Early Prognostic Value of an Algorithm Based on Spectral Variables of Ventricular fibrillAtion From the EKG of Patients With suddEn Cardiac Death: a Multicenter Observational Trial
Early Prognostic Value of an Algorithm Based on Spectral Variables of Ventricular fibrillAtion From the EKG of Patients With suddEn Cardiac Death: a Multicenter Observational Trial
Ventricular fibrillation (VF)-related sudden cardiac death (SCD) is a leading cause of mortality. Patients may survive with neurological damage despite state-of-the-art treatment. Current biological and imaging parameters show significant limitations on early predicting cerebral performance at hospital admission. A spectral-based model was recently suggested to correlate time-dependent VF spectral changes with acute cerebral injury in comatose survivors after cardiac arrest, which opens the possibility to implement early prognostic tools in clinical practice. The AWAKE trial is an investigator-initiated, multicenter, observational trial aiming to validate a spectral-based model to early predict cerebral performance and survival in resuscitated comatose survivors admitted to specialized intensive care units. The primary clinical outcome is favorable neurological performance (FNP) during hospitalization. Patients will be categorized into 4 subsets of NP according to the risk score obtained from the predictive model. The secondary clinical outcomes are survival to hospital discharge, and FNP and survival after 6 months of follow-up. Model-derived categorization will be compared with clinical outcomes to assess model sensitivity, specificity and accuracy. Eligible patients will be included prospectively and retrospectively, using an electronic Case Report Form to enter data from medical records and in-person interviews. Patients will be divided into: study group (predictive data required) including comatose (Glasgow Coma Scale -GCS- ≤8) survivors undergoing temperature control after return of spontaneous circulation (RoSC), and control group including patients who regain consciousness (GCS=15) after RoSC. VF tracings prior to the first DC shock will be digitized and analyzed to derive spectral data and risk scores.
Trial design:
The AWAKE trial is an investigator-initiated, multicenter, observational trial currently underway in 4 centers in Spain (Hospital Clínico San Carlos, Hospital Universitario La Paz, Hospital General Universitario de Ciudad Real, and Hospital Universitario Gregorio Marañón), and open to more participant hospitals. Only tertiary care hospitals with extensive expertise in the management of SCD and post-resuscitation care will participate in the trial. All of them are required to have on-site 24 hours a day, 7 days a week availability for percutaneous coronary intervention, and critical care units staffed by cardiologists and/or critical care specialists. This selective center recruitment is supported by recent data suggesting that direct admission to invasive heart centers is associated with lower mortality in patients with out-of-hospital cardiac arrest (OHCA).
Consecutive patients admitted to the hospital after out or in-hospital cardiac arrest due to VF will be registered and screened for potential inclusion in the study. Patients will be classified into either of two groups:
Ethics:
The study protocol has been approved by the ethics committee of the Hospital Clínico San Carlos, which in is in accordance with the Helsinki Declaration. Every patient will be requested to sign the informed consent before inclusion in the trial. If patient's condition makes it unfeasible to obtain a signed informed consent, it will be requested to their next of kin. Prospective patients will be requested to sign the informed consent during hospital admission, whereas retrospective patients/relatives will sign the informed consent at the time of follow-up contact. The main investigator at each participating center will be responsible for its on-site obtainment and storage according to current Spanish law.
Decisions to withdraw advanced life support will be led by the physician in charge and discussed with representatives according to institutional standards, with no intervention whatsoever from the investigating team.
Eligible patients will need to fulfill all of the inclusion criteria. Exclusion criteria will also apply.
Temperature management in comatose survivors after cardiac arrest due to ventricular fibrillation:
Patients included in the study group must have undergone therapeutic mild hypothermia (target temperature 32 to 34ºC) or controlled normothermia (target temperature 36ºC) for at least 24 hours according to the institution protocol, as long the it is in agreement with current requirements. Routine use of prehospital or in-hospital cooling during cardiopulmonary resuscitation with rapid infusion of large volumes of cold intravenous fluid is discouraged, due to lack of evidence to improve neurological outcome or mortality. Rewarming strategies (e.g. passive or active rewarming) are open depending on institutional standards. Sedation, drug-induced paralysis during mechanical ventilation, the use of cardiovascular drugs, support devices or any other intensive care therapy is left at the criteria of the treating physicians.
Data collection and spectral analysis of VF tracings:
All data, either from the prospective or the retrospective cohort, will be entered in an electronic Case Report Form (eCRF) by individual investigators at each center. Demographic and clinical variables will be obtained from all available medical records and by in-person interview, if possible. Clinical variables and biomarkers as neuron-specific enolase (highest recorded value), electroencephalogram, somatosensory evoked potentials (N20 component with median nerve stimulation) and brain imaging, are encouraging to be collected in the prospective cohort, if available. VF recordings prior to the first DC shock will be scanned using commercial desktop scanners (600 dots per inch) and uploaded to the eCRF. Data consistency and quality will be weekly monitored by one of the investigators (M.M.B).
