Author + information
- Received December 3, 2013
- Accepted December 17, 2013
- Published online March 1, 2014.
- Benjamin J.W. Chow, MD∗,†∗ (, )
- Sharmila Dorbala, MD, MPH‡,
- Marcelo F. Di Carli, MD‡,
- Michael E. Merhige, MD§,
- Brent A. Williams, PhD‖,
- Emir Veledar, PhD¶,
- James K. Min, MD#,
- Michael J. Pencina, PhD∗∗,
- Yeung Yam, BSc∗,
- Li Chen, MSc∗,††,
- Sai Priya Anand, BSc∗,
- Terrence D. Ruddy, MD∗,†,
- Daniel S. Berman, MD‡‡,
- Leslee J. Shaw, PhD¶ and
- Rob S. Beanlands, MD∗
- ∗Department of Medicine (Cardiology), University of Ottawa Heart Institute, Ottawa, Ontario, Canada
- †Department of Radiology, University of Ottawa, Ottawa, Ontario, Canada
- ‡Division of Cardiovascular Medicine and Division of Nuclear Medicine, Brigham and Women's Hospital, Boston, Massachusetts
- §Departments of Cardiology, Internal Medicine, and Nuclear Medicine, Niagara Falls Memorial Medical Center, Buffalo, New York
- ‖Department of Center for Health Research, Geisinger Medical Center, Danville, Pennsylvania
- ¶Department of Medicine, Emory University School of Medicine, Atlanta, Georgia
- #Department of Radiology and Department of Imaging, Weill Cornell Medical College, New York, New York
- ∗∗Department of Biostatistics, Boston University Biostatistics and Harvard Clinical Research Institute, Boston, Massachusetts
- ††Cardiovascular Research Methods Centre, University of Ottawa Heart Institute, Ottawa, Ontario, Canada
- ‡‡Department of Imaging, Cedars-Sinai Medical Center, Los Angeles, California
- ↵∗Reprint requests and correspondence:
Dr. Benjamin J. W. Chow, University of Ottawa Heart Institute, 40 Ruskin Street, Ottawa, Ontario K1Y 4W7, Canada.
Objectives This study sought to determine and compare the prognostic and incremental value of positron emission tomography (PET) in normal, overweight, and obese patients.
Background Cardiac rubidium 82 (Rb-82) PET is increasingly being used for myocardial perfusion imaging (MPI). A strength of PET is its accurate attenuation correction, thereby potentially improving its diagnostic accuracy in obese patients. The prognostic value of PET in obese patients has not been well studied.
Methods A total of 7,061 patients who had undergone Rb-82 PET MPI were entered into a multicenter observational registry. All patients underwent pharmacologic Rb-82 PET and were followed for cardiac death and all-cause mortality. Based on body mass index (BMI), patients were categorized as normal (<25 kg/m2), overweight (25 to 29.9 kg/m2), or obese (≥30 kg/m2). Using a 17-segment model and 5-point scoring system, the percentage of abnormal myocardium was calculated for stress and rest patients categorized as normal (0%), mild (0.1% to 9.9%), moderate (10% to 19.9%), and severe (≥20%).
Results A total of 6,037 patients were followed for cardiac death (median: 2.2 years) and the mean BMI was 30.5 ± 7.4 kg/m2. A total of 169 cardiac deaths were observed. PET MPI demonstrated independent and incremental prognostic value over BMI. Normal PET MPI conferred an excellent prognosis with very low annual cardiac death rates in normal (0.38%), overweight (0.43%), and obese (0.15%) patients. As well, both moderately and severe obese patients with a normal PET MPI had excellent prognosis (0.20% and 0.10%, respectively). The net reclassification improvement of PET was 0.46 (95% confidence interval [CI]: 0.31 to 0.61), and appeared similar in the moderately and severe obese patients which were 0.44 (95% CI: 0.12 to 0.76) and 0.63 (95% CI: 0.27 to 0.98), respectively.
Conclusions Rb-82 PET has incremental prognostic value in all patients irrespective of BMI. In the obese population, where other modalities may have reduced diagnostic accuracy, cardiac PET appears to be a promising noninvasive modality with prognostic value.
Obesity is a growing pandemic in both developed and developing countries (1). Diagnosis and assessment of the severity of coronary artery disease (CAD) in obese patients are important for reducing mortality and morbidity. Investigating obese patients for CAD continues to be a challenge because the diagnostic accuracy of noninvasive imaging is often limited by poor exercise tolerance, poor acoustic windows, attenuation artifact, and/or poor signal-to-noise ratios (2–4). Obese patients are also at higher risk of complications from invasive investigations (5). Thus, it is important to identify noninvasive modalities that have preserved diagnostic and prognostic value in patients irrespective of body mass index (BMI).
