CLINICAL RESEARCH: CARDIAC IMAGING
A prognostic score for prediction of cardiac mortality risk after adenosine stress myocardial perfusion scintigraphy
Rory Hachamovitch, MD, MSc, FACC*,
Sean W. Hayes, MD ,
John D. Friedman, MD, FACC,
Ishac Cohen, PhD and
Daniel S. Berman, MD, FACC,*
* Division of Cardiovascular Medicine, Department of Medicine, Keck School of Medicine, University of Southern California
Department of Imaging (Division of Nuclear Medicine), Department of Medicine (Division of Cardiology), the CSMC Burns & Allen Research Institute, Cedars-Sinai Medical Center, and Department of Medicine, University of California at Los Angeles, School of Medicine, Los Angeles, California
Manuscript received November 26, 2002;
revised manuscript received August 23, 2004,
accepted August 30, 2004.
* Reprint requests and correspondence: Dr. Daniel S. Berman, Cedars-Sinai Medical Center, 8700 Beverly Boulevard, Room T1254, Los Angeles, California 90048 (Email: bermand{at}cshs.org).
 |
Abstract
|
|---|
OBJECTIVES: We sought to derive and validate a score to estimate risk after adenosine stress.
BACKGROUND: Maximizing the prognostic information extracted from adenosine stress myocardial perfusion scintigraphy, a commonly performed test, is often challenging for referring physicians.
METHODS: A split-set validation of a score predicting cardiovascular mortality was performed in 5,873 consecutive patients studied by adenosine stress, dual-isotope single-photon emission computed tomography (SPECT; follow-up 94% complete, mean 2.2 ± 1.1 years).
RESULTS: On follow-up, 387 cardiac deaths occurred (6.6%). The Cox proportional hazards model most predictive of cardiac death included age, % myocardium ischemic, % myocardium fixed, early revascularization, dyspnea, diabetes mellitus, rest and peak stress heart rates, abnormal rest electrocardiogram (ECG), and an interaction between % myocardium ischemic and early revascularization (chi-square = 376). The final prognostic score was calculated as follows: (age [decades] x 5.19) + (% myocardium ischemic [per 10%] x 4.66) + (% myocardium fixed [per 10%] x 4.81) + (diabetes mellitus x 3.88) + (if patient treated with early revascularization, 4.51) + (if dyspnea was a presenting symptom, 5.47) + (resting heart rate [per 10 beats] x 2.88) – (peak heart rate [per 10 beats] x 1.42) + (ECG score x 1.95) – (if patient treated with early revascularization, % myocardium ischemic [per 10%] x 4.47). Scores of <49, 49 to 57, and >57 identified low, intermediate, and high risk (0.9%, 3.3%, and 9.5% cardiac death/year, respectively). Score results further risk stratified patients with respect to cardiac death in all categories of SPECT abnormality.
CONCLUSIONS: We derived and validated a score incorporating data available after adenosine stress perfusion SPECT. This score maximizes the prognostic information extracted from this test and may enhance the application of this test as part of an overall strategy.
|
Abbreviations and Acronyms
| | CAD = coronary artery disease | | ECG = electrocardiogram | | ETT = exercise treadmill test | | MPS = myocardial perfusion single-photon emission computed tomography | | SPECT = single-photon emission computed tomography | | Tc-99m = technetium-99m | | Tl-201 = thallium-201 |
|
Numerous previous studies have ascertained the independent and incremental prognostic value of pharmacologic stress myocardial perfusion single-photon emission computed tomography (SPECT), or MPS, in various patient subsets (1–4). Although the prognostic implications of adenosine MPS have been shown to be equivalent to those obtained from exercise MPS (1), the results of the former are often more challenging for clinicians to apply to clinical management decisions, due to the absence of information on exercise tolerance and stress-induced symptoms, as well as the altered accuracy of stress-induced electrocardiographic (ECG) changes. Patients referred to pharmacologic stress are at greater baseline clinical risk, have more comorbidities, and are more frequently older, diabetic, female, and with previous coronary artery disease (CAD), further obfuscating the interpretation of adenosine stress results and their incorporation into clinical strategies (1–5). The finding that for any perfusion defect's extent and severity, patient risk varies widely as a function of multiple historical, clinical, and stress test characteristics further complicates matters (6–8).
