Author + information
- Received December 30, 2015
- Revision received March 15, 2016
- Accepted April 4, 2016
- Published online October 1, 2016.
- Rachel Bastiaenen, PhDa,b,
- Hanney Gonna, MBa,b,c,
- Navin Chandra, MDb,
- Ahmed Merghani, MBb,
- Oswaldo Valencia, MDa,
- A. John Camm, MDa,b and
- Mark M. Gallagher, MDa,b,∗ ()
- aDepartment of Cardiology, St. George’s University Hospitals NHS Foundation Trust, London, United Kingdom
- bInstitute of Cardiovascular and Cell Sciences, St. George’s University of London, London, United Kingdom
- cNational Heart and Lung Institute, Imperial College London, London, United Kingdom
- ↵∗Reprint requests and correspondence:
Dr. Mark M. Gallagher, Department of Cardiology, Director of Cardiac Electrophysiology, St. George’s Hospital NHS Foundation Trust, Blackshaw Road, London SW17 0QT, United Kingdom.
Objectives The purpose of this study was to determine the potential value of a novel marker for the severity of structural heart disease and the risk of arrhythmia.
Background The ventricular ectopic QRS interval (VEQSI) has been shown to identify structural heart disease and predict mortality in an unselected population. In ischemic heart disease (IHD), risk stratification for sudden death is imperfect. We hypothesized that VEQSI would identify patients with prior myocardial infarction (MI) compared with healthy subjects and distinguish IHD patients who have suffered life-threatening events from those without prior significant ventricular arrhythmia.
Methods The 12-lead Holter recordings from 189 patients with previous MI were analyzed: 38 with prior ventricular tachycardia/ventricular fibrillation (MI-VT/VF) (66 ± 9 years; 92% male); 151 without prior significant ventricular arrhythmia (MI-no VT/VF) (64 ± 11 years; 74% male). These were compared with 60 healthy controls (62 ± 7 years; 70% male). All ventricular ectopic beats were reviewed and maximal VEQSI duration (VESQI max) was recorded as the duration of the longest ventricular ectopic beat.
Results VEQSI max was longer in post-MI patients compared with normal controls (185 ± 26 ms vs. 164 ± 16 ms; p < 0.001) and in MI-VT/VF patients with prior life-threatening events compared with MI-no VT/VF patients without prior life-threatening events (214 ± 20 ms vs. 177 ± 22 ms; p < 0.001). Multivariate analysis established VEQSI max as the strongest independent marker for prior serious ventricular arrhythmia. VEQSI max >198 ms had 86% sensitivity, 85% specificity, 62% positive predictive value, and 96% negative predictive value for identifying patients with prior life-threatening events (odds ratio: 37.4; 95% confidence interval: 13.0 to 107.5).
Conclusions VEQSI max >198 ms distinguishes post-MI patients with prior life-threatening events from those without prior significant ventricular arrhythmia. This may be a useful additional index for risk stratification in IHD.
- implantable cardioverter-defibrillator
- ischemic heart disease
- sudden cardiac death
- ventricular ectopic beat
- ventricular ectopic QRS interval
In patients with ischemic heart disease (IHD), reduced left ventricular ejection fraction (LVEF) remains the best established predictor of sudden cardiac death (SCD) (1–3). However, in primary prevention trials that selected individuals for implantable-cardioverter defibrillator (ICD) therapy predominantly on the basis of reduced LVEF, only one-third had appropriate device therapy over the 3- to 5-year follow-up period (1,2). This raises concern that many patients are exposed to the risk and expense of ICD therapy from which they receive no benefit. The converse is of greater concern: as most IHD-related SCD occurs in patients with LVEF >35%, many who might benefit are denied the protection of an ICD if this criterion is used alone (4).
Of many electrocardiographic indices proposed as markers of risk for SCD, only the conducted QRS interval has shown consistent predictive value in survivors of myocardial infarction (MI) (1). With an intact conduction system, however, the QRS remains narrow even in the presence of ventricular dilatation and impairment. Ventricular ectopic beats (VEBs) are usually conducted through ventricular myocardium with limited participation of specialized conduction tissue and should therefore provide a better index of the state of the myocardium and risk of SCD (5). In an unselected population attending for Holter monitoring, we have shown that the ventricular ectopic QRS interval (VEQSI) and number of VEB morphologies correlated with the presence of structural heart disease and predicted all-cause mortality (6). Fragmentation of the conducted QRS and paced ventricular electrogram fractionation have also been shown to identify patients at risk of SCD (7,8). By extrapolation, fragmentation of the QRS complex of VEB may therefore also serve as a marker of risk.
