Prognostic Value of Signal Abnormalities on Brain MRI in Post‐Anoxic Super‐Refractory Status Epilepticus: A Single‐Center Retrospective Study

ABSTRACT

Background

Epileptiform activity, including status epilepticus (SE), occurs in up to one‐third of comatose survivors of cardiac arrest and may predict poor outcome. The relationship between SE and hypoxic–ischemic brain injury (HIBI) is not established.

Methods

This is a single‐center retrospective study on consecutive patients with post‐anoxic super‐refractory SE. HIBI was graded as non‐widespread (group 1) or widespread (group 2) by qualitative analysis of DWI/ADC and T2w‐FLAIR. Between‐group differences in the rate of poor neurological outcome at 6 months (primary outcome), SE resolution and consciousness recovery before discharge, and mortality at 6 months (secondary outcomes) were investigated.

Results

From January 2011 to February 2023, 40 patients were included. HIBI was widespread in 45% of patients and non‐widespread in 55%. The rate of poor neurological outcome at 6 months was 27% in group 1 and 83% in group 2 (OR 12.8, CI 95% [2.5–64.3], p = 0.002). The rate of consciousness recovery before discharge was 73% in group 1 versus 22% in group 2 (OR 8.8, CI 95% [1.9–40.3], p = 0.005). SE resolved in 95% of patients in group 1 versus 67% in group 2 (OR 10.5, CI 95% [1.1–97.9], p = 0.039). Mortality rate at 6 months was 27% in group 1 versus 50% in group 2 (OR 0.4, CI 95% [0.1–1.9], p = 0.303).

Conclusion

Patients with widespread HIBI had higher odds of poor outcome at 6 months, lower probability of SE resolution and of consciousness recovery before discharge compared to those with non‐widespread HIBI. Mortality at 6 months did not differ significantly between the two groups.

1. Introduction

Epileptiform EEG activity—including definite status epilepticus (SE), generalized periodic discharges (GPDs), and other rhythmic patterns—occurs in up to 30% of survivors of cardiac arrest (CA) with post‐anoxic encephalopathy [1, 2, 3].

EEG characteristics, such as suppressed background with or without periodic discharges and burst‐suppression, are mostly associated with poor neurological outcome and were included in the European Resuscitation Council's multimodal neuroprognostication algorithm [4]. The TELSTAR trial [5] demonstrated the futility of treating rhythmic and periodic EEG patterns; however, treatment of definite SE remains debated, given the uncertain impact on prognosis, raising ethical issues about continuing versus withdrawing care [6, 7]. SE likely results from a disturbed excitation–inhibition ratio due to the high vulnerability of excitatory synapses to hypoxia [8]. Although SE may be associated with poor outcome [6], good prognosis is reported in some cases [9, 10, 11, 12].

Certain brain regions are more susceptible to hypoxic–ischemic brain injury (HIBI), which can be detected early on MRI [13]. However, their prognostic significance remains unclear; additionally, a consensus on quantifying HIBI and defining its severity has not been established. Prognostic assessment via MRI becomes more complex in cases of SE, due to its potential impact on signal irregularities. Currently, it is uncertain whether the likelihood of developing SE and its refractoriness are directly linked to the degree of HIBI.

This study describes the distribution of HIBI on brain MRI in consecutive survivors of CA that developed super‐refractory status epilepticus (SRSE). The association between distinct radiological patterns and neurological outcome at 6 months, evolution of the SE over time, consciousness recovery before discharge from the Intensive Care Unit (ICU), and mortality at 6 months was also investigated.

2. Methods

2.1. Study Protocol Approvals and Patient Consents

This retrospective, single‐center study was approved by the Ethics Committee Lombardia 3 on September 22, 2022 and conducted at the Cardiac ICU in collaboration with the Epilepsy Center and Neuroradiology Unit of the Fondazione IRCCS San Gerardo dei Tintori, Monza, Italy. The IRCCS Mario Negri Institute for Pharmacological Research, Milano, provided statistical support. Demographics and clinical data were collected by reviewing medical records. Neurological outcome at 6 months was assessed by telephone encounter. Informed consent was waived for patients unable to consent or deceased.

