1 INTRODUCTION

The relationship between anesthesia and the electrical activity of the brain has been described as early as in the 1940's.¹ Development in this field has been extended to the investigation of the relationship between the brain activity and depth of anesthesia.² Research of the brain activity as a surrogate to monitor the depth of hypnosis during anesthesia has mostly focused on the spontaneous electroencephalogram (EEG). However, the spontaneous EEG does not mature before adult age, which could influence its reliability in children.³ Although Mid-Latency auditory evoked potentials (MLAEP) are electrical brain activity, they are fundamentally different from the spontaneous EEG. These potentials are electric responses of the brain to an auditive stimulus. They occur at about 8–50 ms after the stimulus, which is after the auditory brainstem response (0–8 ms) and before the late cortical response (>50 ms). Needless to say, an impaired auditory pathway and/or inadequate delivery of the acoustic stimulus (poor fit of the earbuds in children) will make the recording of a MLAEP unreliable or even impossible.

A typical MLAEP waveform consists of 3 troughs (labelled with N for negative polarity) and 2 peaks (labelled with P for positive polarity) of a few microvolts (Figure 1). To put in perspective, the EEG is about ten times stronger and the ECG is about 100 times stronger. The weak signal of the MLAEP makes it susceptible to noise and diminishes its signal quality. The peaks and troughs are commonly labeled as N₀, P₀, Na, Pa and Nb and occur, in the awake adult, at about 9 ms, 12 ms, 16 ms, 25 ms and 36 ms, respectively, after the application of an auditive stimulus.⁴ The MLAEP is believed to be generated from the medial geniculate body and primary auditory cortex.⁵ In anesthetized adult patients, the peaks and troughs decrease in amplitude and the interval when they occur, which are called latencies, increases.^5-9 Myelination of the brain in de developing child strongly influences evoked potentials, this makes it difficult to elicit these in infants having an age of less than 1 year and very difficult in infants less than 6 months of age. MLAEP mature after the first decade of life, around 10–12 years.^10-13

Mid-Latency auditory evoked potentials could be a parameter to measure the depth of hypnosis in anesthetized children. This narrative review summarizes the current literature concerning the use of MLAEP and its implications in anesthetized children.

2 METHODS

This study adheres to the Preferred Reporting Items for Systematic Reviews and MetaAnalysis (PRISMA) guidelines, since it was initiated as a systematic review. A literature search was performed with the assistance of an experienced librarian of the medical library (W.B.) at Erasmus University Medical Center in Rotterdam. The databases of Embase, Medline, Web-of-science, CINAHL, Cochrane, PubMed publisher and Google Scholar were consulted on April 30 2020. The query consisted, but not solely, of the following search terms: “evoked potential”, “anesthesia” and “children”. A detailed description of the search query per database can be found in Appendix S1.

Manuscript titles were screened for relevance by two authors (Y.C. and I.H.). Any type of clinical study investigating MLAEP during general anesthesia in children with any type of outcome measurements was considered relevant. Manuscripts published in languages other than English or Dutch, case reports and review manuscripts were excluded from analysis. When no consensus could be reached between the two authors about the relevance of a study, the abstract and/or full text was reviewed. When needed a third author (F.W.) was consulted to resolve the matter.

The quality of the studies was assessed with the “Quality Assessment of Diagnostic Accuracy Studies v 2” (QUADAS-2) tool.¹⁴ This tool systematically screens the risks of bias and applicability of the studies included in a systematic review in four key domains: patient selection, index test, reference standard, and flow and timing of the study. As recommended in the background document, the tool was tailored to suit our review question.¹⁵ The assessment for the “reference standard” was omitted, since all types of outcome measurements were marked as relevant (eg, MLAEP compared to clinical hypnotic depths, anesthetic use, hypnosis index of other types of monitors).

