Predicting the Success of Defibrillation and Cardiopulmonary Resuscitation

Ventricular fibrillation (VF) remains the primary rhythm in many instances of sudden cardiac death, and defibrillation by electrical counter-shock represents the treatment of choice for this otherwise lethal arrhythmia. There is no doubt that the duration of VF remains one of the principal determinants for the likelihood of successful defibrillation. When the interval between the estimated onset of VF and the delivery of the first shock is less than 5 min, there is evidence that an immediate electrical shock would be successful [1]. When the duration of untreated VF exceeds 5 min, however, both human and animal studies demonstrate that initial CPR, with chest compression, prior to delivery of a defibrillation attempt, improves the likelihood of restoration of spontaneous circulation (ROSC) [2,3].
Despite major efforts to improve outcomes of cardiac arrest, including new technologies to guide rescuers through CPR maneuvers and new real-time approaches to assess the effectiveness and quality of such interventions, survival from cardiac arrest remains disappointing. For this reason, continued effort is directed toward evaluating new methods that might provide substantial information to the rescuers which would enable them to predict the success of defibrillation and CPR. Specifically, information is needed regarding the duration of the untreated cardiac arrest and the priority of intervention to be performed, namely defibrillation or chest compression.
Accordingly, the different approaches to assess the effectiveness of CPR and predict the success of defibrillation can be summarized in standard assessment of hemodynamics, or in methods based on electrocardiographic parameters measurement.
Ventricular fibrillation is characterized by three time-sensitive electrophysiological phases, including
(1) the electrical phase of 0–4 min,
(2) the circulatory phase of 4–10 min, and
(3) the metabolic phase of >10 min.
During the electrical phase, immediate defibrillation is likely to be successful. As ischemia progresses, the success of attempted defibrillation diminishes without CPR. This phase is characterized by transition to slow VF wavelets during accumulation of ischemic metabolites in the myocardium. Type II VF often fails defibrillation attempts because of re-entry and recurrence of VF. In the metabolic phase, there is no likelihood of successful restoration of a perfusing rhythm [4]. During untreated VF, progressive and severe energy imbalance develops associated with increasing intramyocardial hypercarbic acidosis and depletion of high-energy phosphates. Early CPR, such as to restore coronary perfusion pressure (CPP) and myocardial blood flow, delays onset of ischemic myocardial injury leading to the well-known condition of “stone heart” and facilitates defibrillation [5].
The development of a noninvasive and real-time monitoring that allows prediction of whether or not a shock would achieve return of spontaneous circulation is of great importance in order to prioritize interventions, chest compression or defibrillation, such as to reduce the number of failed defibrillation attempts, the interruptions in CPR, and ultimately improve final outcome of cardiac arrest.
More than 50% of all patients initially resuscitated from cardiac arrest subsequently die before leaving the hospital and the majority of these deaths are due to impaired myocardial function [6–8]. The severity of postresuscitation myocardial dysfunction has been recognized to be related, in part, to the magnitude of the total electrical energy delivered with defibrillation [9]. Increases in the defibrillation energy were associated with decreased postresuscitation myocardial function [9,10]. The ability to predict defibrillation success may therefore minimize the damaging effects of repetitive and unnecessary electrical shocks.
Adverse outcomes also follow interruptions of precordial compression. Accordingly, uninterrupted chest compression would be expected to and, in fact, does produce better 24-h survival and neurological recovery [11]. Substantial interruptions of chest compressions have, instead, detrimental effects on the success of CPR [12], reducing the likelihood of success of defibrillation because of immediate declines of coronary perfusion [11,13]. Even minimal interruptions of only 4 sec cause decreases in aortic diastolic pressure and CPP, which require as many as seven chest compressions before achieving a return to maximal effect. In response thereto, new approaches, able to instruct the rescuers for the best timing for defibrillation, might further reduce interruptions of chest compression due to delivery of probably nsuccessful defibrillation attempts by prompting the delivery of the electrical shock only when there is high likelihood of their success.