2.2. Study Population

Consecutive comatose survivors of CA that developed SRSE, admitted between January 2011 and February 2023, were screened. Inclusion criteria: adult comatose survivors of intra‐ or extra‐hospital CA, patients presenting with SRSE >24 h after CA, brain MRI performed 2–30 days after CA, collection of ≥ 1 prognostic indicator among neuron‐specific enolase (NSE) within 72 h, cortical N20 of the median nerve somatosensory evoked potential (SSEP) within 72 h, pupillary and/or corneal reflexes anytime during hospitalization. Exclusion criteria: patients presenting with suppressed background with or without GPDs on EEG performed > 24 h after CA and not preceded by SE, patients presenting with spontaneous burst suppression on EEG performed > 24 h after CA and not preceded by SE, death before performing brain MRI, concomitant or pre‐existing intracranial pathologies that could prevent adequate evaluation of HIBI (e.g., ischemic or hemorrhagic stroke, CNS infections, etc.), pre‐existing epileptic encephalopathy, unavailability of follow‐up data.

2.3. Standardized Management of Post‐Anoxic Encephalopathy

Post‐resuscitation care followed current guidelines [4] and included targeted temperature management between 33°C and 36°C for 24 h, followed by rewarming. Simplified 4‐channel continuous EEG monitoring (cEEG) began within 24 h of CA. Prognostic indicators were collected: pupillary and corneal reflexes were tested throughout hospitalization and considered indicators of poor prognosis when bilaterally absent. NSE levels were measured at 48 and 72 h and deemed poor prognosis indicators if > 60 μg/L. Median nerve SSEPs were performed 72 h after CA, with bilaterally absent cortical N20 indicating poor prognosis. All patients underwent brain CT and/or MRI, with MRI performed in patients that were not too unstable from a critical care standpoint to tolerate it. Standard 18‐channel EEG was obtained 24–36 h after ROSC and repeated as needed. Standard EEGs were recorded off sedation and neuromuscular blockade for SE diagnosis. cEEG monitoring was performed while patients were under anesthetics/neuromuscular blockade, using a simplified 4‐channel EEG monitor integrated in the ICU monitor. cEEG was used to manage the dose of anesthetics and/or alert the staff to abnormal patterns, prompting standard EEG. cEEG alone was never used to diagnose SE. Electroencephalographic investigations utilized the 48‐channel EEG system BRAIN QUICK and SystemPLUS Evolution 1.4 (Micromed SpA, Treviso, Italy). As previously described [12], patients developing SRSE received antiepileptic drugs (AEDs) and anesthetics (for 24–48 h, per cycle) to completely suppress epileptic activity on cEEG. Benzodiazepines, levetiracetam, valproate, perampanel, topiramate, lacosamide, and phenytoin were used alone or in combination. Propofol, ketamine, thiopental, and isoflurane were used as single agents. All except phenytoin were used off‐label for SE treatment adhering to doses specified in the summary of product characteristics. Medications were administered intravenously except for perampanel and topiramate, which were administered via nasogastric tube. Withdrawal of life‐sustaining therapy (WLST) was applied if multimodal neuroprognostication predicted poor outcome likely [4] and after considering clinical aspects (e.g., age, comorbidities, general organ function) and the patient's health care advance directory when available.

2.4. Classification of EEG Patterns

A neurophysiologist blinded to the patient's clinical course retrospectively reviewed EEG recordings; an experienced epilepsy specialist performed a second review. Discrepancies were solved by discussing doubtful cases. Diagnosis of SE was based on the Salzburg Criteria for non‐convulsive status epilepticus (NCSE) [14]. In accordance with the ILAE Classification [15] and the consensus definitions for new‐onset refractory status epilepticus [16], SE was considered as refractory (RSE) after the failure of benzodiazepines and a first AED and as super‐refractory (SRSE) after failure of a 24‐h cycle of sedative agents. Myoclonic SE was defined as continuous and generalized myoclonic jerks lasting > 30 min. GPDs were diagnosed according to the ACNS terminology [17] in the presence of bilaterally synchronous and symmetric patterns of periodic waveforms lasting < 0.5 s, regardless of number of phases, or waveforms ≥ 0.5 s with no more than three phases standing out of the background activity, with a clearly discernible inter‐discharge interval between consecutive waveforms and recurrence of the waveform at nearly regular intervals, with a frequency < 2.5 Hz.

