in survival curves and odds ratios. The net reclassification improvement and the integrated discrimination improvement were calculated as additional tools.
Four groups were generated: I, normal SSS <4 and normal MFR ≥2; II, normal SSS <4 and abnormal MFR <2; III, abnormal SSS ≥4 and normal MFR ≥2; and IV, abnormal SSS ≥4 and normal MFR <2. Hard cardiac events and MACE across different subgroups were evaluated.
Baseline patient characteristics
Among 957 scans performed, 243 were excluded: 48 with dobutamine or exercise stress; 144 13NH3 scans; 20 repeat scans; 40 inadequate or incomplete dynamic data to enable quantification; and 1 patient with previously known dilated cardiomyopathy. Thus, 704 consecutive patients were enrolled: 677 (96%) had successful follow-up, with 58 patients censored due to early revascularization. Median follow-up was 387 days (interquartile range: 375 to 416 days).
The demographic and 82Rb PET imaging characteristics of patients with follow-up are given in Table 1. The characteristics of the patients who were lost to follow-up (n = 27) were similar.
During follow-up, among the 677 patients in this study, there were 27 hard events (4%), 12 cardiac deaths (1.8%), and 16 nonfatal MIs (2.4%) (1 patient had MI and then cardiac death). For MACE, there were 71 first events (71 of 677; 10.5%). Among patients with events, 20 patients (20 of 71; 28%) underwent late revascularization as first events (13 percutaneous coronary interventions, 7 coronary artery bypass grafts), and 29 patients were admitted: 13 (13 of 71; 18%) because of congestive heart failure; 15 patients (15 of 71; 21%) due to acute coronary syndrome, and 1 patient (1 of 71; 1.4%) due to other cardiac causes (syncope and chest pain and subsequently acute coronary syndrome).
The results of univariate analysis of baseline demographics, standard PET imaging features, and 82Rb MBF quantification parameters for hard cardiac events are summarized in Table 2 and those for MACE are summarized in Table 3.
Patients were divided into 4 subgroups according to the SSS and MFR (Fig. 1). The distribution of patients across the 4 subgroups and the frequency of hard cardiac events and MACE in each of these are provided in Table 4.
For those with a normal SSS and impaired MFR compared with those with a preserved MFR, there was higher incidence of hard events (2% vs. 1.3%, p = 0.029) and a higher incidence of MACE (9% vs. 3.8%, p = 0.003). Among patients with an abnormal SSS, those with MFR <2 compared with those with a preserved MFR had a higher incidence of hard events (11.4% vs. 1.1%, p = 0.05) and a higher incidence of MACE (24% vs. 9%, p < 0.001).
All cardiac deaths occurred in patients with an abnormal MFR (1 [1%] in group II; 11 [6.5%] in group IV). All patients who experienced cardiac death had a severely impaired MFR (MFR <1.5).
Figure 2 shows adjusted event-free survival curves for hard events and MACE in the different subgroups.
Multivariable Cox models
The 82Rb MFR was an independent predictor of cardiac hard events (hazard ratio [HR]: 3.3, 95% confidence interval [CI]: 1.1 to 9.5); p = 0.029). The incremental prognostic value of the MFR over the SSS was also shown by comparing the adjusted SSS models without and with the MFR (p = 0.0197). When only the SSS (HR: 3.1, 95% CI: 1.2 to 8.1; p = 0.018) and previous MI (HR: 6.0, 95% CI: 2.0 to 18.1; p = 0.002) were considered in the model, both were independent predictors. When stress LVEF was added (HR: 0.82, 95% CI: 0.72 to 0.93; p = 0.002), it was significant. At this point, only MI and stress LVEF were significant in the adjusted model. The SSS and stress LVEF were collinear (ρ = 0.6), which resulted in the SSS not being significant. Adding the MFR to the model resulted in the best fit (likelihood ratio test, p = 0.0197) and confirmed the added independent prognostic value of this parameter (Table 5). With 3 individuals classified up and 0 classified down with the MFR in the model, the net reclassification improvement was estimated at 0.11 and showed a trend toward significance (p = 0.092). The integrated discrimination improvement estimated at 0.009 was significant (p < 0.001).
The 82Rb MFR was also an independent predictor of MACE (HR: 2.4, 95% CI: 1.4 to 4.4; p = 0.003). In the model without the MFR, the SSS ≥4, diabetes, Canadian Cardiovascular Society (CCS) angina class ≥II and stress LVEF were independent predictors of MACE. The effect of adding the MFR resulted in better fitting of the model (p = 0.002), after controlling for the significant covariate SSS (HR: 1.9, 95% CI: 1.03 to 3.6; p = 0.041), diabetes, CCS angina class, and stress LVEF (Table 6). This incremental value is also supported by the significant net reclassification improvement (0.112, p = 0.048) and integrated discrimination improvement (0.014, p < 0.001). Among the flow parameters, the MFR provided the most significant independent prognostic value for both hard events and MACE.