Standards for Predicting Success of Defibrillation and Return to Spontaneous Circulation
The evidence is certain that the quality of chest compressions is a major determinant of successful resuscitation. Established predictors of good quality CPR therefore can predict the success of defibrillation and thereby successful resuscitation. For this purpose, invasive hemodynamic measurements and especially coronary perfusion pressure [14–16], and end-tidal CO2 (EtCO2) are therefore employed [17,18]. Blood flows generated by chest compressions are dependent on the pressure gradient between the aortic and the venous pressures. CPP, defined as the difference between simultaneously measured minimal aortic pressure and right atrial pressure during compression diastole, is highly correlated with coronary blood flow during cardiac resuscitation and is currently recognized as the best single indicator of the likelihood of successful defibrillation and ROSC [16,19]. Based on both experimental and clinical observations, ROSC can be predicted when CPP is maintained above 15 mmHg during chest compressions [16]. Resuscitative strategies that increase CPP, including high quality chest compressions as well as the use of vasopressors, have therefore been supported and considered more effective in restore circulation (Pict. 1).

Picture 1 Coronary Perfusion Pressure (CPP) and end-tidal CO2 (EtPCO2) during the first 3 min of precordial compressions (PC). CPP and EtPCO2 were significantly greater in resuscitated animals. * p <0.05, ** p <0.01 and † p <0.0001. Modified from [20]

Expired CO2 is determined by the body’s production of CO2 and the relationship between minute ventilation and pulmonary perfusion. When the circulatory status is normal, pulmonary perfusion is in the physiologic ranges and EtCO2 is determined by minute ventilation. Under settings of cardiac arrest and CPR, cardiac output is usually less than one third of normal and therefore pulmonary flow and EtCO2 is dramatically reduced. EtCO2 is therefore an indirect measurement of pulmonary blood flow and cardiac output produced by chest compressions [17,18]. End-tidal CO2 is highly correlated with CPP during CPR, and may therefore serve as a noninvasive surrogate for CPP and has emerged as another valuable tool for monitoring the effectiveness of chest compressions during CPR [17,18,20–22]. When EtCO2 exceeds the threshold level of approximately 10–15 mmHg during CPR, greater likelihood of successful ROSC has been reported [23,24]. EtCO2 values after 20 min of CPR have been recently recognized as even more reliable in predicting success of resuscitation [25].
Experimentally, in a porcine model of cardiac arrest and CPR, CPP and EtCO2 above the threshold levels were the only predictors for successful resuscitation, independently from the priority of intervention, chest compression, or defibrillation [20]. Although the importance of blood pressures during CPR is clear, invasive measurements, including aortic and right atrial pressures are only available or feasible at the time of resuscitation in a very small minority of patients in critical care settings. The use of EtCO2 measurements is also not widely available, especially because of the need of endotracheal intubation.
These restraints are in contrast, however, with the routine availability of the electrocardiogram (ECG) available in current external defibrillators. The attention, with the intent to identify a better predictor of defibrillation and ROSC, has been therefore focused on the analyses of electrocardiographic features of VF waveform. Filters and algorithms to reduce and eliminate ECG artifacts and noise due to chest compressions or ambient interferences have been developed. Among these, wavelet transform technique constitutes one of the most promising methods [26]. It allows for more reliable real-time ECG analyses to be used to predict defibrillation success and ultimately guide CPR interventions.
Analyses of ECG Features During Ventricular Fibrillation and CPR
We now recognize the importance of ECG signal analysis for predicting whether an electrical shock would successfully reverse VF. Real-time ECG analysis is therefore predictive of patient outcome. Moreover, the ECG analysis is dependent only upon the patient’s condition at the time of treatment, with no need for knowledge of the response interval, which may be difficult to estimate [27]. The initial approaches to ECG analysis included measurements of VF amplitude [28], and then frequency [29].
Picture 2. Increases in VF amplitude after 5 min of CPR in comparison to untreated ventricular fibrillation
Earlier investigations using ECG analysis focused on “amplitude or voltage” of VF wavelets as a predictor of the likelihood of successful defibrillation. VF voltage, or signal amplitude, is defined as the maximum peak-to-trough VF amplitude in a given time window of the ECG signal. Mean VF voltage is the average of VF voltage over the same time interval. It has been observed that VF amplitude declines over time, and greater amplitudes, especially after an interval of CPR, as shown in Picture.2, are associated with correspondingly greater success of defibrillation [28,30–33]. This is because this ECG feature reflects myocardial blood flow and energy metabolism [28,32,33]. Based on the study of Weaver et al. [28], VF amplitude greater than 0.2 mV was recognized as a predictor of significantly greater likelihood of resuscitation.