2.5. Assessment of Hypoxic–Ischemic Brain Injury on MRI

Brain MRI was performed by 1.5 T Philips Ingenia scanner (Philips Medical Systems, Nederland B.V.). The protocol included: turbo spin echo 3D fluid‐attenuated inversion recovery fat saturation, TR/TE/TI = 8000/n350/2350 ms, 265 contiguous sections, 200 × 200 matrix, slice thickness/gap = 1.2/−0.6 mm; spin echo planar diffusion‐weighted imaging, TR/TE = 3658/69 ms, 34 contiguous sections, 104 × 106 matrix, FOV = 230 × 230 × 149 mm, slice thickness/gap 4/0.4 mm, diffusion‐encoding along x‐, y‐, z‐axes averaged, b = 0 and 1000 s/mm². Two experienced neuroradiologists, unaware of the patient's clinical course, independently and retrospectively reviewed brain MRIs. The adjudicators were instructed to only score MRI abnormalities that could be attributed to HIBI. Discrepancies were solved by discussion. For each patient, a qualitative analysis of DWI, ADC, and T2w‐FLAIR was performed. Bilateral, symmetrical, or asymmetrical signal abnormalities attributable to HIBI were assessed in basal ganglia (caudate nucleus, putamen, globus pallidum), thalamus, cerebellum, hippocampus and cerebral cortex of the frontal, parietal, occipital, and temporal lobes. Cerebral cortex involvement was defined as present in case of diffuse signal abnormalities involving at least three adjacent gyri. Based on topography of signal abnormalities, three mutually exclusive patterns were defined: HIBI 0, no signal abnormalities in basal ganglia and cerebral cortex; HIBI 1, signal abnormalities in basal ganglia but no diffuse signal abnormalities in cerebral cortex; HIBI 2, diffuse signal abnormalities in cerebral cortex, regardless the involvement of other areas. Patients were grouped as follows: group (1) patients with either HIBI 0 or HIBI 1 that were referred to as patients with non‐widespread HIBI; group (2) patients with HIBI 2 and referred to as patients with widespread HIBI (Figure 1).

FIGURE 1.

Assessment and grading of hypoxic–ischemic brain injury on brain MRI. DWI, ADC, and FLAIR images from three representative comatose survivors of CA with SRSE. HIBI 0: No signal abnormalities in either the basal ganglia or the cortical gray matter. HIBI 1: Signal abnormalities are present in the basal ganglia, where they can be early detected as diffusion restriction in DWI corresponding to abnormally low ADC values (white arrow) or as increased signal intensity on T2 FLAIR (white arrow) in the early subacute phase. HIBI 2: Diffuse (≥ 3 adjacent gyri) signal abnormalities in the cortical gray matter can be seen in the early phase as diffusion restriction in DWI and abnormally low ADC signal in the frontal, parietal, and occipital lobes (black arrowhead), and in the subacute phase as increased signal intensity on T2 FLAIR; in this example, concomitant analogous signal alterations can also be found in the basal ganglia (white arrow). ADC, apparent diffusion coefficient; CA, cardiac arrest; DWI, diffusion weighted imaging; FLAIR, fluid‐attenuated inversion recovery; HIBI, hypoxic–ischemic brain injury; SRSE, super‐refractory status epilepticus.

2.6. Study Outcomes

The primary outcome was neurological outcome at 6 months, assessed by the Cerebral Performance Category (CPC) scale [range 1–5, with 1 representing intact function and 5 brain death] [18] and dichotomized into good (CPC of 1–2) and poor (CPC ≥ 3). Secondary outcomes included awakening and resolution of SE by the end of the hospitalization and mortality at 6 months.

Awakening was defined as the ability to obey commands before ICU discharge.

