Paper accepted for presentation at 2002 Joint Meeting of the IEEE EMBS and the BMES, Houston, Texas, ... this effect âvagal fatigueâ for the following reason.
Prediction of Tachyarrhythmia Episodes Tran Thong1, Brahm Goldstein2 OGI School of Science & Engineering, Oregon Health & Science University, Beaverton, Oregon , USA 2 Depts of Pediatrics, Oregon Health & Science University, Portland, Oregon, USA
Abstract – Ventricular tachyarrhythmia, in particular ventricular fibrillation, may be partially enabled by an imbalance within the autonomic nervous system. Decreased parasympathetic (or vagal) cardiac regulatory activity may be responsible for the development of conditions that allow ventricular fibrillation to develop and sustain itself. An analysis of R-R intervals immediately prior to episodes of ventricular fibrillation has yielded a modulation pattern of the R-R intervals, which was termed “vagal fatigue,” that appears in a majority of ventricular fibrillation episodes detected and treated by an implantable defibrillator. When absent, this modulation pattern is typically associated with an aborted defibrillation shock, indicating that parasympathetic activity is sufficiently strong that it was able to overcome the tachyarrhythmia prior to therapy delivery. The tachyarrhythmia detector compares the instantaneous R-R interval with a target R-R. The resulting error signal was used to set the “vagal fatigue” i ndicator. Keywords – ventricular fibrillation, VF, tachyarrhythmia, R-R intervals, vagal fatigue, tachycardia, prediction. I.
INTRODUCTION
In the past, efforts aimed at predicting impending episodes of tachyarrhythmia were based on studies of heart rate variability [1], HRV. While traditional statistical or frequency domain based HRV analyses were successful in long-term prediction of patient mortality, they have not been useful in short-term prediction of tachyarrhythmia episodes. Through the use of the nonlinear correlation dimension PD2 of the heartbeat, Skinner and Vybrial[2] were able to link the reduction in PD2 with an imminent episode of ventricular fibrillation. However, this is a complicated measure that is not easily amenable to inclusion in cardiac monitoring or implantable devices. Furthermore, PD2 reduction appears over a long time period (>12 hours) [2]. Thus, PD2 is not useful as an indicator of imminent tachyarrhythmias.
ventricular fibrillation (VF), or ventricular tachycardia (VT), and delivery of programmed therapy. We looked only at records that were > 1000 intervals. In our analyses, we searched for an R-R interval modulation pattern that correlated with imminent ventricular fibrillation. An example of a such a record is illustrated in Figure 1. Since we were interested in the short-term trend of the heart rhythm, ectopic beats (e.g. premature ventricular contractions, PVCs) were removed with a median nonlinear filter. This resulted in the trend illustrated in Fig. 2. These trended data were used for the analyses. II.
Results
We found the staircase effect illustrated in Fig. 2 from 600 s to 500 s prior to VF to indicate an imminent VF. We termed this effect “vagal fatigue” for the following reason. Normally, as non-vagal effects reduce the R-R interval (i.e. increase heart rate), parasympathetic (or vagal) cardiac activity increases with a resultant increased R-R interval. This mechanism is partially evident at 585 s in Fig. 2. However, decreased or fatigued parasympathetic activity then became apparent. This pattern was repeated at 570s and 550s and resulted in the staircase-like modulation of the R-R intervals. While instances of vagal fatigue occurred frequently during sinus rhythm, they rarely exceed a duration of 50 intervals (~ 40 s.) We found a strong correlation between vagal fatigue events of duration greater than 50 intervals and imminent VF episodes (Figs. 3 and 4). In our analyses, vagal fatigue was detected as follows. A target R-R interval was calculated from long-term averaging of earlier intervals. The difference between the actual(filtered) R-R interval and the target interval was 1600
1400
1200
1000 RR (ms)
1
Our goal was to derive a reliable predictor of tachyarrhythmia that can be easily computed within an implantable device. II.
