Combined evaluation of ambulatory‐based late potentials and nonsustained ventricular tachycardia to predict arrhythmic events in patients with previous myocardial infarction: A Japanese noninvasive electrocardiographic risk stratification of sudden cardiac death (JANIES) substudy

Abstract Background Noninvasive electrocardiographic (ECG) markers are promising arrhythmic risk stratification tools for identifying sudden cardiac death. However, little is known about the usefulness of noninvasive ECG markers derived from ambulatory ECGs (AECG) in patients with previous myocardial infarction (pMI). We aimed to determine whether the ECG markers derived from AECG can predict serious cardiac events in patients with pMI. Methods We prospectively analyzed 104 patients with pMI (88 males, age 66 ± 11 years), evaluating late potentials (LPs), heart rate turbulence, and nonsustained ventricular tachycardia (NSVT) derived from AECG. The primary endpoint was the documentation of ventricular fibrillation or sustained ventricular tachycardia. Results Eleven patients reached the primary endpoint during a follow‐up period of 25 ± 9.5 months. Of the 104 patients enrolled in this study, LP positive in worst values (w‐LPs) and NSVT were observed in 25 patients, respectively. In the arrhythmic event group, the worst LP values and/or NSVT were found in eight patients (7.6%). The positive predictive and negative predictive values of the combined assessment with w‐LPs and NSVT were 56% and 94%, respectively, for predicting ventricular lethal arrhythmia. Kaplan–Meier analysis demonstrated that the combination of w‐LPs and NSVT had a poorer event‐free period than negative LPs (p < .0001). In the multivariate analysis, the combined assessment of w‐LPs and NSVT was a significant predictor of arrhythmic events (hazard ratio = 14.1, 95% confidence intervals: 3.4–58.9, p < .0001). Conclusion Combined evaluation of w‐LPs and NSVT was a powerful risk stratification strategy for predicting arrhythmia that can lead to sudden cardiac death in patients with pMI.

Although the ICD implantation rate has increased over the last two decades, SCD has not reduced (Benjamin et al., 2017). Recent reports indicated that the left ventricular ejection fraction (LVEF), nonsustained ventricular tachycardia (NSVT), and programmed stimulation on electrophysiology studies provide a high level of evidence for identifying arrhythmic SCD risk, necessitating ICD implantation (Al-Khatib et al., 2018;Priori et al., 2015;Scirica et al., 2010). However, these markers do not cover patients at risk for aborted SCD.
Ventricular late potentials (LPs) on a signal-averaged electrocardiogram (SAECG) are useful for identifying the risk of arrhythmia leading to SCD in patients with previous myocardial infarction (pMI) who have reduced cardiac function (Gomes et al., 1987;Malik et al., 1992;Steinberg, Regan, Sciacca, Bigger, & Fleiss, 1992). Since the 2000s, primary percutaneous coronary intervention (PCI) has been performed during the earlier stages of cardiac dysfunction and drug-eluting stents have been considered the gold standard for PCI. Therefore, more patients with pMI who performed PCI after 2000s have a preserved LVEF rather than 1980-1990s. However, most investigations regarding risk stratification for fatal arrhythmic events using SAECG were done before 2000s. Recently, LPs have been measured for 24 hr using high-resolution ambulatory electrocardiogram (AECG) systems (Amino et al., 2019;Hashimoto et al., 2018). The use of ambulatory-based LPs for identifying lethal arrhythmias in patients with pMI is not fully understood.
Noninvasive electrocardiographic (ECG) markers, such as heart rate turbulence (HRT) and NSVT, are useful for predicting cardiac events in various clinical settings. However, the predictive value of LPs, HRT, and NSVT alone is limited, especially if the positive predictive value (PPV) is low (Ikeda et al., 2000;Miwa et al., 2012).
We hypothesized that a combined assessment of these indices, including ambulatory-based LPs, could predict lethal arrhythmias. Therefore, this study aimed to prospectively evaluate the usefulness of ambulatory-based LP measurements and determine whether simultaneous assessment of noninvasive ECG risk stratification factors, such as ambulatory-based LPs, HRT, and NSVT, can be used to predict fatal arrhythmic events, such as sustained ventricular tachycardia (VT) or ventricular fibrillation (Vf) in patients with pMI.