SE was considered resolved when epileptic activity disappeared, fulfilling Salzburg criteria and restoring continuous background activity. Evolution from SE to GPDs or brain death at any time during hospitalization was not deemed resolution of SE.

2.7. Study Size

Assuming that a diffuse HIBI is associated with poor outcome in approximately 75% of survivors of CA at 6 months [19] and that this proportion is estimated around 30% in patients with favorable multimodal prognostic indicators [12], a population of 40 patients with SRSE should yield a power of 80% to detect a 45% difference in poor neurological outcome at 6 months between patients with widespread as opposed to non‐widespread HIBI. Significance level was set at p < 0.05, two‐sided.

2.8. Statistical Methods

This study follows the STROBE Statement [20]. Descriptive statistics was summarized. Weighted Cohen's Kappa was used for inter‐rater agreement between the neuroradiologists.

For the primary analysis, a binary logistic regression model was used to assess neurological outcome at 6 months in groups 1 and 2. Secondary outcomes were analyzed analogously. Model adjustment was performed using the decision on WLST as covariate to analyze the primary outcome, awakening before discharge and mortality at 6 months. No model adjustment was applied for analyzing SE resolution, since maximal treatments for SE were provided prior to WLST decision.

The Fisher's exact test was used for an exploratory subgroup analysis assessing the association between each MRI patterns (HIBI 0, HIBI 1, HIBI 2) and neurological outcome at 6 months, SE resolution, awakening before discharge and mortality at 6 months, whenever a significant between‐group difference in the primary analysis was found. Odds ratios and 95% CI were calculated according to Altman (1991) [21]. A Goodman constant [22] was added for analyzing contingency tables in case of cells with zero observations. The Kruskal–Wallis test was used to compare MRI execution times between subgroups. Statistical significance level was set at 95% (p = 0.05), two‐tailed.

3. Results

3.1. Study Population

From January 2011 to February 2023, 538 patients with CA were admitted to the Cardiac ICU and screened for the study; 114 patients were excluded due to death (77 patients) within 24 h of CA or consciousness recovery (37 patients) after rewarming. Of the 424 patients unconscious beyond 24 h, 91 developed SE. Seventy‐eight out of 91 developed SRSE; however, 36 were excluded because they either died due to multiorgan failure or heart failure prior to performing brain MRI (25 patients) or could not perform brain MRI due to prolonged instability from a critical care standpoint (11 patients). Forty‐two patients with SRSE underwent brain MRI and all but two were included in the study (Figure 2). Baseline characteristics are summarized in Table 1.

FIGURE 2.

Study population. CEEG, continuous electroencephalogram; GPDs, generalized periodic discharges; HF, heart failure; MOF, multiorgan failure; MRI, magnetic resonance imaging.

TABLE 1.

Baseline patients' characteristics.

Population (N = 40)	Unit	Median or n (%)	[IQR]
Age	Years	60	[54–70]
Gender	n (%)	23 (57)
Male	n (%)	17 (43)
Female
No flow time	Minutes	0	[0–3]
Low flow time	Minutes	23	[13–31]
Out‐of‐hospital cardiac arrest	n (%)	28 (70)
In‐hospital cardiac arrest	n (%)	12 (30)
Presenting heart rhythm	n (%)	25 (62)
Shockable rhythm (ventricular fibrillation or pulseless ventricular tachycardia)	n (%)	15 (38)
Un‐shockable rhythm (Pulseless electrical activity or asystole)
Super‐refractory status epilepticus
Non convulsive status epilepticus (NCSE)	n (%)	38 (95)
Status myoclonus	n (%)	2 (5)
Pupillary reflex bilaterally absent	n (%)	1 (2.5)
Corneal reflex bilaterally absent	n (%)	8 (20)
Neuron specific enolase (NSE)	μg/L	48	[29–70]
Neuron specific enolase (NSE) at 48‐72 h > 60 μg/L	n (%)	13 (32.5)
Cortical N20 bilaterally absent	n (%)	2 (5)