3 RESULTS

The search resulted in 1539 manuscripts of which 891 remained after removal of duplicates. Manual screening of titles, keywords and abstracts resulted in 45 manuscripts. A careful review of the remaining studies revealed that 15 included only adult patients, 6 were written in a language other than Dutch or English, four concerned patients from the pediatric intensive care unit, one was a case report, one investigated children without pharmacological sedation and in three cases the full text was unavailable. This resulted in the inclusion of 15 studies which were considered relevant for analysis in this review (Appendix S2). Due to the limited number of included studies and varying (primary) objectives, a meta-analysis was deemed futile.

3.1 Quality assessment of studies

Table 1 summarizes the quality of the studies according to the QUADAS-2 tool. Since none of the manuscripts described a randomized or blinded patient selection method, the risk for bias due to patient selection was rated “unclear” for all studies. The study by Alvarez et al. was only published as an abstract and was therefore missing details concerning “patient selection” and “flow and timing”.¹⁶ Accordingly, these domains were rated unclear for risk as well as for applicability concerns. Depth of hypnosis was assessed by three studies with clinical parameters, for example, movement and/or vocalization, without the use of validated tools.^17-19 In one study the same researcher recorded the index values and assessed the depth of hypnosis.²⁰ For these four studies, the risk of bias in the reference standard were rated as unclear. Two studies were rated unclear for their risk of bias in the domain of “Flow and Timing”, because one study was terminated prematurely due to change in the anesthesia practice ¹⁸ and the other failed to generate data from 4 of the 14 patients.²¹ Seven studies allowed premedication for their patients and were therefore rated as “unclear” for applicability concerns of patient selection.^{17, 18, 20-24} Studies were conducted with 3 different commercially available MLAEP hypnosis monitors or unprocessed MLAEP for analysis. Because each MLAEP derived hypnosis monitor has its own unpublished algorithm, it is unclear how different devices relate to each other. Therefore, the applicability concerns of the index test (MLAEP monitor) were rated “unclear” for all manuscripts.

TABLE 1. Quality evaluation of studies according to Quality Assessment of Diagnostic Accuracy Studies v 2 (QUADAS-2) tool

Study	Risk of bias				Applicability concerns
	Patient selection	Index test	Reference standard	Flow and timing	Patient selection	Index test
Liao et al. 2001
Weber et al. 2004
Alvarez et al. 2004
Weber et al. 2005
Ironfield et al. 2007
Disma et al. 2007
Daunderer et al. 2007
Blussé van Oud-Alblas et al. 2008
Blussé van Oud-Alblas et al. 2008
Feuerecker et al. 2011
Kuhnle et al. 2013
Cheung et al. 2013
Cheung et al. 2014
Cheung et al. 2018
Blussé van Oud-Alblas et al. 2019

Note

• Low Risk Unclear Risk High Risk.

3.2 MLAEP extraction

Unprocessed MLAEP (measurement of the actual latencies), were used as the index test by three studies.^{17, 21, 22} The remaining 12 studies used commercially available depth of hypnosis monitors. Three of the studies used the A-line monitor (Danmeter A/S, Odense, Denmark).^{19, 20, 23} To extract a reliable MLAEP from the EEG, measurements must be repeated several times (usually 250 to 1000 times). The A-line monitor uses the already collected data from the same patient to compute the MLAEP waveform faster (after 15 repetitions), that is, an autoregressive with exogenous input model.²⁵ This results in the A-line Autoregressive Index (AAI) as a measurement for the depth of hypnosis.²⁶

The AEP monitor/2 (Danmeter A/S, Odense, Denmark) is the successor of the A-line monitor and was used in 6 studies.^{16, 18, 24, 27-29} It circumvented situations in which the depth of hypnosis could not be measured due to absent or low-quality MLAEP (due to interference or (too) deep anesthesia) by analyzing the spontaneous EEG instead. The computed index value was called the composite A-line Autoregressive Index (cAAI).²⁸

