T-Wave Alternans Predicts Mortality in a Population Undergoing a Clinically Indicated Exercise Test
Tuomo Nieminen; Terho Lehtimäki; Jari Viik; Rami Lehtinen; Kjell Nikus; Tiit Kööbi; Kari Niemelä; Väinö Turjanmaa; Willi Kaiser; Heini Huhtala; Richard L. Verrier; Heikki Huikuri; Mika Kähönen Eur Heart J. 2007;28(19):2332-2337. ©2007 Oxford University Press
Abstract and Introduction
Aims: As a part of the Finnish Cardiovascular Study, we tested
the hypothesis that T-wave alternans (TWA) predicts mortality in a general
population of patients referred for a clinical exercise
The merits of a clinical exercise test as a prognostic means are well recognized. Exercise capacity, levels and changes of blood pressure, as well as heart rate (HR) profile and certain electrocardiographic (ECG) parameters have been shown to predict all-cause and cardiac mortality. T-wave alternans (TWA) is a relatively novel ECG index representing beat-to-beat alternation in the shape, amplitude, or timing of the ST-segment and the T-wave. The main contemporary use of TWA is based on the spectral analysis of microvolt-level T-wave amplitude during exercise. TWA is considered to represent spatial or temporal variations in ventricular repolarization, and it has been linked to both inducible and spontaneous ventricular tachyarrhythmias as well as to the mechanisms leading to their initiation.
The accuracy and usefulness of exercise-induced TWA in predicting arrhythmic events and death have been investigated in several studies with, however, relatively small numbers of participants,[8-20] with exception of two larger studies.[21,22] These studies have mostly included patients with a high risk of life-threatening arrhythmias, such as those with a documented near-fatal arrhythmic event, impaired left ventricular function,[10,12,19] congestive heart failure,[8,13-15,17,18] ischaemic cardiomyopathy, or lower risk patients with prior MI.[9,20,21] Surprisingly, despite a decade of research, there have been no studies of the prognostic significance of TWA in more general populations undergoing routine exercise testing. This gap in our knowledge may be attributable to the inherent restrictions of the spectral approach, which, because of data stationarity, dictates the use of a non-standard exercise test to elevate and fix HR for a sustained period and requires specialized electrodes. To circumvent these difficulties, we employed the time-domain modified moving average (MMA) analysis of TWA. The method has undergone extensive testing in the laboratory and has been shown in smaller studies with post-MI, implantable cardioverter defibrillator patients, and in those at increased risk for impending ventricular tachyarrhythmias to be capable of detecting increased cardiac electrical instability and risk for arrhythmia. The present study was designed to test the hypothesis that TWA has prognostic power in a more general population undergoing a clinically indicated exercise test as a part of the Finnish Cardiovascular Study (FINCAVAS).
As described in the detailed study protocol of FINCAVAS, all consecutive patients undergoing exercise stress test at Tampere University Hospital between October 2001 and January 2003 and willing to participate in the study were recruited. A total of 1037 patients (673 men and 364 women) with technically successful exercise tests (96.6% of all the tests) were included in the study ( Table 1 and Table 2 ). A test was technically adequate if storing the haemodynamic data and continuous digital ECG signal was successful. Patients with atrial fibrillation were not excluded, as this condition does not hinder TWA assessment by the MMA method.
The main indications for the exercise test were diagnosis of coronary heart disease (CHD, frequency 46%), testing vulnerability to arrhythmia during exercise (18%), and evaluation of work capacity (19%) and adequacy of the CHD treatment (24%), as well as obtaining an exercise test profile prior to an invasive operation (13%) or after an MI (10%); some patients had more than one indication. The study protocol was approved by the Ethics Committee of the Tampere University Hospital District, Finland, and all patients gave informed consent prior to the interview and measurements as stipulated in the Declaration of Helsinki.
After an informed consent was signed, the medical history of each patient was collected with a computer-based questionnaire form. Thereafter, the exercise test was performed.
Exercise Test Protocol
Prior to the routine exercise stress test, the subject lay down in the supine position for 10 min, and the resting ECG was digitally recorded. The upright exercise test was performed using a bicycle ergometer with electrical brakes. The lead system used was the Mason–Likar modification of the standard 12-lead system. The initial workload varied from 20 to 30 W, and the load was increased stepwise by 10–30 W every minute. Continuous ECGs were digitally recorded at 500 Hz with CardioSoft exercise ECG system (Version 4.14, GE Healthcare, Freiburg, Germany) and analysed fully automatically by the GE Healthcare-released version of the MMA method.
During the test, HR was continuously registered with ECG, while systolic arterial pressure and diastolic arterial pressure were measured with a brachial cuff every 2 min.
