- Email: [email protected]
Circadian variation of human ventricular fibrillation dominant frequency
Resuscitation 81 (2010) 950–955
Contents lists available at ScienceDirect
Resuscitation journal homepage: www.elsevier.com/locate/resuscitation
Clinical paper
Circadian variation of human ventricular fibrillation dominant frequency夽 Philip Langley a,∗ , Guy A. MacGowan b , Alan Murray c a
Institute of Cellular Medicine, Newcastle University, Newcastle upon Tyne, NE7 7DN, UK Department of Cardiology, Freeman Hospital, Newcastle upon Tyne, UK c Medical Physics Department, Freeman Hospital, Newcastle upon Tyne, UK b
a r t i c l e
i n f o
Article history: Received 27 January 2010 Received in revised form 2 March 2010 Accepted 14 March 2010 Keywords: Ventricular fibrillation Dominant frequency Circadian variability
a b s t r a c t Aim: Circadian variation in human ventricular fibrillation (VF) dominant frequency is unknown. If present this would provide evidence of physiological influence on VF. The objective was to quantify the circadian variation in human VF dominant frequency. Methods: Eight-lead Holter ECG recordings were obtained from a patient with severe myocarditis and chronic VF who was supported by a biventricular assist device. Recordings of up to 24 h duration were obtained on 6 days with an average interval between recordings of 7 days. Dominant frequency and amplitude were obtained using spectral analysis and assessed for (i) circadian (ii) inter-recording and (iii) inter-lead differences. Results: There was a significant circadian variation in amplitude (night: 0.027 ± 0.004 mV Hz vs day: 0.044 ± 0.006 mV Hz, p < 0.0001) but not dominant frequency (night: 7.85 ± 0.62 Hz vs day: 7.93 ± 0.54 Hz, p > 0.05). There were significant differences between recordings in dominant frequency which ranged from 6.80 ± 0.29 Hz to 8.36 ± 0.38 Hz (p < 0.0001) and dominant frequency spectral amplitude which ranged from 0.033 ± 0.014 mV Hz to 0.043 ± 0.017 mV Hz (p < 0.0001). Histograms of dominant frequencies in leads exhibited strikingly different distributions, particularly in V2 that was characterised by a bimodal distribution, while the other leads were characterised by predominantly unimodal distributions. Conclusion: VF dominant frequency spectral amplitude exhibited circadian variability. In a patient with severe myocarditis, supported with a biventricular assist device and in chronic VF, these results provide evidence for modulation of VF, probably induced by changes in posture and physical activity. © 2010 Elsevier Ireland Ltd. All rights reserved.
1. Introduction The cardiovascular system exhibits circadian variability that is thought to promote a healthy balance between the sleep/wake cycle.1 For example, impaired circadian patterns in blood pressure and heart rate have prognostic significance.2,3 The autonomic nervous system plays a key role in mediating this cycle.4 On one hand, circadian variability is an essential characteristic of a healthy cardiovascular system, yet it also seems to provide a window of opportunity for adverse cardiovascular events at certain times of the cycle. Particularly, incidence of sudden cardiac death increases in the morning,5 as does onset of ventricular arrhythmias.6–9 These increases probably relate to the circadian pattern of myocardial infarction.10 It has been suggested that circadian changes in cardiac electrophysiological properties, for example, ventricular refrac-
夽 A Spanish translated version of the summary of this article appears as Appendix in the online version at doi:10.1016/j.resuscitation.2010.03.026. ∗ Corresponding author at: Medical Physics Department, Freeman Hospital, Newcastle upon Tyne, NE7 7DN, UK. Tel.: +44 0 1912231664; fax: +44 0 1912130290. E-mail address: [email protected] (P. Langley). 0300-9572/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.resuscitation.2010.03.026
toriness that shortens during the morning, also contributes to the circadian cycle of arrhythmogenicity.11 Additionally, Venditti et al.12 demonstrated in patients with implantable cardioverterdefibrillators significantly increased defibrillation threshold during the morning and increased numbers of failed VF terminations in the morning compared to other periods of the day. The study did not present data for defibrillation thresholds or failures at night. Reestablishment of circulatory rhythm following defibrillation has been associated with higher VF dominant frequency and larger ECG waveform amplitude.13–15 A number of additional features of the ECG and their frequency bands have been studied in relation to prediction of VF defibrillation success and features related to the ECG gradient, for example median slope, have high predictive power.16 However, dominant frequency remains the most widely reported feature of VF and is the focus of this study. Diurnal variations in features such as dominant frequency might help to explain the observed diurnal patterns of defibrillation success. The aim of our study was to assess circadian variability of dominant frequency and amplitude in a unique model of chronic human VF. Dominant frequency is an important parameter related to the frequency of underlying re-entrant circuits or triggers that sustain the arrhythmia.17,18 The dominant frequency spectral amplitude
P. Langley et al. / Resuscitation 81 (2010) 950–955
951
is a measure of the power contained in the ECG at the dominant frequency. We have already reported the spatial and temporal variability in dominant frequency in short duration ECG recordings of human VF.19 However, due to the difficulty of obtaining recordings of sufficient duration, the circadian variation of dominant frequency is unknown.
rank sum test. Differences between recordings and leads were assessed by Kruskal–Wallis test. A p value less than 0.05 was used to indicate statistical significance. Summary values are shown as mean ± SD unless otherwise stated. Statistical analysis was carried out using Matlab Statistical Toolbox R2009b.
