Author + information
- Received December 23, 2018
- Revision received April 17, 2019
- Accepted May 6, 2019
- Published online August 19, 2019.
- Abhishek Maan, MD, ScMa,
- Lou Sherfesee, PhDb,
- Daniel Lexcen, PhDb,
- E. Kevin Heist, MD, PhDa and
- Alan Cheng, MDb,∗ ()
- aCardiac Arrhythmia Service, Massachusetts General Hospital, Boston, Massachusetts
- bMedtronic, Mounds View, Minnesota
- ↵∗Address for correspondence:
Dr. Alan Cheng, Medtronic, 8200 Coral Sea Street, MVS-2, Mounds View, Minnesota 55112.
Objectives The aim of this study was to assess the variations in ventricular arrhythmia (VA) occurrence according to seasons, months of the year, days of the week, and the time of day in a large implantable cardioverter-defibrillator patient population.
Background Limited data exist on how VA occurrence varies as a function of time.
Methods Data from 6 prospective studies were pooled to assess VA frequency over time. All adjudicated episodes of VAs ≤500 ms were included. VA distribution as a function of hour, day, month, and season were assessed through the construction of 4 negative binomial models. The models included a random patient effect and offset for days spent in each time period.
Results Among 3,969 patients, 7,126.8 cumulative device-years with an average follow-up duration of 1.8 ± 1.4 years/patient were analyzed. VA occurrence was higher in the spring than the summer (0.86% vs. 0.70%; p = 0.009) but not significantly different from the fall (0.74%; p = 0.069) or winter (0.84%; p = 0.732). The estimated probability of occurrence of at least 1 VA episode in each 1-h block during the hours of 8 am to 10 pm over 365 days (0.10% to 0.12%) was higher (estimated 35% to 63% higher) than the referent period of midnight to 1 am (0.07%). No significant variations in VA occurrence were observed according to weekday and individual months of the year.
Conclusions Significantly higher VA occurrence in the spring and during the hours of 8 am to 10 pm were observed. Additional studies are needed to further understand the reasons for these observations, which may involve variations in temperature or differences in catecholamine triggers.
Implantable cardioverter-defibrillators (ICDs) continue to be the cornerstone in the management of patients at risk for sudden cardiac death (SCD). Over the past 15 years, we have gained greater understanding of the incidence of ICD therapies and have also appreciated the deleterious effects of ICD shocks on short-term and long-term outcomes among ICD recipients. Although changes in the way we program ICD devices have improved outcomes (1,2), a fundamental understanding of when ventricular tachyarrhythmic events occur remains elusive. Improvements in understanding this could enhance the way in which devices are programmed to further avoid unnecessary and inappropriate ICD therapies.
Prior studies have investigated temporal trends in SCD and have used information from patient interviews, death certificates, and electrocardiographic recordings. Unfortunately, these data shed little light on the distribution of ventricular arrhythmias (VAs) over time (3–6). We aimed to leverage continuously recorded information from ICDs to assess variations in frequency of VAs according to the time of day, weekday, months, and seasons in a pooled cohort of patients with ICDs.
We pooled data from 6 prospective clinical studies of patients undergoing ICD implantation (PainFree Rx II [Pacing Fast Ventricular Tachycardia Reduces Shock Therapies] ; EMPIRIC [Comparison of empiric to Physician-Tailored Programming of Implantable Cardioverter-Defibrillators Trial] ; WAVE [Prospective registry of single-chamber ICDs using the Wavelet feature: Worldwide Wave Investigators]; EnTrust [Clinical Investigation of the Medtronic EnTrust™ Implantable Cardioverter Defibrillator (ICD), Model D 153 ATG] [NCT00273195] ; MVP [Managed Ventricular Pacing versus VVI 40 Pacing Trial] , and OMNI [Assessing Therapies in Medtronic Pacemaker, Defibrillator, and Cardiac Resynchronization Therapy Devices] ). Most of the devices that were featured in our study had at least 2 leads (i.e., atrial and ventricular leads) (Table 1). The WAVE study was a single-chamber ICD study that included the WAVELET episode discrimination features. The OMNI registry included single-, dual-, and triple-chamber devices. Other studies included dual-chamber devices.
