Outcome prediction in comatose cardiac arrest patients with initial shockable and non‐shockable rhythms

Prognosis after out‐of‐hospital cardiac arrest (OHCA) is presumed poorer in patients with non‐shockable than shockable rhythms, frequently leading to treatment withdrawal. Multimodal outcome prediction is recommended 72 h post‐arrest in still comatose patients, not considering initial rhythms. We investigated accuracy of outcome predictors in all comatose OHCA survivors, with a particular focus on shockable vs. non‐shockable rhythms.


Editorial Comment
In a selected ICU cohort of patients still comatose 72 h after cardiac arrest, the authors found signs of more severe hypoxic/ischemic brain injury in those having presented with nonshockable rhythm than in those with shockable first rhythm.This was reflected by lower survival among patients with non-shockable first rhythm whereas prognostic neurological examinations performed equally well in both groups.

| INTRODUCTION
Outcome prediction in comatose patients after cardiac arrest (CA) has become a cornerstone of post-resuscitation care.It is currently based on the 2021 European Resuscitation Council (ERC) guidelines, 1 which recommend a multimodal approach to neuroprognostication including clinical assessment, biomarkers, neurophysiology, and imaging initiated in patients still comatose 72 h after CA. 1 This multimodal approach is essential to provide an accurate prognosis based on the severity of hypoxic-ischaemic brain injury.As withdrawal of life-sustaining therapy (WLST) due to presumed poor neurological outcome is the leading cause of in-hospital death after outof-hospital CA, 2,3 the decision to withdraw intensive care treatment must be soundly based.
5][6] Non-shockable rhythms are usually related to noncardiac causes of CA or a consequence of prolonged periods without blood flow, leading to more severe brain injury.As a non-shockable rhythm is considered a predictive prehospital factor for poor outcome, 7 and associated with early WLST, 8 self-fulfilling prophecies are likely to be expected. 8However, the prevalence of good neurological outcome in patients with nonshockable rhythms can vary considerably, depending on factors like age and conversion to shockable rhythm during resuscitation, among others. 9The current guidelines do not distinguish between shockable and non-shockable rhythms in terms of prognostication, 1 and it remains uncertain how the recommended diagnostic predictors apply to OHCA patients with these different rhythms.
We, therefore, aimed to investigate the prognostic accuracy of the recommended predictors in patients still comatose 72 h after OHCA, stratified by initial shockable vs. non-shockable rhythms.In addition, we assessed the capacity of these predictors in the entire cohort.We hypothesized that outcome prediction would be comparable between patients with shockable and non-shockable rhythms, and that the prognostic predictors would reveal signs of severe brain injury relatively more often in patients with non-shockable rhythms.

| Study population
This was a sub-study of the prospective observational NORCAST trial, which included 259 adult OHCA patients still comatose after return of spontaneous circulation (ROSC), who were admitted to Oslo University Hospital, Ullevål between September 2010 and January 2014.Patients with OHCA following trauma or intracerebral hemorrhage were excluded.All patients received protocol-based intensive care treatment with targeted temperature management at 33 C for 24 h (TTM33) and were deeply sedated with mainly fentanyl and midazolam, regardless of the initial rhythm. 10r the present sub-study, patients were classified as shockable or non-shockable according to the first rhythm recorded at the scene.
Ventricular tachycardia or fibrillation were defined as shockable, pulseless electrical activity (PEA), and asystole as non-shockable rhythms.
For outcome prediction, only patients still comatose (Glasgow Coma Scale [GCS] <9) 72 h after OHCA were included (Figure 1).Patients with an unknown initial rhythm (including ROSC prior to arrival of the emergency medical services) or who died within 72 h after OHCA were excluded (Figure 1).
To assess outcome, we used the cerebral performance category (CPC), 11,12 with a score of 1-2 (none to moderate cerebral disability) as good outcome and 3-5 (severe cerebral disability to death) as poor outcome 6 months after OHCA. 10,13The highest CPC score within the first 6 months was used as "best CPC score" for patients who regained consciousness but later died from other causes prior to the six-month endpoint.Time to awakening was defined as time from OHCA to GCS ≥9, with late awakening defined as awakening later than 6 days after OHCA based on data from the NORCAST trial where median time to awakening in patients with good outcome was 6 days. 10WLST was registered when the WLST decision was documented in the medical records.All decisions regarding diagnostics, treatment and WLST were made exclusively by the attending physician.