2. Methods
3. Results
2.1. Patient characteristics
3.1. Spectral parameters
The patient was a 29-year-old female admitted to the Freeman Hospital with cardiogenic shock due to viral myocarditis. A Berlin Heart Biventricular Assist Device was implanted, and the patient underwent heart transplantation three months later, though subsequently died in the early post-operative period. Histological examination of the explanted heart revealed appearances of a fibro-inflammatory cardiomyopathy in keeping with post-acute lymphocytic myocarditis though with no morphological features as to the aetiology. There was serological evidence of a recent adenoviral infection at the time of presentation. Shortly after implantation of the ventricular assist device the patient developed sustained VF. The patient was not prescribed any cardiac medication or anti-arrhythmics. There were no significant serum electrolyte abnormalities. Despite the VF the patient recovered well on biventricular support, enough to allow home visits. The settings of the device (rate, systolic and diastolic pressures) were kept constant throughout the assist device support period. The cause of death after transplantation was primary graft failure. Informed consent and ethical approval were obtained to conduct this study.
Fig. 1, left panel, shows the minute-by-minute dominant frequency and amplitude for each recording. The right panel shows histograms of dominant frequency for each recording. The most notable observations were that (i) there was clear circadian variation in amplitude, and (ii) dominant frequency changed between recordings.
2.2. Data recording and dominant frequency analysis Eight-lead Holter ECG (H12+, Mortara, Italy) recordings were obtained from the patient on 6 separate days. The first recording (day 1) was obtained 26 days after implantation of the assist device. The mean interval between recordings was 7 days (range 5–10). Five recordings were 24 h duration and one was limited to 18 h due to patient request. No recordings were available at VF onset and VF did not terminate. ECGs were recorded at a sample rate of 1000 Hz and stored on computer for subsequent processing and analysis. ECGs were filtered using a 6th order Butterworth filter with pass band of 0.5–30 Hz to minimise baseline and movement artefact while preserving the fibrillation signal. Spectral analysis using the Fast Fourier Transform was applied to contiguous 1 min segments of the ECG recordings, and dominant frequency and amplitude were determined for each segment. Spectra were visually inspected to see if they exhibited harmonic peaks sometimes observed in highly organised ventricular arrhythmias that might have hindered the detection of the dominant spectral peak but there was none. Herein dominant frequency refers to the frequency with the largest spectral amplitude measured in Hz and amplitude refers to the amplitude of the spectrum at the dominant frequency measured in mV Hz. Amplitude is an indirect measure of the power of signal components at different frequencies. Frequency resolution of the spectra was 0.015 Hz. Signal processing and analysis was carried out using Matlab R2009b (The Mathworks Inc., USA). Differences in dominant frequency and amplitude were quantified (i) between day and night (ii) between recordings, and (iii) between ECG leads. 2.3. Statistical analysis Circadian variation in dominant frequency and amplitude was assessed by comparing differences in mean day (08:00–00:00, comprising morning: 08:00–12:00, and afternoon/evening: 12:00–00:00) and night (00:00–08:00) values using the Wilcoxon
3.2. Circadian differences in spectral parameters Fig. 2 summarises the mean day and night dominant frequency and amplitude for each recording. Amplitude was significantly reduced at night compared to day (0.027 ± 0.004 mV Hz vs 0.044 ± 0.006 mV Hz, p < 0.0001). This indicated that the amplitude during the day is on average 1.63 (0.044/0.027) times that at night. In a sub-analysis of morning and afternoon/evening values, there was a significant reduction in amplitude from morning to afternoon/evening (0.055 ± 0.011 mV Hz vs 0.041 ± 0.006 mV Hz, p < 0.01) and a further significant reduction from afternoon/evening to night (0.041 ± 0.006 mV Hz vs 0.027 ± 0.004 mV Hz, p < 0.003). However, dominant frequency did not change significantly from night to day (7.85 ± 0.62 Hz vs 7.93 ± 0.54 Hz, p > 0.05). To illustrate the circadian variation of the ECG Fig. 3 illustrates representative sections of day and night ECGs of 10 s duration alongside their associated spectra. Both ECG and spectral amplitude are clearly increased in the day compared to night. 3.3. Differences between recordings Fig. 4 summarises dominant frequency and amplitude from recordings on different days. There were significant differences in both dominant frequency (p < 0.0001) and amplitude (p < 0.0001). Dominant frequency ranged from 6.80 ± 0.29 Hz on day 1 to 8.36 ± 0.38 Hz on day 22. Amplitude ranged from 0.033 ± 0.014 mV Hz on day 22 to 0.043 ± 0.017 mV Hz on day 36. 3.4. Differences between leads Fig. 5 shows histograms of dominant frequency from each lead. The histogram distributions were not equal for all leads. Particularly V2 for all recordings exhibited a bimodal distribution whereas the other leads predominantly exhibited a unimodal distribution. Table 1 describes the similarity of distributions between leads by quantifying the percentage overlap (area under the curves) in the distributions. Values are average for the six recordings. V2 distributions exhibited the least similarity with the other lead histograms with approximately 50% overlap. Fig. 6 compares by lead the dominant frequencies with the greatest histogram count for each recording. For V2 the dominant frequencies with the greatest histogram count for each peak in the bimodal histogram distribution were identified as indicated by the two boxplots in Fig. 6. There were significant differences between leads for the lowest V2 dominant frequency (p < 0.02) but no significant difference for the highest V2 dominant frequency (p > 0.05).