| Study design and ethics
This study was a substudy of the Japanese Noninvasive

Electrocardiographic Risk Stratification for Prediction of Sudden
Cardiac Death (JANIES) study (Kinoshita et al., 2020). The JANIES study was designed as a multicenter, observational, prospective cohort study and evaluated noninvasive ECG indices, such as LPs, HRT, and NSVT, which can be analyzed simultaneously using a 24-hr digital AECG, to predict cardiac mortality and/or lethal arrhythmias in highrisk patients. Participants in the substudy were enrolled between April 2012 and March 2015. Inclusion and exclusion criteria used in this study are reported elsewhere (Kinoshita et al., 2020

| AECG recordings
Data derived from the 24-hr digital AECG system were analyzed to detect NSVT and specific noninvasive ECG markers, such as the 24-hr SAECG (24-hr SAECG) and HRT. All patients underwent 24-hr digital AECG with ordinary daily activities. The presence of NSVT

| Measurement of ambulatory-based LPs
Late potentials were recorded for all patients with PMI using the AECG system (Spider View; Ela, Paris, France). ECG data were obtained at a sampling rate of 1,000 Hz and 16-bit A/D conversion. For LPs  TA B L E 1 Baseline characteristics of the study patients (n = 104) measurement, ECG data were filtered and ranged from 40 to 250 Hz.

| HRT measurements
Heart rate turbulence was measured from the AECG, according to a previously established protocol (Bauer et al., 2008). It was measured if more than one determinate VPC occurred. HRT was evaluated by turbulence onset (TO) and turbulence slope (TS). TO reflects the early-phase sinus rhythm acceleration after VPC and is followed by TS, which represents the compensatory deceleration phase after VPC. TS was calculated as the steepest regression over any five consecutive sinus rhythm RR intervals after VPC and within 15 sinus rhythms (Bauer et al., 2008). TO ≥ 0% and TS ≤ 2.5 ms/RR intervals were considered abnormal. HRT was considered positive when the TO and TS were abnormal; it was considered negative when TO and/ or TS were abnormal or when HRT could not be calculated due to a lacking VPC.

| Assessment of other clinical parameters
We prospectively evaluated age, sex, hypertension, dyslipidemia, diabetes mellitus history, and EF using the modified Simpson method (Lang et al., 2015), and left ventricular dimension diameter (LVDD) using echocardiography. A reduced EF was defined as an EF < 40%.
Also, QRS duration and QTc intervals were measured from 12-lead ECG at the period near when the 24-hr AECG recordings were performed.

| Follow-up and study endpoints
The primary endpoint was the occurrence of fatal arrhythmic events, such as Vf or sustained VT. The causes of death were confirmed from medical or autopsy records, and from the primary doctors or those who had witnessed the death. For fatal arrhythmic events, the records were verified from ECG monitoring detected in the hospital room, 24-hr Holter electrocardiograms recorded from the hospital, or by confirming the use of an ICD. In patients with ICDs, an appropriate shock delivery for VF events and antitachycardia pacing for sustained VT were included in the fatal arrhythmic events.
Follow-up data were collected at 6-month intervals until September 2015.

| Statistical analyses
Data are presented as mean ± standard deviation for normally dis-

| Baseline characteristics
During the mean follow-up period of 25 ± 9.5 months, 11 patients reached the primary endpoint. Baseline characteristics of this study population are described in Table 1. In echocardiographic data, LVEF was lower in arrhythmic event group; conversely, LVDD was larger in nonarrhythmic event group (Table 1).