3.2. Assessment of Hypoxic–Ischemic Brain Injury on MRI

Brain MRI was performed after a median of 15 days (IQR [8–19]) from CA. The median time to perform brain MRI was 15 days (IQR [7–17]) in the HIBI 0 subgroup, 10 days (IQR [8–16]) in the HIBI 1, and 16 days (IQR [9–23]) in the HIBI 2, with no significant differences in the median execution times between the subgroups (p = 0.472). Signal abnormalities in either or both T2w‐FLAIR and DWI/ADC were detected in the basal ganglia (57.5%), thalamus (22.5%), hippocampus (28.7%), cerebellum (46.2%), and cerebral cortex (45.0%). Sixteen patients (40%) were classified as HIBI 0, 6 (15%) as HIBI 1, and 18 (45%) as HIBI 2 (Table 2). Inter‐rater agreement was almost perfect (κ = 0.866). Disagreement occurred in four cases.

TABLE 2.

Assessment of hypoxic–ischemic brain injury (HIBI) on brain MRI.

	Unit	Population (N = 40)
Patients undergoing brain MRI	n (%)	40 (100)
Time from cardiac arrest to brain MRI	Days
Median		15
Interquartile range		8–19
Signal abnormalities in either of both FLAIR and/or DWI/ADC
Basal ganglia a	%	57.5
Thalamus	%	22.5
Hippocampus	%	28.7
Cerebellum	%	46.2
Cerebral cortex b	%	43.7
Post‐anoxic brain injury pattern
HIBI 0	n (%)	16 (40)
HIBI 1	n (%)	6 (15)
HIBI 2	n (%)	18 (45)

Abbreviation: HIBI, hypoxic–ischemic brain injury.

^a Putamen and/or caudate nucleus and/or globus pallidus and/or substantia nigra in either hemisphere.

^b Bilateral symmetric or asymmetric signal abnormalities involving ≥ 3 adjacent gyri of the frontal, parietal, temporal, or occipital lobes.

The severity of the HIBI was used to categorize patients: group 1 included the 22 patients with either HIBI 0 or HIBI 1 that were considered as having non‐widespread HIBI; group 2 was represented by the remaining 18 patients that had widespread HIBI coinciding with the HIBI 2 pattern. Prognostic indicators in group 1 and 2, together with the proportion of patients with likely poor outcome according to them, are detailed in Table 3. Electroclinical features of patients in each group are shown in Table 4.

TABLE 3.

Prognostic indicators and treatment restriction in patients with non‐widespread (group 1) and widespread (group 2) hypoxic–ischemic brain injury.

Prognostic indicators	Group 1 (N = 22)	Group 2 (N = 18)	p
Pupillary reflex bilaterally absent—n (%)	0 (0.0)	1 (5.5)	0.476
Corneal reflex bilaterally absent—n (%)	5 (22.7)	8 (44.4)	0.185
Neuron specific enolase (NSE) at 48–72 h > 60 μg/L – n (%)	3 (13.6)	10 (55.5)	0.007
Cortical N20 bilaterally absent—n (%)	0 (0.0)	2 (11.1)	0.192
Status myoclonus—n (%)	0 (0.0)	2 (11.1)	0.131
Poor outcome likely a — n (%)	0 (0.0)	13 (72.2)	p < 0.00001
Withdrawal of life‐sustaining treatment—n (%)	1 (4.5)	3 (16.6)	0.252

^a Unconsciousness lasting ≥ 72 h without confounders and presence of at least two indicators of poor prognosis.

TABLE 4.

Electroclinical features in patients with non‐widespread (group 1) and widespread (group 2) hypoxic–ischemic brain injury.

Electroclinical features	Group 1 (N = 22)	Group 2 (N = 18)	Total (%)
Electrographic features (Salzburg criteria)
Epileptiform discharges > 2.5 Hz (with or without ictal motor phenomena)	16	10	26 (65)
Epileptiform discharges < 2.5 Hz with typical spatiotemporal evolution (with or without ictal motor phenomena)	5	3	8 (20)
Epileptiform discharges < 2.5 Hz with subtle clinical ictal phenomena	1	5	6 (15)
Clinical motor features
Subtle ictal motor manifestations (including facial myoclonus)	4	6	10 (25)
Generalized myoclonus	0	2	2 (5)
Any ictal motor manifestation	4	8	12 (30)

3.3. Primary Outcome

The odds for poor neurological outcome at 6 months for a patient in group 2 were about 13 times higher than the odds for a patient in group 1 even after adjusting for WLST decision (27% in group 1 vs. 83% in group 2; OR 12.8, CI 95% [2.5–64.3], p = 0.002).

