Many surgical procedures would not be possible without the patient entering a state of general anaesthesia (GA). The essential features of a successful GA, displayed by the patient, are a reversible loss of consciousness with a lack of movement, a lack of awareness, unresponsiveness to painful stimuli and a lack of recall of the surgical intervention. Inadequate GA may lead to intraoperative awareness with recall (due to patient underdosage) or to prolonged recovery and an increased risk of postoperative complications for the patient (due to overdosage). An important contributing factor to inadequate GA is our current limited ability to evaluate the levels of consciousness [1].

The incidence of awareness has been reduced from about 1–2% in unselected patients in the 1980s [2] to about 0.1% at present [3–6]. However, certain surgical procedures (for example caesarean section or cardiac surgery) or high risk patients have a substantially increased risk of awareness [7]. The consequences of intraoperative awareness range from an absence of prolonged after-effects to a post-traumatic stress disorder [7]. In comparison, worldwide anaesthesia-related mortality rates during the 1990 to 2006 period ranged between 75 (in Thailand) and 1 (in the USA or Japan) per 100,000 anaesthetic procedures. The main causes of anaesthesia-related mortality were problems with airway management and cardiovascular events related to anaesthesia and drug administration [8].

Important signs of inadequate patient GA (Figure 1), signs which develop in response to stress or painful stimuli, are movement, increased breathing or heart rate and increased blood pressure [9]. These warning signs are often attenuated since many surgical procedures require administration of a muscle relaxant or the patient receives medication that affects the heart rate and blood pressure [9]. Avoidance of muscle relaxants is no guarantee of avoiding awareness. Studies have reported patients with awareness under non-relaxant anaesthesia [3–5]. Fortunately, such awareness is very rare and very rarely reported as unpleasant [3].

It is generally believed that awareness will not occur if GA is maintained with a volatile agent delivered to the patient’s blood at a concentration of at least 0.5 minimum alveolar concentration (MAC) [9]. However, for some patients, the use of total intravenous anaesthesia techniques is more appropriate [3,10–13]. Individual variations in patient drug requirements will also lead to occasional over dosage or under dosage with both volatile-based and intravenous-based techniques [9,11]. Failure to adjust the anaesthetic requirements to individual variations in patient drug requirements and to the intensity of pain stimulation during a specific surgery procedure will also lead to over dosage or under dosage with both volatile-based and intravenous-based techniques [9,11].

The potencies of general anaesthetics range over at least five orders of magnitude [14]. However, for any given anaesthetic, the concentration at which consciousness is lost is well defined. For intravenous or volatile anaesthetics, the transition from the conscious to the unconscious state occurs abruptly over a change in concentration of only a factor of less than 1.6 [14]. The physiological mechanism for this abrupt switch between consciousness and unconsciousness was explained by the intrinsic bi-stability of thalamocortical neurons coupled with the reciprocal inhibitory connections between hypothalamic sleep-promoting centres and the arousal nuclei in the midbrain and the brainstem, which would tend to favour rapid transitions between high and low states of ascending arousal [14].

Administration of at least 0.5 MAC of a volatile anaesthetic agent should prevent awareness [15] and most anaesthetists employ this approach to prevent awareness in their daily practice [9]. Similar results can be obtained with total intravenous infusion devices that incorporate predicted plasma and effect site concentrations [16]. However, since we measure the volatile anaesthetic’s concentration in the lungs and only simulate the concentration of intravenous anaesthetic drugs, we do not know exactly and accurately what anaesthetic concentration is reaching the brain at all times throughout the surgical procedure. In other words, volatile-based and intravenous-based anaesthetic devices measure or predict drug concentration not drug effect [9].

The transition from a state of wakefulness to a state of GA is accompanied by profound changes in the brain’s spontaneous electrical activity recorded from electrodes placed on the scalp (an electroencephalograph or EEG). The EEG reflects the compound synaptic activity of excitatory and inhibitory post-synaptic potentials generated by cortical neurons [17]. However, to achieve an optimal level of GA, it is not possible, nor practical, to adjust the delivery of volatile or intravenous anaesthetics on the basis of an on-line EEG [17]. Only the advances in computer hardware and signal processing algorithms have enabled the processing of EEG signals (Table 1) and the development of monitors that estimate the depth of general anaesthesia (DGA) on a near to on-line basis [9,17–20]. To understand how DGA monitors work, their respective advantages and disadvantages, it is necessary to understand the theoretical foundations that contributed to their design and this also includes a basic understanding of the relation between GA and consciousness.

