Article Text
PDF
Abstract
Background It is unclear if there are predictors of survival, including ECG characteristics, that can guide resuscitative efforts in pulseless electrical activity (PEA) cardiac arrests. We studied the predictive potential of presenting prehospital ECGs on survival for patients with out-of-hospital cardiac arrest (OHCA) with PEA.
Methods We studied prehospital ECGs of patients with OHCA prospectively enrolled between June 2007 and November 2009 at the Ottawa/OPALS (Ontario Prehospital Advanced Life Support Study) site of the Resuscitation Outcomes Consortium PRIMED study (Prehospital Resuscitation using an IMpedance valve and Early versus Delayed analysis). We included adult non-traumatic OHCA with PEA rhythm where resuscitation was attempted. We measured HR, QRS interval and presence of P waves, and determined their impact on return of spontaneous circulation (ROSC) and survival to hospital discharge (SHD) using multivariate regression analysis.
Results The demographic characteristics of the 332 included cases were the following: mean age 71.8, male 58.4%, SHD 5.4% and ROSC at ED arrival 26.5%. Survivors had similar HR (56.8 vs 52.0 beats per minute (bpm), p=0.53) and QRS intervals (128.7 vs 129.6 ms, p=0.95) compared with non-survivors. Prehospital ECG characteristics did not predict SHD or ROSC on multivariate analyses. Patients with initial HR <30 bpm had a 3.8% survival rate; those with both HR <30 bpm and QRS≥120 ms had a 3.7% survival rate. Location of arrest predicted SHD (adjusted OR (AdjOR)=1.49, 1.11 to 1.99; p=0.007). Atropine use negatively predicted SHD (AdjOR=0.06, 0.02 to 0.22; p<0.001). Predictors of ROSC ALS paramedic on scene (AdjOR=8.90, 1.11 to 71.41; p=0.04) and successful intubation (AdjOR=3.35, 1.75 to 6.39; p<0.001). Atropine use negatively predicted ROSC (AdjOR=0.27, 0.14 to 0.50; p<0.001).
Conclusions Presenting prehospital ECG characteristics did not predict SHD or ROSC in OHCA PEA victims and should not be used to guide termination of resuscitation. Location of arrest was a positive predictor for SHD; atropine use was a negative predictor. ALS paramedic on scene and successful intubation were positive predictors of ROSC; atropine use was a negative predictor.
Trial registration number NCT00394706; post-results.
- cardiac arrest
- emergency departments
Statistics from Altmetric.com
Key messages
What is already known on this subject
The proportion of patients with cardiac arrest with an initial pulseless electrical activity (PEA) rhythm is rising and the overall survival is low.
There is a spectrum of aetiologies in patients with PEA arrest, some of which are potentially reversible.
It is unclear if predictors of survival exist or whether such predictors can guide resuscitative efforts, including presenting ECG characteristics.
What this study adds
Initial ECG characteristics (HR, QRS width and presence of P waves) have no clear predictive abilities on survival and should not guide resuscitation decisions.
Survival was noted in a few patients with unfavourable initial ECG characteristics, suggesting that termination of resuscitation decisions should not be made based solely on such presenting ECG characteristics.
Introduction
Pulseless electrical activity (PEA) in cardiac arrest is defined as coordinated electrical activity of the heart not due to ventricular fibrillation (VF) or ventricular tachycardia (VT) without a palpable pulse.1 2 PEA includes electromechanical dissociation (EMD or true PEA), where no myocardial contractions occur, and pseudo-PEA (or pseudo-EMD), where myocardial contractions occur without a palpable pulse. Typically treated equally, they are essentially different clinical entities. True PEA represents a complex and more severe pathology. It usually manifests as bradycardia with a wide QRS complex and is often seen in prolonged cardiac arrest, following defibrillation of prolonged VF or with other conditions such as hyperkalaemia, hypothermia and drug overdose.3 4 In contrast, pseudo-PEA may be a transient state in the progression to true PEA due to global myocardial dysfunction or it may be caused by a host of extracardiac causes. These include hypovolaemia, papillary and myocardial wall rupture, tension pneumothorax, pericardial tamponade, and massive pulmonary embolism.5 6 In most cases of pseudo-PEA, the heart continues to make purposeful contractions, but forward flow is greatly diminished, resulting in absence of consciousness and pulse. Pseudo-PEA often manifests as a narrow complex tachycardia, which can progress to true PEA, VF or asystole.