Patients needed to fulfill the following 2 criteria to be included: a primary prevention ICD indication and performance of at least 1 device interrogation. There was variability in how settings pertaining to number of intervals to detection were programmed in the studies included. To ensure greater homogeneity in this analysis while maximizing the number of patients included, we included patients whose devices were programmed to 12/16 or 18/24 only, the 2 most commonly used settings. Categorical variables are reported as numbers and proportions, and continuous variables are reported as mean ± SD.
VAs were defined using the definitions of the respective studies and included VAs ≤500 ms. An episode did not require the need for device therapy to be included in the cohort. For example, if the device had a ventricular tachycardia zone programmed to monitor only, those episodes were included in the analysis. All episodes of ventricular tachycardia and ventricular fibrillation included in this analysis were adjudicated by an independent panel of physicians and had a tachycardia cycle length of ≤500 ms (rate at least 120 beats/min). Each episode was independently reviewed by an electrophysiologist at the study site. For some studies, the episode was also independently reviewed by a Medtronic representative. In cases of disagreement, either a second independent physician or an episode review committee member would review the episode for final classification. Any episode of VA that could not be adjudicated because the intracardiac electrogram was not available, including such episodes that received device therapy, was excluded from the analysis.
We constructed 4 separate negative binomial models to assess whether a patient’s risk for experiencing an episode of VA was influenced by the time of day (hourly intervals), day of the week, individual months, or seasons. For each of these models, the outcome variable was the number of episodes of VA in a given time period. Indicator variables were used in each model for the referent category. In our “seasonal” model, we used indicator variables for the seasons of summer, fall, and winter and established spring as baseline. The estimated chance of at least 1 episode of VA in a given 91-day season is the estimated probability on the basis of the negative binomial fit. We used the modeling equation to model the probability of at least 1 VA episode during a 91-day season (p[≥1 VA episode in a 91-day season]).
The modeling of the hourly analysis was done in the same manner as the seasonal model. Each day for every patient was split into 24 segments. If a patient had 365 days of follow-up, the patient contributed 365 data points for a given hour. In this model, we created 23 indicator variables for each hour and considered the hour of midnight to 1 am as the reference variable. In the weekday model, indicator variables were created for each of the weekdays, and Sunday was considered the reference category. Similarly, in a “monthly” model, indicator variables were created for each individual month, with January as the reference category.
We limited the contribution of each patient to the number of VA episodes the patient experienced in each time period. The models also included a random effect for each patient to account for variability within a patient across the various time periods. A p value of <0.05 was considered statistically significant. We also evaluated observed rates of VA episodes across time periods in subgroups defined by sex and age group (i.e., >55, 55 to 64, 65 to 74, and at least 75 years of age).
A total of 3,969 subjects were included in this analysis. The average follow-up period per participant was 1.8 ± 1.4 years, and a total of 7,126.8 device-years of data were analyzed. The mean age of participants was 65 ± 12.5 years, and the average ejection fraction of participants was 28.0 ± 10.2%. The remaining variables are in line with prior ICD studies (Table 1).
In our analysis of episodes of VA according to the seasons, we observed that the estimated chance of occurrence of at least 1 episode of VA in a given 91-day season was 0.86% for spring and 0.70% for summer. This difference in the incidence of VA was statistically significant (p = 0.009); in comparison, the corresponding estimated chances of VA occurrence for fall (0.74%) and winter (0.84%) were not significantly different from that for spring (p = 0.07 and p = 0.73, respectively). The observed annualized rates of VA per patient-year according to season on the basis of the patients’ first year of follow-up data (Figure 1). Figure 2 shows these annualized rates across various studies that were included in this analysis.
In the monthly model, we observed that none of the coefficients for the indicator variables (individual months) were observed to be statistically significant. Similarly, upon analysis of VAs according to individual days of the week (with Sunday being considered the baseline), no significant variation in the occurrence of VAs was observed.
