This was a retrospective, single-centre observational study and did not affect routine care of participants. Baseline demographics included the age and gender of each participant. The current study utilised the Strengthening Reporting of Observational Studies in Epidemiology (STROBE) guidelines to optimise quality of reporting (Supplementary Material) [26].

There was no age restriction for patients. Data from EEG reports were retrospectively collected for patients who had undergone both a routine EEG and sleep-deprived EEG study during a five-year period (January 2018 to January 2023). The choice of test for each patient was pre-specified by the referring practitioner and may have originated from an inpatient, emergency department or outpatient setting. Referrers were not necessarily required to use the same referral form. Participants must have had both studies performed at Liverpool Hospital, a major tertiary referral and teaching hospital in south-western Sydney, in the state of New South Wales, Australia.

In situations where patients had more than two EEG studies, the routine and sleep-deprived EEG studies with the shortest inter-study interval were selected.

EEGs were recorded in accordance with the International 10–20 system. Duration of routine and sleep-deprived EEG recording time was 20 minutes and 60 minutes respectively. A Natus brain monitor EEG amplifier PK1274 (Natus Medical Incorporated, Middleton, MI, USA) was utilised and electrode placement followed the international 10–20 system. The protocol, at this institution, for sleep deprivation was for the patient to have undertaken 24 hours of sleep deprivation—any deviation from this protocol was noted. Patients were excluded if they had experienced at least four hours of sleep prior to the sleep-deprived EEG.

Routine triggering manoeuvres, including hyperventilation and photic stimulation, were utilised in both the routine and sleep-deprived studies. Major tertiary-referral hospitals in Sydney with epilepsy services were contacted to assess their protocols.

EEGs were reported by experienced clinical neurologists from the same institution who had level 2 or 3 Australian and New Zealand Association of Neurologists (ANZAN) certification for EEG training and education. Each of the reporting neurologists had at least 10 years of experience in clinical and EEG reporting practice. Inter-ictal epileptiform activity was identified by the reporting neurologist in accordance with previously specified criteria [27]. Topography and clinical phenotype of seizure semiology were not analysed. Neurophysiology technologist staff assisted the primary author with locating EEG data from the Liverpool Hospital Neurophysiology Department.

The primary author de-identified and collected data from each EEG report including: date of study; whether the study was normal or abnormal; presence and type of epileptiform activity; state of patient during epileptiform activity; details regarding non-epileptiform activity; any prescribed anti-seizure medications (ASMs); time elapsed between the study and seizure; speciality of the referrer; amount of sleep deprivation; and time elapsed between routine and sleep-deprived studies. In situations where there was suspicion of possible epileptiform activity, rather than confirmed epileptiform activity, this was also noted. If available, referral paperwork was analysed to provide these data.

The primary outcome was the yield of epileptiform activity for the same patient when comparing the routine to the sleep-deprived EEG. Sub-group analyses were performed based on age, gender, referrer status and order of study requests. If available, comparisons were made with reference to other collected data points.

The primary author analysed the data and applied statistical tests with the “R” statistical package (4.4.2, R Foundation for Statistical Computing, Vienna, Austria). Descriptive statistical analyses were performed to summarise the data. For continuous variables, median and interquartile range (IQR) were reported. McNemar’s test was used to compare between routine and sleep-deprived EEG studies with regards to the presence of epileptiform activity and undertake subgroup analysis. Fisher’s Exact Test was used to evaluate the effect of prespecified subgroups on diagnosis of epileptiform activity for each type of EEG. A p value of less than 0.05 was chosen as the significance level for all statistical tests.

There were 116 patients included in the study. Fifty four percent of patients were male and 80% of patients were aged at least 18 years of age. There were no patients under 13 years of age who underwent both a routine and sleep-deprived EEG during the study period. A total of 114 patients (98%) completed the prescribed 24 hours of sleep deprivation. In this study, 97 patients (84%) underwent the routine EEG prior to the sleep-deprived EEG. The group demographics are further outlined in Table 1.

Documentation as to whether a patient was prescribed an ASM was present more often in sleep-deprived EEG referrals, compared to routine EEG referrals (11% versus 3%). There were few EEG referrals that documented the time since the last suspected seizure (5% of the routine EEGs and 3% of the sleep-deprived EEGs). Documentation as to the identity of the referring physician was available for 33% of the routine EEGs and 42% of the sleep-deprived EEGs.

EEGs were reported as being within normal limits in 54% of routine and 60% of sleep-deprived EEGs. Non-epileptiform abnormalities were identified in 36% of routine and 28% of sleep-deprived EEGs.

Epileptiform activity was identified in 24 patients (21%) when the results of both routine and sleep-deprived EEG studies were included. There was no significant difference in the number of EEGs with epileptiform activity when comparing the routine and sleep-deprived group studies (12% versus 13% respectively, p = 1.0). Five patients (4.3% of the whole cohort) had epileptiform activity present on both routine and sleep-deprived EEGs.

Test order was found to significantly affect the overall diagnostic yield of undertaking both tests. Sleep-deprived EEGs were significantly more likely to demonstrate epileptiform activity if undertaken prior to the routine EEG (7/19 (37%) versus 8/97 (8%), p = 0.003). Routine EEGs were significantly more likely to demonstrate epileptiform activity if undertaken following a sleep-deprived EEG (6/19 (32%) versus 8/97 (8%), p = 0.011).

All eight patients with epileptiform activity on initial routine EEG subsequently had a sleep-deprived EEG undertaken. All seven patients, with epileptiform activity on an initial sleep-deprived EEG, subsequently had a routine EEG undertaken. Amongst patients who had epileptiform activity identified on their initial routine or sleep-deprived EEG, the proportion of subsequent EEGs with epileptiform activity did not differ significantly between the groups (2/8 (25%) versus 3/7 (43%) respectively, p = 0.61).

