Author + information
- Received September 8, 2008
- Revision received April 7, 2009
- Accepted April 30, 2009
- Published online July 1, 2009.
- Sharmila Dorbala, MD⁎,†,⁎ (, )
- Rory Hachamovitch, MD, MSc‡,
- Zelmira Curillova, MD⁎,†,
- Deepak Thomas, MD⁎,
- Divya Vangala⁎,
- Raymond Y. Kwong, MD† and
- Marcelo F. Di Carli, MD⁎,†
Reprint requests and correspondence:
Dr. Sharmila Dorbala, Brigham and Women's Hospital, Cardiovascular Faculty Offices, Shapiro 5, Room 128, 70 Francis Street, Boston, Massachusetts 02115
Objectives This investigation sought to study the incremental value of gated rubidium (Rb)-82 positron emission tomography (PET) myocardial perfusion imaging (MPI) over clinical variables for predicting survival and future cardiac events.
Background The prognostic value of Rb-82 PET-MPI and left ventricular ejection fraction (LVEF) reserve (stress minus rest LVEF) is not well defined.
Methods 1,432 consecutive patients undergoing gated rest/vasodilator stress rubidium-82 PET were followed up for at least 1 year. Of these, rest and peak stress LVEF and LVEF reserve were available in 985 patients. Cardiac events (CE) including cardiac death or nonfatal myocardial infarction and all-cause death were assessed.
Results Over a mean follow-up of 1.7 ± 0.7 years, 83 (5.8%) CE and 140 (9.7%) all-cause death were observed. There was an increase in risk for both end points with an increasing percentage of abnormal and ischemic myocardium. With normal, mild, moderate, or severely ischemic scans, the observed annualized rates of CE were 0.7%, 5.5%, 5%, and 11% and of all-cause death were 3.3%, 7.2%, 6.9%, and 12.5%, respectively. In 985 patients with peak stress gated data, the observed annualized rates of CE (2.1% vs. 5.3%, p < 0.001) and all-cause death (4.3% vs. 9.2%, p < 0.001) were higher in patients with an LVEF reserve <0% compared with those with an LVEF reserve ≥0%. On Cox proportional hazards analysis, after consideration of clinical, historical, and rest LVEF information, stress PET results and LVEF reserve yielded incremental prognostic value with respect to both CE and all-cause death.
Conclusions Vasodilator stress Rb-82 PET-MPI provides incremental prognostic value to historical/clinical variables and rest LVEF to predict survival free of CE and all-cause death. An increasing percentage of ischemia on PET-MPI is associated with an increase in the risk of CE and all-cause death. Left ventricular ejection fraction reserve provides significant independent and incremental value to Rb-82 MPI for predicting the risk of future adverse events.
Myocardial perfusion imaging (MPI) with dedicated positron emission tomography (PET) (1) and integrated PET/computed tomography (CT) (2) are highly accurate for detection of obstructive coronary artery disease (CAD) (1,3). Also, MPI with dedicated PET and PET/CT seem to be more accurate than single-photon emission tomography (SPECT) in the diagnosis of obstructive CAD, especially among those undergoing pharmacologic stress (4–6). Because of the increased availability of PET scanners, as well as improved access to the generator-produced radiotracer rubidium (Rb)-82, the clinical use of MPI with PET is increasing.
Although the incremental prognostic value of SPECT-MPI is well established, data regarding the prognostic value of PET-MPI are limited (7–10). In the early studies, the patient cohort was comprised of high-risk patients with known CAD, reflecting the pattern of PET use in the early 1990s (9). A more recent study (8) reported the incremental value of PET but was limited in size and statistical power (cardiac death or myocardial infarction [MI], n = 17), precluding optimal multivariable modeling. Therefore, the incremental prognostic value of Rb-82 PET over clinical variables remains largely undefined.
Our objectives were to define the incremental value of MPI and left ventricular (LV) function assessed with Rb-82 PET for the prediction of survival free of cardiac events (CE) and all-cause death in a large cohort of patients referred for rest/vasodilator stress Rb-82 PET.