The clinical outcomes for out-of-hospital cardiac arrest (OHCA) cases are grim. Recent literature suggests that the proportion of cardiac arrests presenting with PEA is rising and present in up to 20%–25% of cases. Meanwhile, the proportion of cardiac arrests in ‘shockable rhythms’, such as VF or VT, which are typically associated with higher survival, is decreasing over time.7–11 Unlike out-of-hospital VF/VT arrests where the survival rate can be 20% or more, the survival rate in PEA is typically 2%–5%, which is only slightly higher than the survival rates seen with asystolic OHCA.9 11–13
Unfortunately, the literature surrounding prognosis of PEA arrest is older, limited by small sample sizes, and is lacking in survival data beyond hospital admission or return of spontaneous circulation (ROSC). A study is required to evaluate if ECG characteristics are helpful in predicting survival to hospital discharge (SHD) and ROSC for OHCA with PEA as initial rhythm. We hypothesise that higher HRs, narrower QRS complex widths and the presence of P waves are more likely to represent pseudo-PEA. Since the differential diagnosis of pseudo-PEA includes several reversible causes, we therefore hypothesise that these three factors may predict a higher rate of survival. In contrast, we hypothesise that lower HR, wider QRS complex widths and absence of P waves are more likely to represent true PEA and a lower rate of survival.
Therefore, our objective was to study the potential of presenting prehospital HR, QRS width and presence of P waves to predict SHD and ROSC among patients with out-of-hospital PEA arrest.
Methods
Design
We performed a secondary data analysis of patients with OHCA prospectively enrolled in the Ottawa/Ontario Prehospital Advanced Life Support Study (OPALS) paramedic sites of the Resuscitation Outcomes Consortium (ROC) PRIMED study (Prehospital Resuscitation using an IMpedance valve and Early versus Delayed analysis) database between June 2007 and November 2009.10 The Ottawa/OPALS paramedic sites cover a population of nearly 3 000 000 people in an area over nearly 14 000 km2; they include the following cities/regions in the province of Ontario, Canada: London, Niagara, Ottawa, Prescott-Russell, Sudbury, Thunder Bay, Windsor and Waterloo. The PRIMED study method and rationale and its study findings have been published elsewhere.10 14 15
Selection of participants
Patients 18 years of age or older who had non-traumatic OHCA, had PEA as initial presenting rhythm and who had chest compressions delivered by EMS providers were eligible for the study. The initial rhythm was determined by the EMS provider and confirmed by authors MH and MG during ECG analysis.
We excluded patients for the following reasons: their mortality data were unavailable, ECG data were unavailable, their injuries were due to trauma or burns, the arrest was due to exsanguination, they were pregnant, they were prisoners, they had ‘do not attempt resuscitation’ orders, the rhythm analysis was performed by police or a lay responder, they had received initial treatment by an EMS agency that was not in the ROC (Ottawa/OPALS) group, the rhythm was determined to be non-PEA by research staff, or they had a pacemaker-generated rhythm. In addition, patients were excluded if they had ‘opt-out’ bracelets, which were made available through community consultation and public advertisements in anticipation of the PRIMED study.