Maximizing the prognostic information extracted from testing mandates incorporation of multiple, complementary data elements. In daily practice, however, combining clinical, historical, and stress test data remains a challenge for clinicians, particularly with pharmacologic stress testing. A number of previous studies have examined the value of aggregate or composite variables summarizing clinical and/or exercise treadmill test (ETT) information for predicting patient risk (9,10). Further, previous studies have shown that by combining stress test and clinical data, estimates of risk with alternative therapeutic approaches can be generated, thus estimating, a priori, a therapeutic survival benefit (11). It would enhance the clinical value of this approach if estimates of patient benefit with revascularization versus medical therapy could be added to an overall prognostic score.
A validated prognostic adenosine stress score incorporating clinical, historical, and nuclear variables may enhance physicians' understanding of adenosine stress results and improve their incorporation into clinical strategies. Thus, the goal of this study was to derive and validate a prognostic adenosine score incorporating clinical, stress test, and MPS data for prediction of cardiovascular mortality and treatment benefit.
|
Methods
|
|---|
Patients.
All patients who underwent rest thallium-201 (Tl-201)/adenosine technetium-99m (Tc-99m) sestamibi dual-isotope MPS at Cedars-Sinai Medical Center for evaluation of suspected or known CAD between 1991 and 1998 (each patient considered once) were identified (Fig. 1). Patients with known valvular heart disease or nonischemic cardiomyopathy were excluded, and follow-up was 94%. Of this population, 4,784 (81%) underwent adenosine stress without walking as an adjunct to testing, constituting the cohort for primary analyses of this study. In addition, 3,294 patients (56%) were identified who underwent adenosine stress during the period when patients who were able to walk did so as an adjunct to stress (July 1995 to May 1998). This patient group was used in a secondary analysis modifying the prognostic score to include a factor ability to walk. This study was approved by Cedars-Sinai Medical Center's Institutional Review Board. The patients included in the current study have been included in previous studies from our group (2,4,6–8,12).
View larger version (36K):
[in this window]
[in a new window]
|
Figure 1 Flow chart defining the number of patients included/excluded for various criteria and the definition of cohorts for the primary and secondary analyses.
|
|
Stress testing, image acquisition, processing, and interpretation.
Throughout this study, we used protocols that we have described at length elsewhere (2,4,6–8,12–16). All patients were requested to be withdrawn from nitrates for 6 h, calcium blockers for 24 h, and beta-blockers for 48 h at the time of MPS and were instructed not to consume caffeine-containing products for 24 h before testing.
Dual-isotope adenosine MPS protocol.
Thallium-201 (3.0 to 4.5 mCi) was injected intravenously at rest (with dose variation based on patient weight), and SPECT acquisition started 10 min after radioisotope injection. All SPECT acquisitions were performed, as previously described (circular or elliptical 180° acquisition for 64 projections at 25 s/projection for Tc-99m or 35 s/projection for Tl-201) without attenuation or scatter correction (16).
Adenosine was infused at a rate of 140 µg/kg body weight per minute for 5 to 6 min with Tc-99m sestamibi (25 to 40 mCi) injected intravenously at the end of the second or third minute of infusion (for 5- and 6-min protocols, respectively). Patients undergoing adenosine stress with walking (33% of patients tested from July 1995 to May 1998) walked at a 0 to 10° grade and 1 to 1.7 mph. During adenosine infusion, the 12-lead ECG was recorded each minute, and leads aVF, V1, and V5 were continuously monitored. Semiquantitative visual interpretation of MPS images was done using our standard 20-segment, 5-point scoring system (16).
Derived variables.
Summed stress, rest, and difference scores derived from 20-segment scores (16) were converted to percent total myocardium (% myocardium) involved with stress, ischemic, or fixed defects (summed scores/80 [maximum potential score = 4 x 20] x 100). A summed stress score <4 (<5% myocardium) defined normal scans, 4 to 8 (5% to 10% myocardium) represented mildly abnormal scans; and >8 (>10% myocardium) indicated moderately to severely abnormal scans.
An abnormal rest ECG was defined as the absence of any abnormality, except sinus tachycardia, sinus bradycardia, early repolarization abnormality, and isolated, single premature atrial or ventricular contractions.
Follow-up data.
Follow-up was performed at 2.2 ± 1.1 years (all >1 year) by trained personnel. The end point examined was cardiac death, confirmed by a review of the death certificate, hospital chart, or physician's records. Early revascularization was defined as performance of percutaneous coronary intervention or coronary artery bypass graft surgery  60 days after MPS (17).
Statistical analysis.
Baseline characteristics were described in terms of median (25th, 75th percentiles) for continuous variables and frequencies for categorical variables. The former was compared using the Wilcoxon rank-sum test and the latter using a chi-square test for comparisons of discrete variables. A value p < 0.05 was considered statistically significant.