The proportion of patients with poor outcome at 6 months increased across the three mutually exclusive MRI patterns. A significantly higher probability of poor functional outcome at 6 months in patients with HIBI 2 compared to those with HIBI 0 was found, whereas no statistically significant difference between patients with HIBI 1 and HIBI 0 was observed (25% in HIBI 0, 33% in HIBI 1, 83% in HIBI 2; OR: 1.00 (reference) vs. 1.5 CI 95% [0.2–11.5] p = 0.69 vs. 15 (2.8–80.3), p < 0.001) (Table 5).

TABLE 5.

Primary outcome and subgroup analysis.

Cerebral performance category at 6 months c
	CPC 1–2 N (%)	CPC 3–5 N (%)	OR [95% CI]	p
Patient group
Group 1 a (reference)	16 (73)	6 (27)	12.8 [2.5–64.3]	0.002
Group 2 b	3 (17)	15 (83)
Brain MRI pattern
HIBI 0 (reference)	12 (75)	4 (25)	1.00	0.69 < 0.001
HIBI 1	4 (67)	2 (33)	1.5 [0.2–11.54]
HIBI 2	3 (17)	15 (83)	15 [2.8–80.3]

^a Group 1 includes both the HIBI 0 and HIBI 1 MRI patterns.

^b Group 2 refers to the HIBI 2 MRI pattern.

^c The odds ratio of the primary outcome was adjusted using withdrawal of life‐sustaining treatment as covariate.

3.4. Secondary Outcomes

The odds for awakening before discharge from ICU were significantly higher in group 1 compared to group 2 even after adjusting for WLST decision (73% in group 1 vs. 22% in group 2; OR 8.8, CI 95% [1.9–40.3], p = 0.005). A significantly lower rate of consciousness recovery in patients with HIBI 2 compared to those with HIBI 0 was observed, but no statistically significant difference was found between patients with HIBI 1 and those with HIBI 0 (75% in HIBI 0 vs. 66% in HIBI 1 vs. 22% in HIBI 2, OR: 1.00 (reference) vs. 1.5 CI 95% [0.2–11.5] p = 0.69 vs. 10.5 [2.1–51.2], p = 0.002).

SRSE resolved in 82.5% (33/40) of cases, evolved to GPDs in 10% (4/40) and persisted despite treatment in 7.5% (3/40). Group 1 had more than 10 times higher odds for SE resolution during hospitalization compared to group 2 (95% in group 1 vs. 67% in group 2, OR 10.5, CI 95% [1.1–97.9], p = 0.039). A significantly higher rate of SE resolution was found in patients with HIBI 0 compared to those with HIBI 2, but no statistically significant difference was observed between patients with HIBI 1 and HIBI 0 (100% in HIBI 0 vs. 83% in HIBI 1 vs. 67% in HIBI 2, OR: 1.00 (reference) vs. 9.0 CI 95% [0.3–254.0] p = 0.13 vs. 17.0 [0.8–334.0], p = 0.018). Of the seven patients with either persistent SRSE or GPDs, one had persistent vegetative state and six died within 6 months (3/6 died after the decision on WLST due to irreversible respiratory failure despite maximal treatments; 2/6 died due to systemic infection after ICU discharge; 1/6 died due to cardiac arrhythmia before ICU discharge) (Data not shown).

Mortality at 6 months occurred in 15/40 patients. WLST was applied to three patients in group 2 and to one patient in group 1 due to refractory hemodynamic instability despite maximal support with amines (1 patient) or respiratory failure despite maximal respiratory support (3 patients) (data not shown). No significant between‐group difference in the odds for mortality at 6 months was found even after adjusting for WLST decision (27% in group 1 vs. 50% in group 2, OR 0.4, CI95% [0.1–1.9], p = 0.303) (Table 6).