There are two types of general anaesthetics: (a) intravenous agents (e.g., propofol), generally administered together with sedatives or narcotics and (b) volatile agents (e.g., sevoflurane). Both types of anaesthetics modulate the permeability of ion channels that regulate synaptic transmission and membrane potentials in key regions of the CNS [14,23,24]. All general anaesthetic agents are relatively apolar, to be able to cross the blood-brain barrier, and interact with their target (i.e., receptor) through weak polarization forces and hydrogen bonding [25]. The binding of a general anaesthetic to its receptor leads to neuron hyperpolarisation due to increased inhibition or to decreased excitation thus altering neuronal activity [26]. The actions of a general anaesthetic on the molecular level are reflected on the brain’s electrical activity as a transition from the low voltage, high-frequency pattern of wakefulness (known as activated EEG), to the slow-wave EEG of the deep, non-rapid eye movement (NREM) sleep, and finally to an EEG burst-suppression pattern [27].

General anaesthetics act on different targets, i.e., receptors, for example γ-aminobutyric acid (GABA) type A, nicotinic acetylcholine, and glutamate receptors (e.g., N-methyl-d-aspartate receptors) in the brain, as well as glycine receptors in the spinal cord [28,29]. A clear distinction has to be made between the actions of general anaesthetics and the mechanism of general anaesthesia. The actions of general anaesthetics on the brain receptors are well known; the process by which these actions translate into general anaesthesia (i.e., the mechanism of general anaesthesia) is not well understood [21].

Distinct anaesthetic properties are correlated with distinct sites in the CNS; for example, loss of consciousness is associated with the cerebral cortex, loss of memory with the limbic system and immobility and analgesia with spinal cord effects [21]. General anaesthetics modulate the activity of several CNS structures including the spinal cord, brain stem, pons, thalamic nuclei and specific regions of the brain cortex (Figure 2). The effects of general anaesthetics on the brain, which lead to loss of consciousness, are best understood at the level of the cortex and the thalamus.

The thalamus has numerous connections to the cortex (i.e., the thalamo–cortical pathway, TCP) and individual thalamic nuclei modulate the activities of specific cortical regions. In addition, the cortex sends connections back to the thalamus (i.e., the corticofugal pathway, CFP) which greatly outnumber the thalamo–cortical connections [14]. The corticofugal connections provide a continuous depolarization of the thalamo–cortical neurons, during the brain’s wakeful state, preventing them from entering a synchronized, oscillatory state. The activity of the TCP is in balance with the activity of the CFP through a positive feedback loop [14]. A reduced activity of the CFP promotes the activity of the TCP, the TCP neurons enter into an oscillatory mode which further reduces the activity of the CFP and the subject enters into a sleep-like state (i.e., hypnosis). Most anaesthetics (ketamine is an important exception) produce EEG patterns that are reminiscent of spindle and δ waves, the EEG landmarks of falling asleep and deep, dreamless NREM sleep, respectively. Most anaesthetics initially produce high-frequency oscillations followed by a lower frequency and a higher amplitude EEG pattern at or beyond the point at which consciousness is lost [14]. It has been suggested that loss of consciousness occurs when information transfer through the thalamus is disrupted during GA following a reduction of thalamic metabolism and blood flow [14,30,31]. It has also been suggested that the thalamus could behave as a consciousness switch [30]. However, thalamic activity does not decrease with all anaesthetics; some general anaesthetics (e.g., ketamine) increase thalamic metabolism and some general anaesthetics (e.g., sevoflurane) substantially reduce thalamic activity without loss of consciousness when applied in low doses and lead to unconsciousness at high doses [32,33]. All anaesthetics will produce sedation at moderate doses. Therefore, it is more likely that the effects of general anaesthetics on the thalamus reflect global cortical activity rather than the thalamus acting as a consciousness switch [1,22]. It has been suggested that efficient communication among cortical areas (especialy among the long range cortico–cortical connections) requires a functional thalamic connection [34,35]. A functional thalamic disconnection of the brain cortex, during GA, could still lead to cortical arousal, but not to consciousness [22].