The studies by Cheung et al. were conducted with the aepEX PLUS monitor (Medical Device Management Ltd., Braintree, Essex, UK).^30-32 Instead of an autoregressive model, it applied the more conventional “moving time averaging” technique to extract the MLAEP and to compute the aepEX index. This technique requires 256 repeated measurement to extract an entirely new MLAEP waveform, which takes about 37 s. Instead of computing an entirely new MLAEP waveform every 37 s, the MLAEP and aepEX index is updated every 0.3 s by discarding the first (oldest) measurement and adding the last (newest) measurement.³⁰

Seven studies using a commercially available MLAEP monitor reported the average index values observed during different depths of hypnosis.^{20, 23, 27, 28, 30-32} These values are illustrated in Figure 2. The primary objective of these seven studies was to evaluate the performance of a MLAEP monitor to detect different levels of depth of hypnosis in children (Table 2).^{20, 23, 27, 28, 30-32}

TABLE 2. Summary of manuscripts investigating the relationship between MLAEP and depth of hypnosis

Study	Outcome	Results	Missing data due to monitor failure
Weber et al. 2004	pka of AAIb during awake versus LMAc insertion (mean ± SE)	0.99 ± 0.00	Not reported
	pk of AAI during awake versus eye closure (mean ± SE)	0.77 ± 0.08
	pk of AAI during eye closure vs LMA insertion (mean ± SE)	0.83 ± 0.07
Disma et al. 2007	Correlation between AAI and UMSSd	r: −.8237 (p: <.0001)	Not reported
	Correlation between AAI and heart rate	r: .2382 (p: <.05)
Blussé van Oud-Alblas et al. 2008	pk of cAAIe versus UMSS (mean ± SE)	0.90 ± 0.08	26 data points (total 325 included)
	pk during induction of cAAI versus UMSS (mean ± SE)	0.84 ± 0.08
	pk during emergence of cAAI versus UMSS (mean ± SE)	0.74 ± 0.02
Blussé van Oud-Alblas et al. 2008	pk during induction of cAAI versus UMSS (mean ± SE)	0.61 ± 0.1	1 patient out of 21 patients
	pk during Wake-up test of cAAI versus UMSS (mean ± SE)	0.79 ± 0.05
	pk during Emergence of cAAI versus UMSS (mean ± SE)	0.78 ± 0.03
Cheung et al. 2013	pk aepEX versus UMSS (mean [95% CI])	0.64 (0.55–0.72)	43 patients out of 69 patients
	AUCf (mean [95% CI])	0.77 (0.68–0.85)
Cheung et al. 2014	pk of aepEX versus UMSS (mean (95% CI))	0.68 (0.44–0.92)	59 patients out of 69 patients
	AUC (mean [95% CI])	0.72 (0.62–0.82)
Cheung et al. 2018	pk of aepEX versus UMSS (mean [95% CI])	0.68 (0.53–0.82)	32 patients out of 45 patients
	AUC (mean [95% CI])	0.89 (0.80–0.95)

• ^a Prediction probability.
• ^b A-line autoregressive index.
• ^c Laryngeal mask airway.
• ^d University of Michigan sedation scale.
• ^e Composite A-line ARX index.
• ^f Area under the receiver operator characteristics curve.

Six studies investigated the MLAEP during different amounts of anesthetics or during certain anesthetic procedures, such as intubation and extubation^{17-19, 21, 22, 29} and 3 studies investigated the effect of MLAEP monitoring on efficiency of the anesthetic regime, such as recovery time from anesthesia and total anesthetics used (Table 3).^{16, 19, 24}

TABLE 3. Summary of manuscripts investigating the relationship between mid-latency auditory evoked potentials, anesthetics, anesthetic procedure and efficiency of anesthetic regime

