Author + information
- Received March 24, 2008
- Revision received December 1, 2008
- Accepted December 23, 2008
- Published online March 31, 2009.
- Michael P. Slawnych, MD, PhD*,
- Tuomo Nieminen, MD, PhD†∥,
- Mika Kähönen, MD, PhD‡,††,
- Katherine M. Kavanagh, MD*,
- Terho Lehtimäki, MD, PhD§,¶,
- Darlene Ramadan, RN*,
- Jari Viik, PhD‡‡,
- Sandeep G. Aggarwal, MD*,
- Rami Lehtinen, PhD‡,††,§§,
- Linda Ellis, BSc*,
- Kjell Nikus, MD**,
- Derek V. Exner, MD, MPH*,* (, )
- REFINE (Risk Estimation Following Infarction Noninvasive Evaluation),
- FINCAVAS (Finnish Cardiovascular Study) Investigators
- ↵*Reprint requests and correspondence:
Dr. Derek V. Exner, 3330 Hospital Drive NW, Room G208, Calgary, Alberta, Canada T2N 4N1
Objectives We sought to evaluate the utility of T-wave alternans (TWA) assessment in the immediate post-exercise period to identify and validate cutpoints for the modified moving average (MMA) assessment method.
Background The presence of TWA is associated with an increased risk of cardiovascular death (CVD). The immediate post-exercise period, where increased physiologic stress and minimal surface artifact coexist, appears ideal to implement the MMA method.
Methods A test (n = 322) and validation cohort (n = 681) provided 1,003 patients with coronary artery disease (CAD). We assessed TWA immediately after exercise. The outcomes, CVD and mortality, were adjudicated independent of the TWA results.
Results During 48 months of follow-up 85 deaths, 54 categorized as CVD (64%), were observed. A linear relationship between the magnitude of TWA and the risk of CVD was identified. As a continuous measure TWA voltage was equivalent to ejection fraction in predicting the risk of CVD. To facilitate clinical application, a sensitive, modest predictive accuracy (20 μV) and a specific, greater predictive accuracy MMA cutpoint (60 μV) were identified and validated. Each cutpoint was associated with a 2.5-fold greater risk of CVD, independent of other important variables, including ejection fraction.
Conclusions Post-exercise assessment of TWA using the MMA method is a strong, independent predictor of risk in patients with CAD. The 20-μV cutpoint (87% sensitivity) appears to be most suitable in higher-risk patients, whereas the 60-μV cutpoint (95% specificity) appears more appropriate when TWA is used as a single screening test in those at lower risk. (Assessment of Noninvasive Methods to Identify Patients at Risk of Serious Arrhythmias After a Heart Attack; NCT00399503)
Although a low left ventricular ejection fraction (EF) often is used to identify patients with coronary artery disease (CAD) who are at risk of serious outcomes, the majority of patients who experience cardiovascular death (CVD) or cardiac arrest after myocardial infarction (MI) do not have a low EF (1). Further, only a minority of patients with low EF values will experience a life-threatening arrhythmia during a 5-year period (2,3). Thus, risk stratification methods beyond solely relying on a low EF are required. The assessment of dynamic changes in cardiac repolarization has been advocated for this purpose. Visible beat-to-beat alterations in cardiac repolarization, or T-wave alternans (TWA), have been reported in patients with ischemia, inherited primary electrical disorders, and electrolyte imbalances (4–6). The presence of lower-amplitude TWA also has been independently associated with an increased risk of serious outcomes (7–11). Multiple approaches of assessing TWA have been described (12) but 2 methods, the spectral (8,10,13–16) and the modified moving average (MMA) methods (1,11,17,18), have been best studied. Both predict an independently greater risk of serious outcomes.
This analysis sought to further evaluate the utility of TWA assessment in the immediate post-exercise period, with a goal of identifying and validating cut-off values for TWA by use of the MMA method. An important objective of this analysis was to identify MMA cutpoints that could be used in future studies.