Score derivation and validation.
Patients were randomly assigned to training and validation groups (n = 2,369 and 2,415, respectively). Rest ECG findings predictive of cardiac death were modeled to define an ECG score, prognostically summarizing the ECG data. In the training set, a Cox proportional hazards model (18) was used to assess the relationship between individual clinical, historical, and adenosine stress data and cardiovascular death. This model was bootstrapped (400 iterations, 50% of population per sample) to identify variables for the final Cox model (19,20). Beta coefficients from this final Cox model were used as weighting factors (each covariate weighted in proportion to its predictive value) to create a score. Cox modeling was repeated using the combined training plus validation. This score was tested in the training and validation sets, and the relationship between the score and risk of cardiovascular death was assessed.
Using this same approach, the prognostic score was further adjusted by incorporating data from a subset of patients who underwent testing after initiation of walking adenosine protocols. Similarly, we also derived and validated a simplified score that would require less information and make for easier calculation. In this, we did not consider post-MPS patient management nor any interactions, but only elements available at the time of interpretation (MPS results, age, presenting symptoms, ECG). Of note, because this simplified score could not be adjusted for treatment, patients who underwent revascularization in the first 60 days after the nuclear study were censored from analyses.
At all steps, care was given to examination of the assumptions of proportional hazards, linearity, and additivity, as appropriate (19). The S-PLUS 2000 (MathSoft, Inc., Seattle, Washington) was used for all analyses. Post-hoc sample size calculation was performed (21).
|
Results
|
|---|
Patient characteristics.
Patient characteristics were virtually identical in the training and validation sets (Table 1). Similarly, there were no differences between training and validation sets undergoing MPS at the time walking adenosine protocols were performed.
Outcome events.
During follow-up, 387 cardiac deaths occurred (6.6% mortality rate, 3.0%/year). In patients undergoing non-walking adenosine stress, 352 cardiac deaths occurred (7.4%), with an additional 35 cardiac deaths in patients undergoing walking adenosine protocols (3.2%). Early revascularization occurred in 9.0% of patients undergoing non-walking and 11.3% of patients undergoing MPS when walking adenosine protocols were performed.
Multivariable survival analysis.
Electrocardiographic score
A number of rest ECG variables were found to be prognostically important by univariable Cox modeling (Table 2). The presence of right or left bundle branch block, anterior or posterior hemiblock, QRS interval >0.12 ms, and second- or third-degree atrioventricular block were combined into a single variable ("any block"), which was more predictive than any of its constituents. The most predictive multivariable model of ECG variables for cardiac death included "any block" (beta = 0.628), left ventricular hypertrophy with repolarization abnormalities (beta = 0.724), premature ventricular beat(s) (beta = 0.832), and nonspecific ST-T wave changes (beta = 0.331). Based on these results, the weighted ECG score was 0.628 (if "any block" was present) + 0.724 (if left ventricular hypertrophy with repolarization) + 0.832 (if premature ventricular contraction[s]) + 0.331 (if nonspecific ST-T wave changes).
Overall score
The model most predictive of cardiac death in the training set is shown in Table 3 (global chi-square = 376). On the basis of these beta coefficients, relative weights and directions were assigned to each covariate, when combined to define a prognostic score. Annualized cardiac death rates by score quartile reveal similar observed mortality rates in the training and validation sets (Fig. 2).
View larger version (17K):
[in this window]
[in a new window]
|
Figure 2 Rates of cardiac death (expressed per year of follow-up) in quartiles of prognostic score (p < 0.001 for all three groups across quartiles). Solid bars = training set; open bars = validation set; cross-hatched bars = overall set with recalculated scores.
|
|
The final weights for the adenosine score were determined by recalculating the coefficients from the Cox model using the covariates listed in Table 3 in the overall patient set (n = 4,784, 372 cardiac deaths). The results of this model reveal minor changes in the coefficients for early revascularization and an abnormal ECG. After these modifications, the annualized cardiac mortality rates in prognostic score quartiles (Fig. 2) did not materially change in comparison to the mortality rates in initial prognostic score quartiles in the training and validation sets.