| Prognostic predictors
In the NORCAST study, pupillary light reflexes (PLR), corneal reflexes, and GCS were assessed daily.Standard electroencephalography (EEG) and somatosensory evoked potentials (SSEP) were performed in all eligible patients still comatose ≥72 h after sedation withdrawal (Table S1).EEG and SSEP recordings were interpreted independently by two experienced neurophysiologists who were unaware of the patients' medical history.In case of disagreement, a third specialist was consulted.Neuron-specific enolase (NSE) was drawn at admission, 24, 48, 72 h, and 5 and 7 days after OHCA.Results for EEG, SSEP, and NSE were blinded to the treating physicians but could be revealed on request.
For the present sub-study, we collected the following data from the NORCAST database, adjusted according to the 2021 ERC guidelines 1 ; PLR and corneal reflexes 72 and 96 h after OHCA, and 72 h after sedation withdrawal, NSE 24, 48, and 72 h after OHCA, and EEG and SSEP at least 72 h after sedation withdrawal.EEG grading was used as defined in NORCAST. 10,14Noteworthy, our EEG definition differed slightly from the current guidelines recommendation. 14,15G grade 1-predominantly post-central alpha-activity mixed with theta-activity.
EEG grade 3-dominating/substantial delta-activity, or low-amplitude irregular and non-reactive delta-activity.
EEG grade 4-burst suppression, general epileptic activity, status myoclonus, non-reactive activity with low amplitude, alpha-coma and theta-coma.
EEG grade 5-no visible EEG activity during high-sensitivity registration.
Computed tomography (CT) of the brain was not part of the study protocol but could be requested at the discretion of the attending physician.It was therefore performed at different time points.Findings were described as oedema and/or loss of gray-white matter differentiation, which were used to define brain injury.A gray-white-matter ratio was not calculated.
The following predictors were used to assess poor outcome: NSE values >60 μg/L at 24, 48, or 72 h, absent PLR and corneal reflexes 72 and 96 h after OHCA, and 72 h after sedation withdrawal, bilaterally absent SSEP N20 response and EEG grade 4-5 at least 72 h after rewarming, and signs of brain injury on the CT.We used the same predictors to assess good outcome: Present PLR, corneal reflexes, and SSEP N20 response, an EEG grade 1-2, and "no evidence of brain injury" for brain CT.For NSE, values within normal range (<17 μmol/L) were considered to indicate good outcome, in accordance with previous studies. 1,16,17

| Ethics
The Regional Committees for Medical Research Ethics South East Norway approved the NORCAST study (REK-S-O/A-2010/1116a).
Informed consent was obtained from relatives shortly after hospital admission, and from all patients who regained decision-making capacity within 6 months.