Data collection
In addition to collecting elements suggested in the Utstein template for resuscitation registries, we also collected digital electronic recordings of prehospital ECGs and chest compressions, otherwise known as cardiopulmonary resuscitation (CPR) process data.16 We collected data from the Ottawa/OPALS paramedic regions involved in the PRIMED study. ECG and CPR process data monitors varied by region and included devices produced by Medtronic (LP-12, LifePak 500 automated external defibrillator (AED)), Philips and Laerdal (MRx ALS/BLS, HeartStart Home, Onsite AED) and Zoll (M Series, AED Pro BLS).15
For ECG analysis, we reviewed three-lead ECGs collected through defibrillation pads. We analysed the first six discernible ECG complexes or the first 15 s of the rhythm tracing, whichever was longer. We chose this interval for measurement to simulate a similar rhythm recognition interval used by a resuscitation provider in the field. Measurement of ECG characteristics was performed by a single investigator (MH).
To measure inter-rater reliability, the ECG characteristics of 60 randomly selected cases (18% of total) were measured by two independent reviewers (MH and MG). Kappa statistics were calculated for each ECG variable. P wave presence was measured as a categorical measure. HR and QRS were also evaluated as categorical measures using HR <60 bpm and QRS >120 ms as decision thresholds. Disagreement between reviewers was resolved by consensus. An example ECG rhythm tracing is shown in figure 1.
Example of ECG tracing and measurement of HR, QRS duration and presence of P waves. RR, respiratory rate.
Predictive variables
Predictive variables were defined as the following: (1) HR in beats per minute (bpm) for each case was recorded as reported in the CPR process data during the rhythm analysis of PEA. HRs were checked against computer-reported HR by manual measurement of the first six QRS complexes. (2) QRS width in milliseconds for each case was recorded as the average of the first, second and third QRS complexes appearing in the rhythm tracing. Premature ventricular complexes were not used for measurement of QRS width. (3) P waves were deemed present if there was at least one discernible P wave within the first 15 s of the ECG.
Outcome measures
The primary outcome measure in our study was SHD. The secondary outcome measure was ROSC present in the ED.
Analysis
All statistical analyses were performed with a commercially available statistical package (SAS V.9.2). We report descriptive statistics with 95% CIs. We determined the association of ECG characteristics (HR, QRS width, P waves) with SHD and ROSC using multivariate regression analysis. We modelled our data using a stepwise analysis of variables, in which each forward selection step (using p value threshold of <0.15) can be followed by one or more backward elimination steps; the selection process terminates if no further effect can be added or the model remains identical to the previous. First, we performed a series of univariate analyses and only selected those variables with two-tailed levels of significance of p≤0.20 for further consideration. Second, we used a stepwise approach to determine which remaining variables should be included in the final logistic regression models. Finally, before reporting on the adjusted influence of each variable included in the final models, we verified that basic model assumptions were met, that residuals were normally distributed, that there were no significant outliers or confounders, and that acceptable homoscedasticity was present.
Sample size
For practical reasons, we enrolled all cases with CPR process data that were available from the Ottawa/OPALS site of the PRIMED study.
Results
A total of 2840 OHCA cases were enrolled between June 2007 and November 2009. Five hundred and three of these cases were identified as PEA and met our inclusion criteria (figure 2). Of these, 171 cases had CPR process data that were not provided, leaving 332 cases with available ECG data for final inclusion and analysis. Characteristics of the 503 PEA cases are shown in table 1. Overall, the demographic and clinical data for the group with available ECG data were similar to those with missing ECG data. Among the 332 included cases, 58% were male with a mean age of 71.7. Seventy-seven per cent of arrests occurred at the patient’s home residence, while 56.3% of all arrests were witnessed. Overall, 5.4% survived to hospital discharge.
Characteristics of 503 eligible PEA cases: 332 included study patients (ECG tracing available) and 171 excluded patients (ECG tracing missing)
Study cohort and exclusions. OHCA, out-of-hospital cardiac arrest; PEA, pulseless electrical activity; PRIMED, Prehospital Resuscitation using an IMpedance valve and Early versus Delayed analysis.