Cardiac rubidium 82 (Rb-82) positron emission tomography (PET) is increasing in clinical use for the assessment of myocardial perfusion imaging (MPI). Although it appears to have superior diagnostic accuracy and lower patient radiation exposure, its most recent rise in favor may have been brought about by the recent technetium Tc 99m shortage. Some of the advantages of PET (accurate scatter and attenuation correction) may allow it to maintain diagnostic accuracy in the obese population (6). Though PET has prognostic value over routine clinical predictors for all-cause mortality and major adverse cardiac events, its prognostic value in obese patients has not been well studied (7–9).
The objective of this large multicenter cohort study is to understand whether the incremental prognostic value of Rb-82 PET MPI is maintained in normal, overweight, and obese patients.
A total of 7,061 patients from 4 centers who had undergone Rb-82 PET MPI were entered into a multicenter observational registry. All patients underwent pharmacologic Rb-82 PET as per their local clinical protocol (8) and were followed for cardiac death and all-cause mortality (7–11). The study was approved by each center's institutional human research ethics board.
At the time of PET, a medical history was recorded for all patients (8). Based on BMI, patients were categorized as normal (<25 kg/m2), overweight (25.0 to 29.9 kg/m2), or obese (≥30 kg/m2). Obese patients were further categorized as moderately obese (30.0 to 34.9 kg/m2) and severely obese (≥35 kg/m2).
PET image analysis
All patients refrained from caffeine ingestion for ≥12 h prior to MPI. Patients underwent pharmacologic stress Rb-82 PET MPI using center-specific protocols (7,9,12–14). Rb-82 was infused intravenously and images were acquired at rest and following pharmacologic stress using center-specific imaging and stress protocols. Rb-82 MPI was acquired using a dedicated PET (ECAT ART; Siemens-CTI, Knoxville, Tennessee; Posicam HZL/R, Positron Corporation, Houston, Texas) or a hybrid PET/computed tomography (CT) scanner (Discovery Rx or STE Light Speed [16, 64 slice CT], GE Healthcare, Milwaukee, Wisconsin; Biograph 64, Siemen's, Knoxville, Tennessee). Attenuation correction was performed using a radionuclide line source or using a low-dose chest CT scan.
Using a 17-segment model and 5-point scoring system (normal , mild , moderate , severe , and absent  radiotracer uptake), summed stress score (SSS), summed rest score, and summed difference score (SDS) were calculated. At 3 sites, the percentage (%) of myocardium was calculated using summed scores divided by 68 and multiplied by 100, and 1 site used software as described previously (15). Patients were categorized as normal (0%), mild (0.1% to 9.9%), moderate (10% to 19.9%), and severe (≥20%).
The primary outcome measure of cardiac death was available for 6,037 patients from 3 centers and the secondary outcome of all-cause mortality was available for all 7,061 patients (4 centers). Trained study coordinators, who were supervised by the site clinical investigators, performed the determination of mortality status. Follow-up was determined using scripted telephone interviews and a review of electronic medical records. For patients who died, source documents (i.e., patient's medical record, verbal confirmation by the patient's primary care physician, or review of death certificates) were used for confirmation. At all U.S. centers, the National Death Index was applied for follow-up and confirmatory purposes.
Statistical analyses were performed using SAS (version 9.3, SAS Institute Inc., Cary, North Carolina) and SPSS (version 21.0, IBM, Armonk, New York). Statistical significance was defined as p < 0.05. Continuous variables were presented as mean ± SD or median (interquartile range [IQR]), and categorical variables were presented as frequencies with percentages. The 2-sided Wilcoxon rank sum test was used to compare continuous variables and chi-square test was used for categorical variables.
Prognostic value of MPI was assessed for both univariable and multivariable associations with cardiac or all-cause death. Because noncardiac death competes with cardiac death for mortality, a competing risks analysis was performed for cardiac death with noncardiac death. Cumulative incidence function was used in estimating the probability of cardiac death. For unadjusted comparisons of event rates, a test proposed by Gray (16) was used for the competing risk analysis of cardiac death and a log-rank test was used for the all-cause death. Any variables with a p value <0.10 in a univariable analysis was included in a subsequent multivariable models. For cardiac death, the subdistribution hazard model proposed by Fine and Gray (17) was used to assess the independent prognostic value of MPI by adjusting for baseline clinical characteristics. For all-cause death, the Cox proportional hazard model was used. Model overfitting was considered and the proportional hazards assumption was met. The incremental prognostic value of the model with MPI was evaluated by a significant increase in the global chi-square value and compared by the global model fit using the likelihood ratio tests. The discrimination ability of the model including clinical predictors and MPI to predict cardiac or all-cause death was measured by the C-index of Harrell and compared in different BMI categories (18). The improvement of reclassification using the MPI was also assessed by calculating the categorical free net reclassification improvement (19).