Descriptions of the REFINE (Risk Estimation Following Infarction Noninvasive Evaluation) and the FINCAVAS (Finnish Cardiovascular Study) studies and their primary results have been reported (1,11). In brief, REFINE participants were enrolled from 6 Canadian hospitals between May 2001 and July 2004, and were considered eligible if they had a confirmed MI and at least mild left ventricular dysfunction in the initial week after MI. The 322 REFINE participants underwent serial noninvasive risk assessment, including post-exercise TWA assessment with the MMA method from a 20- to 30-min high-resolution digital electrocardiographic (ECG) recording. The final (3-month) assessment in the REFINE study was found to provide the most prognostic information (1) and was used in this analysis. Participants of the FINCAVAS study were enrolled from a consecutive series of patients undergoing clinical exercise stress testing at Tampere University Hospital between October 2001 and December 2004. A total of 2,103 patients were enrolled. Of these, 681 (32%) had a history of CAD. These 681 patients were used to provide a population of patients with CAD in whom we could validate the results obtained in the REFINE population.
Adjudicated outcome data were available for both cohorts. Outcomes were assessed independent of TWA results. We assessed CVD and all-cause mortality as principle outcomes because of their unambiguous nature (19). Sudden death was assessed as a nonprinciple outcome due to the smaller number of events and challenges in categorizing deaths as arrhythmic versus nonarrhythmic.
Assessment of TWA
The MMA algorithm was used to identify and quantify TWA. The MMA algorithm separates odd from even beats to allow direct comparison of their average morphologies. A series of beats, ranging from 16 (weighting or update factor of 1/8) through 128 (weighting factor of 1/64), can be used. The MMA value is calculated for every incoming beat. This results in continual moving averages of the odd and even beats. This approach has been demonstrated to be intrinsically robust (17,20,21), and algorithms to reduce the influence of noise and respiratory artifact have been incorporated to enhance diagnostic accuracy (22). Because we sought in our analysis to identify and validate cutpoints in the immediate post-exercise period, the highly accurate approach used in the FINCAVAS study was used (no subtraction of noise and update factor of 1/8) (11).
Test cohort (REFINE)
The post-exercise assessment period was selected as an optimal time point to implement the MMA method, because stress-induced TWA would be detectable but surface noise and skeletal artifact could be minimized (1). Three-month post-exercise data were available for 306 of the 322 (95%) REFINE participants (1). Data from the initial test period (initial month after MI) were used in the remaining 16 participants to provide a complete dataset (322 patients).
All REFINE participants underwent a submaximal exercise treadmill assessment (target heart rate of 120 to 130 beats/min or 85% of the maximum predicted for age). A high-resolution (1,000 samples per second), 3-channel, digital ECG recording (leads V1, V5, and Z) was obtained at 2 to 3 min into the recovery stage. Recordings were obtained for 20 to 30 min to provide a series of sufficiently long windows free of artifact and ventricular ectopy to apply the MMA method. Data were analyzed with the use of the GE Healthcare (Wauwatosa, Wisconsin), version 5.2, MMA algorithm, adapted from the method described by Nearing and Verrier (17). The tracings were assessed with the use of an automated system that identified TWA at any point during the recording. All identified TWA episodes were manually verified. The amplitude of the largest verified MMA voltage was recorded. No requirements for heart rate or the duration of TWA were used because previous research indicated that the MMA method appears to be less heart rate dependent than the spectral TWA method (17).
Validation cohort (FINCAVAS)
The FINCAVAS study was designed to evaluate TWA assessment during exercise. However, post-exercise data also were available on all 681 participants with CAD in the FINCAVAS study. The method by which MMA data were collected has been described previously (11). In brief, an upright bicycle was used and the workload increased from 20 to 30 W in a step-wise manner (10 to 30 W/min). Continuous digital ECGs were recorded at 500 Hz. A modified Mason-Likar 12-lead system was used. We calculated MMA continuously during the recovery period (3 to 6 min). For this analysis, MMA was calculated from all 12 leads and from a similar lead set as that used in the REFINE study (V1, V5, and aVF). Similar analytic methods to those described for the REFINE cohort were used in the FINCAVAS study, as was the same GE Healthcare version 5.2 MMA algorithm. The FINCAVAS investigators performed an independent analysis of their post-exercise data by using the pre-identified cutpoints identified in the REFINE study before combining these datasets.
After independent validation of the cutpoints, data from the 2 cohorts were combined to provide a more reliable estimate of risk, to assess the linear relationship between TWA voltage using the MMA method with outcome, and to evaluate the prognostic utility of TWA voltage using the MMA method versus EF as a continuous measure in a large (n >1,000) population.