All cortical areas are not equally important for the induction of unconsciousness by general anaesthetics. A complex comprising the lateral temporo–parieto–occipital junction and the mesial cortical core is most likely the final common target for anaesthetic-induced unconsciousness [22]. Anaesthetics disrupt two essential cortical functions: cortical integration and cortical information capacity. Anaesthetics may disrupt cortical integration by acting on structures that facilitate long-range cortico-cortical interactions [22]. Anaesthetics disrupt synchronization of brain areas by delaying feedback interactions among distant (not adjacent) brain areas within the cortex [36] or between small-world networks with predominant local connectivity and a few long-range connections—for example the cortico–thalamic system [37]. The cortico–thalamic system produces global responses during deep anaesthetic unconsciousness. However, these responses are limited to a stereotypic burst-suppression pattern, with a corresponding loss of information, thus creating a system having only two possible states (i.e., on or off) [22,38,39].

General anaesthetics also inhibit the excitatory arousal pathways originating in the brain stem and pons or potentiate the sleep pathways that control them [14]. The brain stem and pontine nuclei have been known to be essential in maintaining cortical arousal and form the so called ascending reticular formation [14].

The concept of unity of consciousness was introduced by the 18th century German philosopher Immanuel Kant [21,40]. According to Kant, for a perception to have a meaning its distinct qualities must be brought together into a single experience. Kant postulated the existence of a mechanism that synthesized the diverse and complex processes of the mind to generate a single perception i.e., the unity of consciousness [21]. The visual system is the best understood model of how the brain processes and integrates sensory information. The key question, also known as the cognitive binding problem [41] that arose from these studies was: “How does the brain synthesize the elements of visual processing, located at diverse locations of the cortex, to generate a unified visual perception?”.

Cognitive binding is necessary, but not sufficient, for the brain to sustain the unity of consciousness. The proposed mechanisms for cognitive binding in the brain [42,43] are binding by convergence (transmission of information from primary processing areas to another region of the brain for integration), binding by assembly (binding information in a group of inter-related neurons whose connections grow stronger when repeatedly firing together) and binding by synchrony (coordination of neural firing in time associated with neural events at the frequency of about 40 Hz). The evidence from GA and sleep studies suggests that loss of consciousness is associated with a breakdown of cortical connectivity, a diminished cognitive binding, a reduction of the repertoire of cortical activity patterns (i.e., the predominance of stereotypic responses) and a loss of information [1,21].

Recently, a “cognitive unbinding” paradigm of GA was proposed to explain the actions of general anaesthetics [44–46]. To summarize, consciousness is not an all-or-none property, but is graded; it increases in proportion to a system’s repertoire of distinguishable states. The shrinking of the field of consciousness during GA is proportional to the reduction of the brain’s repertoire of distinguishable states. A transition point is reached during GA when a nonlinear collapse of cognitive binding leads to loss of consciousness [1,21]. A practical application of the concept of cognitive unbinding to GA, the “anaesthetic cascade” theory, was proposed by John and Prichep [47]. The anaesthetic cascade theory attempts to link progressive suppression of consciousness during GA with the action of general anaesthetics on specific brain regions.

The initial signs of GA (stage 1 anaesthesia) can be a state similar to drunkenness, analgesia (the inability to feel pain while still conscious), amnesia (loss of memory), distorted time perception or increased sleepiness. With higher anaesthetic doses, a patient fails to move in response to a verbal command and is considered to be unconscious. However, unresponsiveness during GA can occur without unconsciousness. For example, paralyzing agents used to prevent unwanted movements during anaesthesia do not remove consciousness [48]. Most general anaesthetics will cause a global deactivation of the brain at higher doses [1]. The exceptions are dissociative anaesthetics, like ketamine, that increase global brain metabolism, especially in the thalamus [32]. The ketamine induced changes in global brain metabolism are reflected in the responses of the anaesthetised patient. Ketamine causes loss of motivation to follow commands at lower doses [49]; at higher doses ketamine causes a characteristic state in which the eyes are open and the face takes on a disconnected blank stare consistent with the deactivation of executive circuits in anterior cingulate cortex and basal ganglia [1]. At doses near the unconsciousness threshold, some anaesthetics block working memory and patients may fail to respond because they immediately forget what to do [50]. Also, some patients under general anaesthesia can carry on a conversation using hand signals, but postoperatively deny ever being awake during the surgical procedure [51]. Surgery is performed when the patient is perceived to reach stage 3 general anaesthesia. Increasing doses of modern general anaesthetics (i.e., an overdose) do not cause brain death [52] directly but lead to cessation of spontaneous ventilation and impairment of cardiovascular function (stage 4 anaesthesia). This stage is lethal without proper cardiovascular and respiratory support, which is standard in a modern operating theater. Therefore, monitoring the DGA is essential to tailor the optimal dose of general anaesthetic to the individual patient thus reducing the incidence of intraoperative awareness or preventing the effects of an overdose of general anaesthetic. Although there are several “brain-function monitors” available they have a limited ability to detect directly the presence or absence of consciousness, especially around the transition point when loss of consciousness occurs [1]. This issue will be discussed in detail in Section 4.