TABLE 6.

Secondary outcomes and subgroup analysis.

	Awakening c			Resolution of status epilepticus			Mortality at 6 months c
	N (%)	OR [CI 95%]	p	N (%)	OR [CI 95%]	p	N (%)	OR [CI 95%]	p
Patient group
Group 1 a	16 (73)	8.8 [1.9–40.3]	0.005	21 (95)	10.5 [1.1–97.9]	0.039	6 (27)	0.4 [0.1–1.9]	0.303
Group 2 b	4 (22)			12 (67)			9 (50)
MRI pattern
HIBI 0	12 (75)	1 (reference)		16 (100)	1 (reference)		4 (25)
HIBI 1	4 (66)	1.5 [0.2–11.5]	0.69	5 (83)	9 [0.3–254]	0.13	2 (66)	—	—
HIBI 2	4 (22)	10.5 [2.1–51.2]	0.002	12 (67)	17 [0.8–334]	0.018	9 (50)	—	—

^a Group1 includes both the HIBI 0 and HIBI 1 MRI patterns.

^b Group 2 refers to HIBI 2 MRI pattern.

^c The odds ratio of this outcome was adjusted using withdrawal of life‐sustaining treatment as covariate.

4. Discussion

Treatment of definite SE in post‐anoxic patients remains debated due to its uncertain impact on prognosis. This dilemma is crucial for patients with SRSE, where epilepsy specialists face decisions on continuing or withdrawing treatment. In a previous study of our group [12], sustained pharmacological treatment was provided as long as multimodal prognostic indicators were not unfavorable; among treated patients, good neurological outcome was achieved in 44.4% of cases, justifying prolonged treatment whenever multimodal indicators converge against poor prognosis. However, more than a half had poor neurological outcome, highlighting the need of further characterizing this patient population's prognostic profile.

A diffuse HIBI is considered an indicator of poor prognosis. Due to the absence of a widely accepted classification for HIBI, we proposed a grading system based on signal abnormalities in the cortical and subcortical gray matter and investigated the relationship between mutually exclusive MRI patterns and clinical outcomes. The aim was to clarify the prognostic relevance of brain signal abnormalities in post‐anoxic SRSE to aid in multimodal prognostication and decision‐making.

The grading system was inspired by studies showing that cortical gray matter abnormalities best correlate with neurological outcome [23] and that cytotoxic edema is frequent in the deep gray nuclei but rare in cortical gray matter of good outcome patients [19]. We hypothesized a link between spatial distribution of signal abnormalities and HIBI severity, expecting injury to be limited to the deep gray nuclei in non‐severe cases and extending to the cortical gray matter in severe ones.

The highly variable extent of HIBI in patients with post‐anoxic SRSE is a major finding of our study. More than a half had non‐widespread HIBI and 40% had no abnormalities, indicating that SRSE is not always associated with extensive injury. While cytotoxic edema is early visible on MRI, complementary phenomena undetectable on conventional MRI may contribute to post‐anoxic encephalopathy, such as synaptic failure [24, 25, 26]—which is responsible for electrographic abnormalities—and altered connectivity [27, 28, 29, 30]. These phenomena may vary in combination, resulting in heterogeneous instrumental findings. Signal abnormalities in the hippocampus were infrequently detected, supporting studies [31] indicating no preferential distribution of HIBI in the hippocampus, despite its high vulnerability to hypoxia [32].

Nearly half of the patients had a good neurological outcome at 6 months, indicating little prognostic value of SE itself. We demonstrated that brain MRI provides additional prognostic information: grouping by the extent of HIBI shows that the proportion of patients with a good outcome at 6 months is about two‐and‐a‐half times higher in case of non‐widespread injury. This observation aligns with studies based on semiquantitative imaging analysis [23, 27, 33, 34].

The proportion of patients with poor outcome at 6 months increased across the three mutually exclusive MRI patterns; however, the rate of poor outcome did not differ significantly between patients with only subcortical signs of HIBI and those without abnormalities, supporting the results by Mlynash et al. which found that up to 50% of patients with good outcome had deep gray nuclei signal abnormalities [19]. The under‐representation of the latter category in our cohort may affect results.