The final prognostic score was calculated as follows: (age [decades] x 5.19) + (% myocardium ischemic [per 10%] x 4.66) + (% myocardium fixed [per 10%] x 4.81) + (if diabetes mellitus, 3.88) + (if patient treated with early revascularization, 4.51) + (if dyspnea was a presenting symptom, 5.47) + (resting heart rate [per 10 beats] x 2.88) – (peak heart rate [per 10 beats] x 1.42) + (ECG score x 1.95) – (if patient treated with early revascularization, % myocardium ischemic [per 10%] x 4.47). The prognostic score cut points that correspond to widely used risk categories (low, intermediate, and high risk defined as <1%, 1% to 3%, and >3% risk, respectively) were <49, 49 to 57, and >57 (0.9%, 2.8%, and 6.7% cardiac death/year, respectively). The relationship between two-year Kaplan-Meier survival free of cardiac death and the final prognostic score is shown in Figure 3.
View larger version (11K):
[in this window]
[in a new window]
|
Figure 3 Kaplan-Meier two-year survival as a function of prognostic score in the overall population (p < 0.0001 [log-rank] across scores). Bars = 95% confidence intervals. These bars are centered at the mean prognostic score values of seven subgroups (mean value and number of patients): 1) 34.4, n = 387; 2) 42.3, n = 710; 3) 49.1, n = 1,528; 4) 55.3, n = 909; 5) 60.2, n = 601; 6) 65.9, n = 458; and 7) 74.9, n = 163.
|
|
This score identified 388 patients (8.1% of cohort) as having >5% predicted improvement in survival with revascularization over medical therapy, and 679 patients (14.2%) as having >2.5% predicted improvement in survival with revascularization over medical therapy.
Subgroup analysis.
The observed annualized cardiac death rates increased between low, intermediate, and high prognostic score categories in multiple patient subsets (Table 4). Similarly, this score further risk stratified patients with normal, mildly abnormal, and moderate to severely abnormal MPS results, also indicating the presence of incremental value of this score over MPS results alone (Fig. 4).
View larger version (27K):
[in this window]
[in a new window]
|
Figure 4 Rates of cardiac death (expressed per year of follow-up) in patients with low (<49), intermediate (49 to 57), and high (>57) prognostic scores (*p < 0.001 across three prognostic score categories). Solid bars = training set; open bars = validation set; cross-hatched bars = overall set with recalculated scores.
|
|
Simplified prognostic score.
Using a methodologic approach similar to that described subsequently, we also derived and validated a simplified score considering only elements that would be available at the time the scan was interpreted, and incorporating neither post-MPS patient management nor any interactions. This model included the covariates shown in Table 5 (overall model chi-square = 345). The simplified final prognostic score was calculated as follows: (age [years])+ (% myocardium ischemic) + (% myocardium fixed) + 10 (if dyspnea was a presenting symptom) + (resting heart rate x 0.5) – (peak heart rate x 0.5) + 30 (if rest ECG is abnormal). Figure 5 shows the relationship between Kaplan-Meier two-year survival free of cardiac death and this simplified prognostic score.
View larger version (11K):
[in this window]
[in a new window]
|
Figure 5 Kaplan-Meier two-year survival as a function of the simplified prognostic score in the population examined (p < 0.0001 [log-rank] across scores). These bars are centered at the mean prognostic score values of five subgroups (mean value and number of patients): 1) 46.5, n = 408; 2) 67.3, n = 506; 3) 82.9, n = 807; 4) 97.4, n = 1,161; and 5) 123.9, n = 1,902.
|
|
Patients were risk stratified by the simplified score as low, intermediate, and high risk with respect to the annual risk of cardiac death. A low score (<80) defined 1,143 patients as low risk (0.6% cardiac death/year), intermediate scores (80 to 100) defined 1,380 patients as intermediate risk (2.1% cardiac death/year), and high scores (>100) defined 2,261 patients as high risk (5.6% cardiac death/year). Within each of these three risk strata defined by the simplified score, the more complex score was able to successfully restratify significant numbers of patients (low-risk subgroup: 14% of all low-risk patients as >1% risk [1.9%]; intermediate-risk subgroup: low complex score identified 51% of intermediate-risk patients as lower risk [p < 0.03]; high-risk subgroup: complex score reclassified 31% as intermediate risk and 1.2% as low risk [p < 0.001 vs. high-risk complex score]). Expressed statistically, the overall model (Table 3) had a c-index of 0.79, whereas the simplified score model had a c-index of 0.76. Finally, the simple score is generalizable only to patients treated medically after MPS and cannot identify post-MPS therapeutic differences.
Walking adenosine protocols.