A total of 6,037 patients (mean age: 62.4 ± 12.9 years, female: 48.1%) were followed for cardiac death with a median follow-up of 2.2 years with an IQR of 1.3 to 3.4 years (Table 1). Mean BMI was 30.5 ± 7.4 kg/m2 and 1,303 patients (21.6%) were classified as normal, 2,047 (33.9%) as overweight, and 2,687 (44.5%) as obese. Of the 2,687 obese patients, 1,343 (50.0%) were moderately obese and 1,344 (50.0%) were severely obese. In our study cohort, BMI was inversely related to age, and the prevalence of many cardiac risk factors and symptoms increased with BMI severity. The prevalence of normal MPI decreased with BMI (Table 1). Similar findings were observed in the all-cause mortality cohort (Online Table 1).
Cardiac death and all-cause mortality
Cardiac death occurred in 169 (2.8%) patients, and all-cause death occurred in 570 patients (8.1%), respectively (Table 2). The annualized rate of cardiac death was 1.2% and all-cause mortality was 3.1%. There was a trend toward cardiac death being inversely rated to BMI (p = 0.072), whereas a statistically significant result was observed for all-cause mortality (p < 0.001) (Table 2).
Clinical parameters (age, sex, symptoms, cardiac risk factors, history of revascularization, and BMI) and PET MPI results (SSS, summed rest score, and SDS) were significant predictors of cardiac death and all-cause mortality (Table 3).
Stress perfusion defect and cardiac death and all-cause mortality
The clinical predictors identified in the univariable analysis were used to determine the incremental value of PET MPI measures (Table 3). The models of cardiac death controlling for the competing risk of noncardiac death showed that SSS category had incremental prognostic value to clinical variables including BMI. Worsening SSS category was associated with increased rates of cardiac death in all BMI categories (Fig. 1). Mild, moderate, and severe SSS was associated with greater hazard ratio (per category increase in SSS) for cardiac death and was incremental to clinical predictors and BMI (Table 4). The same pattern was observed with all-cause mortality (Online Table 2).
The ability of the model (clinical predictors and SSS) to predict cardiac or all-cause death was measured by the C-index of Harrell in each of the BMI categories. C-statistics for cardiac death were similar for each BMI category: normal = 0.45 (95% confidence interval [CI]: 0.1 to 0.74); overweight = 0.53 (95% CI: 0.28 to 0.78); and obese = 0.54 (95% CI: 0.28 to 0.78) (Table 5). The same observation was made for all-cause mortality (Table 5).
The addition of PET MPI resulted in a category-free net risk reclassification improvement (NRI) of 0.46 (95% CI: 0.31 to 0.61) for cardiac death and 0.20 (95% CI: 0.11 to 0.28) for all-cause mortality. These values were similar in moderate and severely obese patients for cardiac death (NRI: 0.44 [95% CI: 0.12 to 0.76] and 0.63 [95% CI: 0.27 to 0.98], respectively) and all-cause death (NRI: 0.20 [95% CI: 0.00 to 0.41] and 0.28 [95% CI: 0.06 to 0.51], respectively).
PET MPI and annualized all-cause mortality and cardiac death
A normal PET MPI conferred an excellent prognosis (annual cardiac death) in normal (0.38%), overweight (0.43%), and obese (0.15%) patients. As well, both moderately and severe obese patients with a normal PET MPI had excellent prognosis (0.20% and 0.10%, respectively) (Table 6). As expected, patients with a severely abnormal PET MPI fared worse (Table 6). Similar results were observed with all-cause mortality (Online Table 3).
Severity of ischemia, prior MI, and left ventricular ejection fraction
The severity of ischemia (SDS) and prior myocardial infarction (MI) had incremental prognostic value over clinical variables, but both lost their incremental value when SSS was added to the multivariable model. This may be explained by the fact that SSS is a measure of both ischemia (SDS) and scar (prior MI). A subanalysis was performed for all-cause mortality in the 3,117 patients with left ventricular ejection fraction data. SSS (per categories) was incremental to clinical measures, prior MI, BMI, and left ventricular ejection fraction (hazard ratio: 1.17 [95% CI: 1.02 to 1.25]). However, we were underpowered to detect the incremental value of each SSS category.