| Statistics
Patient characteristics and prognostic predictors are described using frequencies and percentages for categorical variables and mean and standard deviations for continuous variables.Differences between shockable and non-shockable groups were assessed by independent samples t-test for continuous and χ 2 -test, or Fisher's exact test in case of violated assumptions for χ 2 -test, for categorical variables.Two-sided tests were used, and results with p-values <.05 were considered statistically significant.
To predict prognostic accuracy, false positive rate (FPR, 1-specificity) for poor outcome and sensitivity for good outcome were calculated with "best CPC score" as outcome parameter.The corresponding 95% confidence interval (CI) was calculated using Wilson's method. 18Differences in outcome prediction between shockable and non-shockable patients were considered non-significant if there was no overlap in their respective CIs.
For all binary outcome predictors (PLR, corneal reflexes, CT, and SSEP), the accuracy of the FPR in predicting poor outcome is reciprocal to its sensitivity in predicting good outcome.
Compared to the shockable group, non-shockable patients were characterized by more females, fewer witnessed arrests, and more non-cardiac causes of arrest (Table 1).Six-month survival with good outcome was higher in patients with shockable than non-shockable rhythms (64% vs. 24%, p < .001;Table 1), which is comparable to the results in all NORCAST patients (64% vs. 22%, respectively, not tabulated).There was no significant difference in good outcome between patients with initial PEA (29%) and asystole (19%) ( p = .411,not tabulated).Time from arrest to sedation withdrawal, and to awakening were similar in both groups ( rhythms (Table 2).Significantly more late awakeners in the shockable vs. the non-shockable group had a good outcome (83% vs. 45%, p = .017)(Table 2).WLST was initiated twice as often in non-shockable than in shockable patients (50% vs. 24%, p < .001).However, there was no significant difference in median time from OHCA to WLST between the groups (13 vs. 10 days) (Table 2).
The outcome predictors indicated a higher proportion of patients with severe brain injuries in non-shockable than shockable patients (Table 3, Figure 2).NSE, for example, showed a rise and fall pattern in the shockable group, whereas it was continuously increasing from 24 to 72 h with significantly higher mean values at 48 and 72 h in the non-shockable group (Figure 2).

| Good outcome prediction
There were no substantial differences in sensitivity for predicting good outcome between the shockable and non-shockable groups ( shockable and non-shockable rhythms, while FPR seemed higher in shockable rhythms (0.73 vs. 0.5).Sensitivity for EEG seemed lower in patients with shockable than non-shockable rhythms (54% vs. 83%).
The sensitivity of brain CT was high (94%) for shockable rhythms only.
As for poor outcome prediction, no significant differences were found between the groups, as all predictors had overlapping CIs.

| Outcome prediction for the entire cohort
Poor and good outcome prediction for the entire cohort is shown in low FPR (0-0.07)showed an upper CI limit >10% (Table 5a).
Good outcome prediction was similar to that of the stratified groups with sensitivities from 88% for corneal reflexes 96 h after OHCA, to 95%-100% for brain CT, SSEP, and PLR and corneal reflexes 72 h after sedation withdrawal.Although sensitivity of NSE increased from 21% 24 h after OHCA to 64% 72 h after OHCA, they were still low (Table 5b).