The complex score was reassessed in 3,294 patients who underwent adenosine stress during the period when walking was used when possible as an adjunct (July 1995 to May 1998). Using the training set of this subset (1,634 patients, 99 cardiac deaths), the 10 variables in the complex score were entered into a Cox model along with the dichotomous variable "walk" (0 = non-walking adenosine; 1 = walking adenosine). The results revealed that the coefficients for the 10 variables in the score were stable, and the use of a walking protocol identified patients at a lower risk (coefficient: –0.7803, p < 0.002; global chi-square = 142, p < 0.00001). Annualized cardiac death rates for patients in the lower quartiles of the training and validation sets were 0.1% and 0.3%, 0.9% and 1.1%, and 3.0% and 2.7%, and in the highest quartile, 7.8% and 9.3%, respectively The final prognostic score incorporating the option for walking adenosine studies was calculated as follows: (age [decades] x 5.19) + (% myocardium ischemic [per 10%] x 4.66) + (% myocardium fixed [per 10%] x 4.81) + (diabetes mellitus x 3.88) + (if patient treated with early revascularization, 4.51) + (if dyspnea was a presenting symptom, 5.47) + (resting heart rate [per 10 beats] x 2.88) – (peak heart rate [per 10 beats] x 1.42) + (ECG score x 1.95) – (if patient treated with early revascularization, % myocardium ischemic [per 10%] x 4.47) – (if patient underwent walking as an adjunct to adenosine stress, 7.80). Figure 6 shows the relationship between Kaplan-Meier two-year survival free of cardiac death, and this modified the prognostic score.
View larger version (12K):
[in this window]
[in a new window]
|
Figure 6 Kaplan-Meier two-year survival as a function of prognostic score in the population examined during the time period that walking adenosine stress was performed (p < 0.0001 [log-rank] across scores). Bars represent 95% confidence intervals. These bars are centered at the mean prognostic score values of seven subgroups (mean value and number of patients): 1) 30.5, n = 445; 2) 39.8, n = 620; 3) 45.6, n = 551; 4) 50.5, n = 556; 5) 57.2, n = 792; 6) 65.6, n = 218; and 7) 73.5, n = 112.
|
|
Sample size considerations.
Post-hoc sample size calculations revealed that a sample size of 2,084 patients per subset would have been required to achieve 80% power to detect a 3% difference in absolute event rates from the independently split samples. The current study had >90% power with a two-sided test at alpha = 0.05 to detect a 3% difference in absolute event rates from the independently split samples.
|
Discussion
|
|---|
In this study, we demonstrate that multiple sources of prognostic information derived from MPS can be summarized by a single score potentially utilized in daily practice. To the best of our knowledge, this is the first such score incorporating MPS results, as well as the first comprehensive score to be applied to pharmacologic stress. This score is closely associated with the risk of cardiac death and significantly risk stratifies patients with any MPS result, yielding incremental information over MPS results alone. By including a term for survival with revascularization early after MPS, separate estimates for survival with revascularization and medical therapy after adenosine stress MPS can be calculated, enhancing post-test patient management. Further, we extended this score to also include walking adenosine protocols, a recent, commonly used enhancement of vasodilator stress protocols, thus extending the application of this score to a wider patient group and, for the first time, demonstrating the prognostic value of MPS for this protocol.
Comparison with previous studies.
Several previous studies have derived and validated prognostic composite variables to summarize clinical or ETT data using either multivariate (10,22–24) or bayesian (9) methods that have subsequently been revalidated in external cohorts after the initial derivation and validation. Other scores have also been derived for anatomic end points and later validated toward prognostic end points (5). We have also extended the concept of predictive instruments that incorporate a therapeutic term (to predict not only risk but also differential treatment benefit) from the therapeutic realm into a diagnostic testing milieu (11).
Role of predictive instruments.
Predictive instruments serve a variety of roles and applications in health care. Important issues in their development and planning include the target audience (which type of physician will employ a particular score), the complexity of the score, and the cohort and disease to be evaluated. The scores derived and validated in the current study are designed for use by physicians interpreting and reporting adenosine MPS. With further validations, these scores can be incorporated into reporting, thus potentially benefiting patients undergoing MPS and assisting physicians in making better medical decisions. The data elements these scores require and the need to code and collect these data in a particular fashion are potential limitations to the use of these scores in daily practice.
Composition of the prognostic adenosine score: Parsimony versus accuracy.
This score addresses a perceived need of clinicians by providing a validated prognostic score that may enhance physicians' understanding of adenosine MPS results and optimize their incorporation into clinical decisions. An early question was whether to derive a parsimonious score, hence simple to use and more likely to be incorporated in practice, or to derive a more complex score that would require software for calculation.