We hypothesized that maximal VEQSI duration (VEQSI max), the number of VEB morphologies, and maximal VEB fragmentation (VEB fragmentation max) would identify patients with prior MI compared with healthy subjects. We hypothesized that these VEB indices would distinguish IHD patients who have suffered ventricular tachycardia/ventricular fibrillation (VT/VF) from those without a history of significant ventricular arrhythmia, independent of LVEF and conducted QRS interval.
We recruited 189 patients with previous MI, identified from coronary care records and the ICD clinic of St. George’s Hospital, London. Acute MI was defined as symptoms and electrocardiogram (ECG) changes consistent with infarction and elevated cardiac troponin. Inclusion criteria were MI at least 3 months before enrollment and cardiac catheterization followed by revascularization where appropriate. There were 151 patients (age 64 ± 11 years; 74% male) without prior significant ventricular arrhythmia (MI-no VT/VF cohort) and 38 patients (66 ± 9 years; 92% male) with secondary prevention ICD implantation for prior life-threatening ventricular arrhythmia (MI-VT/VF cohort). Qualifying ventricular arrhythmic events in the MI-VT/VF cohort had occurred at least 3 months post-MI. Clinical assessment comprised documentation of medical history and medications; physical examination including blood pressure, pulse, height, and weight; and blood sampling for renal function, brain natriuretic peptide (BNP), and inflammatory markers (C-reactive protein and erythrocyte sedimentation rate).
Patients were compared with 60 healthy controls (62 ± 7 years; 70% male). These were individuals without known cardiac risk factors, history of cardiac disease, or family history of inherited heart disease. These healthy volunteers had no significant abnormality on ECG and transthoracic echocardiography.
The study had previously been given ethical approval by the Outer West London ethics committee, and it complied with the Declaration of Helsinki.
Digital 10-s 12-lead ECGs were acquired using laptop-based software (Cardiosoft, GE Healthcare, United Kingdom) and reviewed at 10 mm/mV and 25 mm/s. Intervals including PR, RR, QRS, and QT were recorded in milliseconds. The QT interval was corrected using Bazett’s formula. Pathological Q waves and QRS fragmentation (fQRS) were considered present when observed in ≥2 ECG leads in the same coronary artery territory. A Q-wave was defined as ≥40 ms in duration or >25% of the following R-wave in voltage. fQRS included various RSR patterns, as previously described (9). Ventricular paced QRS complexes were excluded from Q-wave and fQRS analysis.
Holter monitoring was performed for a 24-h period. Digital 10-electrode 12-channel recording devices with a sampling frequency of 1,024 Hz (CardioMem CM 3000-12, Getemed, Germany) were applied in the Mason-Likar configuration. Analysis was performed on a workstation using commercial Holter analysis software (Cardioday, Getemed).
All recordings were analyzed by the same physician, blinded to the clinical diagnosis, who performed careful manual over-reading to eliminate artifact and correct the automated identification of VEB and their classification by morphology. Eleven traditional Holter ECG variables were selected for evaluation: VEB frequency; ventricular couplets; episodes of nonsustained ventricular tachycardia (NSVT); maximum heart rate during NSVT; minimum, mean, and maximum heart rate; time domain indices of heart rate variability (HRV) (SD of NN intervals and HRV triangular index), and frequency domain indices of HRV (high- and low-frequency power). NSVT was defined as ≥3 consecutive VEBs. Frequent VEBs were defined as VEB >1/min (10). Recordings with persistent atrial arrhythmia, persistent pacing, high frequency of ectopic beats, and/or poor quality were excluded from HRV analysis.
All VEBs in each recording were inspected. Differences in VEB morphology were identified with reference to bundle branch block pattern, QRS axis, and R-wave progression (11). The number of different VEB morphologies was counted and recorded. VEQSI and VEB fragmentation were measured for each VEB morphology from a single representative QRS complex chosen for the clarity of its onset and termination (Figure 1). Fusion beats, couplets, and NSVT were excluded from analysis. VEQSI measurements were made using electronic callipers on a simultaneous 12-derivation ECG segment at 20 mm/mV and 100 mm/s. We measured from the start of the QRS showing the earliest onset to the end of the QRS showing the latest termination. The duration of the broadest VEB was considered to be the VEQSI max of that patient (6). Fragmentation measurements were made on a simultaneous 12-derivation ECG segment at 10 mm/mV and 25 mm/s. VEB fragmentation was defined as >2 notches in the R′ or S waves and/or 2 notches separated by >40 ms (12). We recorded the total number of fragmented leads for each VEB morphology (excluding lead aVR). The VEB with the maximum number of fragmented leads was considered to be the VEB fragmentation max for that patient.