| DISCUSSION
In this NORCAST sub-study, we found that 66% of patients still comatose 72 h after OHCA with initial shockable rhythms and 24% with non-shockable rhythms, survived with good outcome to 6 months.
Prediction of poor and good outcome was comparable in both groups.
T A B L E 3 Prognostic predictors for brain injury in still comatose patients ≥72 h after OHCA, stratified by initial shockable (n = 128) and nonshockable (n = 50) rhythms.PLR and corneal reflexes had better precision 72 h after sedation withdrawal and 96 h after OHCA than 72 h after OHCA.However, the generalisability of this assumption is undoubtedly reduced by the low number of non-shockable survivors.As expected, the prognostic predictors indicated a higher proportion of patients with severe brain injuries among those with non-shockable than shockable rhythms.
1][22] The wide range (<0.1%-29%) of good outcome in OHCA patients with nonshockable rhythms reflects the heterogeneity of underlying causes for CA in this group. 9As non-shockable rhythms are considered predictive for poor outcome, it is more likely to lead to early WLST. 8,23Decisions on WLST taken as late as median 13 days in the non-shockable group might have contributed to the relatively high survival with good outcome in the present study.
WLST will always interfere with outcome prediction in CA patients. 3,6,24Interestingly, 20% of patients with good outcome woke up late in a recent study where WLST was not performed. 6Therefore, early WLST must be based on a reliable prediction of poor prognosis.In the present study, poor outcome prediction showed low FPRs in almost all predictors for the entire cohort.Nevertheless, none of them reached an FPR ≤0.001 which was considered acceptable for WLST according to a survey among health care workers. 25However, very low FPRs are at the expense of sensitivity, reducing the predictor's clinical usability.
As an example, for NSE the current guidelines therefore suggest an FPR of 0.01-0.02as acceptable in poor outcome prediction. 1 Nonshockable rhythms had higher FPRs at all time points and only NSE 72 h after OHCA in the shockable group had an acceptable FPR of 0.01 for poor outcome prediction, with a relatively low sensitivity of 46%.This is consistent with findings in the TTM sub-study by Stammet et al. 26 They also described a "rise-and-fall" pattern for a good outcome (comparable to our findings for shockable rhythms) and increasing values for poor outcome (comparable to our findings for non-shockable rhythms). 26Importantly, among patients with non-shockable rhythms with favorable outcome at 6 months, 17% had NSE values >60 μg/L at 48 h.In contrast, a larger retrospective multicentre study described equal outcome prediction for NSE in patients with shockable and nonshockable rhythms. 27Although the threshold of 60 μg/L at 48 h indicating poor outcome seems to be too low for patients with nonshockable rhythms in the present cohort, this may be solely due to the small sample size and should be interpreted with caution.
In contrast to previous studies, 28 the FPRs for clinical predictors remained high up to 48-72 h after OHCA.[31][32] Consequently, with contributing factors for late awakening, clinical parameters are more reliable in the later course, such as 72 h after sedation withdrawal, where we found FPR values of zero.Again, wide CIs due to few study patients clearly limit the general validity of the results.
Although brain CT in the present study was predominantly performed when brain injury was suspected, the FPR was slightly higher in shockable patients than in previous studies (0.06 vs. 0-0.02). 1 Noteworthy, Beekman et al. showed that early CT findings indicating hypoxic brain injury often led to de-escalation of intensive care treatment. 33Thus, a "biomarker-guided" indication for brain CT could increase its specificity and contribute to a more robust prognostication. 34esent PLR and corneal reflexes 72 h after sedation withdrawal and 96 h after OHCA, all were reliable predictors of good outcome prediction.Again, higher accuracy in the later course may be due to sedation effects and TTM33 the first days after OHCA.Sensitivity for EEG and NSE was lower in the present than in previous studies. 16,17,35,36While different EEG classifications 14,15 or sedation may explain differences for EEG, different NSE levels may be explained by different storage procedures or analysis methods. 17,37,38EP is generally considered a reliable predictor for both poor and good outcome with high specificity but lower sensitivity. 39,40ile poor outcome prediction was good (with a broad CI, however), we found a lower specificity with a surprisingly high sensitivity in good outcome prediction, compared to previous studies. 40,41This may be due to a cautious interpretation of bilaterally present N20 responses, with 24% of registrations considered inconclusive in the non-shockable group (Table S1) and the relatively small number of examinations.Recent proposals using SSEP signals as a continuous measure and peak-to-peak analysis instead of the binary absent/ present may further improve accuracy. 40e analyzes of the entire cohort indicated that none of the predictors were reliable enough to be used alone to predict good outcome.A multimodal approach, as recommended for poor outcome prediction, 1 would clearly increase prediction accuracy, as Vanat et al.
recently showed by presenting a new multimodal good outcome prediction score. 42This may prevent early WLST in late awakeners with potentially good outcome. 6,34,43e present study has important limitations.Due to the relatively low number of patients with good outcome in the non-shockable group, the validity of the results is limited.Notably, a broad 95% CI with an upper limit >5% 19 reduces the overall significance considerably, even with an FPR of zero.Thus, these results are not conclusive and should be interpreted with caution since the true FPR value for the population is more uncertain.Nevertheless, there is currently no generally accepted standard for an optimal FPR value.The EEG classification used 10,14 is no longer recommended.Some EEG patterns would be classified differently using the current terminology. 15EEG classifications according to current guidelines recommendations could have improved accuracy.Of note, sedation may also impact EEG patterns. 44Brain CT was not a part of the study protocol but requested on clinical indication.This could have resulted in selection bias if performed more often when brain injury was suspected.For bilaterally absent SSEP N20 responses, blinding was lifted in four cases but likely contributed to WLST in only one case.We cannot rule out that this might influence the results due to the low number of SSEPs performed.