Epileptiform activity was identified on routine EEG but not sleep-deprived EEG in nine patients (7.8%) and the reverse was true for 10 patients (8.6%). Patients with an initial negative sleep-deprived EEG were more likely to have epileptiform activity on the subsequent EEG in comparison with patients who had an initial negative routine EEG (3/12 (25%) versus 6/89 (7%)), although this difference was not significant (p = 0.072). These data are further categorised by age group in Fig. 1.

There was a significantly higher proportion of patients aged 18 years with epileptiform activity on routine EEG compared to those aged $\ge$18 years (8/23 (35%) versus 6/93 (6%) respectively, p = 0.001). Sleep-deprived EEGs were also significantly more likely to demonstrate epileptiform activity in patients aged 18 years compared to $\ge$18 years (8/23 (35%) versus 7/93 (8%) respectively, p = 0.002). These results are further outlined in Table 2.

In a subgroup analysis, when comparing routine and sleep-deprived EEGs, there was no significant difference in the identification of epileptiform activity based on age group, gender and known identity of referrer (Table 3). More details as to the nature of epileptiform activity are stratified according to age group in Table 4.

Amongst all patients, the median time elapsed between each pair of EEGs was 53 days (IQR 16–154 days). Regarding patients who underwent a routine EEG as the first study, the median time elapsed until a subsequent sleep-deprived EEG was 45 days (IQR 10–111 days). Regarding patients who underwent a sleep-deprived EEG as the initial study, the median time elapsed until a subsequent routine EEG was 192 days (IQR 49–414 days) (Fig. 2). Due to the relatively low sample size, any effect of variation in time between paired studies was difficult to further characterise.

Of the tertiary-referral hospitals that responded with written protocols, heterogeneity existed amongst instructions for sleep-deprivation. Variations included: staying awake the entire night prior; losing at least half of normal sleep; allowing sleep between 1 and 5 AM; and age-specific criteria. The duration of sleep-deprived recordings varied from at least 40 minutes and up to 90 minutes in some cases.

The estimation of the yield of sleep-deprived EEGs was optimised in the present study in that 98% of patients achieved 24 hours of sleep deprivation. The inclusion criteria for less than four hours of sleep-deprivation was established to reduce any effect of poor adherence to sleep-deprivation instructions: the criteria for sleep-deprivation in the present study was equal to or more strict than multiple sleep-deprived protocols of some other nearby tertiary-referral centres. In this way, although two patients in the present study did not achieve a full 24 hours of sleep-deprivation, they had sufficient sleep-deprivation to allow for optimised external validity of our results.

The present study is potentially limited in that it did not explore the time spent awake versus asleep. Association of epileptiform activity with stages of sleep was not analysed. The added value of this information is questionable in that there is mixed evidence as to whether sleep-deprivation or sleep itself is the trigger for epileptiform activity in sleep-deprived patients [10].

There is a possibility that the longer duration of sleep-deprived EEG studies, in comparison to the routine studies, has contributed to an overestimation of their yield by virtue of more prolonged monitoring. The external validity of these results is supported by the typically longer duration of sleep-deprived recordings as noted in the literature [10].

Adding external validity to these results is the fact that Liverpool Hospital is a large tertiary centre, with a significant multicultural population. In a government publication, 49% of people, within the local health district, reported to speak a language other than English at home [31]. The most spoken languages, other than English, in the district were Arabic, Vietnamese, Mandarin, Cantonese and Greek [31]. This may have affected the level of adherence to the performance of the study but reflects real-world experience.

The significant methodological heterogeneity in the literature represents a challenge in comparing these results to other studies [10]. Although the current study only included patients who had undergone both routine and sleep-deprived EEG testing, some other studies in the literature have not necessarily required both EEG techniques to be undertaken on the same participant [16, 17, 21]. These studies also did not specify the seizure date in relation to the date of the EEG study [16, 17, 21]. A retrospective study, of adult patients who underwent emergency EEG, following a first unprovoked seizure, concluded that the diagnostic yield was highest when performed within 16 hours after the onset of the event [28]. One study, of 300 consecutive adults and children with unexplained seizures, found epileptiform abnormalities in 51% of patients who had an EEG within 24 hours versus 34% of those with a later EEG [32].

The retrospective nature of data collection in this study has potentially introduced confounders to the results. EEG abnormalities were identified from their respective reports so the threshold for reporting of epileptiform and non-epileptiform findings may have differed between individual reporting neurologists, thus introducing bias. The relatively small sample size and large variability in time intervals between pairs of EEG studies may also impact upon the results. Limited data were available as to the date of last known seizure and prescribed ASMs, factors that can influence the presence and nature of epileptiform activity on EEG [33].

Aside from sleep deprivation, there are other techniques outside the scope of this study that may increase the yield of identifying epileptiform activity. One such technique is known as “ambulatory-EEG monitoring” [34], in which prolonged EEG testing, over hours to days, can occur in an outpatient setting. This has its own limitations including accessibility and higher costs to patients. One prospective study [35] assessed patients with a historical suggestion of epilepsy but a normal or non-diagnostic initial routine EEG and found a greater benefit with ambulatory EEG, when compared with sleep-deprived EEGs. Contrasting this was a retrospective cohort study that found a similar diagnostic accuracy of a subsequent sleep-deprived or ambulatory EEG (sensitivity 45% versus 63% respectively) in patients who had a normal routine EEG after experiencing a first unprovoked seizure [36]. The 2022 National Institute for Health and Care Excellence guidelines recommend for consideration for a subsequent ambulatory EEG in some cases, after both a normal routine and sleep-deprived EEG, where the diagnosis of epilepsy remains obscure but probable [25].