Effect of coupling interval on VEQSI
A subset of 10 Holter recordings with frequent VEB was reviewed to determine the effect of coupling interval on VEQSI. The predominant VEB morphology in each recording was identified and VEQSI was measured for the maximum and minimum coupling intervals and 4 additional coupling intervals within this range.
Echocardiography was performed using standard views from the parasternal and apical windows to acquire 2-dimensional, color Doppler, and color tissue Doppler images (VIVID 7 with 4S-MHz probe, GE Vingmed Ultrasound, Horten, Norway). Three consecutive cardiac cycles were recorded for each view at end expiration. Left ventricular end-diastolic diameter (LVEDD), left ventricular end-systolic diameter, and left ventricular wall thickness were derived from conventional 2-dimensional and M-mode images in the parasternal long- and short-axis views. LVEF was calculated by Simpson’s biplane method using apical 4- and 2-chamber views. Results were compared with American Society of Echocardiography/European Society of Cardiology (ASE/ESC) guidelines to derive normal and abnormal values and to quantify the degree of abnormality present (13).
Patients in the MI-VT/VF cohort were followed up for death and/or further life-threatening events using patient records, ICD records, and stored intracardiac electrograms. Events were considered life threatening when appropriate shock therapy was delivered for VF or rapid sustained VT (rate >200 beats/min).
Statistical data analysis was performed with SPSS version 21.0 (SPSS Inc., Chicago, Illinois). Univariate analysis of dichotomous, categorical, and continuous data was performed to determine their influence or relationship with prior ventricular arrhythmia. The distribution of continuous variables was assessed for normality using the Shapiro-Wilk test. Comparison between groups of continuous data was performed via independent samples Student's t test after controlling for equality of variance using Levene’s statistic, or the Mann-Whitney U test where appropriate. The chi-square test or Fisher exact test were used for categorical data. Several models of multivariate regression analysis were made using all available variables, and the most significant markers of prior life-threatening ventricular arrhythmia were established using forward stepwise (likelihood ratio) logistic regression analysis. The multivariate analysis was validated using a bootstrap method with 1,000 repeat samples from the dataset. Receiver operator characteristics curve analysis was used to determine an optimal cutoff value for VEQSI max. Correlations between distributions were made using the Spearman method. A 2-tailed p value <0.05 was considered significant.
Comparison of normal controls and patients with prior myocardial infarction
The VEB indices were all greater in patients with previous MI compared with normal controls: VEQSI max (185 ± 26 ms vs. 164 ± 16 ms; p < 0.001); number of VEB morphologies (3 ± 3 vs. 2 ± 2; p < 0.001); and VEB fragmentation max (7 ± 5 vs. 2 ± 4; p < 0.001) (Table 1).
Patients with prior myocardial infarction: Comparison of those with and those without prior life-threatening ventricular arrhythmia
At the time of assessment, timing of the initial MI was more remote for the cohort with prior life-threatening events (MI-VT/VF) than the cohort without prior significant ventricular arrhythmia (MI-no VT/VF). Patients in the MI-VT/VF cohort included more men, with a higher New York Heart Association functional class, BNP, urea and creatinine levels, and more frequent use of antiarrhythmic medications (Table 2).
Electrocardiogram and echocardiogram characteristics
The conducted QRS duration was longer in MI-VT/VF patients than MI-no VT/VF patients (112 ± 45 ms vs. 94 ± 14 ms; p < 0.001). Other ECG characteristics were similar. LVEF was lower (40 ± 17% vs. 55 ± 17%; p < 0.001), and LVEDD was higher (58 ± 1 mm vs. 49 ± 1 mm; p < 0.001) in the MI-VT/VF cohort compared with the MI-no VT/VF cohort (Table 2).