A total of 1,598 consecutive patients underwent gated rest and vasodilator stress Rb-82 MPI between November 2003 and October 2006 at the Brigham and Women's Hospital. Patients with prior cardiac transplantation, moderate or severe valvular heart disease (n = 45, 2.8%), atrial fibrillation (n = 74, 4.6%), and those with a follow-up period <365 days (n = 47, 2.9%) were excluded. Patients with coronary revascularization after the index PET scan were included. The remaining 1,432 patients, including 985 with rest and peak stress LV ejection fraction (EF), comprised the study cohort for this analysis. A subset of these patients (n = 510) comprised the patient population of a prior publication on the relation between ischemic burden and left ventricular ejection fraction (LVEF) reserve (3). The Human Research Committee of Brigham and Women's Hospital approved this study.
Rb-82 PET/CT MPI protocol
Patients were instructed to fast for 6 h and withhold caffeine-containing beverages (24 h before the test) and antianginal medications (beta-blockers, calcium blockers, and nitrates on the morning of the test) and were studied using a whole-body PET/CT scanner (Discovery ST, GE Healthcare, Milwaukee, Wisconsin) (3). The majority of patients were referred for pharmacological stress testing. In 90 patients (6%) who were able to exercise, PET-MPI served as a follow-up for an equivocal SPECT MPI. Pharmacological stress was performed using standard dipyridamole (142 μg/kg/min for 4 min in 1,212 patients, 85%) or adenosine (140 μg/kg/min for 6 min in 220 patients, 15%).
Emission images were obtained after intravenous administration of 1,480 to 2,220 MBq (40 to 60 mCi) of Rb-82 at rest and stress, starting at 90 to 120 s after completion of the radionuclide infusion (120 s with known LVEF of <25%) and continued for 5 min. Gated images were acquired using 8-frame gating. The CT portion of the PET-CT examination in this analysis was used solely for correction of photon attenuation by soft tissues. Commercial software (ACQC, GE Healthcare) was used for appropriate registration of the transmission and emission images when available. Images were reconstructed using ordered subsets expectation maximization (OSEM) (30 iterations and 2 subsets), and a 3-dimensional PET filter was used (Butterworth filter cutoff frequency 10, order of 5).
Analysis of myocardial perfusion and gated images
Images were interpreted semiquantitatively by 2 experienced observers using a standard 17-segment model and a 5-point scoring system. Global summed stress score (SSS), summed rest score (SRS) and summed difference score (SDS) (the difference between SSS and SRS) were computed. Although vasodilator stress testing produces flow heterogeneity and only rarely true ischemia, we refer to reversibility as ischemia, and use it as an operational term throughout the article. The SSS, SRS, and SDS were converted into percentages to compute total percent myocardium abnormal, scarred, or ischemic determined as: (100/68) × SSS, (100/68) × SRS, or (100/68) × SDS, respectively (11). Based on the percent myocardium abnormal, the overall scan results were categorized into normal (0%), mild (1% to 10%), moderate (11% to 20%), or severely abnormal (>20%). Similarly, scans were categorized as having no (0%), mild (1% to 5%), moderate (5.1% to 10%), or severe (>10%) ischemia or scar, respectively. Rest and stress LV volumes and EF were calculated using commercially available software as described previously (3). The LVEF reserve was computed as stress minus rest LVEF.
Patient follow-up and outcome measures
All patients were followed up for a minimum of 1 year after the index PET study. Outcomes were determined by a review of electronic medical records, mailed questionnaire, or a pre-scripted telephone interview. The end points of this study included CE (a composite of cardiac death or nonfatal MI) and all-cause death. In patients with multiple events (n = 4), only the first event was considered for survival analysis. Cardiac-specific end points were verified by a review of the electronic medical records and death certificates. Event rates were annualized by dividing the observed events by the mean duration of follow-up for each group.
Continuous variables are reported as mean ± SD and were compared using 1-way analysis of variance. Categorical variables are reported as a proportion and compared using a chi-square test. A 2-tailed value of p < 0.05 was considered to be significant.
