Periodic discharges and status epilepticus: A critical reappraisal

(cid:1) This study supports the utility of differentiating between periodic discharges (PDs) during and after Status Epilepticus (SE) (cid:1) Interictal PDs during SE are not related to outcome; PDs occurring 24–72 h from SE resolution strongly correlate with poor prognosis. (cid:1) Considering only PDs after SE cessation improves EMSE (Epidemiology-based mortality score in SE) overall accuracy.


Introduction
Status epilepticus (SE) is a life-threatening neurological emergency with a poor outcome in approximately 10-30% of patients (Leitinger et al., 2019;Leitinger et al., 2020).SE prognosis relies on patient-related variables such as age and comorbidities, as well as SE features such as etiology, duration and clinical presentation (Sutter et al., 2013).
EEG is crucial in the diagnosis and management of SE (Sutter et al., 2011), but its contribution in predicting SE outcome is more difficult to outline, mainly due to the intrinsic temporal evolution of EEG patterns according to SE duration and treatment (Alkhachroum et al., 2020).
Two pivotal studies (Nei et al., 1999) (Jaitly et al., 1997) indicated that PDs both during and after SE may be associated with unfavorable outcome, setting the base for including PDs in the EMSE EEG features.However, Nei and colleagues based their retrospective findings on univariate analysis without adjusting for etiology and other outcome-related variables (Nei et al., 1999).Conversely, the presence of PDs after SE cessation was associated with poor outcome after adjusting for etiology (Jaitly et al., 1997), with a cohort composed for more than 60% of patients presenting SE following decreased level of antiseizure medications (ASM) Àmostly associated with good outcome-, or cardiac arrest (CA) Àquite invariably associated to bad prognosis.Finally, a more recent study incorporating the 2012 ACNS terminology (Hirsch et al., 2013) and assessing the prognostic significance of continuous EEG (cEEG) after SE resolution, revealed conflicting results (Alvarez et al., 2015).While EEG background was found to be independently associated with outcome even adjusting for other prognosis-related findings, no such association was observed for rhythmic and periodic patterns (RPPs) (Alvarez et al., 2015).In this context, our study aimed to clarify the association between PDs during and after SE resolution and in-hospital mortality, after correction with other outcome-related variables.

Study population
This single-center study was performed at the University Hospital of Parma, Italy.The STROBE-guidelines were followed to improve and assure the quality of our study (Von Elm et al., 2007).
Clinical and EEG data of all consecutive adult (age !18 years old) patients with an EEG pattern of SE from January 2020 to October 2023 were retrospectively reviewed.Our institution has a 24/7 EEG service and routinely performs spot or more prolonged EEG recordings.
SE in hypoxic-ischemic encephalopathy after cardiorespiratory arrest were excluded.In case of relapsing SE episodes in the same patient, only the first SE episode was considered.
The following clinical data were collected from the patients' digital charts: age, sex, modified Rankin scale (mRS) at admission, history of seizures, SE etiology, type, duration and outcome.Etiology was defined as recommended by the current guidelines of the International League Against Epilepsy (ILAE) (Trinka et al., 2015).According to literature (Alvarez et al., 2015) (Rossetti et al., 2006), we defined as potentially fatal etiology: acute (<7 days) large vessel ischemic stroke, acute cerebral hemorrhage, acute central nervous system infection, severe systemic infection, malignant brain tumor, chronic renal insufficiency requiring dialysis, systemic vasculitis, metabolic disturbance or acute intoxication sufficient to cause coma in the absence of SE, and intracranial surgery.
SE duration was defined as the time between SE diagnosis and the performance of EEG proving termination of seizure.Because patients were monitored by EEG recordings every 12-24 h, duration of SE represents an approximation with a maximal inaccuracy of 12-24 h.
Primary outcome was in-hospital death, also including patients transferred to palliative care with imminent death (within 15 days).SE was classified as refractory (RSE) if the status persisted despite the administration of two ASM, and as super refractory (SRSE) if it persisted or recurred even after more than 24 h anesthetic treatment (Trinka and Leitinger, 2022).
The study was approved by the local ethical committee and was conducted according to the ethical principles for medical research involving human subjects in the Declaration of Helsinki and its later amendments.

EEG
For each patient, all EEG recordings during SE and between 24 to 72 h after SE resolution were considered.SE cessation could be determined by EEG at the time of termination, or by the next spot EEG that was available to patient.Patients were followed by repetitive spot EEGs (30 min) (at least once daily); relying on clinical judgement, nearly 20% of cases had more prolonged EEG monitoring during SE (>60 min of consecutive recording, usually less than 120 min).After SE resolution, only daily spot EEGs were performed.
All EEGs were reported by trained electroencephalographers (FM, IF, LZ) applying the 2021 ACNS terminology (Hirsch et al., 2021).For the study's purposes, two authors (FM, IF) retrospectively reviewed all recordings to ensure accurate classification, blinded to outcome.
PPs qualify as seizures or SE, both during and after SE, as recommend by the 2021 ACNS Terminology (Hirsch et al., 2021), were considered ictal patterns and not included in the analysis: -PPs with a typical frequency > 2.5 Hz or with definite evolution lasting ! 10 s, -PPs with a clinical correlate time-locked with the EEG pattern, -PPs consistent with ictal interictal continuum (IIC) (with a frequency > 1 and < 2.5 Hz or !0.5 and 1 Hz and a plus modifier or fluctuation lasting ! 10 s) if electro-clinical resolution after ASM medications was documented (Leitinger et al., 2023).
Non-ictal PPs were defined as Interictal SE PDs if they occurred at any time during SE, and as post SE PDs if they occurred 24 to 72 h after SE resolution (Fig. 1).If a patient presented both ictal and non-ictal PPs, we assessed the patient as having interictal SE PDs.At risk of lowering the detection ratio of post SE PDs, patients exhibiting only PPs within the first 24 h after SE were excluded from analyses, to ensure data standardization and to more clearly differentiate between PPs during and after SE resolution.PPs were recorded as present if they occurred any time during an individual EEG monitoring session.Because pattern frequency may fluctuate within a given recording session, we selected the more representative recorded frequency for any given pattern.

Statistical analysis
Categorical variables were presented as absolute values and percentages, while continuous variables as means and standard deviations (SD) or interquartile range (IQR).The Student's t-test, Pearson chi-squared and the Mann-Whitney test were applied for analyzing continuous and categorical variables considering inhospital mortality as the dependent variable.Univariate significances were evaluated throw logistic stepwise regression analysis.
Receiver operating characteristic (ROC) curve was generated to determine EMSE capacity in outcome prediction.Sensitivity (SE), specificity (SP), positive and negative predictive values (PPV and NPV) and number of accurately classified patients (accuracy) for EMSE previously suggested cutoff (EMSE-64) were calculated.
Through an exploratory analysis we recalculated EMSE score incorporating univariate analysis results and considering only post SE PDs.The new EMSE score (modified EMSE, mEMSE) performance was evaluated through ROC curve analysis, assessing the area under the curve (AUC), as well as SE, SP, PPV, NPV and accuracy, for mEMSE-64 cutoff.
All analyses were executed utilizing the Jamovi software (Jamovi, 2022 version), with statistical significance set at P 0.05.

Baseline characteristics and univariate comparisons
Of a cohort of 204 patients with available EEG during SE, we excluded 15 patients with SE following CA.We finally included 189 SE patients.In-hospital death was seen in 62 (32%) patients.Univariate comparisons of demographics, clinical and SE features of in-hospital survivors and non survivors groups are presented in Tables 1 and 2.
Around half of patients were females, the median age was 69.4, and 33% had a history of seizure.Age, mRS > 2 and mean STESS were positively associated with unfavorable outcome (P = 0.001, P < 0.001, P < 0.001, respectively).
33% of patients presented SE with prominent motor symptoms, while 67% were NCSE (24/127 with coma).SE etiology was acute symptomatic in 57% of patients, remote symptomatic in 21% and progressively symptomatic in 14%.Five patients suffered from SE within epileptic syndromes and in 8 cases SE etiology was undetermined.Mean SE duration was 3 days (IQ: 1-4), and a potentially fatal etiology was found in 33% of cases (8 acute large vessel ischemic strokes, 8 acute cerebral hemorrhages, 7 acute central nervous system infections, 11 acute severe systemic infections, 16 malignant brain tumors, 8 metabolic disturbances or acute intoxications and 4 intracranial surgeries).

PDs during and after SE
All patients received at least 1 EEG/day, and 44 (22%) received at least one more prolonged EEG recording.Duration was 30 min (mean) (median: 30 min) for spot EEG and 86 min (mean) (median: 80 min) for more prolonged EEG recording.The mean cumulative duration of recording for each patient was 48 min/day during SE and 30 min/day for the post SE period.
PPs were present during SE in 75 (39%) patients: 23% patients showed interictal SE PDs (10 together with ictal PPS) (40 Lateralized PDs ÀLPDs, 4 generalized PDs ÀGPDs).16% patients presented PPs only fulfilling the criteria for an ictal pattern (17 time-locked with major or subtle motor symptoms, 8 with a typical frequency > 2.5 Hz or with definite evolution and 6 with clinical and EEG resolution after ASM) and were excluded from analysis.
79% patients had at least 1 EEG between 24 and 72 h after SE resolution.Among them, post SE PDs were present in 21% (22 LPDs, 9 GPDs).The association between interictal SE PDs and outcome did not reach statistical significance, and this held true also when considering LPD and GPD separately.
Post SE PDs were significantly associated to bad outcome (P < 0.001), even after correction for age, mRs, SE duration and refractoriness, potentially fatal etiology and STESS score.Both LPDs (P = 0.004) and GPDs (P < 0.001) subgroups were significantly associated to poor prognosis.Finally, the association with bad outcome remained even when PDs were initially present during SE and persisted at its resolution (14/31) (P < 0.001), as well as cases where PDs first appeared after SE resolution (17/31) (P < 0.001).Table 3 synthesizes main PDs features during and after SE.

Discussion
The main finding of our retrospective study is that interictal PDs during SE do not seem to be related to outcome, whereas PDs occurring 24-72 h from SE cessation strongly correlate with poor prognosis.In our cohort, considering only PDs after SE cessation improved EMSE overall accuracy, striking a better balance between sensitivity and specificity for the (m)EMSE-64 cutoff.
PDs are commonly observed in critically ill patients, and their presence is associated with a variable but generally heightened risk of unfavorable outcome (Husari and Johnson, 2020;Li et al., 2017).PDs are significant predictors of poor prognosis after correction for other outcome-related variables in acutely ill patients (Pedersen et al., 2013) and in more selected cohorts of subjects with subarachnoid hemorrhage (SAH) (Claassen et al., 2007) or sepsis (Ferlini et al., 2023).Furthermore, PDs are independently associated with short-and long-term functional decline in patients without brain injury (Sainju et al., 2015) as well as in post-stroke epilepsy patients (Fukuma et al., 2023).Conversely, some studies propose that the underlying illness itself plays a more significant role on outcome than the presence of PPs (Orta et al., 2009;  Kate et al., 2012) and a comprehensive case-control study found no significant association between PDs and prognosis after correction for age, level of consciousness, and etiology (Foreman et al., 2012).
PPs during SE were historically described as a pattern of terminal stages (Treiman et al., 1990), but later evidence challenged this association (Garzon et al., 2001).From a prognostic standpoint, evidence on the association between outcome and PPs during SE is blurred.The studies by Nei (Nei et al., 1999) and Jaitly (Jaitly et al., 1997) set the basis for EMSE EEG parameters (Leitinger et al., 2015), even if they have recognized methodological limitations, as discussed above.Additionally, a retrospective study on a little cohort of RSE patients suggested a correlation between PDs during SE and poor prognosis (Liberalesso et al., 2013).Conversely, according to other studies, SE prognosis might rely more on etiology and age than on the presence of specific ictal EEG patterns, including PDs (Garzon et al., 2001).The same holds true considering RPPs appearing in the 24 h following SE resolution (Alvarez et al., 2015).It is noteworthy that the paper by Alvarez and colleagues was the only on the topic incorporating the 2012 ACNS standardized critical care EEG terminology, which, for the first time, provided a consensus definition for RPPs (Hirsch et al., 2013).Moreover, there was considerable diversity in EEG recordings and interpretations across studies.Nei et al. (Nei et al., 1999) monitored patients with spot or 24-hour EEG without specifying the distribution among subgroups.The two studies by Jaitly and Alvarez assessed continuous EEG (cEEG) (Jaitly et al., 1997) (Alvarez et al., 2015).However, in the first one, cEEG was used exclusively in the intensive care unit (ICU), with spot EEGs upon request after ICU discharge.Additionally, SE resolution was defined as ''control of clinical SE," likely encompassing several instances of non-convulsive SE stage detection rather than post SE EEG assessment (Jaitly et al., 1997).
Our retrospective cohort study was limited by the absence of cEEG monitoring, likely resulting in a lower detection rate of PPs along with less accurate SE duration estimation.While we acknowledge the importance of cEEG in the management of SE patients, our study reflects the current clinical practice in many    centers.It is worth noting that our analysis was based on the latest ACNS version (Hirsch et al., 2021) which offers a rigorous definition of PDs, excluding patterns that can be deemed ictal based on the current EEG criteria for seizure or SE.Notably, in our cohort, even when accounting for all PPs (including those within an ictal pattern), there was no observed association between PPs during SE and outcome.On the contrary, PDs after SE cessation were significantly associated to prognosis after correction for other outcomerelated features.
The EMSE score originates from an exploratory approach in which several prognostic SE features, gathered from the literature, were combined to create a score that performed best in a defined cohort of 92 patients (Leitinger et al., 2015).Since then, along with STESS (Rossetti et al., 2008), EMSE is the most widely used and externally validated score (Brigo et al., 2022), with the leading EMSE version including Etiology, Comorbidity, Age and EEG (EMSE-EACE).Among EEG features, EMSE score included, together with SB and ASIDs, LPD and GPD, without specifying the timing (during or after SE).Notably, in the original EMSE study, only GPDs were related to prognosis, persisting after correction for multiple comparisons (Leitinger et al., 2015).
EMSE performed well in our cohort in predicting in-hospital mortality, with an overall AUC of 0.751.However, consistent with findings from a recent meta-analysis, the suggested cutoff (EMSE-64) showed a low specificity compared to sensitivity and a notably low positive predictive value (PPV), with a potential misclassification of patients with a favorable prognosis as being at high risk for mortality (Yuan et al., 2023).
In our study, similar to the EMSE one in terms of epidemiological and demographic variables (Leitinger et al., 2015), both LPD and GPD after SE were significantly associated to outcome, while neither of the two was related to prognosis if present during SE.
By combining PDs presence after SE cessation with other EMSE EEG variables, EMSE overall AUC increased with a better balance between sensitivity and specificity and greater PPV for the 64cutoff.This is probably due to the substantial specific weight of PDs within EMSE EEG features, given the low percentage of patients with ASIDs and spontaneous SB, both in our cohort and in the original one (Leitinger et al., 2015).
Considering patients with SE resolution, the presence of post SE PDs may reflect a more severe cerebral injury, potentially accounting for their association with poorer prognosis.However, if PDs presence after SE cessation might directly contribute to further secondary neuronal injury is still an unresolved question.Treating such patterns might pose the risk of over-treatment, potentially yielding more adverse than beneficial effects.Probably, the relationship between PDs and brain damage relies on the frequency of PDs (Subramaniam et al., 2019).Vespa demonstrated that seizures and PDs were associated to metabolic crisis and reduced oxidative metabolism, as measured by cerebral microdialysis in traumatic brain injury patients (Vespa et al., 2016).Additionally, the presence of PDs was associated with secondary damage after SAH, leading to normal compensatory mechanisms failure and tissue hypoxia with frequencies > 2 Hz (Witsch et al., 2017).Higherfrequency PDs, along with other EEG features aligning with the IIC spectrum (Rodriguez Ruiz et al., 2017), are associated with seizures and likely contribute to further neuronal death (Hirsch et al., 2021).Therefore, therapeutic interventions are warranted (Rodriguez Ruiz et al., 2017).In contrast, PDs at lower frequencies as those observed in our cohort, are more likely expression of preexisting brain damage, secondary to SE cause, directly related to SE, or its complications.The observation that SE episodes with comparable duration and features only occasionally lead to PDs after SE resolution indicates that there might be an individual intrinsic vulnerability leading to the appearance of this pattern.From a prognostic standpoint, PDs presence after SE cessation might thus identify a population with a heightened risk of adverse outcome (almost 2/3 with in-hospital death in our cohort), warranting strict clinical and multiparametric monitoring to target management and therapeutic strategies.

Limitations
There are several limitations to the current study.First, this is a retrospective study, based on an EEG registry, leading to underrepresentation both of convulsive SE and ''more benign" SE cases, resolved before EEG recording.
As discussed above, another limitation is the lack of cEEG monitoring.Although most patients underwent more than one standard EEG within 24 h, with at least one EEG per day for all, this likely led to RPPs underestimation (Rossetti et al., 2020) and contributed to an approximation error in SE duration (ratio between the maximal inaccuracy to SE mean duration: mean 38%, median 27%).However, it reflects common clinical practice, particularly in the context of SE episodes in occurring in emergency wards and in non-intensive care departments.
Additionally, patients presenting PPs only within the first 24 h after SE resolution were not analyzed, thus reducing the number of patients with post SE PDs.PPs disappearing in the hours following SE resolution might be more closely associated with interictal SE PDs than with PDs occurring further after SE resolution, with their potentially different prognostic implications.However, while some results might support this hypothesis (Alvarez et al., 2015), further studies (ideally using cEEG) are needed to draw definitive conclusion on this topic.
It would be of great value to assess prognosis according to the amount and duration of PD persistence after SE cessation.However, the lack of cEEG and the variability in EEG recording hinder the assessment of the relationship between PD burden and patient outcome.
Finally, it would be relevant to evaluate parameters such as disability among survivors and return to patient's baseline in relation to PDs presence but, due to retrospective study design, it was not possible to extract this information consistently and reliably.

Conclusion
Our study shows how PDs prognostic value in SE patients is strongly time dependent and that, after adjusting for other prognosis-related variables, only PDs after SE cessation are related to outcome.Through an exploratory analysis we incorporated our results into the EMSE score, creating a modified (mEMSE) displaying better accuracy in the short-term SE prognosis in our cohort.
EM and SL revised the manuscript.LZ acquired and interpreted the data and revised the manuscript.CM draft and revised the manuscript.LP revised the manuscript.PDS interpreted the data, draft and revised the manuscript.IF acquired and interpreted the data, planned and designed the study, draft and revised the manuscript.All authors read and approved of the final version of the manuscript.

Ethical publication statement
We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

FM
is supported by the 2023 International Federation of Clinical Neurophysiology (IFCN) Research Fellowship Grant.PDS is supported by the Swiss National Science Foundation (163398, CRS115-180365) and is supported by the 2022 Swiss League Against Epilepsy Research Support Prize.AZ, EM, SL, CM, LZ and IF report no disclosures.

Table 1
Main patients' data.

Table 3
PDs features during and after SE.