Study	Outcome	Results
Alvarez et al. 2004	Recovery time: controls vs cAAIa vs BIS guided anesthesia respectively (seconds)	419 vs 411 vs 396 (p = .993)
	Mean Etsevob: controls vs cAAI vs BIS guided anesthesia (vol%c)	1.71 vs 1.50 vs 1.87 (p = .442)
Weber et al. 2005	Propofol requirement of cAAI vs control group (mg kg−1 h−1: mean ± SD)	4.2 ± 1.7 vs 6.4 ± 1.3 (p < .01)
	Emergence time: cAAI vs control (minutes: mean ± SD)	5.1 ± 3.7 vs 13.2 ± 8.2 (p < .01)
Daunderer et al. 2007	Latencies (Na, Pa, Nb, and P1) “Awake” vs “Intubation” or “Steady state”	Lower during “Awake” (p < .05)
	Latencies (Na, Pa, Nb and P1) “Awake” vs “Extubation”	p > .05
Ironfield et al. 2007	pkd of AAIe versus EtSevo of 0–1 year old (mean ± SE)	0.53 ± 0.14
	pk of AAI versus EtSevo of 2 to11 years old (mean ± SE)	0.53 ± 0.10
Feuerecker et al. 2011	Latency of Na vs MACf sevoflurane of 2 months to 3 years old	r = 0.237 (p = .163)
	Latencies vs MAC sevoflurane of 2 months to 3 years old (Spearman's correlation: Pa, Nb and P1 respectively)	0.663, 0.783, 0.752 (p < .0001 for all)
	Latency of Na vs vol% sevoflurane of 2 months to 3 years old	r: .214 (p = .208)
	Latencies vs vol% sevoflurane of 2 months to 3 years old (Spearman's correlation: Pa, Nb and P1 respectively)	0.688, 0.768, 0.735 (p < .0001 for all)
	Latency of Na vs MAC sevoflurane of 6 to 14 years old	r = 0.523 (p = .0003)
	Latencies vs MAC sevoflurane of 6–14 years old (Spearman's correlation: Pa, Nb and P1 respectively)	0.734, 0.853, 0.868 (p < .0001 for all)
	Latency of Na vs vol% sevoflurane of 6–14 years old	r = .513 (p = .0003)
	Latencies vs vol% sevoflurane of 6–14 years old (Spearman's correlation: Pa, Nb and P1 respectively)	0.728, 0.845, 0.860 (p < .0001 for all)
	Latencies (Na, Pa, Nb or P1) preopioid vs postopioid application	p > .05
Liao et al. 2011	Correlation between AAI and EtSevo	R2 .01 (p = .01)
	Time to spontaneous movement: AAI vs control (minutes: mean ± SD)	3.9 ± 3.7 vs 6.1 ± 5 (p = .02)
	Time to fit for discharge: AAI vs control (minutes: mean ± SD)	60.5 ± 10.0 vs 66.8 ± 9.0 (p = .03)
Kuhnle et al. 2013	Latencies (Na, Pa, Nb or P1) during serum propofol 0, 3, 6 and 9 µg ml−1	p < .05 (except Na: 3 vs 6 µg ml−1)
Blussé van Oud-Alblas et al. 2019	Model to predict cAAI with serum propofol concentrations (mean [CV%])	E0: 63.4 (14.9) Emax: 0.786 (6.1) γ: 6.85 (46.4)

• ^a Composite A-line ARX index.
• ^b Endtidal sevoflurane.
• ^c Volume percentage.
• ^d Prediction probability.
• ^e A-line autoregressive index.
• ^f Minimum alveolar concentration.

Two studies investigated the effect of age on the MLAEP as the primary objective^{17, 18} and four manuscripts investigated age as their secondary objective.^{22, 30-32} A brief summary of these studies can be found in Table 4.