With respect to the hour of the day, we observed that the coefficients for the indicator variables for the time periods of each 1-h increment from 8 am to 10 pm were all statistically significant. The results of the hourly analysis showed that patients were more likely to experience episodes of VA during the hours of 8 am to 10 pm in comparison with the period from 10 pm to 8 am. Over the course of 1 year, the estimated probability of experiencing at least 1 episode of VAs during a given hour of the daytime period of 8 am to 10 pm was 35% to 63% higher than the period of midnight to 1 am (Table 2). We did not observe any differences in the seasonal variation of VAs when we calculated seasonal VA rates according to sex. In our analysis stratified according to age group, we observed similar seasonal variations in patients older than 55 years.
This study on variations in occurrence of VAs over various time periods in a pooled patient population with ICDs had 3 major findings and is articulated in the Central Illustration: (1) there was a significantly greater incidence of VAs in the spring compared with the summer, although it was inconclusive whether fall and winter had lower rates compared with spring; (2) there was a significantly higher estimated chance of occurrence of VAs during each 1-h block during the daytime hours of 8 am to 10 pm in comparison with midnight to 1 am, while other 1-h blocks between 10 pm and 8 am were not significantly different from that baseline time period; and (3) no significant differences in the occurrence of VAs according to the months of the year as well as during the days of the week were observed.
Using ICD patients to assess the temporal behavior of VAs provides a unique opportunity to truly understand the exact timing of these events given the devices’ ability to continuously record. Although other device data, including average heart rate and physical activity, would have helped delineate a mechanism for what was observed, we speculate that the findings are not random and may reflect variations in autonomic tone, adrenergic stimulation, or other factors that may be modulated by normal circadian rhythm variations. Why these specific findings were observed is unclear, but one could conjecture that peak periods of VAs during the spring, for example, might be due to the transition of environmental conditions from winter to spring rather than specific thresholds in environmental conditions that were crossed. In other words, the degree of change in temperature, for example, might have a greater influence on the biological rhythm and mechanisms that play a role in the occurrence of VAs than the absolute temperature alone. It is also possible that other factors, including air pollution (12,13), barometric pressure, or humidity (14,15), which also differ across the seasons, could have played a role either directly or indirectly in the propensity for premature ventricular beats. All these factors have been suggested to be related to an increased risk for arrhythmias. In this regard, our findings are similar to the observations from the TEMPEST (Temperature-Related Incidence of Electrical Storm) study. In the TEMPEST study, most episodes of electrical storm occurred in association with an increase in temperature compared with the month prior. There is some degree of similarity in this observation and our finding of increased risk for VAs during the spring. Although speculative, we believe that the variation in temperature across seasons might have been the causative factor for this observation (16).
In addition, we also speculate that the change in the degree of physical activity and the consequent variation in autonomic tone might be an additional contributory mechanism for the higher rate of VAs during spring compared with other seasons (17,18). Apart from this change in physical activity, behavioral changes during the spring, such as changes in fluid and diet intake with subsequent fluctuations in electrolyte balance, might also contribute to our findings of seasonal variations in the occurrence of VAs. Data from prior studies have also implicated the role of thermal stress due to variations in temperature as a modulating factor toward alterations in VAs (19,20). Although the mechanism by which thermal stress might contribute to variations in VAs is less clear, it is possible that variations in sympathetic tone and endogenous catecholamine levels might be the mediating factors. Considering that at least some VAs might be induced by myocardial ischemia, it is conceivable that seasonal variations in ischemia and risk factors such as platelet reactivity and fibrinogen levels might also play a role (21–23). Another factor that could also contribute to our findings is the amount of follow-up data available across different seasons. The WAVE study required follow-up of only 6 months, and the EnTrust study did not require 12-month follow-up data. Hence, more patient follow-up data were available in the seasons of spring and summer compared with fall and winter.