Subsequently, other parameters were computed utilizing Fourier transformation analyses in a selected ECG interval. Specifically, VF median frequency, peak power frequency, edge frequency, and spectral flatness measure were introduced. The starting point for all these calculations is the “power spectrum,” defined as the square of Fourier amplitudes. Brown et al. [29,34] specifically developed this technique that analyzed VF voltage and VF frequency such to obtain the so-called VF median frequency. Median frequency of VF served as a predictor of the success of electrical defibrillation [35,36]. Experimentally, a median frequency of more than 9.14 Hz had 100% sensitivity and 92% specificity in predicting the success of defibrillation. Frequency analysis of VF wavelets and, specifically, median frequency was also correlated with CPPs in animal models as well as human victims and therefore became the preferred ECG feature to be used as a predictor of outcome [35–38].
To improve sensitivity and specificity of the ECG predictors for defibrillation success and ROSC, more sophisticated methods of VF waveform analyses have recently been introduced and investigated, including wavelet decomposition, nonlinear dynamics methods, and a combination of different ECG parameter analyses. Since then, several new approaches have been proposed and their effectiveness proved in predicting defibrillation outcomes. One of these employed the N(a) histograms analysis, which was demonstrated to be superior to mean VF frequency analysis [39]. A combination of spectral features in the VF waveform was anticipated as a monitor of efficacy of interventions during CPR, evolving into a “probability of successful defibrillation” function [40]. This is a combination of two decorrelated spectral features based on a principal component analysis of an original feature set with information on centroid frequency, peak power frequency, spectral flatness, and energy. The function “probability of defibrillation success” discriminated between shocks followed by ROSC and No-ROSC. Methods employing entropy measure have also been shown to provide more optimal prediction of ROSC after electrical shock in human VF recordings [41], and methods of filtered ECG features from higher ECG subbands, instead of features derived from the main ECG spectrum, have improved accuracy of shock outcome prediction during CPR [42].
Our Approach: Amplitude Spectrum Area (AMSA)
The Amplitude Spectrum Area (AMSA) is another efficient ECG-derived defibrillation predictor in which mean amplitude and dominant frequency are combined. The amplitude spectrum is obtained by fast Fourier transform of the ECG scalar signal (Pict.3), and is calculated from the resulting amplitude frequency spectrum according to the following equation:
Picture.3 A representative example of the amplitude frequency relationship and the area under the curve that defines the amplitude spectrum area (AMSA)
 AMSA = ε Ai x Fi
where Ai is the amplitude at the ith frequency Fi.
In a porcine model of cardiac arrest and CPR, threshold values of AMSA for defibrillation success have been established and AMSA has confirmed its capability to optimize the timing of ventricular defibrillation [43]. It was, in fact, highly correlated with CPP levels during CPR and, similarly to CPP, significantly greater values were observed in animals that were resuscitated compared to those that were not [44]. Amplitude spectrum analysis demonstrated a negative predictive value for resuscitation of 96% and a positive predictive value of 78%. It is therefore apparent that AMSA represents a measurement that potentially fulfills the need for minimizing ineffective and detrimental defibrillation attempts during resuscitative maneuvers. The high negative predictive value, in particular, would minimize repetitive and ineffective electrical shocks during CPR. The progressive increases in AMSA observed before successful resuscitation further demonstrate that AMSA has the potential of providing an objective guide allowing for better quality control of CPR. Failure to increase AMSA values to near threshold levels prognosticates, instead, failure of defibrillation.
Subsequent validation studies have confirmed that AMSA has an impressively higher specificity and positive predictive value compared with the other predictors, maintaining sensitivity and negative predictive value comparable to the invasively assessed CPP. More importantly, AMSA is not invalidated by artifacts resulting from precordial compression, fulfilling the goal of a predictor that would allow for uninterrupted precordial compression during ECG analyses.
