EEG measures the summation of excitatory and inhibitory postsynaptic electrical discharges generated from pyramidal neuronal activity in the neocortex [6] and consists of waveforms that can be characterized by their amplitude (i.e., the magnitude of waveforms measured in microvolts [μV]) and frequency (i.e., the number of cycles per second [Hz]) [7]. Typical EEG amplitudes range from 10 to 100 μVs, which is 100 times lower than those of an electrocardiography. EEG activity is often described in terms of EEG power, which is the square of the EEG amplitude (μV²). EEG frequencies are grouped into different bands, conventionally called the slow (< 1 Hz), delta (1–4 Hz), theta (5–8 Hz), alpha (9–12 Hz), beta (13–25 Hz), and gamma (26–80 Hz) bands [2].

The relationship between EEG activity in different regions of the neocortex can be described using coherence and phase shift. Coherence measures the degree of correlation between two EEG signals in a specific frequency band; a coherence value of 1 represents two perfectly correlated synchronous signals, whereas a coherence value of 0 represents no correlation [8]. Typically, coherence is used to measure the synchronization between the left and right sides of the brain, and increased coherence is observed with deeper sedation or anesthesia. Phase shift or phase synchrony can be used to determine the temporal relationship (connectivity) between EEG signals in different regions of the brain [9].

Anesthetics alter the electrical activity of the brain and generate distinct age- and dose-dependent EEG patterns, followed by topographic distributions, suggesting that multiple neural circuits contribute to inducing loss of consciousness [2,10]. Gamma aminobutyric acid (GABAergic) anesthetics (e.g., sevoflurane and propofol) increase EEG amplitude while decreasing EEG frequency. At surgical anesthesia levels, most EEG activity exists in the slow/delta and alpha bands. Alpha activity stems from the interaction between the cortical and thalamic pacemakers (corticothalamic network) [6], which can decrease with increased anesthetic doses and may not be present in younger infants, as discussed in the next section [8].

Infants undergo rapid changes in brain myelination and maturation, resulting in changes in EEG amplitude and frequency [8]. EEG in newborn children is dominated by slow oscillations that diminish in amplitude with age [6]. Markus et al. [11] measured frontal EEG in infants aged < 12 months under sevoflurane anesthesia and found that newborns had a much higher relative delta power (80%) than older infants aged 10–12 months (48%). Compared to older infants (4–6 months), newborns had lower relative alpha (4.6% vs. 14.4%) and beta (3.2% vs. 10.9%) power. De Heer et al. [12] found that in infants aged < 6 months under sevoflurane anesthesia, alpha coherence and beta coherence were absent, whereas theta and delta oscillations were present. In infants younger than 3 years under sevoflurane anesthesia, Cornelissen et al. [8] used multichannel EEG recordings to show that (1) slow/delta (0.1–4 Hz) oscillations were present in all ages, (2) theta (4–8 Hz) and alpha (8–12 Hz) oscillations emerged by 4 months, (3) alpha oscillations increased in power from 4 to 10 months, (4) frontal alpha-oscillation predominance emerged at 6 months, (5) frontal slow oscillations were coherent from birth until 6 months, and (6) frontal alpha oscillations became coherent at around 10 months, persisting into older ages. These EEG changes reflect developmental milestones in the maturation of the thalamocortical circuity and “highlight the importance of developing age-dependent strategies to monitor the brain states of children receiving general anesthesia” [8].

In children under sevoflurane anesthesia, Akeju et al. [13] assessed changes in frontal EEG power spectra and coherence as a function of age and found that EEG power significantly increased from infancy through approximately 6 years, subsequently declining to a plateau at approximately 21 years. Alpha (8–13 Hz) coherence, an EEG feature associated with sevoflurane-induced unconsciousness in adults, was absent in patients aged < 1 years. In a study by the same group, Lee et al. [14] showed that under propofol anesthesia, the total EEG power (0.1–40 Hz) peaked at approximately 8 years and declined with increasing age. In children aged > 1 years, propofol-induced EEG was qualitatively similar regardless of age, featuring slow and coherent alpha oscillations. In infants aged < 1 years, frontal alpha oscillations were not coherent.