Using a large multicenter PET registry, we demonstrate that Rb-82 PET MPI has independent and incremental prognostic value that is preserved in patients who are overweight and obese. This is especially relevant because obesity is a growing pandemic and the diagnosis and risk stratification of obese patients can be challenging (1).
Accuracy of PET MPI in the obese population
PET's superior diagnostic accuracy and improved specificity can be attributed to its better spatial resolution, coincidence detection, and accurate attenuation correction. Such advantages may be particularly relevant in the increasingly obese population, which is typically vulnerable to soft tissue attenuation artifact and has poor acoustic echocardiogram windows (20,21).
The accuracy of PET in obese patients has been previously studied, but it could not be determined in our study cohort. Notwithstanding, the results of our analysis suggest that PET has prognostic value and the low event rate in patients with normal PET reassures us that we are unlikely to be missing prognostically important CAD.
Rb-82 PET has been compared with single-photon emission computed tomography (SPECT) (without attenuation correction [AC]) and showed that it has superior diagnostic accuracy (6–9). In patients with a BMI of ≤30 kg/m2, the accuracy by SPECT versus PET was 70% versus 87%, and in obese patients, the accuracy was 67% versus 85%, respectively (6). Furthermore, they showed that image quality was excellent in 78% of PET scans compared with 62% of SPECT studies.
With the adoption of iterative reconstruction, SPECT + AC is feasible, thus potentially improving its diagnostic accuracy and prognostic value in the obese population. Although a direct comparison of PET to contemporary SPECT + AC has been performed in a small cohort, the incremental prognostic value of PET to SPECT + AC is unknown (22). Thus, the results of our study suggest that PET is a reasonable alternative in centers without direct access to SPECT + AC.
Prognosis of PET MPI
With improved diagnostic accuracy, it is expected that Rb-82 PET would also have prognostic value. The prognostic value of PET has been previously demonstrated in multiple single-center studies (7,9,11). Only recently, our large multicenter registry, of which the current report is a subanalysis, has confirmed the prognostic value of Rb-82 PET previously observed in the single-center studies (8). Dorbala et al. (8) demonstrated that Rb-82 PET MPI perfusion severity predicted all-cause mortality and cardiac death, which was incremental to clinical predictors.
Prognosis of MPI in obese patients
Though the prognostic value of SPECT in obese patients has been previously studied (23–26), there is limited data supporting the prognostic value of PET in obese patients. A small single-center study suggested that the prognostic value of Rb-82 PET is preserved in a small obese cohort (9). However, data supporting the prognostic value of Rb-82 PET in a large multicenter obese population is limited. Our study confirms that the C-statistics were similar across the different BMI categories and were preserved in the severely obese. Equally important, a normal Rb-82 PET MPI portended excellent prognosis with similarly very low event rates irrespective of BMI.
Given the advantages of Rb-82 PET and the prognostic value of coronary flow reserve, it would be expected that coronary flow reserve would also have prognostic value in the obese patients. Whether Rb-82 and coronary flow reserve offer superior prognostic information over existing modalities remains to be determined.
The inverse relationship between BMI and cardiac death has been referred to as the “obesity paradox” and has previously been observed by several groups (27–29). It is accepted that age is a strong predictor of outcome. Although our analysis adjusted for age, the fact that age was inversely related to BMI may partly account the observed paradox in the larger BMI patients. The observation of the obesity paradox could also indicate the presence of referral bias and could be explained by the fact that obese patients may have lower exercise tolerance or higher prevalence of exertional symptoms. Similarly, physicians may have a lower threshold to investigate patients with a higher BMI. The inverse relationship of SSS and BMI would also support the notion that the prevalence of CAD or ischemia was lower in the obese population.
A cardiac risk factor history was available, but missing variables (such as fasting lipid profiles) precluded the use of the traditional risk prediction engines such as Framingham Risk Score. Other clinical variables such as ejection fraction were not available from all centers and, therefore, were not included in the multivariable analysis. However, previous studies have shown that PET MPI has incremental prognostic value over ejection fraction assessment; thus, we would not anticipate that our results would differ from previous studies. Clinical variables such as study indication, family history, lipid profile, medication use, and prior testing were unavailable and, hence, may hinder the generalizability of our results. Although the majority of patients were consecutively enrolled at each participating center, 1 center enrolled consecutive patients with gated PET undergoing vasodilator stress and another's pooled cohort was composed of consecutive patients with normal PET. Because patient enrolment was independent of BMI, we do not believe that our results are significantly biased.