TABLE 4. Summary of manuscripts investigating the effect of age on mid-latency auditory evoked potentials during anesthesia

Study	Outcome	Results
Ironfield et al. 2007	pka of AAIb vs EtSevoc (age: 0–1 years) (mean ± SE)	0.53 ± 0.14
	pk of AAI vs EtSevo (age: 2–11 years) (mean ± SE)	0.53 ± 0.10
Daunderer et al. 2007	Latencies (Na, Pa, Nb, P1) of <4 years and >4 years	NSd
Feuerecker et al. 2011	Latencies (Na, Pa, Nb and P1) “infants” vs “schoolchildren” during “Awake” or 1.3vol%e or 2.6vol% sevoflurane	Nb at 2.3vol%; p < .5 Na, Pa and P1; NS
Cheung et al. 2013	EC50f and AUCg of aepEX vs UMSSh (age: 1–3 vs 3–6 vs 6–16 years) during propofol anesthesia	NS
Cheung et al. 2014	EC50 and AUC of aepEX vs UMSS (age: 1–3 vs 3–6 vs 6–18 years) during sevoflurane anesthesia	NS
Cheung et al. 2018	EC50 and AUC of aepEX vs UMSS (age: 1–3 vs 3–6 vs 6–18 years) during desflurane anesthesia	NS

• ^a Prediction probability.
• ^b A-line autoregressive index.
• ^c Endtidal sevoflurane.
• ^d Not significant.
• ^e Volume percentage.
• ^f Half maximal effective concentration.
• ^g Area under the receiver operator characteristics curve.
• ^h University of Michigan sedation scale.

4 DISCUSSION

Despite an extensive search and diligent screening of the current literature, we found no more than 15 relevant manuscripts reporting the use of MLAEP monitoring during anesthesia in children. Studies investigating the clinical performance of an MLAEP monitor as a depth of hypnosis monitor demonstrated a moderate to high correlation with the depth of hypnosis and an age-independent performance up to 14 years.^{17, 18, 20, 22, 23, 27, 28, 30-32} However, studies investigating the relationship between MLAEP and anesthetic drug dose showed conflicting results.^{18, 19, 21, 22} These differences might be explained by the algorithm used to compute an index value of the commercially available MLAEP devices.

Studies investigating the MLAEP waveform, observed increasing latencies and decreasing amplitudes when increasing the dose of anesthetics.^{21, 22} One study described an increased amplitude during anesthesia.¹⁷ When administering sufentanil to a child, the MLAEP waveform does not change.²² This might indicate that opioids don't have a clinically relevant effect on the depth of hypnosis and that the MLAEP waveform represents the depth of hypnosis component of anesthesia.

When investigating a commercially available MLAEP monitor, its relationship with the endtidal sevoflurane showed a weak correlation.^{18, 19} Liao et al. found a r² of 0.03 for their linear regression model,¹⁹ while Ironfield et al. demonstrated a p_k value of 0.53.¹⁸

It might be expected that this weak correlation and inconsistent change of the amplitude influences the performance of an MLAEP derived depth of hypnosis monitor. However, when analyzing the clinical performance of the A-line, AEP monitor/2 and aepEX monitor as a predictor for the UMSS during emergence, a p_k value ranging from .64 to .78 was observed during all commonly used anesthetics (propofol, isoflurane, sevoflurane, and desflurane) in children.^{27, 28, 30-32}

An even higher p_k value (.99) was reported by Weber et al. when used to predict the awake state and movement of the patient during LMA insertion.²³ We can assume that an awake person has an UMSS ≤1 and that LMA insertion is a significant physical stimulus; thus, having an UMSS ≥3 when a patient does not react during LMA insertion. Figure 2 illustrates the relationship between the index values and UMSS of each type of MLAEP monitor. Of note, the UMSS is as the name implies a tool to measure the depth of sedation. An UMSS of 4, in which the patient is unarousable, can be considered an adequate depth of hypnosis for anesthesia. However, when an even “deeper” hypnotic depth is needed or a measurement of the anesthetic depth is needed, the UMSS is not validated to measure this. The first commercially available MLAEP monitor designed for use as a depth of anesthesia monitor, the A-line, showed a linear relationship with the UMSS, while the newer models, that is, the AEP monitor/2 (which superseded the A-line) and the aepEX monitor, show a sigmoidal relationship. The sigmoidal relationship could suggest a binary behavior for the depth of hypnosis, that is, consciousness and unconsciousness, instead of a gradual relationship. By performing a Receiver Operating Characteristics (ROC) analysis, Cheung et al. investigated the performance of the aepEX to detect conscious (UMSS ≤1) and unconscious (UMSS ≥2) patients.^30-32 They found an area under the curve (AUC) of 0.72–0.89, which implies that the aepEX is a reasonable device to distinguish between consciousness and unconsciousness.