Cox proportional hazards modeling stratified by early revascularization was used to assess the association of PET information and event-free survival time after adjustment for baseline characteristics (based on statistical or clinical grounds, as well as known confounders of this relationship) (12). Distinct models were determined for each of the end points of CE and all-cause death. Incremental prognostic value was defined as a significant increase in the Wald statistic and likelihood ratio tests or an increase in the area under the curve of receiver-operator characteristics (ROC) analysis, after the addition of imaging data to an optimized model of pre-imaging data alone. The models were carefully examined for proportional hazards assumption, multicollinearity, and the additive value of the terms.
We considered information in the order of clinical evaluation: step 1, clinical + rest EF and step 2, stress test and perfusion data. Step 1 included the clinical risk factors (age, female sex, body mass index, diabetes, hypertension, hypercholesterolemia, family history, cigarette smoking), historical information (prior MI, revascularization, vascular or renal disease) and rest LVEF. Step 2 consisted of the addition of the stress test (peak heart rate, ST-segment depression) and perfusion data. Separate models were developed with percent myocardium ischemic and percent myocardium scarred, or percent myocardium abnormal, respectively. In 985 patients, LVEF reserve was added to the best PET-MPI model including clinical factors, rest LVEF, stress variables, and PET-MPI (percent myocardium ischemic and percent myocardium scarred or percent myocardium abnormal). We maintained a ratio of ≥10 events per degree of freedom of the model whenever possible. The threshold for covariate model entry into was p < 0.05, and the threshold for covariate removal was p > 0.10 (13). The S-PLUS 2000 (Release 2) software package (Insightful Corp., Seattle, Washington) with supplemental libraries was used for all analyses.
The baseline characteristics of the study cohort are shown in Table 1. The mean age of the 1,432 patients was 63 years, with 52% female subjects. The pre-test likelihood of CAD was calculated using the logistic-based formula developed and reported by Pryor et al. (14). The percentage of patients with known CAD (prior coronary artery bypass graft, percutaneous coronary intervention, or Q-wave MI) or high, intermediate, or low pre-test likelihood of CAD was 30.6%, 6%, 47.5%, and 16%, respectively.
Overall, 46% (N = 664) of the study cohort had a normal MPI study and 54% (n = 768) had an abnormal MPI study. Of the entire cohort of patients, 27% (n = 381), 14% (n = 203), and 13% (n = 184) showed mild, moderate, or severely abnormal scans, respectively. No ischemia, mild, moderate, or severe ischemia was noted in 11% (n = 143), 17% (n = 249), 13% (n = 188), and 13% (n = 188) of patients, respectively. No scar, small, medium, or large scar was noted in 17% (n = 238), 18%, (n = 263), 6% (n = 91), and 12% (n = 176) of patients, respectively. The mean LVEF of this cohort was 60 ± 14%.
The mean follow-up was 1.7 ± 0.7 (range 1 to 3.3) years. Among the 1,432 patients, there were a total of 83 CE (5.8%, including 4 patients, had both an MI and a subsequent cardiac death), 43 cardiac deaths (3%), 44 nonfatal MI (3.1%), and 140 all-cause death (9.8%). Among the 985 patients with rest and peak stress gated data, there were a total of 48 CE (including 23 cardiac deaths and 29 nonfatal MIs; 4 patients had both an MI and a subsequent cardiac death) and 81 all-cause death.
Table 1 summarizes the characteristics of patients with CE or all-cause death. Patients who experienced CE were older, less often female, and more frequently had diabetes, a history of smoking, prior coronary revascularization, and Q waves on electrocardiogram. Further, they had a lower mean body mass index, and had a lower peak heart rate and a lower rest and peak systolic blood pressure compared with those who did not experience CE. With respect to the nuclear scan results, patients who experienced events had more frequently abnormal MPI results, with a higher percent of myocardium abnormal, percent myocardium scarred, and percent myocardium ischemic. Also, the mean LV volumes were higher, and the mean LVEF was lower in patients with events compared to those without events. The mean LVEF reserve was lower in patients with events compared to those without those events.