The process of monitoring depth of GA and administration of a general anaesthetic during surgery is a closed-loop control system where the human is responsible for reasoning and action. The anaesthetists play the roles of controller and actuator by deciding on the amount of anaesthetic and when to administer it. On the other hand, the activity of monitoring is performed automatically by commercially available DGA monitors. Together they form a closed-loop control system.

In the last decade several experiments have been performed, where anaesthesia was controlled in a closed-loop control system without human interference [114–117]. Although these experiments were relatively successful and represent a proof of principle, there is still a long way to the autonomous commercially available depth of GA control system. Additionally, these experiments raise new questions, which have been only partially addressed so far. For example, the challenge is now to establish fully the safety, efficacy, reliability and utility of closed-loop anaesthesia for its adoption into the clinical environment [119]. Moreover, these systems still have to go trough rigorous testing and validation. Additionally, while the closed-loop system can maintain a constant target sedation level it cannot adjust the level of anaesthetic delivery in advance to meet the future needs during a surgical procedure. On the other hand, this scenario can easily be predicted and successfully resolved by the anaesthesiologist in communication with the surgeon. Therefore, a closed-loop control system for automatic administration of anaesthetic is feasible, but some important practical issues have to be addressed.

The proposed control systems are most often built around a well established BIS monitor, which is now standard equipment for GA monitoring. However, in the last decade the BIS monitor has undergone rigorous testing, which has shown several drawbacks [118,119]. Therefore the problem of constructing an ideal DGA is still unsolved.

It is a matter of debate how consciousness is formed let alone how it should be measured. In other words, we currently rely on indirect measurable quantities, which give us a relatively reliable measure of the state of awareness. Several algorithms and methods have been developed for assessment of the DGA in the last decade. Their common denominator is the reduction of information, which comes from one or more complex signals, into a single feature, which reflects the state of consciousness. This concept puts DGA monitoring systems in the domain of software sensors [109,110].

The first commercial EEG-mono-parameter was the Bispectral-Index or BIS, which was first introduced in 1992 and after 1999 others followed (e.g., Narcotrend, AEP-Monitor/2, PSA, CSM ...). These monitoring systems analyse the potential fluctuations measured from the patient’s forehead. After the signals are digitalised on the AD converters, the pre-processing unit determines which part of the signal is going to be analysed. Next, the artefact algorithm is applied to remove possible artefacts stemming from eye movement, swallowing or heart activity. The quality of the artefact removal algorithm is therefore crucial for the reliability of the DGA monitor. On the other hand, these artefacts are sometimes used as surrogate parameters, e.g., detection of swallowing determines the steady state of GA. Furthermore, the electromyogram (EMG) can also be used as surrogate parameter since the movements of the facial muscles are clearly visible in the frequency spectrum above 30 Hz. Consequently, the increase of the EMG surrogate parameter is interpreted as a decrease of the anaesthetic’s effect and vice versa. Next, algorithms of commercial monitors were trained and tested against the signals from the databases. The measured signal is then compared against the samples from the database and classified. Consequently, the index is as good as the database against, which it was trained. Finally, the calculated index values are post-processed to minimize fluctuations on the output. This is done by averaging over past index values, which stabilises the index, but adds delay into the index. Therefore, all commercial indices are: (a) constructed abstract quantities that are not directly linked to any physiological parameters and (b) have an inherent time delay (Figure 4). In the next sections, we will present some of the depth of anaesthesia monitors, which are in use in clinical practice and have endured some amount of clinical testing.

The iterative process in development of the index algorithm improves its quality and reliability. In this process the algorithm should be refined against a high quality database with expert interpretation. In this regard several publications exist about the development of the BIS, Narcotrend and PSA monitor algorithms [131,134,144]. BIS was developed in an iterative process, where the discriminating parameters were tested for performance against a database of EEG measurements related to different anaesthesia states [131]. The Narcotrend algorithm was developed in a similar fashion, where the database was visually inspected and classified into stages. The classified data then served to extract the time and frequency domain features to be used in classification function [134,137]. The development of the PSI algorithm was based on three databases of EEG records, which were used for development of artefact detection and removal algorithms, classification algorithms and finally a third database was used for calibrating the index [144]. On the other hand little is known about the inclusion of prior knowledge in the development of other indices (AAI, SE, CSI and IoC). This does not mean that the monitors were not tested and calibrated against real EEG signals, however, the size of the database and the development process are an important property of the DGA monitor and can reflect the discriminating power of the index.