Pairwise comparisons of continuous and categorical data between patients that did versus those that did not suffer CVD in each cohort were compared with the use of the Wilcoxon rank sum and the Fisher exact test, respectively. The capacity of TWA voltage, as assessed using the MMA method, to predict CVD and mortality was assessed with the use of Cox multivariate models from which hazard ratios (HRs) and 95% confidence intervals (CIs) were obtained. The proportional hazards assumption was confirmed to be valid with the use of log-log plots and assessment of weighted residuals (23). Given the potential impact of age, sex, history of diabetes, heart rate, and significant ischemia during exercise and EF, these variables were pre-specified to be included in the multivariate models. Logistic regression models were used to calculate the area under the receiver-operating characteristic (ROC) curves. The areas under ROC curves were compared with use of a nonparametric method (24). The time to development of outcomes was graphically displayed by constructing Kaplan-Meier time to event curves, and differences in survival were assessed using the log-rank test statistic. A test for trend across ordered groups based on the Wilcoxon rank sum test was used to assess the relationship of increasing TWA voltage with outcome (25). Linear regression was used to assess the relationship between TWA voltage and EF. All analyses were performed with Stata 9.2MP (StataCorp LP, College Station, Texas). Two-sided p values ≤0.05 were considered significant.
Characteristics of the 2 cohorts, stratified by whether the participants did or did not experience CVD, are shown in Table 1.The median follow-up was 48 months (47 months in the REFINE and 48 months in the FINCAVAS study). In both cohorts, the majority of subjects were men and the median age of participants was 62 years. Approximately 20% of patients had a history of diabetes, and 45% had a history of hypertension. Characteristics were similar in the 2 cohorts. All REFINE participants had their EF assessed at 8 to 10 weeks after MI. Ejection fraction data were available for 391 of the 681 FINCAVAS participants (57%). The median EF in the REFINE study was 0.47, whereas the median EF in the FINCAVAS study was 0.63.
A total of 85 deaths (8.5%) were observed: 30 in the REFINE and 55 in the FINCAVAS study. Of these, 54 (66%) were categorized as cardiovascular (5.4%): 20 in the REFINE and 34 in the FINCAVAS study. A total of 35 deaths (17 in the REFINE and 18 in the FINCAVAS study) were categorized as sudden.
Characteristics by outcome
Patients in the REFINE study who suffered CVD were more likely to have a history of diabetes and had lower median EF values compared with patients in the REFINE study who did not suffer CVD in follow-up. Patients in the FINCAVAS study who suffered CVD were older, more likely to be men, had lower median EF values, and were more likely to have ischemia provoked during exercise testing compared with patients in the FINCAVAS study who did not experience CVD in follow-up. Similar differences were observed for total mortality and sudden death.
Post-exercise versus stress phase
Both post-exercise and stress data were available in the FINCAVAS study. The median noise level in the immediate post-exercise phase when the 3-lead set (5 μV, interquartile range [IQR] 3 to 9 μV) was used was lower than the median value observed during the stress phase (8 μV, IQR 6 to 11 μV) in the FINCAVAS study (p < 0.001). Only post-exercise data were available for REFINE participants and a similar low noise level was observed (median 3 μV, IQR 2 to 6 μV). Although both the stress and post-exercise periods provided information on risk in the FINCAVAS study, the areas under the ROC curve for the 3-lead set tended to be larger for the post-exercise versus stress phase data for all-cause mortality (area 0.64 vs. 0.56; p = 0.1) and CVD (area 0.67 vs. 0.58; p = 0.1).
Cutpoint identification (REFINE) and validation (FINCAVAS)
Two MMA cutpoints were identified in the REFINE study by evaluating the ROC data for CVD. One cutpoint provided high sensitivity and modest positive accuracy (20 μV), whereas the other provided high specificity and greater positive predictive accuracy (60 μV). Both cutpoints were found to be associated with similar test characteristics and predictive value in the FINCAVAS cohort. Performance of these cut-points in the cohorts individually and combined are shown in Tables 2, 3, and 4.⇓⇓⇓
The 20-μV cutpoint resulted in high sensitivity in each cohort. The combined sensitivity was 87% for CVD. The positive accuracy for predicting CVD with this cutpoint was modest (7% combined) and the negative accuracy was very high (98% combined). Similar results were observed for total mortality (Table 2).
The 60-μV cutpoint resulted in high specificity in each cohort. The combined specificity was 95% for CVD. The positive accuracy associated with this cutpoint was 2-fold larger than the 20-μV cutpoint (14% combined), whereas the negative accuracy remained high (95% combined). Similar results were observed for total mortality (Table 3).