800
600
METHODOLOGY 400
The patient database of the Biotronik (Berlin, Germany) Phylax XM and Biotronik Mycro Phylax implantable cardioverter defibrillators (ICDs) was used for this initial investigation. These two ventricular ICDs have the capability of recording long-range R-R intervals prior to the detection of an episode of tachyarrhythmia, either
200 3000
2500
2000 1500 1000 Time (seconds before detection of arrhythmia)
500
0
1. Example of an R-R sequence prior to an episode of VF. (Record is 3928 intervals long)
Paper accepted for presentation at 2002 Joint Meeting of the IEEE EMBS and the BMES, Houston, Texas, 23-26 October 2002. Copyright IEEE 2002.
1600
1400
predictor. The reason is that an aborted shock is an indicator that the vagal activity was strong enough to return the rhythm back to normal. Thus, in 88% of the VF episodes analyzed, the vagal fatigue predictor appeared to correctly assess the underlying parasympathetic activity.
1200
RR (ms)
1000
800
With VT episodes, the success rate was lower (58%). We believe that this was due to: •
Anti-tachycardia pacing therapy delivered without delay after VT detection resulting in only ~5 s for vagal activity to terminate the VT episode before therapy is delivered. This time was up to 12 s prior to shock therapy for VF. VT is most often a re-entry tachycardia, which is a different mechanism than for VF.
600
400
•
200 700
650 600 550 Time (seconds before detection of arrhythmia)
500
2. R-R interval record after median filtering. computed. When this error signal was negative (actual R-R < target R-R) for >50 intervals, a vagal fatigue was labeled. 50 episodes of VF from the Biotronik database were analyzed using the vagal fatigue predictor. The results are summarized in Fig. 3. Similarly, 81 episodes of VT were analyzed and the results summarized in Fig. 4. III. Discussion In addition to the correctly predicted episodes of VF, we considered the non-detected episodes of VF with aborted shock treatments in Fig. 3 to also validate the vagal fatigue Distribution of VF Episodes VF predicted. 24 - 48%
not detected: aborted shock. 20- 40% not detected: pacing. 1 - 2% not detected: short record. 1 - 2% not detected: unknown reason. 4 - 8%
3. Distribution of VF episodes analyzed
The distribution of the vagal predictors closest to the VT/VF is shown in Fig. 5. Vagal fatigue predictors occurred within 2 hours in 98% of episodes, 1 hour in 88%, and 30 minutes within 71%. III.
Conclusions
Our preliminary analysis indicates that most episodes of VF requiring therapy can be predicted by vagal fatigue from 1 minute to 2 hours prior to the tachyarrhythmia. If this prediction capability is validated in a larger ICD population, we may have a tool for initiating preventive therapy and avoid episodes of VF or VT. ACKNOWLEDGEMENT The authors are grateful to Biotronik for making available the R-R interval database. REFERENCES [1] Heart Rate Variability, M. Malik, A.J. Camm, Ed., Futura Publishing Company, Inc., Armonk, NY, 1995. [2] J.E. Skinner, C.M. Pratt, T. Vybiral, “A reduction in the correlation dimensions of heartbeat intervals preceded imminent ventricular fibrillation in human subjects”, Am Heart J. 1993, Vol. 125, pp. 731-743. 18 16 14
Distribution of VT Episodes
Not detected: aborted therapy: 1 - 1% VT detected: 47 - 58%
Not detected: short record: 8 - 10%
12
Frequency
Not detected therapies: 25 - 31%
10 8 6 4 2 0
20"30"40"50" 1' 2' 3' 4' 5' 6' 7' 8' 9' 10' 20' 30' 40' 50' 1h 2h 3h
Minimum prediction length
5. Histogram of minimum prediction length. 4. Distribution of VT episodes analyzed.
Paper accepted for presentation at 2002 Joint Meeting of the IEEE EMBS and the BMES, Houston, Texas, 23-26 October 2002. Copyright IEEE 2002.