| LP parameters fluctuation throughout 24 hr
All LP parameters (fQRS, LAS 40 , and RMS 40 ) significantly fluctuated over the 24-hr period. The median fQRS was longest between 0:00 and 4:00; conversely, it was shortest between 12:00 and 16:00 (p < .0001) (Figure 1). Additionally, the median LAS 40 was longest between 4:00 and 8:00; conversely, it was shortest between 12:00 and 16:00 (p < .036). Conversely, the median RMS 40 was the largest between 8:00 and 12:00 and the smallest between 0:00 and 4:00 (p = .001) (Figure 1). The rate of w-LPs in the arrhythmic event group was significantly higher than the b-LPs positive rate in the nonarrhythmic event group (63% vs. 15%, respectively; p = .002). However, there was no difference between groups regarding b-LPs-positive rate (36% vs. 14%, in the arrhythmic and nonarrhythmic group, respectively; p = .079) ( Table 1). The worst value of fQRS and LAS 40 was significantly higher than the best value of fQRS and LAS 40 (p < .001, in both groups). The worst value of RMS 40 was significantly smaller than the best value of RMS 40 (p < .001). The noise level of best value of LP was significantly higher than that of the worst value of LP. (Table 2).

| Association between ECG risk stratification markers and end arrhythmic events
The clinical characteristics of the patients are shown in Table 3.
Vf was observed in three patients, and sustained VT was documented in eight patients. Ten of the eleven patients had a NYHA cardiac functional classification of I or II. EF < 40% was documented in four patients (Nos 1, 2, 9, and 11). Five patients died due to lethal arrhythmia in sustained VT and Vf (Nos 2, 3, and 11), or heart failure (Nos 4 and 5). Only three patients (Nos 2, 5, and 7) developed arrhythmic events without w-LPs positive and/or documented NSVT.  (Table S1). Moreover, comparison among combined noninvasive markers including EF was performed. As a result, w-LPs + NSVT was also an independent factor (HR: 12.2 [3.0-49.8], p < .0001) (Table S2). Kaplan-Meier analysis demonstrated that both w-LPs and NSVT positivity were associated with a significantly lower event-free rate than other combinations of w-LPs and HRT results (p < .0001) ( Figure 2). Moreover, we performed the analysis by Kaplan-Meier method between reduced or with preserved EF among three categories (w-LPs(−) and NSVT (−); w-LPs(+) or NSVT (+); w-LPs(+) and NSVT (+)). In all these categories, there were no significant differences between reduced (EF < 50%) or preserved EF (EF ≤ 50%) for cumulative event-free rate by log-rank test ( Figures S1-S3).

| D ISCUSS I ON
There were significant daily fluctuations in LP parameters in patients with pMI. Risk assessment, with consideration to diurnal variations of LP, was useful for predicting serious cardiac events. The combined evaluation of w-LPs and NSVT was a powerful risk stratification strategy for predicting fatal arrhythmic events after adjusting for age and EF. The results of this study included up-to-date information available on noninvasive risk stratification strategy in the contemporary ela, because PCI using drug-eluting stent tended to be performed after 2000s. As such, most patients included in this study had preserved systolic function.

| 24h-LP as a predictor of arrhythmic events
In this study, LP parameters significantly worsened during the night and improved during the daytime (Table 2 and Figure 1). The sensitivity, PPV, negative predictive value (NPV), and PLR among the w-LPs were higher than those among the b-LPs. Therefore, 24h-LP can be a useful Their results are similar to our findings. We also demonstrated that the assessment of w-LPs was useful for predicting arrhythmic events in patients with ischemic heart disease (IHD

| NSVT as a predictor of arrhythmic events
Nonsustained ventricular tachycardia can potentially cause lethal arrhythmia. Lethal arrhythmias, such as VT and/or Vf, finally occurred via VPC (based on electrophysiological substrate) and modulators such as autonomic nerve activity, electrolyte imbalance, and age (Ikeda, Yusu, Nakamura, & Yoshino, 2007). Several studies have demonstrated that NSVT is useful for predicting lethal arrhythmia in patients with pMI with reduced EF (Farrell et al., 1991;Maggioni et al., 1993). Recent AHA (Al-Khatib et al., 2018) and ESC (Priori et al., 2015) guidelines indicated that NSVT derived from AECG