Measurement of T-Wave Alternans
Assessing the relationship between TWA and mortality is one of the original goals of FINCAVAS. The algorithm used in the identification and quantification of TWA is based on the time-domain MMA analysis that has been described in detail earlier. In brief, the MMA algorithm separates odd from even beats. Average morphologies of both the odd and even beats are calculated separately and continuously updated by a weighting factor of 1/8 or 1/32 of the difference between the ongoing average and the new incoming beats. The update is calculated for every incoming beat and results in continual moving averages of the odd and even beats. This approach is intrinsically robust and makes MMA suitable for TWA analysis during the period of activity or fluctuating HRs. In addition, algorithms have been incorporated to reduce the influence of noise and artefacts, such as those caused by pedalling and respiration.
The TWA values were calculated continuously during the entire exercise test from rest to recovery using all standard leads (I, II, III, aVR, aVL, aVF, and V1–V6). The maximum TWA value at HR <125 b.p.m. was derived. TWA values at higher HR were excluded, based on the previous results indicating that inaccuracies in TWA measurement can result at HR exceeding this range.
The TWA values derived by the MMA method are larger by a factor of ∼4–6 than the values reported by the spectral method. This difference is attributable to the fact that the time-domain MMA method determines the maximum difference in the T-wave amplitude between successive beats, while the spectral method derives an average value from its spectra, which are generated across the entire T wave and across 128 beats.
Measurement of ejection fraction (EF) is not routine for patients referred for a clinical exercise test. However, EF was determined for 529 (51%) of the study patients with echocardiography (n = 522) or isotope techniques (n = 7) within 6 months (average, 43 days) of the exercise test.
Death certificates were received from the Causes of Death Register, maintained by Statistics Finland, in January 2006; this source has been shown to be reliable. The certificates included causes of death based on the 10th revision of the International Classification of Diseases (ICD-10). The diagnosis numbers and certificate texts were used to classify the deaths as all-cause, cardiovascular death, and sudden cardiac death (SCD; defined as a cardiac death within 24 h after the onset of symptoms).
The analyses were performed for two different incremental update factors of the TWA detection algorithm: 1/32 and 1/8, the latter of which proved to be superior in the prediction of death, consistent with experimental and clinical[20,24] studies. Several cut points were evaluated, including the value coinciding with the 75th percentile (46 µV), the cut point that proved to predict cardiac arrest or arrhythmic death in a prior study. In addition, the cut points of 50, 60, 65, and 70 µV were tested. The cut points of 46 (75th percentile) and 65 µV (93rd percentile, which yielded the best Cox regression results) were used in subsequent analyses. The MMA algorithm was not tuned to the data.
Continuous patient characteristics were compared between those with TWA <65 and ≥65 µV using the t-test for independent samples ( Table 1 ), and the χ2 test was applied for dichotomous variables ( Table 2 ).
The relative risks of TWA for all-cause and cardiovascular death as well as for SCD were estimated with a Cox proportional hazards model using the following covariates: sex, age, body-mass index (BMI), daily smoking (yes/no), use of ß-blockers (yes/no), functional capacity class according to the New York Heart Association (NYHA), and reached percentage of expected age-adjusted maximal HR (205–age/2), as well as prior diagnoses of CHD (yes/no), MI (yes/no), diabetes (yes/no), hypercholesterolemia (yes/no), and hypertension (yes/no) ( Table 3 ). The NYHA score, a surrogate for heart failure, was transformed into a dichotomous variable by differentiating the classes with good (II or less) or poor (III) functional class. The statistical analyses were performed with the SPSS release 12.0.1 for Windows (SPSS Inc., Chicago, IL, USA). All statistical tests were two-tailed and used an alpha level of <0.05. Sensitivity, specificity as well as positive predictive value (PPV) and negative predictive value (NPV) were calculated ( Table 4 ).
During the follow-up period of 44 ± 7 months (mean ± SD), there were 59 deaths (5.7% of the population). Of those, 34 (3.3%) were classified as cardiovascular death and 20 (1.9%) further as SCD. The causes of death for four (0.4%) patients remained unknown. Patient characteristics and number of deaths for those with TWA <65 µV (n = 950) and TWA ≥65 µV (n = 87) are given in Table 1 and Table 2 . EF for 529 patients was 65 ± 15% (mean ± SD). Only 67 patients (12.7%) had EF < 50%, and only eight patients (1.5%) presented with EF <30%.
The mean values (±SD) for the peak TWA levels were 39 ± 19 µV for patients without events (controls), 47 ± 26 µV for all-cause mortality (P = 0.01 in t-test for independent samples compared with controls), 55 ± 30 µV for cardiovascular deaths (P = 0.006), and 56 ± 34 µV (P = 0.04) for SCD. The prevalence of diabetes mellitus was the only parameter that differed among the groups with statistical significance of P = 0.04.