Although the category-free net risk reclassification appeared to be equally robust in the subgroup of moderate and severely obese patients, our study may have been insufficiently powered for obese BMI group comparisons.
At the time of PET imaging (prior to July 1, 2009), there was no uniformly accepted approach for detection and resolution of misalignment between CT transmission and PET emission data. Therefore, this information is not available from the different centers. With the adoption of methods to detect misalignment artifacts, one might anticipate that the diagnostic and prognostic value of PET would only improve (30).
Whether prognostic value of PET is superior to other modalities in obese patients remains unanswered. Therefore, future studies should consider comparing the diagnostic and prognostic value of different noninvasive modalities in the obese population. Such information would be important to reduce equivocal results, necessitating downstream testing, thereby improving resource utilization and minimizing healthcare costs.
Rb-82 PET has incremental prognostic value in all patients irrespective of BMI. In the obese population where other modalities may have reduced diagnostic accuracy, cardiac PET appears to be a promising noninvasive modality with prognostic value.
This study was supported in part by an unrestricted grant from Astellas Pharma Global Development; Bracco Diagnostics, Inc.; a National Heart, Lung, and Blood Institute grant (K23HL092299); and a program grant from the Heart and Stroke Foundation of Ontario (PRG6242). Dr. Chow has received research grants from GE Healthcare; and educational support from TeraRecon. Dr. Dorbala has received research grants from Astellas Global Pharma Development and Bracco Diagnostics; has served on advisory boards for Astellas Global Pharma Development; has received honoraria from MedXcel; and owns stock in GE Healthcare. Dr. Di Carli has received research grants from Toshiba and Gilead Sciences. Dr. Merhige has served on the Speakers' Bureau for Bracco Diagnostics; has received honoraria from the Positron Corporation; served as medical director for the Positron Corporation; and owns stock in Positron Corporation. Dr. Min has served on the Speakers' Bureau for Bracco Diagnostics; has ownership interest in TC3; and has served on advisory boards for GE Healthcare and Edwards Lifesciences. Dr. Ruddy has received research grants from GE Healthcare, Atreus, and MDS Nordion. Dr. Berman has received research grants from Lantheus Medical Imaging, Siemens, and Cardium Therapeutics, Inc.; has received honoraria from Spectrum Dynamics; has served on advisory board for Bracco Diagnostics; and has received royalties from Cedars-Sinai Software. Dr. Shaw has received research grants from Astellas Global Pharma Development and Bracco Diagnostics. Dr. Beanlands has received research grants from Lantheus Medical Imaging, GE Healthcare, and MDS Nordion; has received consulting fees from Jubilant DraxImage, Lantheus Medical Imaging, and GE Healthcare; and has served on advisory boards for Lantheus Medical Imaging and Jubilant DraxImage. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Joao Lima, MD, served as Guest Editor for this paper.
- Abbreviations and Acronyms
- attenuation correction
- body mass index
- coronary artery disease
- confidence interval(s)
- computed tomography
- interquartile range
- myocardial infarction
- myocardial perfusion imaging
- positron emission tomography
- rubidium Rb 82
- summed difference score
- single-photon emission computed tomography
- summed stress score
- Received December 3, 2013.
- Accepted December 17, 2013.
- American College of Cardiology Foundation
- Mathew B.,
- Francis L.,
- Kayalar A.,
- Cone J.
- Hibbert B.,
- Simard T.,
- Wilson K.R.,
- et al.
- Ziadi M.C.,
- Dekemp R.A.,
- Williams K.A.,
- et al.
- Dorbala S.,
- Di Carli M.F.,
- Beanlands R.S.,
- et al.
- Yoshinaga K.,
- Chow B.J.,
- Williams K.,
- et al.
- Chow B.J.,
- Wells G.A.,
- Chen L.,
- et al.
- Chow B.J.,
- Wong J.W.,
- Yoshinaga K.,
- et al.
- Merhige M.E.,
- Breen W.J.,
- Shelton V.,
- Houston T.,
- D'Arcy B.J.,
- Perna A.F.
- Dorbala S.,
- Hachamovitch R.,
- Curillova Z.,
- et al.
- Sdringola S.,
- Loghin C.,
- Boccalandro F.,
- Gould K.L.
- Freedman N.,
- Schechter D.,
- Klein M.,
- Marciano R.,
- Rozenman Y.,
- Chisin R.
- Elhendy A.,
- Schinkel A.F.L.,
- van Domburg R.T.,
- et al.
- Lavie C.J.,
- Milani R.V.,
- Ventura H.O.