MACE in subgroup categories of SSS and MFR
The percentage of MACE was analyzed at different levels of the SSS: SSS <4, SSS 4 to 7, SSS ≥8. With the aim of understanding the impact of progressive reductions in MFR on outcomes; MACE were evaluated at different degrees of MFR impairment (MFR ≥2, MFR 1.9 to 1.5, MFR <1.5) in different the SSS categories. At any level of the SSS, the incidence of MACE was highest in patients with the lowest MFR (<1.5) (Fig. 3).
We conducted exploratory analyses in patients with a normal global MFR (≥2). We observed that patients with abnormal regional MFR in a single-vessel territory compared with those with normal MFR in all vascular territories had increased MACE (9 of 94; 9.6% vs. 11 of 300; 3.7%; unadjusted p = 0.015, adjusted p = 0.024) and a trend for hard events (3 of 94; 3.2% vs. 2 of 300; 0.7%; unadjusted p = 0.083, adjusted p = 0.135). Among those with normal global MFR and abnormal regional MFR, there was no significant difference in events for those with and without an abnormal SSS. Eight patients with abnormal MFR in 2 territories were not considered in the comparison. Among those with an impaired global MFR, it was uncommonly (17 of 275; 6.2%) attributable to a significant reduction in MFR in 1 vascular territory. The number of events was too small to allow conclusive comparisons.
This study is one of the first to demonstrate the added and independent prognostic value of MFR using 82Rb PET beyond the relative MPI in a large cohort of patients referred for assessment of ischemia. Patients with impaired 82Rb MFR had a higher incidence of hard cardiac events and MACE at approximately 1-year follow-up. In the multivariable model analysis, 82Rb MFR was an independent predictor of hard events and MACE over the SSS.
Comparison with previous PET studies
At least 4 previous studies assessed the prognostic value of standard relative MPI using PET with 82Rb (3–6). These studies are difficult to compare because there are important variations in: 1) population; 2) PET technology and protocol; and 3) endpoints. Studies included 367 to 1,442 patients, with 31% to 70% of patients with known CAD and evaluated hard events, MACE, or all-cause mortality (3–6). The studies each concluded that abnormal relative perfusion is associated with a worse prognosis. Two more recent studies showed that there was added prognostic value using the LVEF (5,6). However, none of these studies measured the absolute myocardial flow. Our study is unique because it is the first to assess the prognostic value of MFR for hard events with 82Rb PET in a large population.
In the setting of ischemic heart disease, Tio et al. (15) showed that MFR measured with 13NH3 PET predicts adverse outcomes in patients with severe CAD and left ventricular dysfunction who were not candidates for revascularization. This is a different population from that in the current study, which evaluated patients referred for assessment of ischemia including patients with and without known CAD. The current study population was more comparable to that of Herzog et al. (14) who demonstrated the incremental utility of the MFR with 13NH3 PET over standard relative MPI for predicting outcomes. Compared with Herzog et al. (14) study, the current study has a larger sample size (N = 677 vs. N = 229), similar percentage of male patients (61% vs. 69%), slightly fewer patients with known CAD (58% vs. 66%), a similar percentage of patients with MFR <2 (41% vs. 44.5%), but fewer patients had abnormal MPI (39% vs. 55%) and shorter follow-up. In line with Herzog et al. (14), our results show that: 1) in patients with normal and abnormal relative 82Rb PET perfusion, subgroups with reduced MFR had a worse prognosis than their normal 82Rb MFR counterparts; and 2) MFR on 82Rb PET was an independent predictor of hard events (HR: 3.3, 95% CI: 1.1 to 9.5; p = 0.029) and MACE (HR: 2.4, 95% CI: 1.4 to 4.4; p = 0.003) after adjusting for relative MPI and other confounding variables. 82Rb MFR improves risk stratification.
The present study also evaluated the effect of reducing MFR in different SSS subgroups. At any level of SSS, the percentage of MACE was highest among patients with the lowest MFR (<1.5) and statistically significant among those with an abnormal SSS (≥4). Among patients with cardiac death (n = 12), all had significantly reduced MFR (<1.5). Also, we assessed and observed that stress LVEF was a strong and independent predictor of adverse outcomes, extending findings of previous studies (5). Finally, we used 82Rb, which, as a generator product, can be more widely distributed; thus, it has greater potential for wide clinical use compared with 13NH3 and H215O.
Routine integration of 82Rb MFR with relative MPI could represent a valuable tool for the clinician to better stratify a patient's risk of adverse cardiac events. Abnormal 82Rb MFR means worse outcomes in any category of relative MPI, and this could affect management decisions for these patients. Even in those with mildly abnormal relative MPI who may be considered for medical therapy, impaired 82Rb MFR had worse outcome (Fig. 3). Identification of impaired 82Rb MFR in this group could have important impact on decisions for invasive angiography and revascularization. In patients with normal standard relative perfusion, reduced 82Rb MFR would also indicate a worse prognosis, and this could also affect management and dictate the need for more aggressive medical therapy and closer follow-up of the patient. On the other hand, because patients with moderate to severe SSS on relative MPI may already be more likely to undergo invasive angiography and revascularization, the added value of impaired 82Rb MFR may be less in this group but may still affect decisions for those who are at high risk of intervention. Because all cardiac deaths occurred in patients with severely reduced 82Rb MFR (<1.5), it may also be that this signifies a particularly high-risk group. Further studies would be required to understand the impact of MFR on directing decision making to affect outcome.