Mortality and T-Wave Alternans
The unadjusted prevalence of the three classes of mortality for the patient groups divided by two different TWA cut points (46 and 65 µV) are shown in Figure 1.
Using Cox regression, the unadjusted relative risk for SCD at the cut point of 65 µV was 6.3 (95% CI, 2.5–15.9; P < 0.001), for cardiovascular mortality 5.6 (95% CI, 2.7–11.4; P < 0.001), and for all-cause mortality 3.3 (95% CI, 1.8–6.1; P < 0.001). After adjustments were made for sex, age, BMI, smoking, use of ß-blockers, reached percentage of expected maximal HR, dichotomous NYHA class, and for prior diagnoses of CHD, MI, diabetes, hypercholesterolemia, and hypertension, the relative risk for SCD was 7.4 (95% CI, 2.8–19.4; P < 0.001; Figure 2A), for cardiovascular mortality 6.0 (95% CI, 2.8–12.8; P < 0.001; Figure 2B), and for death from any cause 3.3 (95% CI, 1.8–6.3; P = 0.001; Figure 2C). The corresponding values using TWA cut point of 46 µV are presented in Table 3 . Sex, BMI, prior MI, and the existence of diabetes were the statistically significant covariates for cardiovascular death, whereas none of the covariates reached significance for SCD ( Table 3 ).
Sensitivity, specificity, as well as PPVs and NPVs are given in Table 4 .
The findings of the present study, obtained in a large cohort of patients undergoing a clinically indicated exercise test, show that exercise-induced TWA is a strong predictor of cardiovascular mortality, especially of SCD. These observations widen the potential clinical applications of TWA analysis to a more general population of patients not suffering from congestive heart failure and/or depressed left ventricular function.
Previous studies using spectral analysis have consistently shown that positive TWA during the exercise test indicates an increased risk of mortality. The relative risks of mortality for patients with a positive TWA in the present study are comparable with the summary relative risk of 3.8 (95% CI, 2.4–5.9) in 19 studies relating TWA to cardiac arrhythmic events or deaths. In the previous studies, 2398 of the 2608 patients had congestive heart failure, a prior MI, or an implantable cardioverter defibrillator. The studies were therefore performed on populations consisting mostly of high-risk patients, which is also evident in higher cardiovascular mortality, typically 4–10% per year,[8-13,15,22] compared with the 0.9% per year in the current study. The EF data provide another indication that our patients were in a low-risk category compared with the populations studied previously. EF is normal for the great majority of the present cases. It is probable that those in whom EF was not measured had even better cardiovascular health, because there was no need for EF determination. The literature indicates that EF is an arrhythmia risk stratifier only when the EF levels are below normal.
In our study, elevated TWA specifically identified patients at increased 3–4-year risk of SCD. The SCD in a general population with no congestive heart failure is most commonly caused by ventricular fibrillation triggered by an ischaemic event. Therefore, it is plausible to speculate that the presence of TWA during an exercise test reflects the existence of abnormal repolarization making the heart vulnerable to ventricular fibrillation at the time of an ischaemic event. The majority of animal and clinical studies have shown that TWA is caused by underlying regional inhomogeneities of ventricular repolarization predisposing to ventricular arrhythmias. Prior clinical studies have mostly included patients with a specific substrate for ventricular tachyarrhythmia, such as infarct scar or heart failure. In the present study, a large proportion of patients presented with TWA without a prior infarction or congestive heart failure, suggesting that elevated TWA can also be present in hearts without an evident structural substrate for life-threatening arrhythmia.
Previous studies have typically employed special electrodes and the spectral analysis technique in the analysis of TWA. We used standard electrodes and a time-domain TWA method, MMA analysis, which is capable of analysing the data with fluctuating HRs. The methodology incorporates an incremental update factor, which is used in constructing the average beat. We tested the incremental update factor values of 1/32 and 1/8, and the latter proved superior in predicting mortality, when a cut off value of HR <125 b.p.m. was used. However, the methodology was relatively robust in adjusted Cox regression analyses with either of the two incremental update factor values: TWA remained as a significant harbinger of cardiovascular and SCD at any of the cut points applied (data not shown for 50, 60, and 70 µV).