TA B L E 2 The best and worst values of LP parameters
The best value of LP should be considered for ICD implantation. Recently, it was reported that the risk PPV of NSVT for arrhythmic events in reduced EF patients was low (range: 4%-16%); the NPV, on the other hand, was high (82%-97%). In this study, the NSVT was a statistically significant index for predicting arrhythmic events (HR: 3.6, 95% CI: 1.6-11.2, p = .027). Furthermore, the NPV was extremely high (94%), and the PPV was very low (26%) (Table 4). Therefore, in patients with pMI, NSVT alone is limited in its ability to predict arrhythmic events.

| HRT as a predictor of arrhythmic events
In our study, HRT was not a useful risk stratification marker. It has been reported that HRT was evidenced by several papers for identifying cardiac death or arrhythmic events in patients with pMI (Disertori, Masè, Rigoni, Nollo, & Ravelli, 2016). The exact reason why HRT was not a significant factor in this study is difficult to certify. It is speculated that this is due to the patient population in this study which had Abbreviations: HRT, heart rate turbulence; LP, late potential; LVEF, left ventricular ejection fraction; N-LRs, negative likelihood ratios; NPV, negative predictive value; NSVT, nonsustained ventricular tachycardia; PA, predictive accuracy; P-LRs, positive likelihood ratios; PPV, positive predictive value; RH, relative hazard; w-LPs, LP positive in worst value. HRT, deceleration capacity, or heart rate variability) was positive, then they performed electrophysiological test for provoking ventricular arrhythmia. In case of inducible VT, then ICD was implanted to prevent SCD. This algorithm yielded excellent sensitivity, specificity, and NPV.

TA B L E 5 Result of multivariate Cox regression analysis
In subanalysis of this study, LPs and NSVT-positive rate were significantly higher in the with inducible VT group than in the without inducible VT group. However, HRT was not significant between with inducible VT and without inducible VT. Moreover, none of patients was HRT-positive among nine patients who received appropriate ICD shock delivery due to spontaneous lethal arrhythmia. The results of their investigation partially support our data. Indeed, in our study data, HRT was positive in 6 patients out of 11 patients who developed VT or Vf (Table 3). Five of these six patients with arrhythmic events had low EF under 50% (Nos 1, 2, 4, 5, and 11). On the other hand, Bonnemeier et al reported that early reperfusion by PCI from TIMI 2 to TIMI 3 restored HRT parameter (Bonnemeier et al., 2003). Also, they reported that there was positive correlation between EF and TS; then, TS could be restored according to EF improvement. In our study, all of the patients received PCI or coronary artery bypass angioplasty (