952
P. Langley et al. / Resuscitation 81 (2010) 950–955
Fig. 1. Left panel: dominant frequency (DF) and amplitude (Amp) across each recording. Median dominant frequency is shown in red and the range across leads in blue. Amplitude (mV Hz) was scaled by times 50 and is shown in black. Right panel: histograms of dominant frequency for each of the recordings. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Table 1 Similarity of lead dominant frequency distributions shown in Fig. 5. Lead I II V1 V2 V3 V4 V5 V6
I
II
100 – – – – – – –
83 100 – – – – – –
V1 59 65 100 – – – – –
V2 53 56 59 100 – – – –
V3 79 74 58 53 100 – – –
V4 80 77 57 48 85 100 – –
V5 83 83 58 48 76 82 100 –
4. Discussion V6 80 81 55 45 72 77 88 100
Values are the average percentage overlap of area under the curves of lead dominant frequency histograms over the six recordings. The distributions for lead V2 (bold) had least similarity to the distributions of the other leads.
This is the first study to quantify circadian variations in human VF dominant frequency and amplitude. VF is typically sustained for seconds rather than hours so there has been a paucity of data of sufficient duration to allow assessment of circadian variations of VF dominant frequency. Patients developing VF who are supported by ventricular assist devices provide a novel model that facilitates the study of human VF over much longer periods, typically weeks or months, than any other model. Using this model we have collected a unique dataset and quantified the circadian pattern of VF spectral characteristics repeatedly in the same patient over several weeks. The obvious difference of this model to acute
P. Langley et al. / Resuscitation 81 (2010) 950–955
953
Fig. 2. Comparisons of night and day dominant frequency (DF) and amplitude (Amp). There were no significant differences for dominant frequency (p > 0.05) but amplitude was significantly increased during the day (p < 0.0001).
VF is that the assist device ensures the heart remains continuously perfused. This is not a restriction in our study because we were interested in the circadian pattern of VF rather than the effects of ischaemia. However, the aetiology of VF in this model was myocarditis and because of the unique clinical circumstances it is unknown if the results obtained are generaliseable to the wider population of patients presenting with acute VF. We have demon-
strated a circadian pattern that was characterised by significant reductions in VF amplitude at night. Postural changes are the most plausible explanation for this circadian pattern, particularly since daytime amplitudes exhibited large variability reflecting the relatively active state of this patient whereas nighttime amplitudes, when the patient would be expected to be sedentary, were relatively constant. The ECG is susceptible to large amplitude changes
Fig. 3. Examples of ECG leads and corresponding spectra from night and day. For both night and day each row shows an ECG lead of 10 s duration and the power spectrum calculated from a 1 min interval of ECG containing the illustrated ECG section.
954
P. Langley et al. / Resuscitation 81 (2010) 950–955
Fig. 4. Boxplots of dominant frequency (DF) and amplitude (Amp) at each recording. There were significant differences between recordings of both dominant frequency and amplitude. Box represents the interquartile range (IQR: q3–q1) of data and median is shown as a line. Whiskers represent points within q1 − 1.5 × IQR and q3 + 1.5 × IQR.
simply due to changes in the orientation of and proximity of the heart to the recording electrodes.20 However, autonomic influences may also contribute to the circadian pattern of VF. As discussed in Section 1, the effect of the autonomic nervous system on a range of cardiac parameters is well established. Circadian changes in atrial fibrillation dominant frequency have been reported with significantly reduced cycle length (increased dominant frequency) during daytime, but not in patients with already short cycle lengths.21 In
our study dominant frequency of VF did not change significantly from day to night suggesting there is no autonomic influence on VF dominant frequency. The relative stability of dominant frequency over 24 h suggests also that there is little postural influence on this feature. The increased nighttime variability in dominant frequency clearly apparent in most of the recording days but especially from day 22 (Fig. 1) is explained by the relative flatness of the spectra, and hence increased uncertainty in identifying the dominant
Fig. 5. Histograms of dominant frequency for each lead and recording.
P. Langley et al. / Resuscitation 81 (2010) 950–955
955
Conflict of interest statement All authors have no disclosures relevant to this manuscript. Acknowledgements We are grateful to Susan King for recording the body surface potentials for this study. This work was undertaken with the financial support of an EPSRC Advanced Research Fellowship to Philip Langley. References
Fig. 6. Boxplot of dominant frequency for each lead. The dominant frequency with the maximum count for each recording was used to construct the boxplot for each lead. Two boxplots for V2 represent the bimodal distribution of the histogram.