After the pre-processing stage of the data analysis all indices calculate features, which are used for index calculation. The number and type of features vary from monitor to monitor, although, most of the features are established empirically and do not reflect any underlying determinism. Exceptions are BIS, Narcotrend and AEP-Monitor/2 (AAI). The BIS performs a bispectral analysis of the EEG signal and analyses the phase relationships of the EEG components. Here, a weak assumption of the model can be made, that the phase relationships are connected to the number of different oscillatory generators in the brain [131]. The Narcotrend uses autoregressive (AR) modelling to infer different features [134,148]. Also, the AEP-Monitor/2 uses AR models to infer AEP from the EEG [140]. However, AR models are only used for modelling statistical properties of the signal; therefore, it is not possible to infer the underlying dynamics of the signal. At the same time, Narcotrend uses a set of empirical features based on spectral power: spectral edge frequency and β-ratio.

A similar approach was used for the PSA, where a set of spectral features were established from database EEG records [144]. The same approach is also used by the CSM, which is based on α and β power ratios [148]. An alternative approach is used by the Entropy Module (SE and RE) and the IoC Monitor. In the first case the spectral entropy calculation is combined with multi-resolution analysis, which yields an optimal time-frequency resolution. The entropy can be regarded as an estimator of signal complexity and is assumed to be related with the DGA [150,151]. Moreover, it can detect non-linear features in the signal. A similar principle is used by the IoC monitor, which analyses the symbol dynamics of the EEG signal. Finally, while the indices mostly differ in the number and type of features used, the BSR is calculated by all monitors and is included either in the index calculation algorithm, or is displayed as additional information for the anaesthetist [18,148].

After the feature extraction is completed, each monitor uses a different algorithm to calculate or classify the DGA. BIS calculates its value by a simple weighted sum of features [131]. The weighing function is most probably modulated by the state of anaesthesia, since a linear dependence on BSR in deep anaesthesia has been demonstrated in [132]. Narcotrend monitor uses a classification function, which derives the state of anaesthesia from the extracted features. Additionally, the Narcotrend monitor calculates additional surrogate parameters, which are sensitive to atypical patterns of general anaesthesia, e.g., K complexes [134]. These parameters are then used in a plausibility analysis before the index is displayed. This is the probable reason for Narcotrend’s better performance in comparison with BIS regarding the susceptibility to increased EMG activity [152]. Similarly, the AEP-Monitor/2 (AAI) also has a weighting function to calculate the index from the ARX model. The weighting function is modulated by the signal quality and the presence of artefacts. Although, the modulation function implies that the influence of the artefacts on the final index is accounted for, the values of AAI and BIS have decreased significantly in a study where muscle relaxants were administered to the ICU patients [153,155]. The PSA first conducts a plausibility analysis to classify the state of anaesthesia from the values of spectral features. Additionally, it performs surrogate analysis to test against artefacts and BS. A specific approach has been taken by the Entropy Module, where the monitor uses no special inference algorithm but rather relies on two normalised spectral entropy values: state entropy (SE) and response entropy (RE). The authors argue that artefacts like EMG should be treated as a signal and are therefore included in the RE number [79]. The anaesthetist should look for a difference between the SE and the RE, where a difference larger than 10 indicates the presence of significant EMG interference in the EEG signal. The IoC Monitor and CSM both use a fuzzy logic set of rules to produce an index. Although the CSM includes the artefact control function, the features are calculated on the frequency bands which could include EMG activity. Although, a comparative study showed that EMG had no effect on CSM or BIS index in children, an open question remains how much influence the EMG has on the CSM in adults [156]. Additionally, a case report indicates a strong correlation between EMG and CSM in an ICU patient [154]. Finally, the IoC Monitor, which also calculates the index via a fuzzy inference system, excludes the EMG variable from the algorithm. Alternatively it displays the EMG value separately and it is up to the anaesthetist to decide how to react to the increased EMG values. The authors stress the fact that a sudden change in the IoC, accompanied by a rise of EMG, is probably the effect of increased muscle activity [147].