Changes in EEG due to variances in anesthetic levels have been evaluated in children. Rigouzzo et al. [15] described EEG properties in children aged 5–18 years over a wide range of anesthetic concentrations: (1) propofol using target-controlled infusion targets of 2–6 μg/ml in 1 μg/ml increments or (2) expired sevoflurane (eSevo) 1%–5% in 1% increments. Propofol induced dose-dependent EEG suppression of higher-frequency oscillations, whereas eSevo > 3% was associated with an increase in higher-frequency oscillations. Cornelissen et al. [16] used raw EEG data to report the relationship between sevoflurane concentrations and clinical signs of emergence in children aged 0–3 years. Frontal slow-delta (0.1–4 Hz) oscillations were present from 2% eSevo throughout emergence in all children. In children aged > 3 months, frontal alpha EEG oscillations were present at 2% eSevo and disappeared at 0.5%. The time from the disappearance of alpha oscillations and onset of body movement was 2.2 min (95% CI [0.6, 3.7]). In 99% of the patients, body movement occurred within 5 min of the loss of alpha oscillations. In children aged < 3 months with decreasing eSevo, frontal alpha power decreased with a simultaneous but transient increase in beta oscillations (13–30 Hz).

During changes in consciousness under general anesthesia, alterations in local EEG oscillations are accompanied by organized changes in the connectivity patterns between various brain regions, known as functional connectivity. Functional connectivity has been explored as an index of anesthetic state transitions and has been shown to discriminate between anesthetic-mediated changes in states of consciousness. This is important in children aged < 10 months as the power-frequency relationship-based EEG cannot always reliably distinguish between wakefulness and unconsciousness [17].

Puglia et al. [18] recorded a 16-channel EEG in children aged 8–16 years and observed changes in functional connectivity associated with anesthetic state transitions across multiple regions and frequency bands. The baseline period, prior to any sedation or anesthesia, was compared to the anesthetic maintenance period, approximately halfway between surgical incision and cessation of anesthesia, with an age-adjusted MAC value > 0.7 and a change of < 0.1%. Functional connectivity was estimated using a weighted phase lag index, which is a measure of the phase synchronization of two signals. If two EEG signals are completely synchronized (phase-locked), then the weighted phase-lag index is 1. Conversely, if the phase relationship between two EEG signals is random, the weighted phase-lag index would be lower. If there is no phase difference, the value is zero. From baseline to maintenance, an increase in connectivity between prefrontal–frontal regions was noted for the alpha frequency band: median (Q1, Q3) 0.07 (0.05, 0.10) vs. 0.47 (0.29, 0.61), P < 0.001 and theta frequency band: 0.04 (0.03, 0.05) vs. 0.40 (0.25, 0.49), P < 0.001 along with a decrease in parietal–occipital alpha connectivity: 0.17 (0.15, 0.24) vs. 0.09 (0.06, 0.13), P < 0.001. The authors concluded that general anesthesia in children correlates with changes in functional connectivity patterns. Unlike in adults, where functional connectivity can undergo structured transitions during the stable maintenance phase, even after controlling for surgical stimuli [19], functional connectivity in children did not exhibit the same transitions during the maintenance phase and was consistently dominated by theta and alpha prefrontal–frontal and alpha frontal–parietal connectivity.