peak, compared to daytime as illustrated in Fig. 3. Defibrillation success has been related to amplitude and dominant frequency characteristics of VF such that success in restoration of a circulatory rhythm is associated with VF of largest amplitude and highest frequency.13–15 It is interesting that in our patient waveform amplitude was greatest in the morning, when previous studies have indicated that defibrillation success is lowest.12 Median slope of the VF waveform has been proposed as a highly predictive feature of countershock success and relates to the average rate of change of VF signal.16 Applied to our data, since there was no significant circadian variation in dominant frequency, median slope would follow a similar circadian pattern to amplitude. We confirmed this for the daytime and nighttime ECGs illustrated in Fig. 3 for which daytime median slope was approximately 1.5 times that of nighttime. On the other hand, dominant frequency exhibited a pronounced initial decrease after 8 am on several days, particularly days 17, 22 and 29 as illustrated in Fig. 1, and this might account for the reduced defibrillation success rates in the morning hours reported in other studies. We did not attempt defibrillation in this patient, but nonetheless our model of VF might be useful for such studies in the future, although limited because ischaemia is probably a major factor in determining outcome of defibrillation and this is not a feature of the model. Although there was no circadian pattern of VF dominant frequency, there were significant inter-recording and inter-lead differences in dominant frequency. This is the second case of long-term VF data from our group and the results confirm our previously published data which revealed significant inter-recording and inter-lead differences in dominant frequency.19 Inter-lead differences in dominant frequency probably reflect spatial differences in underlying activation patterns that have been shown in invasive human studies.22 Day-to-day variability in dominant frequency might arise due to changes in the cardiac electrophysiological properties due to progression of disease. 5. Conclusion In this novel model of human VF sustained over several weeks we have shown a circadian pattern of VF amplitude but not dominant frequency. In a patient with severe myocarditis, supported with a biventricular assist device and in chronic VF, these results provide evidence for modulation of VF, probably induced by changes in posture and physical activity.
1. Guo Y, Stein PK. Circadian rhythm in the cardiovascular system: considerations in non-invasive electrophysiology. Card Electrophysiol Rev 2002;6:267– 72. 2. de la Sierra A, Redon J, Banegas JR, et al. Prevalence and factors associated with circadian blood pressure patterns in hypertensive patients. Hypertension 2009;53:466–72. 3. Tsuji H, Larson MG, Venditti FJ, et al. Impact of reduced heart rate variability on risk for cardiac events. The Framingham heart study. Circulation 1996;94:2850–5. 4. Van de Borne P, Nguyen H, Biston P, Linkowski P, Degaute JP. Effect of wake sleep stages on the 24 h autonomic control of blood pressure and heart rate in recumbent men. Am J Physiol Heart Circ Physiol 1994;266:H548– 54. 5. Muller JE, Ludmer PL, Wilich SN, et al. Circadian variation in the frequency of sudden cardiac death. Circulation 1987;75:131–8. 6. Zehender M, Meinertz T, Hohnloser S, et al. Prevalence of circadian variations in spontaneous variability of cardiac disorders and ECG changes suggestive of myocardial ischemia in systemic arterial hypotension. Circulation 1992;85:1808–15. 7. Tofler GH, Gebara OCE, Mittleman MA, et al. Morning peak in ventricular tachyarrhythmias detected by time of implantable cardioverter/defibrillator therapy. Circulation 1995;92:1203–8. 8. Behrens S, Franz MR. Circadian variation of arrhythmia events, electrophysiological properties, and the autonomic nervous system. Eur Heart J 2001;22:2144– 6. 9. Englund A, Behrens S, Wegscheider K, Rowland E. Circadian variation of malignant ventricular arrhythmias in patients with ischemic and nonischemic heart disease after cardioverter defibrillator implantation. J Am Coll Cardiol 1999;34:1560–8. 10. Muller JE, Stone PH, Rutherford JD, et al. Circadian variation in the frequency of onset of acute myocardial infarction. N Engl J Med 1985;313:1315–22. 11. Kong TQ, Goldberger JJ, Parker M, Wang T, Kadish AH. Circadian variation in human ventricular refractoriness. Circulation 1995;92:1507–16. 12. Venditti FJ, John RM, Hull M, Tofler GH, Shahian DM, Martin DT. Circadian variation in defibrillation energy requirements. Circulation 1996;94:1607– 12. 13. Strohmenger H, Linder KH, Brown CG. Analysis of the ventricular fibrillation ECG signal amplitude and frequency parameters as predictors of countershock success in humans. Chest 1997;111:584–9. 14. Eftestøl T, Sunde K, Aase OS, Husøy JH, Steen PA. Predicting outcome of defibrillation by spectral characterization and nonparametric classification of ventricular fibrillation in patients with out-of-hospital cardiac arrest. Circulation 2000;102:1523–9. 15. Reed MJ, Clegg GR, Robertson CE. Analysing the ventricular fibrillation waveform. Resuscitation 2003;57:11–20. 16. Neurauter A, Eftestøl T, Kramer-Johansen J, et al. Prediction of countershock success using single features from multiple ventricular fibrillation frequency bands and feature combinations using neural networks. Resuscitation 2007;73:253–63. 17. Chen J, Mandapati R, Berenfeld O, Skanes AC, Jalife J. High-frequency periodic sources underlie ventricular fibrillation in the isolated rabbit heart. Circ Res 2000;86:86–93. 18. Skanes AC, Mandapati R, Berenfeld O, Davidenko JM, Jalife J. Spatiotemporal periodicity during atrial fibrillation in the isolated sheep heart. Circulation 1998;98:1236–48. 19. Langley P, MacGowan GA, Murray A. Spatial and temporal organisation of the dominant frequencies in the fibrillating heart: body surface potential mapping in a rare case of sustained human ventricular fibrillation. Europace 2009;11:324–7. 20. García J, Åström M, Mendive J, Laguna P, Sörnmo L. ECG-based detection of body position changes in ischemia monitoring. IEEE Trans Biomed Eng 2003;50:677–85. 21. Meurling CJ, Waktare JEP, Holmqvist F, et al. Diurnal variations of the dominant cycle length of chronic atrial fibrillation. Am J Physiol Heart Circ Physiol 2001;280:H401–6. 22. Nash MP, Mourad A, Clayton RH, Sutton PM, Bradley CP, Hayward M, et al. Evidence for multiple mechanisms in human ventricular fibrillation. Circulation 2006;114:536–42.