With respect to the observations made regarding the hour of the day, it is possible that these findings might again reflect the variation in autonomic tone between the awake daytime hours and sleeping during the nighttime. The difference in the level of beta-adrenergic stimulation has been observed to increase the level of Ca++ overload and to promote delayed afterdepolarizations (24). In a study based on hourly monitoring of electrocardiographic data on healthy volunteers, Ishida et al. (25) observed an increased degree of heterogeneity of action potential (assessed by QT/QTc dispersion) during the morning hours in comparison with the nighttime hours. These findings are further supported by those of Kong et al. (26), demonstrating greater shortening of the ventricular refractory period during normal waking hours. Variability of ion channels has also been observed to be an additional contributory factor for the differential expression of ion channels and arrhythmogenesis. On the basis of experiments on transgenic mice, Schroder et al. (27) observed circadian variation in the expression of KCNH2, which plays an important role in normal ventricular repolarization. Similar to these findings, Yamashita et al. (28) also observed significant circadian variation in the transcription and gene expression of K+ channels (Kv 1.5 and 4.2 subtypes). In the same study, the daytime versus nighttime variability in the gene expression of K+ channels was observed to be associated with concomitant variation in electrophysiological properties (refractory periods) of cardiomyocytes.
Our findings of a higher incidence of VAs during the spring and during waking hours are in contrast to those of a recently published population-based study by Ni et al. (29), who observed a lack of peak incidence of SCD during morning hours. Although there may be several factors that might explain the difference, the most likely reason is that not all ICD therapies for VAs translate into aborted SCD events (30). It is also worth pointing out that the clinical profile of patients in our study was different from that of the community participants in their study. Considering that the patients included in our study met the indications for primary prevention of SCD, it is intuitive to believe that the risk for the occurrence of VAs in our patient population would be different than among the participants in a community-based study.
Despite this, our findings do add to the existing body of published research on our understanding of VAs. Prior analysis from the MADIT-CRT (Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy) had shown that the incidence of VAs was highest during the morning hours (31). In contradistinction, a post hoc analysis from the ScD-HeFT (Sudden Cardiac Death in Heart Failure Trial) did not observe an early morning peak in the occurrence of VAs and SCD (32). It is conceivable that the alterations in heart rate variability, autonomic tone, and degree of ischemia might be the contributory factors to the timing of VAs (33). The conflicting data on diurnal variations of VAs in our study compared with some earlier studies might be reflective of secular trends in the improvement of medical therapy used for neurohormonal blockade for heart failure. It is also plausible that variations in the occurrence of VAs might be reflective of differences in patient characteristics. Of note, our analysis represents the largest cohort of ICD patients studied, and this may have provided an opportunity to uncover differences in VA timing that smaller studies were unable to elucidate.
Although prior studies, such as analysis from the landmark SCD-HeFT trial, showed a peak incidence of VAs on Mondays, other more recent studies have not demonstrated a peak incidence on Mondays (34,35). In the SCD-HeFT study, the increase in degree of psychological stress was proposed to be mechanism for the Monday peak of VAs. The mechanism for the lack of any daily peak period of VAs during the week in our study remains unclear. We speculate that a redistribution of stress, improvement in the degree of neurohormonal blockage, and widespread use of beta-blockers as part of heart failure therapy might be the contributory factors for our findings (36).
There were limitations of this analysis that should be considered. First, this was a retrospective analysis of prospectively conducted patients across many generations of ICDs. Although some variables, such as ICD indication and number of intervals to detection, were controlled for, data points including heart rate variability and other surrogates of autonomic triggers were not incorporated into the analysis.
Second, only primary prevention ICD patients were included. Third, because ICDs are generally programmed to treat VAs before they result in hemodynamic compromise, it is not entirely feasible to truly know whether a VA episode would have self-terminated.
Fourth, given the inability to resolve the exact location of the patient at the time of an arrhythmic event, suggestions of temperature, physical activity, and environmental factors serving as triggers for these events are purely speculative.
Lastly, we did not evaluate changes in therapy over time after the first VA event. Hence, the incidence of VAs after the first event may have been affected. Despite these limitations, this analysis represents the largest to date on this topic with all VAs independently adjudicated.