Table 2 demonstrates the epidemiological, clinical and ECG characteristics and univariate analysis results of the 332 included cases. There were 18 survivors (survival rate 5.4%) and 314 non-survivors. Survivors differed from the non-survivors in the following epidemiological and clinical characteristics: fewer home arrests (61.1% vs 77.4%), more public location arrests (33.3% vs 6.4), fewer intubation attempts (61.1% vs 75.2%), less atropine use (16.7% vs 76.4%) and less epinephrine use (33.3% vs 82.8%). The 18 survivors had similar mean HR (56.8 vs 52.0 bpm, p=0.53) and similar mean QRS intervals (128.7 vs 129.6 ms, p=0.95) compared with non-survivors. Kappa agreement was 0.69 for HR and 0.74 for QRS interval. In this univariate analysis, when using an a priori dichotomous cut-off for HR of 60 bpm, patients presenting with an initial HR <60 bpm were significantly less likely to survive to hospital discharge (OR 0.371 (0.143 to 0.966) p=0.04). Presence of P waves could not reliably be ascertained between reviewers (kappa=0.35).
Characteristics of survivors (n=18) and non-survivors (n=314) with univariate regression analysis for return of spontaneous circulation and survival to hospital discharge
The survival rates for important ECG characteristic subgroups are described in table 3. SHD was 3.8% in those patients with initial HR <30 bpm and 3.7% in patients with HR <60 bpm and QRS width <120 ms, while it was 8.8% in those with initial HR >60. Survivors had initial HR as low as 6 bpm and QRS as wide as 357 ms.
Survival rates for ECG characteristic subgroups
Table 4 summarises our multivariate analyses. No single ECG characteristic or group of ECG characteristics were found to predict SHD in the multivariate analysis. We found the following predictor for SHD: location of arrest (adjusted OR (AdjOR)=1.49, 1.11 to 1.99; p=0.007). The use of atropine negatively predicted SHD (AdjOR=0.06, 0.02 to 0.22; p<0.001). For ROSC, we found the following predictors: ALS paramedic on scene (AdjOR=8.90, 95% CI 1.11 to 71.41; p=0.04) and successful intubation (AdjOR=3.35, 1.75 to 6.39; p<0.001). The use of atropine negatively predicted ROSC (AdjOR=0.27, 0.14 to 0.50; p<0.001).
Multivariate analysis for ROSC and survival to hospital discharge (n=332)
Discussion
We found no association between ECG characteristics and survival outcomes in OHCA with PEA as initial rhythm. Univariate analysis significantly correlates an initial HR <60 bpm with non-survival, but the significance of this variable did not carry through in multivariate analysis. We explored various cut-offs for HR, QRS interval and a composite of both these variables, but these were not statistically significant. Despite this, we found a number of survivors who had highly ‘unfavourable’ ECGs, namely profound bradycardia and wide complex bradycardia. Inter-rater agreement for HR and QRS interval duration in our study was very good. Conversely, kappa value for presence of P waves was very poor and was therefore not modelled in multivariate analysis. For SHD, multivariate analysis found a positive correlation for location of arrest and a negative correlation with atropine use. For ROSC, multivariate analysis showed statistically significant positive correlations with ALS paramedic on scene and successful intubation and a negative correlation with atropine use.