While the EEG matures in early adulthood, MLAEP mature in the first decade of life.^10-13 This might give the MLAEP a slight advantage over the EEG as a variable to assess the depth of hypnosis in anesthetized children. Six out of 15 studies investigated the effect of age on the MLAEP. Five studies comparing the MLAEP of children from different age groups showed no significant differences in latencies¹⁷ nor index values.^{18, 30-32} Only one study, by Feurecker et al., revealed increasing latencies with increasing age. This was, however, only apparent when comparing infants with elderly patients,²² indicating the consistency of the MLAEP as a parameter for assessing the depth of hypnosis in children up to the age of 14 years.

There are conflicting results concerning MLAEP guided anesthesia to improve the efficiency of the anesthetic regime. Weber et al. demonstrated that the time needed for emergence was reduced by 8 min and less propofol was needed when anesthesia was cAAI guided.²⁴ During a day planned with many short procedures, an 8 min reduction in emergence time could allow more patients. However, in the study of Liao et al. AAI guided anesthesia only saved about 2 min until the patient moved after surgery and saved 6 min until the patient was fit for discharge, of which the clinical relevance can be questionable.¹⁹ Alvarez et al., on the other hand, could not demonstrate these benefits at all with cAAI guided anesthesia.¹⁶ Although Alvarez et al. had a larger number of patients included in their study compared to Weber et al., essential details are lacking from their study since only a published abstract could be found.

In daily practice, application of a MLAEP monitor in children can be challenging. Manufacturers of these monitors usually supply an in-ear headphone with the device, which cannot be adequately placed in the small ear canal of a child. Published manuscripts try to circumvent this problem by securing the earpieces with tape or replacing the headphones with their own over-ear headphones.^{18, 23, 30-32}

The low voltage of the MLAEP waveform (a few hundred microvolts) makes it prone to noise and artifacts. This is especially evident when trying to measure MLAEP in awake young children or during the induction and emergence phase, making its clinical application challenging. This was also illustrated by the study of Daunderer et al. in which they described that only 1645 of the total 8544 AEP segments could be used for analysis.¹⁷ The studies by Cheung et al. reported that paired p_k values during emergence from propofol, sevoflurane and desflurane could only be computed for respectively 26 out of 69, 10 out of 69 and 13 out of 45 patients.^30-32 However, this could also be attributable to the failure of the BIS monitor.

Another consideration about the MLAEP waveform is that it is not fully developed in most children under 10 years.¹² Despite that, MLAEP-based depth of hypnosis monitors can still compute a fairly reliable index value to indicate the depth of hypnosis.

Very few studies concerning the use of MLAEP monitoring in children during anesthesia are conducted. This review yielded only 15 studies concerning this subject. Due to the diversity in defined outcome parameters and methodological differences (eg, different anesthetics, premedication and type of monitor used) a comparative (meta)analysis was considered not meaningful. We therefore lack a statistically grounded recommendation in whether MLAEP monitoring is beneficial during anesthesia in children. This review, however, attempts to systematically highlight the current available literature, so that the reader can make an educated decision on whether MLAEP monitoring could benefit his or her anesthesia.

In conclusion, the current literature shows that MLAEP analysis can be used to reasonably assess the depth of hypnosis in children under anesthesia. Furthermore, its reliability does not seem to depend on the age of the child or the type of commonly used anesthetics according to the current published studies. However, only a few studies investigated the performance of an MLAEP monitor as a depth of hypnosis monitor and conclusions regarding its reliability should be made with caution.

CONFLICT OF INTEREST

All authors have no conflicts of interest to declare.