In an exploration of functional connectivity and network topology in the developing brain, Desowska et al. [17] used a weighted phase lag index and found that sevoflurane-based anesthesia modulated functional connectivity in children aged 0–3 years. Functional connectivity was reduced in delta oscillations between anesthesia and wakefulness. Preserved alpha connectivity, notably in the frontal and frontoparietal connections, emerged at around 10 months. In the youngest infants (0–6 months) who lack robust alpha oscillations and exhibit little difference in power spectra between different anesthetic states, functional connectivity analysis could be used to identify differences between wakefulness and unconsciousness. In another study by the same group that focused on young infants aged 0–4 months, Pappas et al. [20] showed that slow-wave connectivity decreased under sevoflurane administration, particularly between the frontal and parietal areas.

Phase-amplitude coupling (PAC), in which the phase of the slower wave modulates the amplitude of the higher-frequency wave, has been studied as another method for differentiating the depth of consciousness under anesthesia [21,21]. Thalamocortical PAC is believed to be a process by which general anesthesia prevents information transmission across the brain by disrupting neuronal dynamics through coupled rhythmic oscillations [23]. Liang et al. [22] analyzed PAC as a function of age in children aged 0.6–18 years and in adults undergoing propofol induction and sevoflurane maintenance. In infants aged < 1 years, the PAC pattern between slow-alpha and delta-alpha oscillations was absent. In children aged > 1 years, the PAC modulation strength between delta-alpha oscillations increased with age. Zakaria et al. [21] reported similar delta-alpha PAC characteristics in children aged 10 months to 3 years under sevoflurane anesthesia. PAC was also absent in children aged < 1 years, presumably because of a lack of strong alpha oscillations.

The DSA can be monitored in conjunction with raw EEG waveforms and the SEF95 to guide anesthesia management in children [7,10,25,40]. DSA monitoring allows clinicians to directly and accurately visualize the effects of various anesthetic drugs on a patient’s EEG without relying on the pEEG number, enabling them to make better clinical inferences and titrate anesthesia doses more precisely to meet each patient’s requirements at any given time. This is particularly important in children with less predictable anesthetic requirements, such as those receiving total intravenous anesthesia (TIVA), neuromuscular blockers, or a combination of anesthetics with different mechanisms of action; those with atypical neurodevelopment or altered levels of consciousness prior to anesthesia; and those with hemodynamic instability as they need their anesthetic doses titrated more precisely to prevent over- or undersedation [25].

Figs. 3–5 illustrate examples of how to titrate anesthetic doses in response to oversedation, undersedation, and for young infants.

Whereas the dosing examples above were all for a sevoflurane-based anesthetic, propofol-based TIVA has gained popularity recently as an environmentally friendlier alternative to sevoflurane, with a lower associated incidence of laryngospasm, emergence agitation, and postoperative nausea and vomiting. Unlike eSevo concentration monitoring, the propofol concentration (Ce) in the brain cannot be easily monitored, making dosing less accurate and predictable. EEG has been suggested as a reliable dynamic surrogate for brain propofol Ce to guide propofol TIVA dosing [41]. However, EEG-guided TIVA is uncommon in pediatric patients, and EEG curricula are not commonly taught in anesthesia training programs [4]. To take advantage of the clinical benefits of propofol TIVA and improve the dosing accuracy with EEG, Yuan et al. [44] and Jones Oguh et al. [45] initiated a quality improvement project to train their division on using EEG to guide TIVA dosing. The group used a combination of intraoperative teaching, didactic lectures, and quality improvement metrics to develop a sustainable, effective, and scalable program. The group emphasized the use of a combination of raw EEG, SEF95, and DSA data to titrate propofol and provided algorithms for use during induction, maintenance (Fig. 6), and emergence. Over the 12 month period, 79% (62/79) of the clinicians completed the EEG training program, and their knowledge test scores improved from 38% before training to 59% after training (P < 0.001). Four Plan-Do-Study-Act (PDSA) cycles were initiated, and the prevalence of using EEG to guide TIVA cases increased from 5% to 75% [45]. The incidence of perioperative emergency activation did not change significantly, while the emergence time for EEG-guided TIVA cases was significantly longer, the difference was not clinically significant (18 vs. 16 min, P < 0.001).