Risk of Serious Outcomes
The 20- and 60-μV cutpoints each predicted a greater risk of CVD and all-cause mortality in the REFINE and FINCAVAS cohorts individually (Table 4). The sensitive (20 μV) cutpoint was associated with a 2.5-fold greater risk of CVD in the cohorts combined. Similar results were observed when potential confounding variables were adjusted for, including heart rate, ischemia, and EF. The specific (60 μV) cutpoint was associated with a 3.4-fold greater risk of CVD in the cohorts combined. This relationship remained significant despite multivariate adjustment. Similar results were observed for total mortality. The ≥20-μV cutpoint tended to be associated with a greater risk of sudden death before (HR: 1.83, 95% CI: 0.9 to 3.9; p = 0.1) and after adjustment for the variables shown in Table 4(HR: 2.0, 95% CI: 1.0 to 3.5; p = 0.06), whereas the 60 μV cutpoint was associated with a significant, 3.1-fold (95% CI: 1.3 to 7.4; p = 0.01) greater risk of sudden death before and a 2.6-fold (95% CI: 1.0 to 6.6; p = 0.05) greater risk of sudden death after adjustment. The time to development of CVD (Fig. 1,left) and mortality (Fig. 1, right) in the cohorts combined is shown in Figure 1. Patients with the TWA voltages ≥60 μV had the greatest incidence of CVD and total mortality, whereas those with voltages of 20 to 59 μV had an intermediate risk of these outcomes. These MMA cutpoints of 20 to 59 μV and ≥60 μV clearly discriminated among patients who did and did not suffer CVD (log-rank p = 0.0001). These cutpoints also clearly discriminated among patients who died from any cause versus survived (log-rank p = 0.0005).
Linear relationship between TWA voltage and outcome
As a continuous variable, increasing TWA voltage, as assessed with the MMA method in the immediate post-exercise period, was associated with a greater risk of CVD (HR: 1.05 per 5 μV; 95% CI: 1.02 to 1.09; p = 0.004) and mortality (HR: 1.04 per 5 μV; 95% CI: 1.01 to 1.08; p = 0.01) in the cohorts combined. Similar results were observed in each cohort. A clear relationship between the rate of mortality (trend p = 0.002) and the rate of CVD (trend p < 0.0001) was observed (Fig. 2).
Heart rate, ischemia, and beta-blockade
Given the potential influence of heart rate, ischemia, and beta-blocker therapy on the relationship of TWA value with the clinical outcomes, these variables were evaluated. The median heart rate when the maximal TWA value was observed was 82 (IQR 62 to 100) beats/min in the cohorts combined. A weak linear relationship between maximal TWA MMA voltage and the heart rate when this value was recorded was observed (r2= 0.005; p = 0.02). However, no significant association between the heart rate when the maximal post-exercise TWA value was recorded and mortality (HR: 0.94 per 10 beats/min increase, 95% CI: 0.84 to 1.05; p = 0.27) or CVD (HR: 0.91 per 10 beat/min increase, 95% CI: 0.80 to 1.05; p = 0.20) was observed. The heart rate when the maximal TWA value was recorded did not alter the association between increasing TWA MMA amplitude, as a continuous variable, and an increased risk of mortality (adjusted HR: 1.04, 95% CI: 1.01 to 1.08; p = 0.01) or CVD (adjusted HR: 1.05, 95% CI: 1.02 to 1.09; p = 0.002).
A total of 143 patients, 3 in the REFINE and 140 in the FINCAVAS study, had evidence of ischemia on the exercise test immediately preceding the post-exercise TWA MMA assessment (Table 1). The presence versus absence of ischemia was associated with a trend toward a greater risk of mortality (HR: 1.7, 95% CI: 1.0 to 2.7; p = 0.06) and CVD (HR: 1.7, 95% CI: 0.9 to 3.1; p = 0.1), but adjusting for the presence of ischemia on the exercise test did not alter the relationship between increasing TWA MMA amplitude and the risk of mortality (adjusted HR: 1.03; p = 0.03) or CVD (adjusted HR: 1.04; p = 0.01).
In the REFINE study, beta-blockers were held immediately before the exercise test, whereas in the FINCAVAS study these agents were either continued or held 1 to 2 days before the exercise test. Beta-blocker use versus nonuse was not associated with an increased risk of mortality (HR: 1.3, 95% CI: 0.7 to 2.4; p = 0.4) or CVD (HR: 2.0, 95% CI: 0.8 to 4.9; p = 0.2) and did not alter the relationship between increasing MMA amplitude and the risk of mortality (adjusted HR: 1.04; p = 0.01) or CVD (adjusted HR: 1.05; p = 0.003). Adjustment for the presence of significant ischemia, heart rate when the maximal TWA value was recorded, EF, and other variables did not significantly alter the utility of the cutpoints (Table 3).