Frequencies of CE and all-cause death according to MPI results
The observed frequencies of CE and all-cause death increased as a function of the percentage of abnormal (stress perfusion defect), scarred (rest perfusion defect), and ischemic myocardium (reversible perfusion defect) (Table 2). In the mildly ischemic scans, the observed annualized rates of CD, MI, and all-cause death were 2.9%, 2.2%, and 7.8% compared with 2.4%, 3.3%, and 6.6% in moderately ischemic scans. However, when risk adjusted, the predicted rates of CE and all-cause death were 3.5% and 6.2% in mildly ischemic versus 3.9% and 6.1% in moderately ischemic scans. Kaplan-Meier survival curves showed best survival free of CE in the normal group with a worse event-free survival with increasing percent of myocardium abnormal (Fig. 1).
Relation of MPI to rest LVEF and LVEF reserve
A total of 166 patients had abnormal rest LV function (LVEF <40%). An abnormal MPI compared with normal MPI was associated with a higher frequency of CE both in patients with rest LVEF <40% (0% vs. 12%, p = 0.03) and >40% (0.6% vs. 4.5%, p < 0.001). An abnormal MPI compared with normal MPI was also associated with a higher frequency of all-cause death in patients with rest LVEF >40% (3.5% vs. 7.7%, p = 0.001) but not in those with LVEF <40% (6.7% vs. 13.5%, p = 0.3). This was likely caused by the small number (N = 17) of patients with LVEF <40% and normal PET-MPI.
The unadjusted annualized frequency of CE and all-cause death was higher in patients with LVEF reserve <0 (a decrease in LVEF from rest to peak stress) compared to those without a change or an increase in LVEF reserve (Fig. 2).
Multivariable survival analysis. CE
Cox proportional hazards modeling showed that clinical and historical variables and rest LVEF were predictive of this end point (Table 3). The addition of stress PET data and percent myocardium ischemic and percent myocardium scarred (Table 3) incrementally enhanced the above clinical model. The modeling results for CE were very similar when percent myocardium abnormal was substituted for percent myocardium ischemic and scarred.
In 985 patients, the best PET-MPI model, including clinical factors, rest LVEF, stress variables, and PET-MPI, was predictive of CE (Table 3). Addition of LVEF reserve to the best PET model incrementally enhanced the prediction of CE.
Cox proportional hazards modeling identified age, body mass index, insulin use, rest heart rate, and rest LVEF as the optimal clinical model (chi-square 129.2, p < 0.0001) for all-cause death. With the addition of vasodilator stress and PET-MPI data, additionally, peak heart rate, percent myocardium ischemic, percent myocardium scarred and interaction of percent myocardium scarred and peak heart rate, and interaction of rest heart rate and history of prior CAD were significant independent predictors (model chi-square 164.5, p < 0.0001). The vasodilator PET model incrementally enhanced the clinical model with an increase in model chi-square and ROC area under the curve from 69% to 74% (p = 0.006). The modeling results for all-cause death were very similar when percent myocardium abnormal was substituted for percent myocardium ischemic and scarred.
Further, in 985 patients, the model including LVEF reserve provided incremental value to the best PET-MPI model for predicting all-cause death (model chi-square 93.4, p < 0.0001, vs. 85, p < 0.0001 and ROC area under the curve 83% vs. 79%, p = 0.006, respectively). The modeling results for all-cause death were very similar when percent myocardium abnormal was substituted for percent myocardium ischemic and scarred.