Our rubidium generator and elution system (20) are different from those used in other laboratories, but the rubidium 1-tissue-compartment model used for flow quantification (24) should yield similar results with data from other systems. Our low-dose rubidium protocol was developed expressly to limit the peak dead-time losses with 3-dimensional PET while maintaining high-quality images (20). The rubidium activity can be infused over a longer interval to reduce the peak dead-time losses, improving the accuracy and precision of the resulting flow values. Three-dimensional mode PET imaging has been the approach used in our validation studies and assessment of intra- and interoperator variability (16). Three-dimensional imaging is now standard on all new PET scanners.
This study is observational and single centered; thus, there may be selection bias in patients referred for PET MPI. There were only 27 hard events, and overfitting the model for hard events may be a concern. The results need to be confirmed in a larger cohort with more hard events. PET MBF parameters were not available in the clinical report for the referring physicians during the course of this study period; management direction and decision making were not influenced by PET MBF quantification.
82Rb has lower extraction fraction, which may affect the precision at hyperemic flow measurements, a higher positron range, which can reduce image resolution, and a relatively short half-life for imaging perfusion and function in patients with reduced left ventricular function. However, data support that flow quantification with 82Rb is feasible, accurate, and reproducible (27–29) and has been validated against microspheres (30). As such, 82Rb PET flow quantification has promise for risk stratification.
We focused primarily on global MFR that reflects diffuse (12,13) and potentially greater disease burden (7). Previous data in nonischemic and ischemic heart disease do suggest a potential prognostic value of global MFR measurements. We did explore regional MFR in patients with normal global MFR and observed those with abnormal regional MFR had increased MACE, suggesting that there may be added value for regional MFR. Among those with impaired global MFR, it was uncommonly (6.2%) attributable to significant reduction in MFR in 1 vascular territory, thus making conclusive findings on regional MFR difficult. Larger studies will be required.
There is some interest in stress MBF as an independent parameter and indeed stress MBF is useful, but in the current study, MFR was the more significant predictor of outcome.
In a large cohort of patients referred for PET MPI to assess myocardial ischemia, assessment of MFR with 82Rb yields independent and added prognostic information beyond relative MPI. Clinical integration of MFR with relative PET MPI will enhance risk stratification in this patient population.
The project was supported in part by the Molecular Function and Imaging Program (HSFO PRG6242). Dr. Ziadi was a Research Fellow supported by University of Ottawa International Fellowship Award and the Molecular Function and Imaging Program (HSFO PRG6242). Dr. deKemp is a consultant to DraxImage; and has received grant funding from a government/industry program (GE Healthcare and MDS Nordion). Dr. Ruddy received grant support and honoraria from GE Healthcare and MDS Nordion. Dr. Beanlands is a consultant to Jubilant DraxImage and Lantheus Medical Imaging; has received research grants from Lantheus Medical Imaging, GE Healthcare, and MDS Nordion; and is a Career Investigator supported by the Heart and Stroke Foundation of Ontario (HSFO). All other authors have reported that they have no relationships to disclose.
- Abbreviations and Acronyms
- confidence interval
- hazard ratio
- left ventricular ejection fraction
- major adverse cardiac event(s)
- myocardial flow reserve
- myocardial infarction
- myocardial perfusion imaging
- 13N ammonia
- positron emission tomography
- summed stress score
- Received August 2, 2010.
- Revision received January 25, 2011.
- Accepted January 31, 2011.
- American College of Cardiology Foundation
- Hachamovitch R.,
- Kang X.,
- Amanullah A.M.,
- et al.
- Yoshinaga K.,
- Chow B.J.,
- deKemp R.A.,
- et al.
- Dorbala S.,
- Hachamovitch R.,
- Kwong R.Y.,
- et al.
- Neglia D.,
- Michelassi C.,
- Trivieri M.G.,
- et al.
- Olivotto I.,
- Cecchi F.,
- Camici P.G.,
- et al.
- Herzog B.A.,
- Husmann L.,
- Kaufmann P.A.,
- et al.
- Tio R.,
- van Veldhuisen D.J.,
- Zijlstra F.,
- et al.
- Schepis T.,
- Gaemperli O.,
- Adachi I.,
- et al.
- deKemp R.A.,
- Klein R.,
- Renaud J.,
- et al.
- Cerqueira M.D.,
- Weissman N.J.,
- Dilsizian V.,
- et al.
- Gewirtz H.,
- Fischman A.J.,
- Abraham S.,
- et al.
- Gibbons R.J.,
- Balady G.J.,
- Bricker J.T.,
- et al.
- Beanlands R.S.,
- Nichol G.,
- Huszti E.,
- et al.
- El Fakhri G.,
- Dorbala S.,
- Di Carli M.F.,
- et al.