MMA analysis of TWA has previously been employed with ambulatory ECGs to stratify arrhythmia risk in a low-risk population of post-MI patients. Nested case–control analysis revealed 4–7-fold higher odds of life-threatening arrhythmias with TWA cut points of 46–53 µV, depending on the lead selected. The results of that preliminary study are remarkably comparable to those of the present, much larger general population-based investigation. These observations underscore the robustness of the MMA method and its utility in both ambulatory ECG monitoring and exercise testing. Recent findings by Shusterman et al. further underscore the potential utility of MMA and other time-domain-based methods in detecting surges in TWA immediately prior to life-threatening ventricular tachyarrhythmias.
The NPV of the present data ( Table 4 ) is highly comparable with the results of the spectral method, for which the NPV averages 97.2 (with a range of 96.5–97.9). PPV is calculated with reference to the prevalence of the particular disease in the population studied. For this reason, the three studies in lower risk post-MI patients reviewed by Gehi et al. reported a PPV of 6.0 (95% CI, 4.5–7.4) for cardiac arrhythmic events. Thus, the PPV in the present study ( Table 4 ) is comparable or even better than that summarized by Gehi et al.
There are some study limitations that may prevent the generalization of the present results. The definition of SCD is never clear-cut. We used death within 24 h after the onset of symptoms as a definition for SCD. It is possible that some of these deaths are not due to ventricular tachyarrhythmia. However, TWA was a strong predictor of cardiovascular mortality but did not predict well the non-cardiac deaths, showing that the occurrence of TWA during exercise reflects abnormal cardiac electrical or mechanical function predisposing to cardiac death. Another limitation is that we do not have information on changes in parameters affecting mortality risk (e.g. smoking, lifestyles, and medication) during the follow-up.
In conclusion, TWA assessed with the time-domain MMA method and standard electrodes during a routine exercise test is a promising candidate for a clinically useful prognostic marker. Elevated TWA seems specifically to predict an increased 3–4-year risk of SCD. The prediction of SCD in a general population is a challenge, because a large cumulative number of SCDs occur among patients with no evidence of congestive heart failure or prior MI. A combination of positive TWA and other markers validated in low-risk populations could then be used to screen patients for an increased risk of SCD. More aggressive preventive strategies should then be recommended and applied to those patients.
Table 1. Patient Characteristics for All Participants According to T-Wave Alternans 65 µV Cut Point
Table 2. Unadjusted Percentage of Women, Frequency of ß-Blocker Use, as well as Prevalence of Cardiovascular Disease, Symptoms, Risk Factors, and Death for All Participants According to T-Wave Alternans <65 µV (n = 950) and T-Wave Alternans ≥65 µV (n = 87)
Table 3. Adjusted Relative Risks for Sudden Cardiac Death and Cardiovascular Mortality According to T-Wave Alternans, Adjusted by Covariates Used in the Cox Regression Models
Table 4. Sensitivity, Specificity, as well as Positive and Negative Predictive Values for Sudden Cardiac Death and for Cardiovascular and All-Cause Mortality Using Two Different T-Wave Alternans Cut Points
The authors wish to thank the staff of the Department of Clinical Physiology for collecting the exercise test data.
Financial support was received from the Medical Research Fund of Tampere University Hospital, the 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.
Tel: +358 3 311 65394; fax: +358 3 311 65511. E-mail: email@example.com
Tuomo Nieminen,1 Terho Lehtimäki,2,3 Jari Viik,4 Rami Lehtinen,3,5,6 Kjell Nikus,7 Tiit Kööbi,3,5 Kari Niemelä,7 Väinö Turjanmaa,3,5 Willi Kaiser,8 Heini Huhtala,9 Richard L. Verrier,10 Heikki Huikuri,11 Mika Kähönen3,5
1 Department of Pharmacological Sciences, Medical School, University of Tampere, Tampere, Finland
2 Laboratory of Atherosclerosis Genetics, Department of Clinical Chemistry, Tampere University Hospital, Tampere, Finland
3 Medical School, University of Tampere, Tampere, Finland
4 Ragnar Granit Institute, Tampere University of Technology, Tampere, Finland
5 Department of Clinical Physiology, Tampere University Hospital, FI-33520 Tampere, Finland
6 Tampere Polytechnic, University of Applied Sciences, Tampere, Finland
7 Heart Centre, Department of Cardiology, Tampere University Hospital, Tampere, Finland
8 GE Healthcare Information Technologies, Freiburg, Germany
9 School of Public Health, University of Tampere, Tampere, Finland
10 Harvard Medical School, Cardiovascular Institute, Beth Israel Deaconess Medical Center, Boston, MA, USA
11 Department of Medicine, University of Oulu, Oulu, Finland
Disclosure: W.K. is an employee of GE Healthcare Information Technologies, Freiburg, Germany.
R.L.V. is co-inventor of patents for T-wave alternans measurement including by the modified moving average method, which have been licensed to GE Healthcare.
The other authors do not have any conflicts of interest.