Contents lists available at ScienceDirect
Resuscitation journal homepage: www.elsevier.com/locate/resuscitation
Clinical paper
Circadian variation of human ventricular fibrillation dominant frequency夽 Philip Langley a,∗ , Guy A. MacGowan b , Alan Murray c a
Institute of Cellular Medicine, Newcastle University, Newcastle upon Tyne, NE7 7DN, UK Department of Cardiology, Freeman Hospital, Newcastle upon Tyne, UK c Medical Physics Department, Freeman Hospital, Newcastle upon Tyne, UK b
a r t i c l e
i n f o
Article history: Received 27 January 2010 Received in revised form 2 March 2010 Accepted 14 March 2010 Keywords: Ventricular fibrillation Dominant frequency Circadian variability
a b s t r a c t Aim: Circadian variation in human ventricular fibrillation (VF) dominant frequency is unknown. If present this would provide evidence of physiological influence on VF. The objective was to quantify the circadian variation in human VF dominant frequency. Methods: Eight-lead Holter ECG recordings were obtained from a patient with severe myocarditis and chronic VF who was supported by a biventricular assist device. Recordings of up to 24 h duration were obtained on 6 days with an average interval between recordings of 7 days. Dominant frequency and amplitude were obtained using spectral analysis and assessed for (i) circadian (ii) inter-recording and (iii) inter-lead differences. Results: There was a significant circadian variation in amplitude (night: 0.027 ± 0.004 mV Hz vs day: 0.044 ± 0.006 mV Hz, p < 0.0001) but not dominant frequency (night: 7.85 ± 0.62 Hz vs day: 7.93 ± 0.54 Hz, p > 0.05). There were significant differences between recordings in dominant frequency which ranged from 6.80 ± 0.29 Hz to 8.36 ± 0.38 Hz (p < 0.0001) and dominant frequency spectral amplitude which ranged from 0.033 ± 0.014 mV Hz to 0.043 ± 0.017 mV Hz (p < 0.0001). Histograms of dominant frequencies in leads exhibited strikingly different distributions, particularly in V2 that was characterised by a bimodal distribution, while the other leads were characterised by predominantly unimodal distributions. Conclusion: VF dominant frequency spectral amplitude exhibited circadian variability. In a patient with severe myocarditis, supported with a biventricular assist device and in chronic VF, these results provide evidence for modulation of VF, probably induced by changes in posture and physical activity. © 2010 Elsevier Ireland Ltd. All rights reserved.
1. Introduction The cardiovascular system exhibits circadian variability that is thought to promote a healthy balance between the sleep/wake cycle.1 For example, impaired circadian patterns in blood pressure and heart rate have prognostic significance.2,3 The autonomic nervous system plays a key role in mediating this cycle.4 On one hand, circadian variability is an essential characteristic of a healthy cardiovascular system, yet it also seems to provide a window of opportunity for adverse cardiovascular events at certain times of the cycle. Particularly, incidence of sudden cardiac death increases in the morning,5 as does onset of ventricular arrhythmias.6–9 These increases probably relate to the circadian pattern of myocardial infarction.10 It has been suggested that circadian changes in cardiac electrophysiological properties, for example, ventricular refrac-
夽 A Spanish translated version of the summary of this article appears as Appendix in the online version at doi:10.1016/j.resuscitation.2010.03.026. ∗ Corresponding author at: Medical Physics Department, Freeman Hospital, Newcastle upon Tyne, NE7 7DN, UK. Tel.: +44 0 1912231664; fax: +44 0 1912130290. E-mail address: [email protected] (P. Langley). 0300-9572/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.resuscitation.2010.03.026
toriness that shortens during the morning, also contributes to the circadian cycle of arrhythmogenicity.11 Additionally, Venditti et al.12 demonstrated in patients with implantable cardioverterdefibrillators significantly increased defibrillation threshold during the morning and increased numbers of failed VF terminations in the morning compared to other periods of the day. The study did not present data for defibrillation thresholds or failures at night. Reestablishment of circulatory rhythm following defibrillation has been associated with higher VF dominant frequency and larger ECG waveform amplitude.13–15 A number of additional features of the ECG and their frequency bands have been studied in relation to prediction of VF defibrillation success and features related to the ECG gradient, for example median slope, have high predictive power.16 However, dominant frequency remains the most widely reported feature of VF and is the focus of this study. Diurnal variations in features such as dominant frequency might help to explain the observed diurnal patterns of defibrillation success. The aim of our study was to assess circadian variability of dominant frequency and amplitude in a unique model of chronic human VF. Dominant frequency is an important parameter related to the frequency of underlying re-entrant circuits or triggers that sustain the arrhythmia.17,18 The dominant frequency spectral amplitude
P. Langley et al. / Resuscitation 81 (2010) 950–955
951
is a measure of the power contained in the ECG at the dominant frequency. We have already reported the spatial and temporal variability in dominant frequency in short duration ECG recordings of human VF.19 However, due to the difficulty of obtaining recordings of sufficient duration, the circadian variation of dominant frequency is unknown.
rank sum test. Differences between recordings and leads were assessed by Kruskal–Wallis test. A p value less than 0.05 was used to indicate statistical significance. Summary values are shown as mean ± SD unless otherwise stated. Statistical analysis was carried out using Matlab Statistical Toolbox R2009b.