Although many advocate reporting of estimates of risk (25), this is a complex issue. Because non-nuclear data—clinical, historical, stress, and treatment information—add incrementally to MPS data (6–8), the risk associated with any specific MPS defect's extent and severity varies widely with numerous clinical and historical characteristics, such as age, diabetes mellitus, and type of stress performed, which must be adjusted for in risk estimation. The prognostic score demonstrates that various combinations of clinical, stress test, and MPS data result in a similar score, thus a similar short-term risk; a younger patient with severe MPS defects and an older, diabetic patient with less severe MPS may be at similar risk.
Furthermore, a post-test referral bias (usually associated with markedly lowered specificity and a mild increase in sensitivity) also impacts prognostic MPS applications (12,26,27). Aggressive revascularization strategies in patients with extensive ischemia lower the risk of adverse post-MPS outcomes. Although this bias cannot be completely eliminated, the inclusion of patients treated with either medical therapy or early revascularization in prognostic analyses may permit enhanced estimation of risk by means of expressing risks with the two therapeutic approaches.
Although parsimony and simplification would enhance the application of a prognostic score, it would also adversely affect the ability to accurately estimate risk for individual patients. Predicting risk based solely on the relationship between perfusion results and outcomes would result in a mis-estimate of risk in many patients. As the current study shows, risk assessment is enhanced with increasing score complexity. As a compromise, we also include a more simplified score in an attempt to partially overcome this limitation.
Prediction of risk versus benefit.
The use of this score to generate estimates of risk with medical therapy and with revascularization introduces an important concept. Identifying therapeutic benefit may be of greater importance than simply estimating risk after MPS. For example, a patient with severe and extensive MPS abnormalities would clearly be at high risk. However, if these defects are fixed (nonreversible), this patient is not likely to benefit from revascularization. Alternatively, with increasing amounts of inducible ischemia, the potential benefit from revascularization increases (2). By distinguishing patients at high risk with little or no benefit from patients at high risk with substantial potential benefit, we are providing a more sound clinical application of MPS and improving patient care and the role MPS plays in practice.
Composition of prognostic adenosine score.
We identified nine covariates in the prognostic score. With respect to clinical data, patient age, presence of diabetes mellitus, resting ECG abnormalities, and dyspnea as a presenting symptom were included. Both fixed and reversible defects were included in the model. The inclusion of these two variables, rather than the use of the extent and severity of stress defects, was due to the overwhelming strength of fixed defects as a predictor of cardiac death, as well as the statistically significantly interaction between the presence of inducible ischemia and the use of early revascularization as a predictor of lower risk of cardiac death. Both rest and peak heart rates during adenosine stress were predictors in this model. A number of variables were not predictors of cardiac death in the current study—in particular, a history of CAD/revascularization, ST-segment change during adenosine stress, and use of anti-ischemic medications, despite generally being considered incremental predictors of this end point. It is possible that the inclusion of covariates previously not considered, such as use of early revascularization, rest and peak heart rates and fixed and reversible defects as separate variables, provided information similar to these variables.
Study limitations.
Statistical and clinical
We derived a prognostic score in an observational data series using established multivariate statistical techniques. The use of this approach enhanced the likelihood that the most predictive variables and their weights were identified; the split-set approach ensured the reproducibility of the derived score. Nonetheless, this score will only be considered fully validated after testing in other cohorts consisting of geographically and clinically diverse patients.
The impact of various selection biases, spurious observations, and single-site data cannot be ignored as limitations. The patients in this study were referred for MPS, and the results may not be generalizable to all such patients at all centers. To limit the potential bias from referral to early revascularization, we did not censor patients undergoing early revascularization but adjusted for revascularization in the analysis. The loss to follow-up in the current study is greater than that we have previously reported (6% vs. <5%), possibly due to our cohort's greater age (mean 73 years), and may have introduced an additional bias.
Technical
Scintigraphic studies in the current work were interpreted by experienced observers using a standardized, semi-quantitative approach to visual interpretation, which we have developed and have documented to be highly reproducible (14). We have expressed perfusion images results as % myocardium rather than a semiquantitive summed score (7), thereby providing a measure with intuitive implications not possible with the unit-less summed scores, which can be applied to scoring systems using varying numbers of segments (e.g., 20,17,13), and is applicable to quantitative methods that report abnormalities as percent myocardium (28). Unpublished data from our site indicate that semiquantitative percent myocardium estimates closely agree with quantitative analysis (r2 >0.94).
Conclusions.