Most patients with SRSE and no signs of HIBI had a favorable outcome. Our results support findings by Barth et al. [35] showing that patients with a malignant EEG pattern—which included epileptiform activity, according to the definition by Westhall et al. [36]—and no brain MRI abnormalities had a favorable outcome in most cases. In our study, the proportion of patients without signal abnormalities is similar to that reported by Barth et al., despite some differences between the two populations.

Patients with SRSE and non‐widespread HIBI had nearly three times the probability of consciousness recovery before discharge compared to those with widespread HIBI, suggesting a link between consciousness disorders and cortical anoxic brain injury [31]. While the low awakening rate in patients with widespread HIBI may indicate severe compromise, persistent coma occurred approximately in one quarter of patients with non‐widespread injury. The loss of thalamo‐cortical circuits integrity [26] may contribute to persistent coma; since synaptic transmission impairment may occur despite preserved membrane potential [24], technologies measuring cytotoxic edema may have limited negative predictive value.

SRSE resolved after prolonged treatment in most cases; however, patients with widespread HIBI showed a low probability of SE resolution. SE may represent one of the electrographic phenotypes of post‐anoxic synaptic dysfunction [25, 37] resulting from an excitation–inhibition imbalance [38].

This study has limitations. First, patients with post‐anoxic SRSE represent a minor proportion of those with post‐anoxic encephalopathy; however, this group is significant for which brain MRI is ordered for neuroprognostication. Second, only a minority had predictions of poor outcome likely based on multimodal prognostication, suggesting better chances of good outcomes compared to post‐anoxic patients with worse prognostic profiles. These factors, the small size and the single‐center design, limit the generalizability of our results. Nevertheless, we believe relevant insights were provided. Third, EEG review was performed without inter‐rater agreement measurement. However, our previous study [12], which included part of this cohort, showed high inter‐rater agreement for EEG classification in this population, following internationally accepted criteria [15, 16, 17, 18]. Fourth, SE contribution to signal abnormalities could not be established due to the lack of comparison with post‐anoxic comatose patients without SE. Further insights could be gained from CT/MRI perfusion [39, 40] or MRI spectroscopy [41]. Fifth, the definition of HIBI severity is arbitrary. The Neurocritical Care Society [42] guideline defines brain damage as diffuse in the presence of bilateral restricted diffusion in the anterior and posterior circulation, extending beyond vascular territories and involving both cortex and deep gray matter. The European Resuscitation Council and European Society of Intensive Care Medicine guideline [4] does not provide specific radiological criteria, recommending confirmation of generalized and extensive ischemic injury on conventional visual analysis by an experienced neuroradiologist. Both guidelines emphasize the lack of a standardized measurement method. We chose a qualitative method that is easy, reliable, and quick. We also proposed a severity classification; however, the study is not powered to analyze subgroups and a multicenter observational study will be needed to validate the grading system. Sixth, since brain MRI was performed later than the optimal 2‐ to 7‐day window after CA [43], signal abnormalities may be underestimated; however, we believe our timing reflects the real‐world experience of many healthcare centers. Lastly, the study lacks semiquantitative or quantitative methods that could have provided more standardized imaging analysis.

Author Contributions

Susanna Diamanti: conceptualization, investigation, writing – original draft, writing – review and editing, formal analysis. Francesco Pasini: investigation, writing – review and editing, data curation. Cristina Capraro: investigation, writing – original draft, data curation. Mirko Patassini: investigation, data curation. Elisa Bianchi: formal analysis, writing – review and editing, methodology. Matteo Pozzi: writing – review and editing, data curation. Marco Normanno: data curation. Anna Coppo: supervision, data curation. Paolo Remida: supervision. Leonello Avalli: supervision. Carlo Ferrarese: supervision. Giuseppe Foti: supervision. Simone Beretta: supervision, writing – review and editing, data curation.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

Anonymized clinical data will be made available upon reasonable request from a qualified investigator.