| CONCLUSIONS
In OHCA patients still comatose 72 h after OHCA, prediction of poor and good outcome was comparable between shockable and WIMMER ET AL.
a Independent samples t-test.b χ 2 -test; OHCA; out-of-hospital cardiac arrest, CPR; cardio-pulmonary resuscitation, ROSC; return of spontaneous circulation, CPC; Cerebral performance category, CPC 1-2 good outcome, CPC 3-5 poor outcome, #Best CPC within 6 months: In total, 14 patients who obtained CPC score 1-3 within the first 6 months after OHCA died prior to six-month follow-up and are tabulated as CPC 5 under "CPC score at six months," *19 missing values, 9 in shockable and 10 in non-shockable rhythm group, **23 missing values, 11 in shockable and 12 in non-shockable rhythm group.
Clinical outcome data for patients still comatose ≥72 h after OHCA stratified by initial shockable (n = 128) and non-shockable (n = 50) rhythms.Cerebral performance category with 1-2 good outcome and 3-5 poor outcome (best CPC score within 6 months used), Early awakening; GCS ≥9 prior to day six after OHCA, Late awakening; GCS 9 later than day six after OHCA, WLST; withdrawal of life-sustaining therapy.
19 the shockable group (Table4a and 4b).19Inaddition, EEG grade 4-5 and NSE values >60 μg/L at 24-72 h showed lower FPRs for shockable than non-shockable patients, with reasonable narrow CIs.Sensitivity was generally low in both groups (7%-64%).The overlapping CIs for all predictors between the groups indicated comparable outcome prediction for patients with shockable and non-shockable rhythms.T A B L E 2 a Independent samples t-test.b χ 2 -test.c Fisher's exact test.d Independent samples median test, OHCA; out-of-hospital cardiac arrest, GCS (Glasgow Coma Scale), CPC; Table 4c and 4d).Except for corneal reflexes 72 h after OHCA, all clinical parameters predicted good outcome with high sensitivity (86%-100%) with increasing values over time, but with a wide range in FPRs (0.41-0.93).For NSE, sensitivity was varying from 22% 24 h after OHCA to 69% 72 h after OHCA, but with a low FPR of 0.02 72 h after OHCA.Corresponding to the low FPR for poor outcome, SSEP had a high sensitivity for good outcome prediction in both
a χ 2 -test.bFisher'sexacttest, OHCA; out-of-hospital cardiac arrest, PLR; pupillary light reflex, SSEP; somatosensory evoked potential, EEG; electroencephalography (both registered 72 h after discontinuation of targeted temperature management at 33 C), EEG grade 4; burst suppression, general epileptic activity, status myoclonus, non-reactive activity with low amplitude, alpha-coma, and theta-coma, EEG grade 5; no visible EEG activity during high-sensitivity registration; NSE; neuron-specific enolase, Brain injury on CT; hypoxic ischemic brain injury on cranial computed tomography (oedema and/or loss of gray-white matter differentiation).F I G U R E 2 NSE values 24, 48, and 72 h after OHCA in patients with initial shockable vs. non-shockable rhythms.NSE; neuron specific enolase, OHCA; out-of-hospital cardiac arrest.T A B L E 4 a T A B L E 4 b Prediction of poor outcome (n = 38) in patients with initial non-shockable rhythms (n = 50), still comatose ≥72 h after OHCA.Abbreviations: Brain injury on CT, hypoxic ischemic brain injury on computed tomography (oedema and/or loss of gray-white matter differentiation); CI, confidence interval; EEG, electroencephalography; FN, false negative; FP, false positive; FPR, false positive rate; NSE, neuron-specific enolase; OHCA; out-of-hospital cardiac arrest; PLR, pupillary light reflex; SENS, sensitivity; SSEP, somatosensory evoked potential; TP, true positive; TN, true negative.T A B L E 4 c Prediction of good outcome (n = 85) in patients with initial shockable rhythm (n = 128), still comatose ≥72 h after OHCA.