Prognostic value of MMA voltage versus EF
The prognostic value of TWA voltage, as a continuous variable, was similar to EF, as a continuous variable, for the outcome of CVD (each area under the ROC curve 0.69) (Fig. 3).Likewise, TWA voltage and EF, as continuous variables, similarly predicted the risk of mortality (each area under the ROC curve 0.67). Only a weak linear relationship between TWA voltage and EF (r2= 0.005; p = 0.04) was identified.
Incremental prognostic information of the TWA cutpoints
Both TWA cutpoints were independently associated with a greater risk of CVD and all-cause mortality after adjustment for important clinical variables (Table 3). Further, the areas under the ROC curve were larger when the TWA cutpoints were added to clinical variables that included age, sex, history of diabetes, EF, and heart rate (ROC areas of 0.70 for CVD and 0.67 for mortality, including the TWA cutpoints, versus ROC areas of 0.65 for CVD and 0.62 for mortality, including only the clinical variables). In fact, the TWA cutpoints provided substantially more information on risk than did the individual clinical variables (p < 0.001 for CVD and p = 0.004 for mortality when the ROC areas were compared).
This analysis demonstrates that post-exercise assessment of TWA with use of the MMA method is a strong, independent predictor of CVD and total mortality in patients with CAD. A linear relationship was documented, with risk increasing as TWA voltage increased. When assessed as a continuous parameter, TWA voltage had similar predictive utility as EF. Patients with increased TWA values had a 2.5-fold greater risk of CVD independent of other important variables, including EF, the presence of significant ischemia, and the heart rate when peak TWA was observed.
Repolarization alternans assessment
Previous studies have demonstrated that TWA assessment can be used to predict serious outcomes in patients with CAD and/or cardiomyopathy. Many have used the spectral method (26). Two studies have directly compared the spectral and MMA TWA methods in the same patients at the same time. Both evaluated MMA in a setting in which noise was minimized. Cox et al. (27) used atrial pacing, whereas the REFINE investigators (1) used the post-exercise period. Both found similar predictive value with the spectral and MMA TWA methods in these low noise settings. Although a negative spectral TWA test has been shown to predict a low risk of serious events in most analyses (1,26,28,29), it has low predictive accuracy in patients with CAD, making it less attractive for risk assessment as a single test (1,26,28). The MMA method (20,30) has been used in a variety of settings (exercise, post-exercise, atrial pacing, and/or ambulatory ECG) and is shown to predict a greater risk of serious outcomes in each (1,11,18,27). However, the varying cutpoints used in these studies have limited the clinical application of this TWA assessment method.
Post-exercise repolarization alternans assessment
The immediate post-exercise period appears to be an optimal time to implement the MMA method because it strikes a balance between observing stress-induced changes in repolarization while minimizing associated noise and artifact. Low noise levels were observed in both the REFINE and FINCAVAS cohorts during the post-exercise phase. This analysis demonstrates that post-exercise assessment of TWA by use of the MMA method is a reliable way to identify patients with CAD at risk of serious outcomes. These results were observed despite differences in the populations. Patients in the REFINE study had survived a recent MI, had testing performed early post-MI, and had median EF values of 0.47, whereas patients in the FINCAVAS study had established CAD, a clinical indication for exercise stress testing, and a median EF of 0.63. On the basis of this analysis, post-exercise MMA assessment appears to be widely applicable in patients with CAD.
Greater risk with increasing TWA voltage: clinical implications
There is benefit to using a quantitative measure of TWA versus a binary one. Larger TWA values are known to predict a greater risk of serious outcomes versus lower values (31,32). This analysis extends these findings by demonstrating a linear relationship between TWA magnitude and risk. Using TWA magnitude as a continuous value to predict risk may be ideal, but additional work is required to develop a robust strategy to accomplish this. Until this is achieved, the use of validated cutpoints that can be applied to a given patient appears warranted. The 20-μV MMA cutpoint provides a sensitive marker of risk, so is likely to be most applicable in settings where patient risk is already elevated as the result of abnormalities in other parameters such as heart rate turbulence or baroreflex sensitivity (1). In contrast, the 60-μV MMA cutpoint is a more specific marker of risk and may be better suited for use in a low-risk population, where TWA assessment is used as a single screening test (11).