2. Methods
3. Results
2.1. Patient characteristics
3.1. Spectral parameters
The patient was a 29-year-old female admitted to the Freeman Hospital with cardiogenic shock due to viral myocarditis. A Berlin Heart Biventricular Assist Device was implanted, and the patient underwent heart transplantation three months later, though subsequently died in the early post-operative period. Histological examination of the explanted heart revealed appearances of a fibro-inflammatory cardiomyopathy in keeping with post-acute lymphocytic myocarditis though with no morphological features as to the aetiology. There was serological evidence of a recent adenoviral infection at the time of presentation. Shortly after implantation of the ventricular assist device the patient developed sustained VF. The patient was not prescribed any cardiac medication or anti-arrhythmics. There were no significant serum electrolyte abnormalities. Despite the VF the patient recovered well on biventricular support, enough to allow home visits. The settings of the device (rate, systolic and diastolic pressures) were kept constant throughout the assist device support period. The cause of death after transplantation was primary graft failure. Informed consent and ethical approval were obtained to conduct this study.
Fig. 1, left panel, shows the minute-by-minute dominant frequency and amplitude for each recording. The right panel shows histograms of dominant frequency for each recording. The most notable observations were that (i) there was clear circadian variation in amplitude, and (ii) dominant frequency changed between recordings.
2.2. Data recording and dominant frequency analysis Eight-lead Holter ECG (H12+, Mortara, Italy) recordings were obtained from the patient on 6 separate days. The first recording (day 1) was obtained 26 days after implantation of the assist device. The mean interval between recordings was 7 days (range 5–10). Five recordings were 24 h duration and one was limited to 18 h due to patient request. No recordings were available at VF onset and VF did not terminate. ECGs were recorded at a sample rate of 1000 Hz and stored on computer for subsequent processing and analysis. ECGs were filtered using a 6th order Butterworth filter with pass band of 0.5–30 Hz to minimise baseline and movement artefact while preserving the fibrillation signal. Spectral analysis using the Fast Fourier Transform was applied to contiguous 1 min segments of the ECG recordings, and dominant frequency and amplitude were determined for each segment. Spectra were visually inspected to see if they exhibited harmonic peaks sometimes observed in highly organised ventricular arrhythmias that might have hindered the detection of the dominant spectral peak but there was none. Herein dominant frequency refers to the frequency with the largest spectral amplitude measured in Hz and amplitude refers to the amplitude of the spectrum at the dominant frequency measured in mV Hz. Amplitude is an indirect measure of the power of signal components at different frequencies. Frequency resolution of the spectra was 0.015 Hz. Signal processing and analysis was carried out using Matlab R2009b (The Mathworks Inc., USA). Differences in dominant frequency and amplitude were quantified (i) between day and night (ii) between recordings, and (iii) between ECG leads. 2.3. Statistical analysis Circadian variation in dominant frequency and amplitude was assessed by comparing differences in mean day (08:00–00:00, comprising morning: 08:00–12:00, and afternoon/evening: 12:00–00:00) and night (00:00–08:00) values using the Wilcoxon
3.2. Circadian differences in spectral parameters Fig. 2 summarises the mean day and night dominant frequency and amplitude for each recording. Amplitude was significantly reduced at night compared to day (0.027 ± 0.004 mV Hz vs 0.044 ± 0.006 mV Hz, p < 0.0001). This indicated that the amplitude during the day is on average 1.63 (0.044/0.027) times that at night. In a sub-analysis of morning and afternoon/evening values, there was a significant reduction in amplitude from morning to afternoon/evening (0.055 ± 0.011 mV Hz vs 0.041 ± 0.006 mV Hz, p < 0.01) and a further significant reduction from afternoon/evening to night (0.041 ± 0.006 mV Hz vs 0.027 ± 0.004 mV Hz, p < 0.003). However, dominant frequency did not change significantly from night to day (7.85 ± 0.62 Hz vs 7.93 ± 0.54 Hz, p > 0.05). To illustrate the circadian variation of the ECG Fig. 3 illustrates representative sections of day and night ECGs of 10 s duration alongside their associated spectra. Both ECG and spectral amplitude are clearly increased in the day compared to night. 3.3. Differences between recordings Fig. 4 summarises dominant frequency and amplitude from recordings on different days. There were significant differences in both dominant frequency (p < 0.0001) and amplitude (p < 0.0001). Dominant frequency ranged from 6.80 ± 0.29 Hz on day 1 to 8.36 ± 0.38 Hz on day 22. Amplitude ranged from 0.033 ± 0.014 mV Hz on day 22 to 0.043 ± 0.017 mV Hz on day 36. 3.4. Differences between leads Fig. 5 shows histograms of dominant frequency from each lead. The histogram distributions were not equal for all leads. Particularly V2 for all recordings exhibited a bimodal distribution whereas the other leads predominantly exhibited a unimodal distribution. Table 1 describes the similarity of distributions between leads by quantifying the percentage overlap (area under the curves) in the distributions. Values are average for the six recordings. V2 distributions exhibited the least similarity with the other lead histograms with approximately 50% overlap. Fig. 6 compares by lead the dominant frequencies with the greatest histogram count for each recording. For V2 the dominant frequencies with the greatest histogram count for each peak in the bimodal histogram distribution were identified as indicated by the two boxplots in Fig. 6. There were significant differences between leads for the lowest V2 dominant frequency (p < 0.02) but no significant difference for the highest V2 dominant frequency (p > 0.05).