AMSA has also proved its validity as predictor for defibrillation outcomes in the clinical scenario [45,46]. Retrospective analysis of human ECG records, representing lead 2 equivalent recordings, confirmed the efficacy of this tool in predicting the likelihood that any one electrical shock would have restored a perfusing rhythm during CPR. AMSA values were significantly greater in successful defibrillation, compared to unsuccessful defibrillation (Table 15.1). A threshold value of AMSA of 12 mV-Hz was able to predict the success of each defibrillation attempt with sensitivity and specificity of more than 91%.
Table 15.1 Amplitude spectrum area and success of defibrillation (DF) attempts. Modified from [46]

PR, Return of a perfusing rhythm; NR, failure of return of a perfusing rhythm * p <0.0001
New Approach to Predict Success of Defibrillation
In assessing the critical perfusion condition during and after resuscitation from cardiac arrest, both investigators and clinicians focused on pressure and blood flow through large vessels and cardiac output. CPR interventions and especially chest compression focused on increasing and maintaining optimal pressures such as to favor large vessel supplies to the heart, in order to prime the heart to successful defibrillation. With the advent of methods by which microvessels and especially capillaries could be visualized, it has become apparent that large vessel pressures and flows alone may not be predictive of the extent to which microvessels and therefore tissues are perfused. Yet, it is the microvessels and specifically the capillaries which serve as the ultimate exchange sites for vital metabolites. The availability of the Orthogonal Polarization Spectral, first, and now Sidestream Dark Field imaging techniques, allowed direct and real-time visualization of arterioles, venules, and capillaries and therefore provided a tool for assessing the effect of CPR interventions on the microcirculation [47]. Experimentally, we investigated changes in sublingual microcirculation during cardiac arrest and CPR. With the aid of the microcirculation imaging, we observed that microvascular blood flow in the sublingual mucosa was highly correlated with CPP during CPR. Like CPP, the magnitude of microcirculatory blood flow was indicative of the effectiveness of the resuscitation intervention and success of defibrillation attempt and outcome. In animals that were resuscitated, microvascular flow was significantly greater than that assessed in animals in which resuscitation attempts failed.
Tissue hypercarbia measurements might also represent a future tool to assess, noninvasively, hemodynamics generated by chest compressions and therefore predict success of CPR intervention. When oxygen delivery to the tissues is critically reduced during circulatory failure states, anaerobic metabolism is triggered with consequent hydrogen ion production. This excess of hydrogen ions is buffered by tissue bicarbonate such that CO2 is generated. Buccal PCO2 has therefore been confirmed as a useful guide to the diagnosis of circulatory shock, including cardiac arrest. It also provided rapid response for confirmation of the effectiveness of treatments. More recently, it has appeared as a useful tool to predict the duration of untreated cardiac arrest and thereby direct the rescuers to the best initial resuscitation maneuvers, chest compression or defibrillation [48].
CPP and EtCO2 represent useful standard means for monitoring the effectiveness of chest compressions and to predict the success of CPR. These measurements, however, are not feasible in settings of out-of-hospital cardiac arrest. Investigators have therefore now focused their attention on the morphology of the VF waveform in order to predict the success of resuscitation. However, the challenge is to ensure high sensitivity and specificity, especially during precordial compression, in order to identify the ideal moment to deliver the defibrillatory shock. A spectral analysis method such as AMSA is a simple parameter that can be easily obtained by a conventional surface ECG that is part of the routinely current practice of advanced cardiac life support. Experimentally, consistent evidence of the validity of AMSA has been proved in both animal and human data, with the advantage of being employed as a real-time indicator for effectiveness of chest compressions and prediction of the success of defibrillation. Future approaches might also integrate noninvasive measurements of blood flows produced by chest compression, including real-time measurements of sublingual microcirculation and tissue PCO2 during CPR.