Stored ECG traces in a codified digital format will be processed as reported elsewhere. Briefly, digitization will be performed using a supervised semi-automatic approach based on region of interest selection, histogram thresholding and intensity transformations. Up to 5-s long segments are extracted after segmentation and signal codification from artifact-free VF tracings. Signals are band-pass filtered between 1.5 and 40 Hz. Two independent investigators will visually inspect the extraction quality (J.G.Q and M.M.B). Averaged power spectral density will be obtained at each frequency using the non-parametric Welch method for using fast Fourier transform and normalized to the peak power in the 1.5-10 Hz band for each patient. Dominant frequency (DF), the relative power between high and low frequency bands (cut-off: 3.9 Hz) (HL-PSDR), and the relative number of spectral peaks above and below the 3.9 Hz threshold with power above 40% the frequency with the highest power (rHL-pKR), along with the number of DC shocks before ROSC, will be the variables used to obtain a model-derived risk score for outcome prediction. Investigators blinded to clinical outcome will perform all data analysis, extraction and quantification using custom-made scripts of MATLAB software (V. 2016b, The Mathworks). The results of such analysis will not be disclosed to the treating physicians and site investigators, not to interfere in standard care, decision-making and outcome assessment.
Outcome assessment:
The primary clinical outcome will be FNP during hospitalization. All patients will be classified using the Pittsburgh Cerebral Performance Categories (CPC) outcome categorization of brain injury, according to current standards of care for SCD assessment. Patients will be considered to have FNP if they score 1 or 2 in the CPC scale (good performance and moderate disability, respectively). CPCs 3, 4 and 5 (severe disability, vegetative state and brain death, respectively) will be considered as a non-FNP. In the prospective cohort, neurological outcome will also be determined using the mini-mental state examination (MMSE). The threshold value in MMSE for normal cognitive function is 24/30, which will be used as complementary neurological evaluation.
In the prospective cohort, neurological outcome will be established before hospital discharge. In the retrospective cohort, the assessment will be done using medical records and in-person interviews with the patient or relatives. During follow-up, neurological outcome will be prospectively assessed by in-person interview in all survivors. Patients within the prospective cohort will be evaluated 6 months after hospitalization. In the retrospective cohort, the follow-up will take place at the moment of patient enrolment or contact with patient´s relatives, in case of decease.
The secondary clinical outcomes will be survival to hospital discharge, and FNP and survival at follow-up. The control group will be used to test the predictive model against a cohort of patients with known FNP, in which the model should also properly predict FNP.
Sample size and study timeline:
The sample size was estimated using the diagnostic performance obtained from the pivotal study. The spectral-based four-variable model using DF, HL-PkR , HL-PSDR, and the number of shocks delivered before ROSC, showed a diagnostic sensitivity of 0.94. Looking for a 95% confidence interval and 5% precision, and assuming that 61% patients will show FNP according to our previous research, a total 143 individuals (both prospective and retrospective cohorts) suitable for analysis must be included.
It is estimated that each of the aforementioned centers receives ~60 SCD per year, but only around one fourth of them will fulfill the inclusion criteria, namely VF as first documented rhythm and ECG tracing availability. Considering the inclusion of 2 more centers (currently undertaking protocol evaluation), the incidence of annual new cases per hospital and retrospective cases from previous years, the sample size will be achieved in a time frame of 30 months. The latter includes the follow-up period for the last group of patients who will be enrolled in the study. As of the writing of this manuscript, the starting center (Hospital Clínico San Carlos) has already evaluated 60 patients, of who 21 met the inclusion criteria and are enrolled in the trial.
Statistical analysis:
Patients from the study group (prospective and retrospective) will be categorized according to the risk score obtained from the multivariate spectral-based model. More specifically, patients will be categorized within four subsets for FNP as follows: very high and high probability of FNP (expected FNP), low and very low probability of FNP (expected non-FNP). Model-derived categorization will be compared with clinical outcomes to assess model sensitivity, specificity and accuracy during hospitalization and follow-up.
In order to assess the clinical relevance of the spectral-based model the investigators will compare its performance and overall accuracy with a model built using the most associated and clinically relevant variables (e.g. highest recorded value of neuron-specific enolase, somatosensory evoked potentials) to the primary outcome. To develop a clinical-based predictive model, each of the clinical variables will undergo univariate analysis to evaluate its association with in-hospital FNP.
Normal distribution of variables will be assessed with the Shapiro-Wilk test. Statistical significance will be assessed by the parametric Student's t-test or the non-parametric Mann-Whitney test, as appropriate. If necessary, the investigators will use Bonferroni correction for multiple comparisons. Categorical variables and percentile comparisons will be performed using a Chi-squared test or the Fisher's exact test, as appropriate. Two-sided p<0.05 will be considered statistically significant. Variables with statistically significant differences and clinically relevant variables will be regressed out against the in-hospital FNP by using a stepwise backwards multivariate logistic regression approach. The clinical-based model will be developed to predict FNP with the highest sensitivity and specificity achievable by using the minimum number of variables that guarantees the best performance and predictive accuracy.
Inclusion Criteria:
Exclusion Criteria: