Subclinical epileptiform activity and sleep disturbances in Alzheimer's disease

ABSTRACT Introduction Subclinical epileptiform activity (SEA) and sleep disturbances are frequent in Alzheimer's disease (AD). Both have an important relation to cognition and potential therapeutic implications. We aimed to study a possible relationship between SEA and sleep disturbances in AD. Methods In this cross‐sectional study, we performed a 24‐h ambulatory EEG and polysomnography in 48 AD patients without diagnosis of epilepsy and 34 control subjects. Results SEA, mainly detected in frontotemporal brain regions during N2 with a median of three spikes/night [IQR1–17], was three times more prevalent in AD. AD patients had lower sleep efficacy, longer wake after sleep onset, more awakenings, more N1%, less REM sleep and a higher apnea‐hypopnea index (AHI) and oxygen desaturation index (ODI). Sleep was not different between AD subgroup with SEA (AD‐Epi+) and without SEA (AD‐Epi–); however, compared to controls, REM% was decreased and AHI and ODI were increased in the AD‐Epi+ subgroup. Discussion Decreased REM sleep and more severe sleep‐disordered breathing might be related to SEA in AD. These results could have diagnostic and therapeutic implications and warrant further study at the intersection between sleep and epileptiform activity in AD.

with AD often do not tolerate long-and overnight EEG recordings, and mesial temporal lobe epileptiform activity often lacks correlates on scalp EEG or is not associated with clinical symptoms (Lam et al., 2016;Lam et al., 2017;Lieb et al., 1976).
We hypothesized that there is an interaction between the presence of epileptiform activity and sleep disturbances in AD.To study this interaction, we performed 24-h EEG recordings with concomitant polysomnography (PSG).This combined EEG-PSG set-up allowed assessment of respiratory parameters and reliable REM sleep scoring, which can be challenging on scalp EEG alone.

Standard protocol approval, registration, and patient consent
All prospective study procedures were performed under study protocol, approved by the Ethical Committee of University Hospitals Leuven (clinical.trials.govNCT03617497).Written informed consent was obtained from all research participants or their caregivers.

Patient selection
Research participants between 55 and 85 years old were prospectively recruited between September 2019 and December 2022.Patients with AD were selected from the memory clinic of University Hospitals Leuven.They had either probable clinical AD or probable AD dementia with evidence of the AD pathophysiological process according to the National Institute on Aging-Alzheimer's Association diagnostic guidelines (Albert et al., 2011;McKhann et al., 2011)

Clinical measures
For each person with AD, we registered their age, duration of symptoms, body mass index (BMI), medication use, comorbidities, caffeine consumption, and alcohol and tobacco use.The patient completed the MMSE (Folstein et al., 1975) as cognitive screening test and we performed a clinical interview with Pittsburg Sleep Quality Index (PSQI) (Mollayeva et al., 2016), Epworth Sleepiness Scale (ESS) (Johns, 1991;Scharf, 2022), Mayo fluctuation scale (Ferman et al., 2004), Cornell scale for depression in dementia (Alexopoulos et al., 1988;Park & Cho, 2022), CDR (Juva et al., 1995), and Reutens Questionnaire for clinical seizure diagnosis (Reutens et al., 1992).The same clinical variables were collected in control subjects except for the duration of symptoms and the Mayo fluctuation scale.

EEG and PSG recordings
Twenty-four-hour ambulatory EEG and PSG recordings were acquired with Medatec Brainwalker 3 device and software (Medical Data Technology, Braine-le-Château, Belgium), using a sampling rate of 200 Hz.
The EEG and PSG recordings occurred at home for all participants.
Scalp electrodes were placed using the International 10-20 system with additional lower temporal electrodes (F9, F10, T9, T10, P9, P10) without P3 and P4, and Cz was used as ground.We placed two additional chin electrodes (EMG 1 and 2) and two electrodes next to the eyes (EOG 1 and EOG 2) to measure sleep.Additional sensors were nasal flow (NAF2P), thoracic and abdominal resistance bands, finger pulse oximetry, electrocardiogram (ECG) derivation Eindhoven I, and electromyogram (EMG) of one leg.Finally, a camera with audio recording was installed in the patient's bedroom for nighttime recording.

EEG review and annotation
EEG's were visually reviewed and annotated by the first author (A.D.) and discussed with a certified neurologist (W.V.P.).All seizures, interictal epileptiform discharges (IED), and episodes of temporal rhythmic delta activity (TIRDA) were annotated.Persyst 14 software (Persyst, CA, USA) was additionally used for automatic spikes detections, which were visually reviewed by first author (A.D.).All annotations were blindly shown to two certified neurologists (W.V.P. and J.M.), which scored them according to the International Federation for Clinical Neurophysiology criteria (Kural et al., 2020).In case of disagreement, the readers reached a consensus after open discussion.IED location was visually determined based on the electrode with the highest amplitude in common referential or phase opposition in bipolar montages.TIRDA was defined as short trains, between 1 and 10 s, of intermittent rhythmic delta activity (1 and 4 Hz, 50-100 µV) in the temporal chains Brigo, 2011, Normand et al., 1995;Reiher et al., 1989).The EEG was defined as epileptiform if one or more of the following occurred: seizure, IED, or TIRDA (Di Gennaro et al., 2003;Gambardella et al., 1995;Lam et al., 2020).

RESULTS
We included 48 patients with AD.Seven were excluded due to technical failure of the device (n = 3) or behavioral problems (n = 4).Thirty-four control subjects were included, but four were excluded from the analy-sis due to intolerance to the EEG (n = 2) equipment or excessive use of psychotropic drugs or alcohol above the limit of three daily units that could influence sleep (n = 2).For PSG analysis, three additional AD subjects and one control subject were excluded due to insufficient signal quality to obtain a reliable scoring of respiratory events.

Demographics (Table 1)
The mean age of AD patients was 4 years higher than the control group.Three AD patients took a benzodiazepine (7%).Thirteen AD patients (32%) and one control participant (3%) were on antidepressive Twelve AD patients had an MMSE ≥27 and were further classified as having mild cognitive impairment (MCI).Clinical characteristics are summarized in Table 1.Biomarkers were available for 24 AD patients (Appendix A).

EEG parameters (Table 2)
Fourteen of 41 patients with AD (34%) had an interictal epileptiform EEG, with a higher prevalence in the MCI (50%) versus the dementia subgroup (28%).The prevalence was almost three times higher than in controls, where 4 of 30 participants (13%) had one or more spikes on scalp EEG (p-value = .086).Of 14 AD patients with an epileptiform EEG, 13 had one or more spikes, and 1 had TIRDA exclusively during wakefulness.Two hundred and thirty-four spikes were marked on scalp EEG in the AD group versus 12 spikes in the control group.
In subjects with spikes, the median was three

3.3.2
PSG parameters in AD-Epi+, AD-Epi-, and controls (Table 4) We performed PSG analysis in AD-Epi+ and AD-Epi-subgroups.There was no difference in the prevalence of OSAS or OAHI between AD-Epi+, AD-Epi-, and controls (p-value = N.S.).However, compared to controls, the AHI (p-value = .012for AD-Epi+ vs. controls) and ODI (p-value = .002for AD-Epi+ vs. controls) were significantly higher in the AD-Epi+ subgroup.The total REM sleep time was significantly decreased in both AD-Epi+ and AD-Epi-subgroups compared to controls (p-value < .001for AD-Epi+ vs. controls; p-value = .009for AD-Epi-vs.controls).REM% was decreased in the AD-Epi+ subgroup compared to controls (p-value = .006fAD-Epi+ vs. controls).The mean SE was lower in both AD subgroups (p-value = .007for AD-Epi+ vs. controls; p-value = .008for AD-Epi-vs.controls), and the median WASO was longer (p-value = .004for AD-Epi+ vs. controls; pvalue = .005for AD-Epi-vs.controls) in both subgroups compared to controls.The AwI was significantly higher in the AD-Epi-subgroup (pvalue = .012for AD-Epi-vs.controls).The AD-Epi-subgroup spent more absolute time in N1 sleep stage (p-value = .010for AD-Epivs.controls) and more relative time in N1 sleep stage than controls (p-value = .003for AD-Epi-vs.controls).There were no significant differences in sleep measurements between the AD-Epi+ and AD-Episubgroups (p-value = N.S.).

DISCUSSION
We performed a cross-sectional study to evaluate the presence of SEA and sleep alterations in AD.Our results show that patients with AD have (1) a three times higher prevalence of SEA, (2) a decrease in total time in REM sleep and %REM sleep, particularly in AD-Epi+, (3) a longer WASO, (4) a decreased SE, (5) more N1 sleep (6) a higher AwI, and (7) a higher AHI and ODI on TST, particularly in AD-Epi+ subgroup.
In this cross-sectional study with prolonged EEG recording in AD patients without a formal diagnosis of epilepsy, we did not capture seizures.However, SEA was present in 34%, almost three times more than in the control population (13%), congruent with previous work (Horvath et al., 2021;Vossel et al., 2016).Although 34% is a high prevalence, the median frequency of three spikes per night is rather low.
Epileptiform discharges were mainly detected in frontotemporal brain regions, reflecting their possible involvement in AD pathogenesis (Vossel et al., 2017) and mainly occurred during stage N2, which is known to promote epileptiform activity.In our AD cohort, 22% of spikes were during wakefulness, compared with only 8% and 9.9% in previous work by Horvath et al. (2021) and Vossel et al. (2016).There is no gold standard regarding when and how to search for epileptic activity in AD.We performed a prolonged overnight EEG recording, which is more sensitive than standard routine EEG recordings (Horvath et al., 2017;Vossel et al., 2013), but still has limited sensitivity in detecting mesial temporal lobe epileptiform activity as is proven by combined scalp-intracranial recordings (Abou Jaoude et al., 2022;Lam et al., 2017;Lieb et al., 1976).
In fact, one patient of our AD-Epi-subgroup had a positive Reutens Questionnaire for clinical seizure diagnosis, although we did not detect any epileptiform abnormalities on scalp EEG.Thus, some patients in the AD-Epi-subgroup may actually belong to the AD-Epi+ subgroup.
Intracranial and prolonged video-EEG recordings are cost-and laborintensive and often not tolerated in a more advanced disease stage.
The extrapolation of deep learning algorithms for detecting epileptiform activity on scalp EEG (Abou Jaoude et al., 2022) will be of great value in searching SEA biomarkers in AD.
Patients with AD in our study had a lower SE, more awakenings, more N1%, a longer WASO, and shorter REM%, confirming previous reports (D 'Rozario et al., 2020;Zhang et al., 2022).Our cross-sectional study focused on interactions between these sleep disorders and SEA in AD.Although we did not find statistically significant differences in sleep parameters between the AD-Epi+ and AD-Epi-subgroups, our findings on decreased REM sleep in AD warrant further discussion.
REM sleep is the least permissive state for epileptiform activity and is reduced in patients with epilepsy (Bazil et al., 2000, Sadak et al., 2022;Yeh et al., 2022), including patients with refractory epilepsy, compared to those with controlled epilepsy (Yeh et al., 2021)  duration in healthy controls of 98 min, which was significantly reduced with 33% to 66 min in AD.More specifically compared to controls, the mean total REM sleep duration dropped with 44% in the AD-Epi+ subgroup and 26% in AD-Epi-subgroup and REM% dropped with 35% in AD-Epi+ subgroup where this was only 19% in AD-Epi-subgroup.We might speculate that REM sleep in AD might be shorter because of an underlying undiagnosed epilepsy.Intracranial recordings and larger sample size studies are necessary to prove this hypothesis.The reliability of specific sleep staging with wearable devices is still quite low (Miller et al., 2022) and requires improvements before sleep parameters, like decreased REM%, could become a noninvasive, easy-to-measure biomarker to identify those patients at risk for/with SEA.The higher AwI and more time spent in N1 sleep was significantly higher in AD-Epi-subgroup compared to controls.We expected it would be higher in AD-Epi+ subgroup since nighttime temporal lobe seizures are associated with higher N1% (Bazil et al., 2000) and scalpnegative temporal lobe seizures, which occur during sleep, are often associated with awakenings (Lam et al., 2016).However, these values might be influenced by the relatively small and unequal sample sizes, where AD-Epi-subgroup was almost twice as large as the AD-Epi+ subgroup.
AD patients have a five times higher chance of having OSAS than cognitively unimpaired older adults (Emamian et al., 2016).The prevalence of OSAS was not significantly higher in our AD cohort; however, six AD patients had severe OSAS where this was absent in the control sample, which is reflected in higher ODI and AHI values.These observations were not reflected in the questionnaires, where ESS and PSQI scores were similar across patients and controls, and no subjects reported symptoms of severe daytime sleepiness or poor sleep quality.These findings possibly reflect the problem of underreporting of sleep problems in the elderly (Gordon et al., 2022)  intermittent hypoxia on amyloid and tau dynamics in cerebrospinal fluid and brain tissue (Bu et al., 2015;Holth et al., 2017;Holth et al., 2019;Ju et al., 2016;Liguori et al., 2017;Liguori et al., 2019;Motamedi et al., 2018;Osorio et al., 2014;Osorio et al., 2015;Spira et al., 2014).CPAP treatment may positively affect cognitive functions in AD (Ancoli-Israel et al., 2008;Liguori et al., 2021;Troussiere et al., 2014).Furthermore, there is a complex interaction between OSAS and epilepsy.OSAS is more prevalent in patients with epilepsy, and CPAP treatment is associated with better seizure reduction and beneficial effects on interictal EEG (Lin et al., 2017;Pornsriniyom et al., 2014).Moreover, there is evidence that the hippocampus, prone to epileptiform activity, might play a role in breathing patterns (Harper et al., 1998).We were interested if SEA and sleep-disordered breathing (SDB) were related in AD.There was no difference in prevalence of OSAS, AHI and ODI on TST between AD-Epi+ and AD-Epi-subgroups.
The increased AHI and ODI on TST in AD-Epi+ subgroup compared to controls supports a possible bidirectional relation between SDB and SEA in AD pathology, which can have diagnostic and therapeutic implications in the way of screening for SEA in AD patients with OSAS and vice-versa.
Our results point toward a hypothesis of a complex interaction between SEA and sleep in AD.Further research should focus on an integrated approach to the role of SEA and sleep in AD to (1) better understand the role of SEA and sleep/SDB in AD pathogenesis; (2) find biomarkers for identifying those patients who are at risk for SEA; and (3) to find those patients in the heterogeneous AD population where antiseizure medication or CPAP treatment may have a significant positive benefit on their cognitive functions.
Our study has several limitations.First, our sample size was relatively small.More studies are needed before generalizing these results.
Second, biomarker data were available for 16 AD patients, whereby another underlying pathology cannot be excluded.Third, our control population consisted of healthy volunteers without subjective cognitive complaints and MMSE scores between 27/30 and 30/30.No data were available for neuropsychological testing, neuroimaging, or biomarker status of our control sample.Therefore, we cannot rule out that some control participants had preclinical AD.Fourth, since two consecutive nights of PSG were difficult to achieve in the AD population, we recorded only one night of PSG.A first-night effect could have influenced our results.Last, we focused only on sleep macrostructure.
Focal epilepsy disrupts sleep spindles, which are essential for memory and learning (Schiller et al., 2022).Analyzing sleep microstructure could give more insight into the role of SEA and sleep in AD.