The results of the current study describe an internally validated prognostic adenosine stress MPS score incorporating clinical, stress test, and MPS data. This score yields enhanced stratification over MPS results alone and, by maximizing the prognostic information extracted from adenosine stress studies, may enhance the value of these tests as a part of testing strategies.
|
Acknowledgments
|
|---|
The authors thank James R. Johnson, PhD, for his invaluable assistance and advice regarding the statistical analyses performed in this manuscript.
|
Footnotes
|
|---|
This work was supported in part by grants from Bristol-Myers Squibb Medical Imaging, Inc., Billerica, Massachusetts, and Fujisawa Healthcare, Inc., Deerfield, Illinois. Dr. James E. Udelson acted as Guest Editor for this paper.
|
References
|
|---|
1. Klocke FJ, Baird MG, Lorell BH, et al. ACC/AHA/ASNC guidelines for the clinical use of cardiac radionuclide imaging—executive summary J Am Coll Cardiol 2003;42:1318-1333.[Free Full Text]
2. Hachamovitch R, Berman DS, Shaw LJ, et al. Incremental prognostic value of myocardial perfusion SPECT for the prediction of cardiac death: differential stratification for risk of cardiac death and myocardial infarction Circulation 1998;97:535-543.[Abstract/Free Full Text]
3. Heller GV, Herman SD, Travin MI, et al. Independent prognostic value of intravenous dipyridamole with technetium-99m sestamibi tomographic imaging in predicting cardiac events and cardiac-related hospital admissions J Am Coll Cardiol 1995;26:1202-1208.[Abstract]
4. Hachamovitch R, Berman DS, Kiat H, et al. Incremental prognostic value of adenosine stress myocardial perfusion SPECT and impact on subsequent management in patients with or suspected of having myocardial ischemia Am J Cardiol 1997;80:426-433.[CrossRef][Web of Science][Medline]
5. Ho KT, Miller TD, Hodge DO, et al. Use of a simple clinical score to predict prognosis of patients with normal or mildly abnormal resting electrocardiographic findings undergoing evaluation for coronary artery disease. Mayo Clin Proc 2002;77..
6. Hachamovitch R, Hayes S, Friedman JD, et al. Determinants of risk and its temporal variation in patients with normal stress myocardial perfusion scans: what is the warranty period of a normal scan? J Am Coll Cardiol 2003;41:1329-1340.[Abstract/Free Full Text]
7. Hachamovitch R, Hayes SW, Friedman JD, et al. Comparison of the short-term survival benefit associated with revascularization compared with medical therapy in patients with no prior coronary artery disease undergoing stress myocardial perfusion SPECT Circulation 2003;107:2900-2907.[Abstract/Free Full Text]
8. Berman DS, Kang X, Hayes SW, et al. Adenosine myocardial perfusion SPECT in women compared with men: impact of diabetes mellitus on incremental prognostic value and effect on patient management J Am Coll Cardiol 2003;41:1125-1133.[Abstract/Free Full Text]
9. Diamond GA, Staniloff HM, Forrester JS, et al. Computer-assisted diagnosis in the noninvasive evaluation of patients with suspected coronary artery disease J Am Coll Cardiol 1983;1:444-455.[Abstract]
10. Mark DB, Hlatky MA, Harrell FE Jr., et al. Exercise treadmill score for predicting prognosis in coronary artery disease. Ann Intern Med 1987;106:793–800..
11. Selker HP, Griffith JL, Beshansky JR, et al. Patient-specific predictions of outcomes in myocardial infarction for real-time emergency use: a thrombolytic predictive instrument Ann Intern Med 1997;127:538-556.[Abstract/Free Full Text]
12. Hachamovitch R, Berman DS, Kiat H, et al. Value of stress myocardial perfusion SPECT in patients with normal resting electrocardiograms: an evaluation of incremental prognostic value and cost-effectiveness Circulation 2002;105:823-829.[Abstract/Free Full Text]
13. Berman DS, Hachamovitch R, Kiat H, et al. Incremental value of prognostic testing in patients with known or suspected ischemic heart disease: a basis for optimal utilization of exercise technetium-99m sestamibi myocardial perfusion single-photon emission computed tomography J Am Coll Cardiol 1995;26:639-647.[Abstract]
14. Berman DS, Kang X, Van Train KF, et al. Comparative prognostic value of automatic quantitative analysis versus semiquantitative visual analysis of exercise myocardial perfusion SPECT J Am Coll Cardiol 1998;32:1987-1995.[Abstract/Free Full Text]
15. Hachamovitch R, Berman DS, Kiat H, et al. Exercise myocardial perfusion SPECT in patients without known coronary artery disease: incremental prognostic value and use in risk stratification Circulation 1996;93:905-914.[Abstract/Free Full Text]
16. Berman DS, Kiat H, Friedman JD, et al. Separate acquisition rest thallium-201/stress technetium-99m sestamibi dual-isotope myocardial perfusion SPECT: a clinical validation study J Am Coll Cardiol 1993;22:1455-1464.[Abstract]
17. Staniloff HM, Forrester JS, Berman DS, Swan HJ. Prediction of death, myocardial infarction, and worsening chest pain using thallium scintigraphy and exercise electrocardiography J Nucl Med 1986;27:1842-1848.[Abstract/Free Full Text]