Although this analysis indicates that TWA assessed in the immediate post-exercise period using the MMA method provides important prognostic information in patients with CAD, it is premature to extend our observations to other groups. Additional prospective validation of the MMA method is required, particularly data related to the exercise versus post-exercise phases. Data on test-retest reproducibility of the MMA method, as well as the spectral method, are required to determine whether changes in TWA over time are clinically meaningful (26). Further, as with any observational study, it is not possible to draw causal inferences and differences in variables that were not adjusted for or residual confounding may exist. Finally, it is premature to make therapeutic recommendations based on the results of TWA testing at this time. This holds true for both the MMA method and the spectral method. Although each method has been independently associated with an increased risk of serious outcomes in this and other studies, there have been no prospective randomized trials to confirm that using TWA to decide on choice of therapy alters patient outcome. This fact is highlighted by the MASTER I (Microvolt T-Wave Alternans Testing for Risk Stratification of Post-Myocardial Infarction Patients) trial, where a similar rate of appropriate defibrillator therapies in patients with non-negative versus negative spectral TWA results was found (33).
Post-exercise MMA is a strong, independent predictor of CVD and mortality in patients with CAD. The greater the TWA value, the greater the risk. To facilitate clinical application of the MMA method, 2 cutpoints were identified, 1 sensitive (20 μV) and 1 specific (60 μV). Although both cutpoints independently predicted a greater risk of CVD and all-cause mortality, the use of one or the other in a given patient will depend on the setting in which the TWA assessment is used. The sensitive 20-μV MMA cutpoint appears most applicable in patients already thought to be at greater risk, whereas the specific 60-μV MMA cutpoint appears best suited when TWA is used as a single test in a lower-risk population.
The REFINE study was supported by the Canadian Institutes of Health Research, the Alberta Heritage Foundation for Medical Research, and the Heart and Stroke Foundation of Alberta. The FINCAVAS study is supported by the Medical Research Fund of Tampere University Hospital, Finnish Cultural Foundation, the Finnish Foundation for Cardiovascular Research, the Academy of Finland (grant no. 104821), the Emil Aaltonen Foundation, Finland, and the Tampere Tuberculosis Foundation. Dr. Exner is a Clinician Scientist of the Canadian Institutes of Health Research and a Scholar of the Alberta Heritage Foundation for Medical Research. Dr. Exner is a consultant to GE Healthcare and has received unrestricted grants in aid from GE Healthcare and Cambridge Heart Inc. Drs. Slawnych and Nieminen contributed equally to this paper.
- Abbreviations and Acronyms
- coronary artery disease
- confidence interval
- cardiovascular death
- ejection fraction
- hazard ratio
- interquartile range
- modified moving average
- myocardial infraction
- receiver-operating characteristic
- T-wave alternans
- Received March 24, 2008.
- Revision received December 1, 2008.
- Accepted December 23, 2008.
- American College of Cardiology Foundation
- Exner D.V.,
- Kavanagh K.M.,
- Slawnych M.P.,
- et al.
- Kleinfeld M.J.,
- Rozanski J.J.
- Bloomfield D.M.,
- Bigger J.T.,
- Steinman R.C.,
- et al.
- Nieminen T.,
- Lehtimaki T.,
- Viik J.,
- et al.
- Bloomfield D.M.,
- Steinman R.C.,
- Namerow P.B.,
- et al.
- Chow T.,
- Kereiakes D.J.,
- Bartone C.,
- et al.
- Chow T.,
- Kereiakes D.J.,
- Bartone C.,
- et al.
- Nearing B.D.,
- Verrier R.L.
- Nearing B.D.,
- Huang A.H.,
- Verrier R.L.
- Nearing B.D.,
- Oesterle S.N.,
- Verrier R.L.
- Kaiser W.,
- Findeis M.,
- Young B.
- Grambsch P.,
- Thernau T.
- Myles R.C.,
- Jackson C.E.,
- Tsorlalis I.,
- Petrie M.C.,
- McMurray J.J.,
- Cobbe S.M.
- Gehi A.K.,
- Stein R.H.,
- Metz L.D.,
- Gomes J.A.
- Ikeda T.,
- Yoshino H.,
- Sugi K.,
- et al.
- Nearing B.D.,
- Verrier R.L.
- Chow T.,
- Kereiakes D.J.,
- Onufer J.,
- et al.