952
P. Langley et al. / Resuscitation 81 (2010) 950–955
Fig. 1. Left panel: dominant frequency (DF) and amplitude (Amp) across each recording. Median dominant frequency is shown in red and the range across leads in blue. Amplitude (mV Hz) was scaled by times 50 and is shown in black. Right panel: histograms of dominant frequency for each of the recordings. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Table 1 Similarity of lead dominant frequency distributions shown in Fig. 5. Lead I II V1 V2 V3 V4 V5 V6
I
II
100 – – – – – – –
83 100 – – – – – –
V1 59 65 100 – – – – –
V2 53 56 59 100 – – – –
V3 79 74 58 53 100 – – –
V4 80 77 57 48 85 100 – –
V5 83 83 58 48 76 82 100 –
4. Discussion V6 80 81 55 45 72 77 88 100
Values are the average percentage overlap of area under the curves of lead dominant frequency histograms over the six recordings. The distributions for lead V2 (bold) had least similarity to the distributions of the other leads.
This is the first study to quantify circadian variations in human VF dominant frequency and amplitude. VF is typically sustained for seconds rather than hours so there has been a paucity of data of sufficient duration to allow assessment of circadian variations of VF dominant frequency. Patients developing VF who are supported by ventricular assist devices provide a novel model that facilitates the study of human VF over much longer periods, typically weeks or months, than any other model. Using this model we have collected a unique dataset and quantified the circadian pattern of VF spectral characteristics repeatedly in the same patient over several weeks. The obvious difference of this model to acute
P. Langley et al. / Resuscitation 81 (2010) 950–955
953
Fig. 2. Comparisons of night and day dominant frequency (DF) and amplitude (Amp). There were no significant differences for dominant frequency (p > 0.05) but amplitude was significantly increased during the day (p < 0.0001).
VF is that the assist device ensures the heart remains continuously perfused. This is not a restriction in our study because we were interested in the circadian pattern of VF rather than the effects of ischaemia. However, the aetiology of VF in this model was myocarditis and because of the unique clinical circumstances it is unknown if the results obtained are generaliseable to the wider population of patients presenting with acute VF. We have demon-
strated a circadian pattern that was characterised by significant reductions in VF amplitude at night. Postural changes are the most plausible explanation for this circadian pattern, particularly since daytime amplitudes exhibited large variability reflecting the relatively active state of this patient whereas nighttime amplitudes, when the patient would be expected to be sedentary, were relatively constant. The ECG is susceptible to large amplitude changes
Fig. 3. Examples of ECG leads and corresponding spectra from night and day. For both night and day each row shows an ECG lead of 10 s duration and the power spectrum calculated from a 1 min interval of ECG containing the illustrated ECG section.
954
P. Langley et al. / Resuscitation 81 (2010) 950–955
Fig. 4. Boxplots of dominant frequency (DF) and amplitude (Amp) at each recording. There were significant differences between recordings of both dominant frequency and amplitude. Box represents the interquartile range (IQR: q3–q1) of data and median is shown as a line. Whiskers represent points within q1 − 1.5 × IQR and q3 + 1.5 × IQR.
simply due to changes in the orientation of and proximity of the heart to the recording electrodes.20 However, autonomic influences may also contribute to the circadian pattern of VF. As discussed in Section 1, the effect of the autonomic nervous system on a range of cardiac parameters is well established. Circadian changes in atrial fibrillation dominant frequency have been reported with significantly reduced cycle length (increased dominant frequency) during daytime, but not in patients with already short cycle lengths.21 In
our study dominant frequency of VF did not change significantly from day to night suggesting there is no autonomic influence on VF dominant frequency. The relative stability of dominant frequency over 24 h suggests also that there is little postural influence on this feature. The increased nighttime variability in dominant frequency clearly apparent in most of the recording days but especially from day 22 (Fig. 1) is explained by the relative flatness of the spectra, and hence increased uncertainty in identifying the dominant
Fig. 5. Histograms of dominant frequency for each lead and recording.
P. Langley et al. / Resuscitation 81 (2010) 950–955
955
Conflict of interest statement All authors have no disclosures relevant to this manuscript. Acknowledgements We are grateful to Susan King for recording the body surface potentials for this study. This work was undertaken with the financial support of an EPSRC Advanced Research Fellowship to Philip Langley. References
Fig. 6. Boxplot of dominant frequency for each lead. The dominant frequency with the maximum count for each recording was used to construct the boxplot for each lead. Two boxplots for V2 represent the bimodal distribution of the histogram.
peak, compared to daytime as illustrated in Fig. 3. Defibrillation success has been related to amplitude and dominant frequency characteristics of VF such that success in restoration of a circulatory rhythm is associated with VF of largest amplitude and highest frequency.13–15 It is interesting that in our patient waveform amplitude was greatest in the morning, when previous studies have indicated that defibrillation success is lowest.12 Median slope of the VF waveform has been proposed as a highly predictive feature of countershock success and relates to the average rate of change of VF signal.16 Applied to our data, since there was no significant circadian variation in dominant frequency, median slope would follow a similar circadian pattern to amplitude. We confirmed this for the daytime and nighttime ECGs illustrated in Fig. 3 for which daytime median slope was approximately 1.5 times that of nighttime. On the other hand, dominant frequency exhibited a pronounced initial decrease after 8 am on several days, particularly days 17, 22 and 29 as illustrated in Fig. 1, and this might account for the reduced defibrillation success rates in the morning hours reported in other studies. We did not attempt defibrillation in this patient, but nonetheless our model of VF might be useful for such studies in the future, although limited because ischaemia is probably a major factor in determining outcome of defibrillation and this is not a feature of the model. Although there was no circadian pattern of VF dominant frequency, there were significant inter-recording and inter-lead differences in dominant frequency. This is the second case of long-term VF data from our group and the results confirm our previously published data which revealed significant inter-recording and inter-lead differences in dominant frequency.19 Inter-lead differences in dominant frequency probably reflect spatial differences in underlying activation patterns that have been shown in invasive human studies.22 Day-to-day variability in dominant frequency might arise due to changes in the cardiac electrophysiological properties due to progression of disease. 5. Conclusion In this novel model of human VF sustained over several weeks we have shown a circadian pattern of VF amplitude but not dominant frequency. In a patient with severe myocarditis, supported with a biventricular assist device and in chronic VF, these results provide evidence for modulation of VF, probably induced by changes in posture and physical activity.