Our study findings are similar to the findings of a retrospective study of PEA by Hauck et al published in 2015.17 In their study, they found that neither initial HR nor QRS interval was correlated with survival, although their survivors had lower (statistically non-significant) HRs than non-survivors. Furthermore, our finding that survivors had higher (statistically non-significant) initial HRs and univariate analysis showing negative correlation between HR <60 with survival can be seen as consistent with the report from Aufderheide et al.3 They found that patients with PEA OHCA with higher initial HRs and narrower QRS intervals were more likely to survive to hospital admission.3 In contrast to Stueven et al,1 who demonstrated that patients in PEA arrest with narrow QRS intervals had a higher rate of survival to discharge compared with patients with wide QRS intervals, we did not find such correlation in our study.1
In our univariate analysis, we found survival to be negatively correlated with epinephrine and atropine use, while our multivariate analysis found survival to be negatively correlated with atropine use alone. The correlation of atropine with lower survival rates in our study is mirrored in a 2011 OHCA observational study based in Japan. Compared with epinephrine alone, they observed those who received atropine in addition to epinephrine were less likely to survive to 30 days; they found no difference in ROSC and 30-day favourable neurological outcomes. Our findings are also comparable to a 2008 study of OHCA with PEA by Väyrynen et al.18 Among all epidemiological and clinical characteristics, they found that cardiac cause of arrest and defibrillation had positive correlations with survival, while epinephrine use had a negative correlation. Also consistent with our results, they found no significant relationship between bystander witnessed status and bystander CPR status on survival, two characteristics typically associated with higher survival of cardiac arrest victims.19 20 The observations of harm from atropine and epinephrine in these studies could perhaps represent a ‘reverse causation’ effect. That is, longer resuscitations with lower survival likelihoods are more likely to have received these drug therapies as last attempts by resuscitation providers. This is supported by a 2011 study by Arrich et al, 21 which correlates an increasing cumulative dose of epinephrine during resuscitations as an independent risk factor for unfavourable functional outcome and in-hospital mortality with both in-hospital and OHCA with asystole or PEA.21 Other studies in cardiac arrest have generally shown epinephrine to be associated with ROSC, but not with SHD or favourable functional status over all spectra of cardiac arrest.22–24
This study has several strengths. We collected data from a prospective cohort from multiple EMS sites across the Ottawa/OPALS district in the province of Ontario, Canada, and we collected the elements suggested in the Utstein template for resuscitation registries. We individually analysed each available ECG tracing manually for HR (checked against calculation by CPR process data), QRS interval and presence of P waves. To confirm our inter-rater reliability and generalisability of our results, we calculated Cohen’s kappa statistic for each of these variables.
Our study is limited by the following factors. First, our study had a relatively small sample size with a small number of survivors, which could introduce sampling and statistical biases. Nonetheless, we were able to draw some important conclusions from our study as we found a meaningful number of survivors who either had profound bradycardia or wide complex bradycardia. This suggests that important decisions such as termination of resuscitation should not be made on initial ECG alone. Beyond this, we feel that even with a much larger sample size, it would be unlikely that we would find further clinically meaningful ECG differences that could guide the management of a resuscitation team. Second, in a proportion of cases, the ECG tracings were not provided with the CPR data. The groups with and without ECG data do, however, appear to be very similar, and it is unlikely that a systematic bias was causative. Third, our data were extracted from cases dating 2007–2009. More recent data were unavailable as they were embargoed by ongoing studies. Other than the removal of recommendations to use atropine in 2010, the care of patients with PEA arrest has not changed since the study period; there would likely be no differences if our study had been replicated using newer data.25
Conclusions
Initial ECG characteristics do not reliably predict OHCA survival outcomes in patients with PEA arrest; even with the presence of profound bradycardia and wide complex bradycardia, there were a number of survivors. Initial ECG characteristics should not guide clinical decisions to continue or withhold resuscitative efforts in OHCA PEA arrest victims. The only predictor of SHD was location of arrest, while atropine use was a negative predictor. Predictors of ROSC were ALS paramedic on scene and successful intubation, while atropine use was a negative predictor.
Acknowledgments
MH would like to thank the Canadian Association of Emergency Physicians for their support in funding this research via the CAEP Research Grant (2015). He would also like to thank Angela Marcantonio and the agencies, services and medics participating in the ROC PRIMED study for their dedication to this research.
References
Footnotes
Contributors MH and CV conceived the study and obtained research funding. MH, CV and MG supervised the conduct of the study and data collection. MH and VW conducted data collection. IGS provided statistical advice on study design and analysed the data. MH drafted the manuscript, and all authors contributed substantially to its revision. MH takes responsibility for the paper as a whole.
Competing interests None declared.
Ethics approval Ottawa Health Science Network Research Ethics Board. Research ethics board (REB) approval with waiver of informed consent was obtained from each participating ROC site for the initial PRIMED study as well as for the Epistry data collection. REB was also obtained from our institution for this secondary data analysis.
Provenance and peer review Not commissioned; externally peer reviewed.