Since the advent of sevoflurane for clinical use in pediatric anesthesia, reports have described involuntary behaviors suggestive of seizure-like movements [46]. Initial case reports and subsequent studies have described these movements, with some studies recording EEG to detect brain wave correlates [47]. EEG studies have utilized nomenclature to describe patterns that have been defined in the neurology-epileptology literature but are unfamiliar to most anesthesiologists [47]. During induction, a prevalence of 30% (95% CI [27.2, 34.5]) has been reported for epileptiform changes [47], while other studies have reported estimates from 19% to 60% [48]. An important consideration to account for this wide range in prevalence is the use of heterogeneous EEG outcome classifications and variability in the number of channels recorded [47]. A standard definition, which is a simplified version of the International League Against Epilepsy (ILAE) and the American Clinical Neurophysiology Society (ACNS) classifications, has been proposed to help with future classifications. These classifications are presented verbatim below.

While some EEG changes are abnormal and associated with certain clinical epilepsy syndromes, their clinical significance in healthy children under anesthesia with no history of neurological diseases is unclear and warrants further research. Therefore, any extrapolation from the anesthetized state should be done with caution.

Spinal anesthesia is achieved by administering sodium (Na⁺) channel antagonists (e.g., bupivacaine) in the subarachnoid space, which produces regional, infraumbilical surgical antinociception [49] with significantly less apnea, hypoxia, and hypotension than that with general anesthesia [50,51]. Na⁺ channel antagonism in the lower spinal regions is not known or expected to produce sedation or decreased level of consciousness. However, the administration of spinal bupivacaine in young infants is associated with an ostensibly decreased level of arousal in the absence of any sedative-hypnotic agent. The neural basis for these spinal anesthesia-associated sedative effects and their implied central effects at an early age have been investigated. EEG data obtained from infants receiving spinal bupivacaine demonstrate distinct, measurable changes consisting of voltage attenuation, frequency slowing with significantly increased delta (0.5–5 Hz) spectral power, and the appearance of sleep spindle complexes of 12–14 Hz resembling normal physiological sleep [49,52]. Sleep spindles are a hallmark of Stage 2 non-rapid-eye-movement sleep, suggesting that spinal anesthetics may confer a more physiological sleep state than traditional sedative-hypnotic effects during general anesthesia [49,52]. Age-related changes in sleep spindles during spinal anesthesia generally correspond to developmental changes reported in the sleep EEG literature, supporting a sleep-related mechanism for the apparent sedation observed during infant spinal anesthesia [53]. Thalamocortical circuitry maturation is thought to be involved in the generation of sleep spindle oscillations, which begin to appear at approximately 46–48 weeks of gestational age [52,53].

Isoelectric EEG reflects electrical inactivity in the cortex and can be observed in states of brain pathology (e.g., hypothermia and coma). Intraoperative isoelectric EEG is concerning in both adults and children under anesthesia, as it is thought to represent an “over-anesthetized” brain. Isoelectric EEG is associated with worse long-term neurological outcomes in neonates undergoing cardiac surgery [54]. In adults, intraoperative low-voltage EEG has been associated with postoperative delirium [55]. In children aged < 3 years undergoing sevoflurane anesthesia, Cornelissen et al. [56] found that 51% experienced isoelectric/discontinuity EEG, which was mostly observed during the first 30 min of anesthesia and was associated with younger age. High induction doses of propofol were administered to children with discontinuity events. In a related study from the same group, Agrawal et al. [37] found that in children aged > 4 months, a decrease in high-frequency activity occurred tens of seconds before each discontinuous event and a decline in spectral edge frequency occurred in the seconds immediately following each discontinuity.