CONCLUSION
Our study showed a higher prevalence of SEA and sleep disturbances in AD.There were no significant differences in sleep macrostructure between AD patients with SEA and those without SEA on scalp EEG, probably influenced by the small sample size.However, REM sleep was most decreased in AD patients with SEA, which might constitute a potential biomarker for SEA in AD.The AHI and ODI on TST were higher in AD, specifically the subgroup with SEA.These findings suggest a complex interaction between sleep and epileptiform activity in AD.Further research, using concomitant intracranial recordings and wearables, should focus on an integrating approach to epilepsy and sleep in AD pathogenesis.
or antipsychotic medication.Although no formal diagnosis of epilepsy was made at moment of study participation, three patients with AD had a positive Reutens Questionnaire for clinical seizure diagnosis: one patient had episodes of confusional awakenings during sleep and episodes where she became pallor, hyperventilated and fainted; the second patient experienced the past year three episodes of staring with partial loss of the postural tone; and the third patient had experienced three episodes of staring with automatisms and dystonia of the right arm.No patient encountered an episode during study participation.ESS and PSQI scores were similar in AD and control groups.
Abbreviations: AHI = apnea-hypopnea index; AI = arousal index; IQR = interquartile range; N.S. = not significant; OAHI = obstructive apnea-hypopnea index; ODI = oxygen desaturation index; OSAS = obstructive sleep apnea syndrome; SD = standard deviation; SE = sleep efficacy; TIB = time in bed; TST = total sleep time; WASO = wake after sleep onset.Correction for age, medication and AHI on TST.Except for AHI, OAHI, and ODI on TST only for age and medication.*Significant p-values (≤.05).†Original values are represented, but statistical testing on square-root-transformed data.‡Original values are represented, but statistical testing on log-transformed data.
Clinical parameters of study population.
Polysomnography parameters controls versus AD.
. Sadak et al. (2022)reported that the diagnosis of epilepsy based on REM% prediction models performed similar to prediction based on the presence of IED on video-EEG and suggested that reduced REM sleep could be a biomarker for epilepsy.We observed an average REM sleep TA B L E 3 Polysomnography parameters controls versus AD-Epi ± and AD-Epi-subgroups.= not significant; OAHI = obstructive apnea-hypopnea index; ODI = oxygen desaturation index; OSAS = obstructive sleep apnea syndrome; SD = standard deviation; SE = sleep efficacy; TIB = time in bed; TST = total sleep time; WASO = wake after sleep onset.Correction for age, medication and AHI.Except for AHI, OAHI, and ODI on TST only for age and medication.
Abbreviations: AHI = apnea-hypopnea index; AI = arousal index; IQR = interquartile range; N.S. *Significant p-values.† Original values are represented, but statistical testing on square-root-transformed data.‡ Original values are represented, but statistical testing on log-transformed data.§ Significant p-value ≤ .025according to Bonferroni correction for multiple comparisons.