Population
All consecutive adult patients who were admitted between November 2019 and July 2021 to the intensive care unit (ICU) of Cochin University Hospital (Paris, France) in a comatose state (defined as a Glasgow coma scale [GCS] ≤ 8 with a GCS motor < 3 and a Richmond Agitation–Sedation Scale RASS ≤ -4) after resuscitation from CA, regardless of initial rhythm, with SSEP performed, were prospectively considered for inclusion. We excluded patients investigated for brain death diagnosis, patients awake before SSEP, and patients who died within 48 h post CA, before a reliable neurological examination could be performed. Patients’ next of kin were informed that data were collected for clinical research purposes.
Data collection
The following data were recorded: patients' characteristics, pre-hospital care and cardiac arrest management data using Utstein style, in-hospital variables including serum lactate at admission, TTM use, type of sedation, clinical indicators of neurological status (clinical status myoclonus defined according to ERC/ESICM guidelines as generalized, continuous and persisting for 30 min or more of the myoclonic jerks, requiring an anti-epileptic drugs regardless EEG results), EEG patterns, SSEP recording and NSE levels [13]. In the present analysis, we used the first EEG and SSEP performed during the ICU stay; NSE levels were determined at days 1, 2 and 3 after CA. Length of stay and cause of death were also reported. Data collection was approved by the Ethics Committee of the French Intensive Care Society (#CESRLF_12-384 and 20–41) and conducted according to French health authorities’ regulations (French Data Protection Authority #MR004_2209691).
ICU management
The management protocol for patients admitted to our ICU after CA reported in Additional file 1: ESM1 has been previously described and did not change throughout the study period [2, 3, 14]. In the absence of contraindication, TTM was immediately started after ICU admission with a target temperature of 32–36 °C adapted to hemodynamic tolerance and using an external cooling device for 24 h. Sedation protocol, according to guidelines, used short acting drugs including propofol and remifentanil. Sedation protocol was based on the RASS, titrated to obtain a RASS of − 5 (no response to voice or physical stimulation) and was interrupted after rewarming (Additional file 1: ESM1).
Neurological prognostication and WLST
Neurological status was evaluated every 3 h by nurses, and daily by ICU physicians until death or ICU discharge. Awakening was defined as three consecutive RASS scores of at least − 2 (patient briefly awakens with eye contact to voice), as previously reported [15]. In patients who were still comatose 72 h after ROSC and 48 h after sedation discontinuation, a multimodal prognostication protocol was used, unchanged during study period, consistent with the 2015 international ERC/ESICM guidelines [13]. WLST was considered in comatose patients with a GCS motor score 1 or 2 when two or more of the following conditions were present: (1) bilaterally absent pupillary and corneal reflex; (2) bilaterally absent N20 waves on SSEP; or (3) refractory electrical status epilepticus, burst suppression or suppression. Electrical status epilepticus was defined as refractory when it did not improve after treatment with 2 lines of major antiepileptic drugs (among phenytoin, phosphenytoin, valproate, phenobarbital). Amplitudes of SSEP was not used in our prognostication algorithm and did not influence WLST (Additional file 1: ESM2).
SSEP recording
SSEP recordings were made using Deltamed Coherence (Natus, Middleton, USA). SSEP were recorded in patients still comatose 72 h after ROSC and 48 h after sedation discontinuation. The SSEP was measured after stimulation of the right and left median nerve using a bipolar surface electrode at the wrist. Stimulation intensity was adjusted to produce visible thumb twitches; if neuromuscular blocking agents were administered, Erb amplitudes were used instead. Monophasic rectangular-wave 200 ms stimulus pulses were delivered. Stimulus frequency was set at 2–3 Hz. Poststimulus recording lasted 50 ms (bandwidth: 30 Hz–3 kHz, sampling frequency: 50 kHz). Two or three sets of 300–1000 responses were averaged. Surface electrodes were positioned at Erb’s points. Needle electrodes were used for scalp derivations: 2 cm posterior to C3 and C4 (C3’ and C4’). N9 (peripheral), N13 and N20 (cortical) responses were recorded. N9 peripheral responses corresponded to Erb's point ipsilateral to the stimulation vs. reference electrode at contralateral Erb's point. For cortical responses, a bi-parietal montage (active electrode contralateral to the stimulation vs. reference electrode ipsilateral to the stimulation) was used.