1. Valenzuela TD, Roe DJ, Nichol G et al (2000) Outcomes of rapid defibrillation by security officers after cardiac arrest in casinos. N Engl J Med 343:1206–1209
2. Wik L, Hansen TB, Fylling F et al (2003) Delaying defibrillation to give basic cardiopulmonary resuscitation to patients with out-of-hospital ventricular fibrillation. JAMA 289:1389–1395
3. Berg RA, Hilwig RW, Ewy GA et al (2004) Precountershock cardiopulmonary resuscitation improves initial response to defibrillation from prolonged ventricular fibrillation: a randomized, controlled swine study. Crit Care Med 32:1352–1357
4. Chen PS, Wu TJ, Ting CT et al (2003) A tale of two fibrillations. Circulation 108:2298–2303
5. Klouche K, Weil MH, Sun S et al (2002) Evolution of the stone heart after prolonged cardiac arrest. Chest 122:1006–1011
6. Peatfield RC, Sillett RW, Taylor D et al (1977) Survival after cardiac arrest in the hospital.Lancet 1:1223–1225
7. DeBard ML (1981) Cardiopulmonary resuscitation: analysis of six years’ experience and review of the literature. Ann Emerg Med 10:408–416
8. Schenenberger RA, von Planta M, von Planta I (1994) Survival after failed out of hospital resuscitation. Are further therapeutic efforts in the emergency department futile? Arch Intern Med 154:2433–2437
9. Xie J, Weil MH, Sun S et al (1997) High-energy defibrillation increases the severity of postresuscitation myocardial dysfunction. Circulation 96:683–688
10. Tang W, Weil MH, Sun S et al (2004) The effects of biphasic waveform design on post-resuscitation myocardial function. J Am Coll Cardiol 43:1228–1235
11. Berg RA, Sanders AB, Kern KB et al (2001) Adverse hemodynamic effects of interrupting chest compressions for rescue breathing during cardiopulmonary resuscitation for ventricular fibrillation cardiac arrest. Circulation 104:2465–2470
12. Yu T, Weil MH, Tang W et al (2002) Adverse outcome of interrupted precordial compression during automated defibrillation. Circulation 106:368–372
13. Steen S, Liao Q, Pierre L et al (2003) The critical importance of minimal delay between chest compressions and subsequent defibrillation: a haemodynamic explanation. Resuscitation 58:249–258
14. Deshmukh HG, Weil MH, Gudipati CV et al (1989) Mechanism of blood flow generated by precordial compression during CPR, I: studies on closed chest precordial compression. Chest 95:1092–1099
15. Sanders AB, Ogle M, Ewy GA (1985) Coronary perfusion pressure during cardiopulmonary resuscitation. Am J Emerg Med 2:11–14
16. Paradis NA, Martin GB, Rosenberg J et al (1990) Coronary perfusion pressure and the return of spontaneous circulation in human cardiopulmonary resuscitation. JAMA 263:1106–1113
17. Weil MH, Bisera J, Trevino RP, Rackow EC (1985) Cardiac output and end-tidal carbon dioxide. Crit Care Med 13:907–909
18. Falk JL, Rackow EC, Weil MH (1988) End-tidal carbon dioxide concentration during cardiopulmonary resuscitation. N Engl J Med 318:607–611
19. Niemann JT, Criley JM, Rosborough JP et al (1985) Predictive indices of successful cardiac resuscitation after prolonged arrest and experimental cardiopulmonary resuscitation. Ann Emerg Med 14:521–528
20. Ristagno G, Tang W, Chang YT et al (2007) The quality of chest compressions during cardiopulmonary resuscitation overrides importance of timing of defibrillation. Chest 132:70–75
21. Gudipati CV, Weil MH, Bisera J et al (1988) Expired carbon dioxide: a noninvasive monitor of cardiopulmonary resuscitation. Circulation 77:234–239
22. von Planta M, von Planta I, Weil MH et al (1989) End tidal carbon dioxide as an haemodynamic determinant of cardiopulmonary resuscitation in the rat. Cardiovasc Res 23:364–368
23. Grmec S, Klemen P (2001) Does the end-tidal carbon dioxide (EtCO2) concentration have prognostic value during out-of hospital cardiac arrest? Eur J Emerg Med 8:263–269
24. Cantineau JP, Lambert Y, Merckx P et al (1996) End-tidal carbon dioxide during cardiopulmonary resuscitation in humans presenting mostly with asystole: a predictor of outcome. Crit Care Med 24:791–796