VEBs were present in 97% of MI-VT/VF patients and 91% of MI-no VT/VF patients. The 24-h VEB count was higher in the MI-VT/VF cohort than the MI-no VT/VF cohort (244 ± 714 vs. 30 ± 315; p < 0.001). Ventricular couplets (3 ± 11 vs. 0 ± 1; p < 0.001) and NSVT (34% vs. 11%; p = 0.001) were more frequent in MI-VT/VF patients than MI-no VT/VF patients (Table 2).
VEQSI max was longer in MI-VT/VF patients compared with MI-no VT/VF patients (214 ± 20 ms vs. 177 ± 22 ms; p < 0.001) (Table 2, Figure 2). When patients were subdivided according to LVEF (normal/mildly impaired >45%; moderately impaired 35% to 45%; severely impaired <35%) and QRS duration (<120 ms; ≥120 ms), VEQSI max remained longer in the MI-VT/VF cohort within all subdivisions of LVEF and conducted QRS intervals (Table 3).
There was no significant change in VEQSI max within the physiological range of coupling intervals demonstrated during Holter monitoring (Figure 3). In particular, VEQSI max did not prolong at shorter coupling intervals.
Number of VEB morphologies
The number of VEB morphologies was greater in MI-VT/VF patients than MI-no VT/VF patients (6 ± 4 vs. 3 ± 2; p < 0.001) (Table 2).
Maximal VEB fragmentation (fragmentation max)
VEB fragmentation max was greater in MI-VT/VF patients than MI-no VT/VF patients (8 ± 3 vs. 6 ± 4; p = 0.004) (Table 2).
Markers of prior life-threatening events
Several univariate markers of prior significant ventricular arrhythmia were identified including the VEB indices (VEQSI max, number of VEB morphologies, VEB fragmentation max), blood markers (BNP, urea, creatinine), conducted QRS duration, Holter variables (VEB count, couplet count, presence of complex VEB, NSVT), and echocardiographic parameters (LVEDD, LVEF) (Table 2). After multivariate logistic regression analysis, only VEQSI max and LVEDD remained independent markers. VEQSI max demonstrated the strongest association, with a 1-ms increase in VEQSI max increasing the odds of prior life-threatening events by a factor of 1.06 (95% confidence interval [CI]: 1.03 to 1.09; p < 0.001) (Table 4). The bootstrap method confirmed that the magnitude of association for VEQSI max and LVEDD with prior significant ventricular arrhythmia withstood resampling and is unlikely to be incidental.
The probability of prior significant ventricular arrhythmia increased with VEQSI max duration. Receiver operating characteristic curve analysis was used to determine the optimal VEQSI max cutoff value associated with prior life-threatening events. VEQSI max >198 ms had 86% sensitivity, 85% specificity, 62% positive predictive value, and 96% negative predictive value for this (area under curve [AUC]: 0.90; 95% CI: 0.85 to 0.95) (Table 5, Figure 4) with an odds ratio: 37.4; 95% CI: 13.0 to 107.5. VEQSI max was the superior marker compared with LVEDD (AUC: 0.90; SE: 0.028 vs. AUC: 0.81; SE: 0.04, respectively).
Relationships among VEQSI max, the number of VEB morphologies, VEB fragmentation max, and left ventricular structural changes
There was moderate correlation between VEQSI max and LVEDD (rs = 0.59; p < 0.001) and VEQSI max and LVEF (rs = −0.58; p < 0.001). Correlations between number of VEB morphologies and LVEDD (rs 0.46; p < 0.001), number of VEB morphologies and LVEF (rs = −0.42; p < 0.001), VEB fragmentation max and LVEDD (rs = 0.38; p < 0.001), and VEB fragmentation max and LVEF (rs = −0.34; p < 0.001) were less strong but still significant.
Antiarrhythmic drug therapy
In the MI-VT/VF cohort, there were 12 patients receiving long-term amiodarone therapy. As amiodarone use can influence conduction properties, additional analysis was performed following the exclusion of these patients. VEQSI max remained significantly longer in the MI-VT/VF cohort compared with the MI-no VT/VF cohort (210 ± 16 ms and 172 ± 21 ms, respectively; p < 0.001). No patient received any other class III antiarrhythmic medication, and none received any class I antiarrhythmic.
During a mean follow-up period of 48 ± 11 months, 10 patients (26%) in the MI-VT/VF cohort suffered further VT/VF events requiring defibrillation and 7 patients (18%) died. These patients all had VEQSI max duration >198 ms. VEQSI max was longer in MI-VT/VF patients who died or had subsequent VT/VF events requiring defibrillation than in the MI-VT/VF patients who survived event free (221 ± 19 ms and 205 ± 20 ms, respectively; p = 0.028).