VA occurrence among primary prevention ICD recipients varies over time and appears to peak during normal waking hours and in the spring. Understanding variability in the occurrence of VAs could aid in more effective ICD programming to reduce unnecessary and inappropriate therapies in the future.
COMPETENCY IN MEDICAL KNOWLEDGE: Prior studies aimed at better understanding the distribution of VAs in patients at risk for SCD have surprisingly deviated from the expected peak in events during the morning and at the beginning of the workweek. The reasons are unclear but may be due in part to the limited number of patients studied. Our study pooled patients from 6 large ICD studies, resulting in 3,969 ICD patients analyzed. We observed a higher rate of VA occurrence during the spring and during the waking hours of 8 am to 10 pm. Although the underlying VA triggers remain unclear, these findings suggest that physical activity and periods of changing from less active to more active states (winter to spring) might be the contributory mechanisms for variations in VA occurrence.
TRANSLATIONAL OUTLOOK: This was a retrospective analysis of prospective studies in patients with ICDs. A better understanding of the “at risk” periods when patients are more likely to develop VAs could help refine recommendations on improved preventive strategies and inform future public health efforts. Further studies could help investigate the mechanistic pathways to understand the findings we observed.
Dr. Maan has received research grants from Medtronic, Biotronik, and Biosense Webster. Dr. Sherfesee is an employee of Medtronic. Dr. Lexcen is an employee of Medtronic. Dr. Heist has received honoraria from Biotronik, Boston Scientific, Medtronic, Abbott, and Johnson and Johnson (all modest in amount); has received research grants from Boston Scientific and Abbott (all modest in amount); and has received consulting fees from Pfizer (modest in amount). Dr. Cheng is an employee of Medtronic.
The authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Clinical Electrophysiology author instructions page.
- Abbreviations and Acronyms
- implantable cardioverter-defibrillator
- sudden cardiac death
- ventricular arrhythmia
- Received December 23, 2018.
- Revision received April 17, 2019.
- Accepted May 6, 2019.
- 2019 American College of Cardiology Foundation
- Tofler G.H.,
- Gebara O.C.,
- Mittleman M.A.,
- et al.
- Wathen M.S.,
- DeGroot P.J.,
- Sweeney M.O.,
- et al.
- Wilkoff B.L.,
- Ousdigian K.T.,
- Sterns L.D.,
- et al.
- ↵U.S. National Library of Medicine. Clinical Investigation of the Medtronic EnTrust™ Implantable Cardioverter Defibrillator (ICD), Model D153ATG. Available at: https://clinicaltrials.gov/ct2/show/NCT00273195. Accessed November 17, 2018.
- Song X.,
- Liu Y.,
- Hu Y.,
- et al.
- Nguyen J.L.,
- Laden F.,
- Link M.S.,
- et al.
- Guerra F.,
- Bonelli P.,
- Flori M.,
- et al.
- Willich S.,
- Maclure M.,
- Mittleman M.,
- Arntz H.R.,
- Muller J.E.
- Muller J.E.,
- Ludmer P.L.,
- Willich S.N.,
- et al.
- Gerber Y.,
- Jacobson S.J.,
- Killian J.M.,
- et al.
- Keatings W.R.,
- Coleshaw S.R.,
- Cotter F.,
- et al.
- Gardner R.T.,
- Ripplinger C.M.,
- Myles R.C.,
- Habecker B.A.
- Kong T.Q. Jr..,
- Goldberger J.J.,
- Parker M.,
- Wang T.,
- Kadish A.H.
- Yamashita T.,
- Sekiguchi A.,
- Iwasaki Y.K.,
- et al.
- Ni Y.M.,
- Rusinaru C.,
- Reiner K.,
- et al.
- Ellenbogen K.A.,
- Levine J.H.,
- Berger R.D.,
- et al.
- Ruwald M.H.,
- Moss A.J.,
- Zareba W.,
- et al.
- Patton K.K.,
- Hellkamp A.S.,
- Lee K.L.,
- et al.
- Ben Ahmed H.,
- Allouche M.,
- Zoghlami B.,
- et al.