Yuan et al. [57] observed isoelectric EEG events in 63% of children aged 0–37 months undergoing sevoflurane or propofol anesthesia. In a follow-up 15-center international study, the observed occurrence of isoelectric EEG events was 32% in 648 children aged < 37 months, but varied significantly between sites (9%–88%), suggesting that variances in anesthesia practice could contribute to the occurrence of isoelectric EEG [58]. Isoelectric events were associated with younger age, endotracheal tube use, propofol bolus for airway placement, and higher eSevo, and were less likely to occur when muscle relaxants were used for intubation. Importantly, isoelectric events were associated with moderate and severe hypotension between induction and incision (odds ratio [OR]: 3.5–4.6) and during surgical maintenance (OR: 3.6–7.1). Infants aged 0–12 mo who experienced isoelectric events had lower pediatric quality-of-life scores at baseline and postoperatively. The authors concluded that sicker infants were more prone to isoelectric EEG and relying on traditional MAC dosing may be “overdosing” the infant brain.

To determine the sevoflurane concentration associated with the occurrence of isoelectric EEG in 50% of children (MAC_isoEEG), He et al. [59] enrolled children aged 3–8 years and allocated them to one of three groups: (1) sevoflurane in 100% O₂, (2) sevoflurane in 50% N₂O mixed with 50% O₂, and (3) sevoflurane in 100% O₂ with a fentanyl bolus (3 μg/kg). The MAC_isoEEG was determined using Dixon’s up-down method after a 15-min period of steady eSevo. The MAC_isoEEG in the 100% O₂ group was (median [95% CI]): 5.30% (5.12, 5.48), 5.83% (5.67, 5.99) in the 50% N₂O group, and 5.37% (5.21, 5.53) in the fentanyl group. The authors concluded that the addition of 50% N₂O modestly increased the MAC_isoEEG, whereas 3 μg/kg fentanyl had no effect on the MAC_isoEEG.

In older infants and children, propofol and sevoflurane generate an alpha frequency (8–12 Hz) oscillation that represents coordinated firing activity of neurons between the thalamus and cerebral cortex [60–62]. The degree of alpha oscillatory power during anesthesia decreases with age and is correlated with cognitive function, suggesting its capacity as a neurophysiological marker with potential clinical significance [63,64]. Plausibly, an association could be found between the alpha oscillatory power and low-voltage EEG patterns, such as discontinuous or isoelectric EEG, during anesthesia. Low alpha power in adults has been associated with the propensity for burst suppression and could be a neurophysiological phenotype for a “vulnerable brain” [65]. Similarly, in an observational cohort study of 54 infants, the degree of alpha oscillatory power induced by sevoflurane anesthesia was associated with the prevalence of EEG discontinuity, with every decibel increase in alpha power associated with a 49% reduction in the odds of developing low-voltage EEG (OR: 0.51, 95% CI [0.30, 0.89], P = 0.02), even after adjusting for chronological age and propofol administration [66]. No differences in alpha power in the baseline awake state were seen before anesthesia among the infants, and this effect could not be explained by differences in anesthetic dosing. These findings suggest hidden brain circuit properties could be unmasked under anesthesia. Anesthetic-induced alpha power is a putative marker of thalamocortical circuit development during the first year of life and can be useful in identifying young patients who have a greater chance of developing low-voltage patterns, with potential future applications in anesthesia management [66].

EEG monitoring provides insight into individual patients’ brain responses to anesthetic drugs, which are dependent on a variety of factors, including age, neurodevelopmental stage, disease state, concomitant medications, hemodynamic status, and surgical stimulation [10,25]. Multiple studies have demonstrated the utility of EEG monitoring in pediatric anesthesia [40,41,67,68], particularly during cardiac surgery, when cardiopulmonary bypass, induced hypothermia, and hemodynamic instability render the anesthetic requirements much less predictable [69,70]. EEG-guided anesthesia allows for accurate and timely visualization of brain responses and can complement routine monitoring in children to optimize anesthesia delivery and improve patient safety and experience. Recent studies have demonstrated that EEG monitoring may improve periprocedural outcomes in children undergoing anesthesia.