In this study, we have shown that VEB indices (VEQSI max, the number of VEB morphologies, and VEB fragmentation max) identified patients with prior MI compared with healthy subjects. In IHD patients, these VEB indices distinguished those who had suffered life-threatening events (MI-VT/VF) from those without a history of significant ventricular arrhythmia (MI-no VT/VF). VEQSI max was greater in MI-VT/VF patients irrespective of LVEF and conducted QRS interval, and it was the strongest independent marker for prior life-threatening ventricular arrhythmia.
We have previously shown that in unselected patients attending for outpatient Holter monitoring, VEQSI max correlated with presence and severity of structural heart disease, and multiple VEB morphologies predicted all-cause mortality (6). ECG data recorded during cardiac catheterization has shown that broadly notched VEB ≥160 ms are markers of left ventricular dilatation and impairment (14). Slowed conduction through diseased myocardium has been shown to result in longer QRS duration during VT in patients with arrhythmogenic right ventricular cardiomyopathy (ARVC), and data from electrophysiological studies have shown that longer VEB duration is associated with myocardial scar (15–17). Broader VEBs have also been associated with development of nonischemic cardiomyopathy (18–20). The greater VEQSI max demonstrated in our IHD patients with prior life-threatening events likely reflects a greater amount of underlying scar and slowed conduction. The incidence of Q-wave MI was also higher in these MI-VT/VF patients compared with MI-no VT/VF patients, albeit not significantly.
In this study, although increased VEQSI max duration correlated with decreased LVEF, it was an independent variable that distinguished IHD patients with prior life-threatening events from those without significant ventricular arrhythmia. Multivariate logistic regression analysis showed it to be the most significant and consistent marker for this. VEQSI max >198 ms had high sensitivity and specificity for the identification of post-MI patients with a previous life-threatening event. In addition VEQSI max was greater in IHD patients with prior significant ventricular arrhythmia who died or suffered a subsequent life-threatening event compared with those who survived with no further significant ventricular arrhythmia during the follow-up period of 48 months. This suggests that VEQSI max may offer incremental value for risk stratification in patients following MI.
Fragmentation of the conducted QRS has been shown to correlate with myocardial scar and predict risk in ischemic cardiomyopathy and Brugada syndrome (BrS) (7,21). In hypertrophic cardiomyopathy (HCM), increased fractionation of paced right ventricular electrograms has been shown to correlate with the risk of VF (8,22). In a prospective study, paced electrogram fractionation analysis (PEFA) also predicted patients at risk of SCD with greater accuracy than noninvasive techniques (23). Therefore, we may have expected fragmentation of the VEB to serve as a diagnostic and risk stratification tool in cardiomyopathy, but in our multivariate analysis, it did not feature, apparently because of the superior predictive power of VEQSI max.
Comparison of VEQSI max, the number of VEB morphologies, and VEB fragmentation max as indices for previous significant ventricular arrhythmia showed that VEQSI max was superior. It was also more convenient. Although 12-lead Holter monitoring improves the ability to differentiate between VEB morphologies compared with older 3-derivation systems, QRS axis is subject to change with posture, and the laborious manual over-reading required to correct for this is likely to also limit clinical applicability. Measurement of VEB fragmentation by the methods that we have used is not possible using a standard 3- or 5-lead Holter system but requires a 12-lead system, limiting its utility in clinical practice. The automated measurement of VEQSI max is likely to be more robust than that of VEB fragmentation or the number of morphologies, and we have previously demonstrated that this index can be determined using Holter monitoring systems with fewer leads (6). We therefore consider VEQSI max to be the most useful of the 3 indices.
Randomized clinical trials have established that ICD therapy can improve survival for individuals at risk of SCD (1). It is therefore important to correctly identify at-risk individuals for treatment. In patients with IHD, reduced LVEF remains the best established predictor of SCD, but this is imperfect (1,2). The majority of SCD occurs in those with low-, intermediate-, or no risk factors and in primary prevention trials that selected individuals for ICD therapy predominantly on the basis of reduced LVEF, only one-third had appropriate device therapy over the 3- to 5-year follow-up period (1,2,4).