| Combined evaluation of noninvasive risk stratification markers
The combined evaluation of w-LPs and NSVT was a powerful risk stratification strategy for predicting fatal arrhythmic events after adjusting for the age and EF (Table 5). To the best of our knowledge, this is the first study to evaluate the usefulness of combined assessment with LP measured for 24 hr and NSVT for identifying fatal arrhythmias. It was previously reported that a combined assessment of LPs recorded in real time and NSVT was useful for predicting serious arrhythmic events (el-Sherif et al., 1995;Kuchar, Thorburn, & Sammel, 1987). The PPV of these combined assessments was between 24% and 31% (el-Sherif et al., 1995;Kuchar et al., 1987). Real-time LP recordings are usually performed during the daytime and last approximately 15 min. We also demonstrated the usefulness of assessing the fluctuations in LP parameters for arrhythmic risk stratification in patients with pMI. In our study, the PPV of the combined evaluation of w-LPs and NSVT (56%) was superior to the PPVs reported in previous studies (el-Sherif et al., 1995;Kuchar et al., 1987). On the other hand, our study indicated that risk markers, such as w-LPs, EF, HRT, and NSVT, had limited applicability in clinical settings; this was because the PPVs were very low (18%-33%) and the NPVs were very high (91%-95%); this is consistent with previous studies (Table 4) (eL-Sherif et al., 1995;Ikeda et al., 2000;Miwa et al., 2012).
In this study, we demonstrated that the PPV of the combination of w-LPs and NSVT was higher than other single-use or combination ECG indices. Furthermore, the PLRs and PA were also higher than other single-use or combination ECG indices. There were no significant differences between reduced (EF < 50%) or preserved LVEF (EF ≥ 50%) for cumulative event-free rate by Kaplan-Meier method ( Figures S1-S3).
Therefore, there is a possibility that the combination of w-LP and NSVT is useful for risk stratification tool regardless of the EF. Also, multivariate Cox regression analysis demonstrated combination of w-LPs and NSVT had a higher hazard ratio than other single-use or combination indices (Tables S1 and S2). ICD insertion should be considered to prevent SCD when the w-LPs and NSVT are positive in patients with pMI.

| Limitations
This study has a few limitations. First, the sample size was small, which limited the power of the study. The JANIES study used F I G U R E 2 Kaplan-Meier curves for serious cardiac events based on the combination assessment of w-LPs and NSVT. The combination of w-LPs-positive values and NSVT-positive values was associated with a poorer event-free rate (p < .0001). HRT, heart rate turbulence; NSVT, nonsustained ventricular tachycardia; w-LPs, the worst value of late potential three commercially available AECG systems (Spider View; SCM 8000; Fukuda Denshi.; Mars, GE Healthcare Inc.). In this substudy, the enrolled patients were examined using the Spider View only.
Because the algorithm for the LPs filters characteristic differences depended on the device, we did not combine the analyses for the LP parameters.
Second, we did not include the analyses of repolarization abnormality indices, such as the T-wave variability (TWV). It has been reported that TWV is a useful marker for identifying the risk of lethal arrhythmia, which can lead to SCD in patients with IHD (Couderc, Zareba, McNitt, Maison-Blanche, & Moss, 2007). Future prospective studies should include a larger sample size and analysis of the repolarization abnormality index.
Third, our study results demonstrate that the noise level of b-LPs was significantly higher than that of w-LPs. There is a possibil- IHD. As a result, the SAECG parameters were significantly getting better with higher noise levels. Therefore, they mentioned that the noise level determines the QRS offset. When the noise level increased, the offset of QRS is shifted earlier in timing, compared to lower noise levels. Then fQRS could be shortened. Similarly, when the noise level was higher, the LAS 40 was shortened and RAS 40 was increased. In this study, LPs parameters showed daily fluctuation. In patients with IHD, it remains unknown whether SAECG parameters have diurnal variations or only daily fluctuations due to confounders such as a noise. In future studies, it is necessary to perform multivariate analysis for seeking the strongest factor which influences SAECG parameters including the noise level, body position HR, and autonomic nervous activity.
In conclusion, the combined evaluation of w-LPs and HRT derived from digital ambulatory-based ECG systems can be a powerful risk stratification strategy for predicting fatal arrhythmias, such as VT or Vf in patients with pMI with relatively preserved EF.

ACK N OWLED G M ENTS
The first author K.H. would like to express great appreciation to the author's advisor, Hiroaki Shimabukuro, who has been an exceptional mentor and for encouraging the author. We would like to thank Editage for English language editing.

All authors have no conflicts of interest in connection with this
article.

E TH I C A L A PPROVA L
The study was conducted in accordance with the Declaration of

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request, but no information infringing on the privacy of the participants will be given.