Risk-adjusted survival curves based on Cox models showed that PET-MPI was able to achieve enhanced risk stratification with respect to both CE and all-cause death (15). For both the primary and secondary end points, a patient with a normal scan had a lower risk, and predicted risk progressively increased in a patient with 10% and 20% and 30% ischemic myocardium, respectively (Fig. 3). Likewise, risk-adjusted survival curves (Fig. 4A with 0% ischemic myocardium and Fig. 4B with 20% ischemic myocardium) showed a significant separation of survival curves for a patient with LVEF 50%, 35%, and 20%, respectively. As shown (Figs. 4A and 4B), survival is worse in a patient with significant ischemic burden compared to one without ischemia. Finally, there was a significant separation of the risk-adjusted survival curves for patients with a decrease in LVEF during peak stress (LVEF <0%) compared to those without a decrease in LVEF during peak stress (Fig. 5).
In a large number of patients undergoing vasodilator stress PET, our findings show that Rb-82 PET-MPI provides independent and incremental information for the prediction of CE and all-cause death. Risk-adjusted survival analysis showed that both statistical (increased chi-square and ROC analysis) and clinical enhanced risk stratification were achieved by the ischemic burden and scar on PET-MPI. The percentage of ischemic or scarred myocardium on Rb-82 PET-MPI added incremental value to the clinical data and rest LVEF in predicting CE and all-cause death. Importantly, a novel finding in this study is that LVEF reserve is a significant independent predictor of both CE and all-cause death.
Previous studies on PET-MPI
Prior studies have evaluated the prognostic value of Rb-82 PET-MPI (7–10). One study by Marwick et al. (9), from over a decade ago, evaluated patients with established CAD or a high likelihood of CAD undergoing coronary angiography. This high-risk cohort reflects the historical referral pattern to Rb-82 PET, but unfortunately it does not reflect current practice, which is now comprised generally of intermediate-likelihood patients. A more contemporary study by Yoshinaga et al. (8) reported on the prognostic value of Rb-82 PET-MPI in a lower-risk, more representative patient cohort (similar to the SPECT literature). However, this study was limited in size (367 patients), statistical power (only 17 hard events of cardiac death or nonfatal MI), and follow-up success (8.3% lost to follow-up rate) (8). Also, none of the prior studies evaluated the prognostic value of LVEF reserve in addition to PET-MPI.
Our findings confirm and extend the results of these prior studies in several important ways. Firstly, this is the largest study to date, evaluating the prognostic value of MPI with PET. The study cohort is comprised of a large number of patients with known or suspected CAD reflecting current clinical practice patterns in the U.S. Further, by having sufficient power, we can adequately model end points without risk of overfitting or underfitting our models, thus better determine the incremental prognostic value of myocardial perfusion PET imaging over historical, clinical, rest LVEF, and stress testing variables. Also, this study showed for the first time the incremental value of LVEF reserve in addition to clinical, rest EF, and PET-MPI data. Lastly, patient imaging was carried out using contemporary imaging technology with PET/CT.
PET versus SPECT MPI
Rb-82 PET-MPI offers several technical advantages over SPECT, including higher sensitivity as well as spatial and temporal resolution and depth-independent attenuation correction (1). These differences in imaging characteristics result in improved diagnostic accuracy for diagnosing obstructive CAD, particularly in obese individuals and those who are undergoing pharmacological stress myocardial perfusion imaging (1,16,17). In addition, the inherent ability to collect LV function data at rest and during peak stress seems to result in improved detection of multivessel CAD (4). Although not routinely available for clinical use, another advantage of PET-MPI is its ability to quantify myocardial blood flow, thereby allowing a better characterization of underlying CAD (1,2,6). The findings of this and other studies expand the diagnostic potential of this technology and suggest that Rb-82 PET-MPI seems to be an excellent alternative to vasodilator SPECT MPI that is rapid, accurate, and highly specific, and able to achieve similar risk stratification for future CE and survival. Despite these advantages over SPECT, it is unclear at this point whether the enhanced accuracy of PET compared with SPECT MPI is sufficient to outweigh its greater cost, hence its general use cannot yet be justified.