Prospective follow-up data are needed to determine the potential role for combining VEQSI max with LVEF and conducted QRS interval in risk analysis algorithms as well as the optimal cutoff value for VEQSI max. Our dataset does demonstrate overlap within the 180- to 200-ms range between the 2 groups. From a clinical perspective, a cutoff value of 198 ms appears most useful. This affords high sensitivity while maintaining good specificity. When selecting patients for prophylactic ICD implantation, high sensitivity and identification of true positives, those at highest risk of ventricular arrhythmia, would appear to be the more important factors. Longer prospective follow-up will be important to determine outcomes of patients with VEQSI max >198 ms, particularly those in the MI-no VT/VF cohort.
Consecutive patients treated for MI at St. George’s Hospital and all IHD patients undergoing follow-up in the ICD clinic with secondary prevention devices were invited to take part in the study. Only one-third agreed, which introduces a possibility of selection bias. Our data showed VEQSI max was a stronger marker for prior life-threatening events than LVEF, but the MI-VT/VF sample size was modest and included few patients with low LVEF. In addition, all patients were recruited prospectively, but the majority of events occurred before recruitment, and this is a retrospective study. Prospective follow-up data are required to determine the outcome of patients with longer VEQSI max without a history of serious arrhythmia at the time of assessment.
The VEQSI max distinguishes IHD patients who have suffered life-threatening events from those without a history of significant ventricular arrhythmia. VEQSI max shows promise as an additional risk stratification tool for sudden death to be considered for use in combination with existing indices.
COMPETENCY IN MEDICAL KNOWLEDGE: Reduced LVEF is the best established predictor of sudden death in patients with IHD, but the majority of events occur in those with LVEF >35%. In this study, the VEQSI max was the strongest marker of prior life-threatening ventricular arrhythmia in post-MI patients.
TRANSLATIONAL OUTLOOK: VEQSI max shows promise as an additional risk stratification index in IHD. Prospective follow-up data in a larger cohort are required. This will be of particular interest in patients with LVEF 35% to 50%.
The authors thank Kenny McKay and Rosemary Rance of Boston Scientific UK for their help in securing funding for this project.
Supported by an unrestricted educational grant from Boston Scientific. Dr. Gonna has received research funding from Biosense Webster and Boston Scientific. Dr. Gallagher has received research funding from Boston Scientific and Medtronic. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Drs. Camm and Gallagher contributed equally to this work.
- Abbreviations and Acronyms
- area under the curve
- brain natriuretic peptide
- confidence interval
- implantable cardioverter-defibrillator
- ischemic heart disease
- left ventricular ejection fraction
- myocardial infarction
- MI-no VT/VF
- prior myocardial infarction and no significant ventricular arrhythmia
- prior myocardial infarction and life-threatening ventricular arrhythmia
- non sustained ventricular tachycardia
- sudden cardiac death
- ventricular ectopic beat
- ventricular ectopic QRS interval
- ventricular fibrillation
- Received December 30, 2015.
- Revision received March 15, 2016.
- Accepted April 4, 2016.
- American College of Cardiology Foundation
- Yap Y.G.,
- Duong T.,
- Bland J.M.,
- et al.
- Myerburg R.J.,
- Mitrani R.,
- Interian A.,
- Castellanos A.
- Bastiaenen R.,
- Batchvarov V.,
- Gallagher M.M.
- Gallagher M.M.,
- Padula M.,
- Sgueglia M.,
- et al.
- Saumarez R.C.,
- Heald S.,
- Gill J.,
- et al.
- Das M.K.,
- Khan B.,
- Jacob S.,
- Kumar A.,
- Mahenthiran J.
- Lown B.,
- Wolf M.
- Miller J.M.,
- Marchlinski F.E.,
- Buxton A.E.,
- Josephson M.E.
- Lang R.M.,
- Bierig M.,
- Devereux R.B.,
- et al.
- Moulton K.P.,
- Medcalf T.,
- Lazzara R.
- Hoffmayer K.S.,
- Machado O.N.,
- Marcus G.M.,
- et al.
- Wijnmaalen A.P.,
- Stevenson W.G.,
- Schalij M.J.,
- et al.
- Del Carpio Munoz F.,
- Syed F.F.,
- Noheria A.,
- et al.
- Morita H.,
- Kusano K.F.,
- Miura D.,
- et al.
- Saumarez R.C.,
- Camm A.J.,
- Panagos A.,
- et al.
- Saumarez R.C.,
- Pytkowski M.,
- Sterlinski M.,
- et al.