SSEP interpretation
N20 latencies were deemed interpretable if at least 2 peripheral (N9), 2 spinal peak (N13) and cortical recordings (N20–P25) per side were bilaterally reproducible and if a noise level below 0.25 μV in all 4 cortical recordings had been achieved. Noise level was determined 5–10 ms after stimulation to exclude stimulation artifacts. The noise level was determined automatically and visually checked. Digitalized SSEPs were reevaluated blinded to patients’ outcome by an expert electrophysiologist. Figure 1 shows representative examples of SSEP recordings. The N20 wave was identified as the major negative peak (C’3–C’4 montage), while P25 was identified as the major positive peak following N20. Three profiles of SSEP responses were determined according to the presence or absence of N20 response. In patients with no reproducible cortical potential and noise level below 0.25 μV, the recording was classified as «absent-absent/AA». Patients with reproducible cortical potential on both sides were classified as «present present/PP». If there was only a unilateral response, we classified patients as «absent-present/AP». As “AA” pattern was already recognized as a robust marker for poor neurological outcome, we only assessed SSEP amplitudes in case of bilateral responses defined as «PP» patterns. We defined N20-baseline amplitude as the highest difference between N20 peak and the baseline and the N20–P25 amplitude using the peak-to-peak N20–P25 amplitude. Amplitude computation was automated by the analysis software use. In case of asymmetry between right and left N20, we retained the best amplitude value. For prediction of poor outcome, we considered the patterns «AA» and «PP», while for prediction of good outcome, we only considered the «PP» pattern. Concerning the Deltamed device normal SSEP amplitudes, the median N20–baseline and N20–P25 amplitudes were 1.89(1.2–2.55)µV and 2.05(1.29–2.55)µV, respectively.
Electroencephalogram (EEG)
EEG recordings were acquired maximum 1 h before SSEPs’ recording over 20 to 30 min with a Natus Deltamed recording system (Natus, Middleton, USA) using 19 electrodes placed according to the 10–20 international system, with additional ground and reference electrodes. During the recordings, auditory (calling of patient’s first name and family name, hand clapping), somatosensory and pain stimuli were applied at least twice, with a minimum of 10 s intervals. EEG traces were retrospectively analyzed de novo by one board-certified electroencephalographers (AM) blinded to the clinical outcome and the SSEPs results. EEGs were interpreted according to the standardized criteria of the American Clinical Neurophysiology Society [16]. Each EEG was classified into one of the mutually exclusive categories defined by Westhall et al. [17], namely, highly malignant pattern (suppressed background or burst-suppression, with or without superimposed periodic pattern), malignant pattern (presence of at least one of the following: abundant periodic discharges or rhythmic spike-waves, electroencephalographic seizure, discontinuous or low-amplitude background, reversed anterior–posterior amplitude gradient, absence of reactivity) or benign pattern (continuous and reactive pattern, absence of malignant feature) [16, 17].
Outcome assessment
The primary outcome was the neurological status at 3 months using the “best” cerebral performance categories (CPC) scale (Additional file 1: ESM3). CPC is defined as CPC1 = good cerebral performance no or minimal disability; 2 = moderate cerebral disability, i.e., disabled but independent; 3 = severe cerebral disability, i.e., conscious but disabled and dependent; 4 = comatose or vegetative state; 5 = death [2, 18]. Good neurological outcome was defined as CPC 1–2, using structured phone interviews. We used the best CPC to avoid considering patients classified as CPC1–2 and who subsequently died (CPC5) from non-neurological causes as poor neurological outcome (CPC3–4–5).
The secondary outcome was the best CPC at 6 months. Finally, correlation of SSEP amplitude with NSE peak, EEG patterns and clinical status myoclonus was assessed.
Statistical analysis
Continuous variables were summarized using medians and interquartile range (IQR), and categorical variables were reported as proportions. We performed Pearson’s Chi2 test for categorical variables, and Wilcoxon, when appropriate, for continuous variables. We assessed the association between candidate variables including patient demographics, Utstein variables, status myoclonus, SSEP recording, NSE peak at D2 and D3 and EEG patterns; regarding neurologic outcome for which we used a binary outcome (CPC 1–2 vs. 3–4–5).
N20-baseline and N20–P25 SSEP amplitudes thresholds were sought in patients with “PP” cortical responses, setting for the best compromise between the higher specificity (and so the lower FPR), associated with an acceptable sensitivity [19]. All statistical analyses were performed using IBM SPSS version 26.0 (IBM Corporation, Armonk, New York). A two-sided p value < 0.05 was considered to be statistically significant. Figures were drawn using Prism 8.4.3 (GraphPad Software, California, USA) and R software 3.6 (R project, worldwide community software).