18. Cox D. Regression models and life tables(with discussion) J Roy Statist Soc Ser B Method 1972;34:187-220.
19. Harrell Jr FE. Resampling, Validating, Describing and Simplifying the Model: Regression Modeling StrategiesNew York, NY: Springer; 200187–103.
20. Altman DG, Royston P. What do we mean by validating a prognostic model? Stat Med 2000;19:453-473.[CrossRef][Web of Science][Medline]
21. Fleiss JL, Tytun A, Ury HK. A simple approximation for calculating sample sizes for comparing independent proportions Biometrics 1980;36:343-346.[CrossRef][Web of Science]
22. Pryor DB, Shaw L, McCants CB, et al. Value of the history and physical in identifying patients at increased risk for coronary artery disease. Ann Intern Med 1993;118:81–90..
23. Morrow K, Morris CK, Froelicher VF, et al. Prediction of cardiovascular death in men undergoing noninvasive evaluation for coronary artery disease Ann Intern Med 1993;118:689-695.[Abstract/Free Full Text]
24. Morris CK, Morrow K, Froelicher VF, et al. Prediction of cardiovascular death by means of clinical and exercise test variables in patients selected for cardiac catheterization Am Heart J 1993;125:171-176.[CrossRef][Web of Science][Medline]
25. American Society of Nuclear Cardiology Imaging guidelines for nuclear cardiology procedures, part 2 J Nucl Cardiol 1999;6:G47-G84.[CrossRef][Medline]
26. Hachamovitch R, Hayes S, Friedman J, et al. Stress myocardial perfusion SPECT is clinically effective and cost-effective in risk stratification of patients with a high likelihood of CAD but no known CAD J Am Coll Cardiol 2004;43:200-208.[Abstract/Free Full Text]
27. Berman DS, Hachamovitch R, Shaw LJ, et al. Nuclear cardiologyIn: Fuster V, Alexander RW, King S, O'Rourke RA, Wellens HJJ, editors. Hurst's The Heart. New York, NY: McGraw-Hill; 2004. pp. 525-565.
28. Berman DS, Abidov A, Kang X, et al. Prognostic validation of a 17-segment score derived from a 20-segment score for myocardial perfusion SPECT interpretation J Nucl Cardiol 2004;11:414-423.[CrossRef][Web of Science][Medline]
Related Article
-
Quantitative instruments for predicting risk ... and benefit
- James E. Udelson and Harry P. Selker
J. Am. Coll. Cardiol. 2005 45: 730-732.
[Full Text]
[PDF]
This article has been cited by other articles:
|
|
|
|
 
T. Higuchi, S. G. Nekolla, M. M. Huisman, S. Reder, T. Poethko, M. Yu, H.-J. Wester, D. S. Casebier, S. P. Robinson, R. M. Botnar, et al.
A New 18F-Labeled Myocardial PET Tracer: Myocardial Uptake After Permanent and Transient Coronary Occlusion in Rats
J. Nucl. Med.,
October 1, 2008;
49(10):
1715 - 1722.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
J. Whittle, N. R. Kressin, E. D. Peterson, M. B. Orner, M. Glickman, M. Mazzella, and L. A. Petersen
Racial Differences in Prevalence of Coronary Obstructions Among Men With Positive Nuclear Imaging Studies
J. Am. Coll. Cardiol.,
May 16, 2006;
47(10):
2034 - 2041.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
R. J. Gibbons and P. A. Araoz
The Year in Cardiac Imaging
J. Am. Coll. Cardiol.,
August 2, 2005;
46(3):
542 - 551.
[Full Text]
[PDF]
|
|
|
|
|
|
|
N. D. Wong, A. Rozanski, H. Gransar, R. Miranda-Peats, X. Kang, S. Hayes, L. Shaw, J. Friedman, D. Polk, and D. S. Berman
Metabolic Syndrome and Diabetes Are Associated With an Increased Likelihood of Inducible Myocardial Ischemia Among Patients With Subclinical Atherosclerosis
Diabetes Care,
June 1, 2005;
28(6):
1445 - 1450.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
Top
Abstract
Methods