One of the most important parameters of a DGA monitor is its response time (i.e., time delay) to a rapid change in the level of sedation. Unfortunately, not many studies have been performed where this parameter was tested. Zanner and co-workers used recorded EEG data to test the responsiveness of the DGA indices [157]. The data were selected to represent an awake state, a general anaesthesia and a suppression of cortical activity. Then the authors measured the response time of BIS and Narcotrend monitors to these EEG signals. The study reported time delays between 26 ± 7 seconds and 106 ± 23 seconds for the transition between EEG suppression and the awake state for BIS, Narcotrend and CSM (see Table 2 for details). Additionally, the study showed that the time delays were not constant and were dependent upon the depth of GA. Also, the response times were different for the increasing and decreasing states of GA. Next, a similar study was performed, where the authors used simulated EEG signals on the same monitors [158]. The time delays fluctuated between 14 seconds and 155 seconds, also, the rise and fall times did not match for the same DGA index. Therefore, both studies raise concerns about the timely detection of rapid EEG changes concomitant with the transition from general anaesthesia to awareness. Additionally, these findings limit the use of DGA monitors for pharmacodynamic modelling. Similar studies for other DGA indices have not been published to date.

Despite the abundance of studies on commercial DGA monitors, it is still difficult to compare their performance because there is no consensus on the validation methodology [19] and the majority of studies have been performed on the minority of DGA monitors (i.e., BIS, Narcotrend and Entropy module). Despite the aforementioned problems, the fact that the BIS monitor has been present for almost two decades and has been evaluated by many studies, makes it a de facto standard against which all other DGA monitors are compared. The BIS and Entropy monitor are the only DGA monitors that have been routinely used in child anaesthesia [163–168]. However, this does not necessarily mean that the BIS monitor is superior to other DGA monitors in all other aspects of anaesthesia monitoring.

Some conclusions can be made about the performance of individual DGA monitors by analyses of the inferred characteristics of their DGA algorithms (only conceptual descriptions of DGA algorithms were published, no DGA algorithm has been published with full technical specifications so far). The influence of various interferences on DGA index can be studied by checking for the presence of surrogate analysis in the respective algorithm. It can be expected that BIS would have more problems with EMG interference than Narcotrend, which was shown in [169]. This is most likely due to the fact that Narcotrend modulates the index value with artefact surrogate analysis. Susceptibility to EMG interference is expected in monitors that have a DGA algorithm similar to the PSA monitor. The performance of artifact removal algorithms was tested in [170], where it was shown that Narcotrend’s artifact removal algorithm was more likely to exclude the data from further analysis, due to the presence of artifacts. Unfortunately, no such claims could be made for other monitors, due to lack of similar studies. The complexity of the algorithm is an important property, which also influences the time delay or the response time of a monitor to an abrupt change in sedation. Here BIS, IoC and Entropy Module DGA monitors are advantageous, since their algorithms are simple, that is, they include few FD or TD features. This fact was shown in recent articles [171,172], where BIS outperformed the Narcotrend, PSI and CSM monitor. Here, despite the fact that CSM has a simpler algorithm in comparison to Narcotrend, it was the slowest to respond to a change in sedation. Burst suppression is an important indicator of the level of deep sedation. The BSR is therefore included in every DGA monitor. The way it is incorporated into the algorithm might have the influence on the way monitors react different or extreme conditions like oversedation. The monitors like BIS, Narcotrend, IoC, CSM might have an advantage, since the BSR analysis is directly included in the index calculation, while on the other hand, monitors like AEP-Monitor/2, PSA and EntropyModule use BSR analysis in surrogate analysis to modulate the value of the index or in some cases display the value of BSR independently. This fact was reflected in the study [173], where BIS outperformed the PSI (PSA) in the prediction of oversedation.

Recently, two studies compared the monitoring of GA with the BIS and the CSM (marketed as a low cost DGA monitor) during propofol/remifentanil [174] and propofol/fentanil anaesthesia [175]. In the first study, the CSM was a satisfactory alternative to the BIS for monitoring hypnotic effect in 87% of patients; in 13% of the patients, the CSM displayed values indicating an awake state despite clinical sleep, all correctly identified with the the BIS [174]. In the second study, propofol anaesthesia guided with the CSM resulted in 20% higher propofol doses compared with the BIS; there were no clinically relevant differences in recovery times [175].

The DGA monitoring technology is promising and used widely, nevertheless to date no health authority has so far recommended that such monitors should be compulsory during general anaesthesia, but should be considered only on an individual basis [176].