25. Kolar M, Krizmaric M, Klemen P et al (2008) Partial pressure of end-tidal carbon dioxide successful predicts cardiopulmonary resuscitation in the field: a prospective observational study. Crit Care 12:R115
26. Li Y, Bisera J, Geheb F et al (2008) Identifying potentially shockable rhythms without interrupting cardiopulmonary resuscitation. Crit Care Med 36:198–203
27. Snyder DE, White RD, Jorgenson DB (2007) Outcome prediction for guidance of initial resuscitation protocol: shock first or CPR first. Resuscitation 72:45–51
28. Weaver MD, Cobb LA, Dennis D et al (1985) Amplitude of ventricular fibrillation waveform and outcome after cardiac arrest. Ann Intern Med 102:53–55
29. Brown CG, Griffith RF, Van Ligten P et al (1991) Median frequency: a new parameter for predicting defibrillation success rate. Ann Emerg Med 20:787–789
30. Dalzell GW, Adgey AA (1991) Determinants of successful transthoracic defibrillation and outcome in ventricular fibrillation. Br Heart J 65:311–316
31. Callaham M, Braun O, Valentine W et al (1993) Prehospital cardiac arrest treated by urban first-responders; profile of patient response and prediction of outcome by ventricular fibrillation waveform. Ann Emerg Med 22:1664
32. Strohmenger HU, Lindner KH, Brown CG (1997) Analysis of the ventricular fibrillation ECG signal amplitude and frequency parameters as predictors of countershock success in humans. Chest 111:584–589
33. Noc M, Weil MH, Gazmuri RJ et al (1994) Ventricular fibrillation voltage as a monitor of the effectiveness of cardiopulmonary resuscitation. J Lab Clin Med 124:421–426
34. Brown CG, Dzwonczyk R, Werman HA et al (1989) Estimating the duration of ventricular fibrillation. Ann Emerg Med 18:1181–1185
35. Brown CG, Griffith RF, Van Ligten P et al (1991) Median frequency: a new parameter for predicting defibrillation success rate. Ann Emerg Med 20:787–789
36. Strohmenger HU, Lindner KH, Lurie KG et al (1994) Frequency of ventricular fibrillation as predictor of defibrillation success during cardiac surgery. Anesth Analg 79:434–438
37. Brown CG, Dzwonczyk R (1996) Signal analysis of the human electrocardiogram during ventricular fibrillation: frequency and amplitude parameters as predictors of successful countershock. Ann Emerg Med 27:184–188
38. Stewart AJ, Allen JD, Adgey AA (1992) Frequency analysis of ventricular fibrillation and resuscitation success. Q J Med 85:761–769 39. Amann A, Achleitner U, Antretter H et al (2001) Analysing ventricular fibrillation ECG-signals and predicting defibrillation success during cardiopulmonary resuscitation employing N(alpha)-histograms. Resuscitation 50:77–85
40. Eftestol T, Sunde K, Aase SO et al (2001) “Probability of successful defibrillation” as a monitor during CPR in out-of-hospital cardiac arrested patients. Resuscitation 48:245–254
41. Watson JN, Uchaipichat N, Addison PS et al (2004) Improved prediction of defibrillation success for out-of-hospital VF cardiac arrest using wavelet transform methods. Resuscitation 63:269–275
42. Neurauter A, Eftestøl T, Kramer-Johansen J et al (2008) Improving countershock success prediction during cardiopulmonary resuscitation using ventricular fibrillation features from higher ECG frequency bands. Resuscitation 79:453–459
43. Pernat AM, Weil MH, Tang W et al (2001) Optimizing timing of ventricular defibrillation. Crit Care Med 29:2360–2365
44. Povoas HP, Weil MH, Tang W et al (2002) Predicting the success of defibrillation by electrocardiographic analysis. Resuscitation 53:77–82
45. Young C, Bisera J, Gehman S et al (2004) Amplitude spectrum area: measuring the probability of successful defibrillation as applied to human data. Crit Care Med 32:S356-S358
46. Ristagno G, Gullo A, Berlot G et al (2008) Electrocardiographic analysis for prediction of successful defibrillation in human victims of out of hospital cardiac arrest. Anaesth Intensive Care 36:46–50
47. Fries M, Weil MH, Chang YT et al (2006) Microcirculation during cardiac arrest and resuscitation. Crit Care Med 34:S454-S457

48. Ristagno G, Cho JH, Yu T et al (2008) Non-invasive measurements for predicting duration of untreated cardiac arrest. Circulation 118:S664