1. Guo Y, Stein PK. Circadian rhythm in the cardiovascular system: considerations in non-invasive electrophysiology. Card Electrophysiol Rev 2002;6:267– 72. 2. de la Sierra A, Redon J, Banegas JR, et al. Prevalence and factors associated with circadian blood pressure patterns in hypertensive patients. Hypertension 2009;53:466–72. 3. Tsuji H, Larson MG, Venditti FJ, et al. Impact of reduced heart rate variability on risk for cardiac events. The Framingham heart study. Circulation 1996;94:2850–5. 4. Van de Borne P, Nguyen H, Biston P, Linkowski P, Degaute JP. Effect of wake sleep stages on the 24 h autonomic control of blood pressure and heart rate in recumbent men. Am J Physiol Heart Circ Physiol 1994;266:H548– 54. 5. Muller JE, Ludmer PL, Wilich SN, et al. Circadian variation in the frequency of sudden cardiac death. Circulation 1987;75:131–8. 6. Zehender M, Meinertz T, Hohnloser S, et al. Prevalence of circadian variations in spontaneous variability of cardiac disorders and ECG changes suggestive of myocardial ischemia in systemic arterial hypotension. Circulation 1992;85:1808–15. 7. Tofler GH, Gebara OCE, Mittleman MA, et al. Morning peak in ventricular tachyarrhythmias detected by time of implantable cardioverter/defibrillator therapy. Circulation 1995;92:1203–8. 8. Behrens S, Franz MR. Circadian variation of arrhythmia events, electrophysiological properties, and the autonomic nervous system. Eur Heart J 2001;22:2144– 6. 9. Englund A, Behrens S, Wegscheider K, Rowland E. Circadian variation of malignant ventricular arrhythmias in patients with ischemic and nonischemic heart disease after cardioverter defibrillator implantation. J Am Coll Cardiol 1999;34:1560–8. 10. Muller JE, Stone PH, Rutherford JD, et al. Circadian variation in the frequency of onset of acute myocardial infarction. N Engl J Med 1985;313:1315–22. 11. Kong TQ, Goldberger JJ, Parker M, Wang T, Kadish AH. Circadian variation in human ventricular refractoriness. Circulation 1995;92:1507–16. 12. Venditti FJ, John RM, Hull M, Tofler GH, Shahian DM, Martin DT. Circadian variation in defibrillation energy requirements. Circulation 1996;94:1607– 12. 13. Strohmenger H, Linder KH, Brown CG. Analysis of the ventricular fibrillation ECG signal amplitude and frequency parameters as predictors of countershock success in humans. Chest 1997;111:584–9. 14. Eftestøl T, Sunde K, Aase OS, Husøy JH, Steen PA. Predicting outcome of defibrillation by spectral characterization and nonparametric classification of ventricular fibrillation in patients with out-of-hospital cardiac arrest. Circulation 2000;102:1523–9. 15. Reed MJ, Clegg GR, Robertson CE. Analysing the ventricular fibrillation waveform. Resuscitation 2003;57:11–20. 16. Neurauter A, Eftestøl T, Kramer-Johansen J, et al. Prediction of countershock success using single features from multiple ventricular fibrillation frequency bands and feature combinations using neural networks. Resuscitation 2007;73:253–63. 17. Chen J, Mandapati R, Berenfeld O, Skanes AC, Jalife J. High-frequency periodic sources underlie ventricular fibrillation in the isolated rabbit heart. Circ Res 2000;86:86–93. 18. Skanes AC, Mandapati R, Berenfeld O, Davidenko JM, Jalife J. Spatiotemporal periodicity during atrial fibrillation in the isolated sheep heart. Circulation 1998;98:1236–48. 19. Langley P, MacGowan GA, Murray A. Spatial and temporal organisation of the dominant frequencies in the fibrillating heart: body surface potential mapping in a rare case of sustained human ventricular fibrillation. Europace 2009;11:324–7. 20. García J, Åström M, Mendive J, Laguna P, Sörnmo L. ECG-based detection of body position changes in ischemia monitoring. IEEE Trans Biomed Eng 2003;50:677–85. 21. Meurling CJ, Waktare JEP, Holmqvist F, et al. Diurnal variations of the dominant cycle length of chronic atrial fibrillation. Am J Physiol Heart Circ Physiol 2001;280:H401–6. 22. Nash MP, Mourad A, Clayton RH, Sutton PM, Bradley CP, Hayward M, et al. Evidence for multiple mechanisms in human ventricular fibrillation. Circulation 2006;114:536–42.