Value of rest LVEF and LVEF reserve
The prognostic value of rest LVEF and volumes is well known (18). In this study, rest LVEF by Rb-82 PET was inversely and independently related to the risk of CE and all-cause death. The available evidence suggests that post-stress LV functional assessment adds to risk assessment by SPECT MPI (18). Recently, peak dipyridamole stress LVEF by Rb-82 PET was shown to be independently predictive of all-cause death (19). These findings are likely related to the longer duration of vasodilation with dipyridamole, allowing assessment of peak hyperemic LVEF information. To the best of our knowledge, this is the first study showing the added utility of LVEF reserve in addition to rest LVEF and PET-MPI. Because of collinearity between rest LVEF and peak stress LVEF, this analysis was performed using rest LVEF and LVEF reserve. Vasodilator LVEF reserve on Rb-82 PET is a marker of severe and extensive obstructive CAD (3) and is inversely related to ischemic burden (3,20). The results of this study extend prior results by showing that an abnormal vasodilator LVEF reserve is not only a marker of anatomic obstructive CAD and ischemia but also an independent and incremental risk marker for future outcomes.
This is a single-center observational study, with retrospective analysis of prospectively collected data, with all the attendant limitations and biases. The study results are most applicable to patients undergoing dipyridamole vasodilator stress PET imaging and may not be generalizable to patients undergoing adenosine, exercise, or dobutamine PET studies because of inherent differences in baseline patient risk and levels and duration of peak coronary flow achieved.
The percentage of ischemic myocardium on vasodilator stress Rb-82 PET-MPI is a powerful predictor of CE and survival in patients with known CAD or an intermediate to high pre-test likelihood of CAD. Rb-82 PET-MPI provides significant incremental value over the baseline clinical variables, rest LVEF and stress data. The addition of LVEF reserve provides significant independent and incremental value to Rb-82 MPI for stratifying risk of future serious adverse events.
The authors thank Shawn Murphy and Henry Chueh and the Partners Health Care Research Patient Data Registry group for facilitating the use of their database.
Dr. Dorbala is a member of the Speakers' Bureau for Astellas and has received speaking honoraria from Bracco Diagnostics. Dr. Hachamovitch is a member of the Speakers' Bureau for Bracco Diagnostics and has received research grants from GE Healthcare. Dr. Di Carli has received research grants from GE Healthcare, Bracco Diagnostics, Siemens, and Astellas; is a member of the Speakers' Bureau for Bracco Diagnostics; and is a Consultant/Advisory Board member for Bracco Diagnostics.
- Abbreviations and Acronyms
- coronary artery disease
- cardiac events
- computed tomography
- ejection fraction
- left ventricular
- left ventricular ejection fraction
- myocardial infarction
- myocardial perfusion imaging
- positron emission tomography
- receiver-operator characteristic
- summed difference score
- single-photon emission computed tomography
- summed rest score
- summed stress score
- Received September 8, 2008.
- Revision received April 7, 2009.
- Accepted April 30, 2009.
- American College of Cardiology Foundation
- Di Carli M.F.,
- Hachamovitch R.
- Sampson U.K.,
- Dorbala S.,
- Limaye A.,
- Kwong R.,
- Di Carli M.F.
- Dorbala S.,
- Vangala D.,
- Sampson U.,
- Limaye A.,
- Kwong R.,
- Di Carli M.F.
- Go R.T.,
- Marwick T.H.,
- MacIntyre W.J.,
- et al.
- Merhige M.E.,
- Breen W.J.,
- Shelton V.,
- Houston T.,
- D'Arcy B.J.,
- Perna A.F.
- Yoshinaga K.,
- Chow B.J.,
- Williams K.,
- et al.
- Sdringola S.,
- Loghin C.,
- Boccalandro F.,
- Gould K.L.
- Hachamovitch R.,
- Hayes S.W.,
- Friedman J.D.,
- Cohen I.,
- Berman D.S.
- Cox D.R.
- Di Carli M.F.,
- Dorbala S.,
- Meserve J.,
- El Fakhri G.,
- Sitek A.,
- Moore S.C.
- Schwaiger M.,
- Ziegler S.,
- Nekolla S.G.
- Lertsburapa K.,
- Ahlberg A.W.,
- Bateman T.M.,
- et al.