Spectral Entropy as a Measure of Hypnosis and Hypnotic Drug Effect of Total Intravenous Anesthesia in Children during Slow Induction and Maintenance

• Monitors for assessment of depth of sedation/hypnosis are widely validated for use in adults

• Data on the relationship between electroencephalography measures and plasma propofol concentrations are scarce

• In 60 patients (age: 3–16 yr) spectral entropy (SpE) was related to the level of hypnosis as assessed by the University of Michigan Sedation Scale in an age-dependent manner

• Younger children (3–6 yr) showed higher SpE values for deep sedation and surgical anesthesia when compared to older ones

THE need and indications for measurement of the level of hypnosis and hypnotic drug effect in children may be even more important than in adults. First, there is recent evidence that awareness during anesthesia is four to eight times more common in children than in adults.1,–,5Second, there are emerging laboratory animal data that suggest various common anesthetics may be toxic for the developing brain.6,–,8Moreover, the “brain growth spurt” continues for several years after birth,9meaning the ability to optimize and minimize the use of hypnotic anesthetics may also be valuable in older children. Third, the interindividual variation of drug effects is larger because of pharmacokinetic changes during childhood.10This challenges the steering of pediatric anesthesia, which might thus benefit from the use of hypnosis monitors.11,–,14 

A thorough validation of depth of hypnosis (DoH) monitors is necessary before outcome studies in children are justified.15At the moment DoH monitors seem to be valid for use in older children, but not in younger children.15Further validation studies of DoH monitors are still needed in different pediatric age groups and with different anesthetics.15,16 

Spectral entropy (SpE) measures the depth of hypnosis by analyzing the regularity of the electroencephalogram signal with a published algorithm.17This is used in the SpE Module (M-Entropy; GE Healthcare, Waukesha, WI). SpE has been validated in adult patients, but in children this has been limited to investigations made only during inhalation anesthesia, and show only reasonable SpE performance.18,–,20The data concerning DoH monitors during total intravenous anesthesia (TIVA) in pediatric anesthesia are scarce, and data on SpE in children were nonexistent until this study.

The primary goal was to evaluate whether SpE reliably measures the depth of hypnosis and the intravenous hypnotic drug effect in children during TIVA. Second, the pharmacodynamic relationship between entropy and the estimated plasma propofol concentration (C^p) was investigated. Third, we characterized several clinical aspects of target-controlled infusion (TCI) propofol anesthesia and SpE monitoring. The differences of the SpE values between the age groups at the corresponding University of Michigan Sedation Scale (UMSS)21levels were also investigated. Our hypothesis was that response entropy (RE) and state entropy (SE) values would have a strong relationship with the UMSS, C^p, and C^effvalues.

Institutional ethics committee approval (for the Hospital for Children and Adolescents, Helsinki University, Helsinki, Finland), along with written informed consent of parents or patients, when appropriate, were obtained. Sixty children and adolescents, aged from 3 to 16 yr whose weights ranged between 15 to 61 kg, were studied. The age and weight criteria were set by the TCI-pump and the Kataria pharmacokinetic model.22American Society of Anesthesiologists class 1 and 2 patients were scheduled for elective surgery with an estimated duration of 1–5 h and requiring general anesthesia. Patients were excluded when either they had a disease or medication that affected the central nervous system or when the surgery affected the head or the neck.

The anesthetic regimen is illustrated in figure 1and is a representative patient case. All patients received a standardized anesthetic regimen. No regional anesthesia or neuromuscular blocking agents were used in this study. EMLA® local anesthetic cream (AstraZeneca; AstraZeneca AB, Södertälje, Sweden) was topically applied over at least two large predefined superficial veins approximately 1 h before and premedication with 0.3 mg/kg (maximum 15 mg) oral midazolam was administered approximately 30 min before the estimated induction of anesthesia. A well-functioning peripheral intravenous catheter was inserted, which was connected to one-way valves along with drug and infusion lines, and acetated Ringer's solution infusion was then commenced. In order to minimize the injection pain of propofol, a venous stasis was applied and intravenous lidocaine 0.5 mg/kg was administered and allowed to take effect for 45 s before opening the venous stasis. Immediately after this, propofol 20 mg/ml (Propofol-Lipuro 2%; B. Braun Melsungen A G, Melsungen, Germany) TCI at C^pof 1 μg/ml commenced by TCI-pump (Alaris Asena; Alaris Medical Systems, Basingstoke, United Kingdom), according to the pharmacokinetic model described by Kataria et al.  22This C^pwas maintained for 8 min, after which the C^pwas increased to 2 μg/ml for another 8 min and thereafter to 3 μg/ml for a third 8-min period. Increased fraction of inspired oxygen was given when needed during this stepwise propofol induction.

After induction the C^pwas increased to 6 μg/ml and remifentanil 1.5 μg/kg was concomitantly infused for more than 1 min, followed 1 min later by orotracheal intubation. Patients were under controlled ventilation with oxygen-air mixture, to aim at normocapnia until the end of surgery. Approximately 5 min before the start of surgery, the C^pof TCI-propofol was set to 5 μg/ml. Approximately 1 min before the start of surgery, a remifentanil bolus 1.5 μg/kg was given, and remifentanil 0.3 μg · kg⁻¹· min⁻¹was infused until the end of surgery.

During surgery the remifentanil infusion was kept stable. After the induction phase the level of anesthesia was adjusted by changing the C^pas clinically indicated by the anesthesiologist in charge, blinded to the SpE values. The anesthesiologist took the decision to increase or decrease the C^pby 0.5–1 μg/ml over approximately 10-min intervals when clinically possible in order to achieve different levels of hypnotic drug effect for study measurements. Remifentanil 1.5 μg/kg boluses were used as secondary rescue analgesic therapy.

Approximately 10 min before the end of surgery, C^pwas targeted to 2–3 μg/ml. Immediately before the end of surgery, intravenous acetaminophen 20 mg/kg and/or ketoprofen 1 mg/kg were given for postoperative analgesia. At the end of surgery, C^pwas set to 0 μg/ml and remifentanil infusion was discontinued. The patient was manually ventilated, and when spontaneous ventilation was considered sufficient, the patient was extubated.

Before the anesthetic induction, the forehead skin was wiped with an alcohol swab and allowed to dry. A disposable adult self-adhesive entropy sensor (GE Healthcare) was placed according to the manufacturer's instructions. Electrode impedances were checked before the start of induction and automatically during anesthesia. SpE values (RE, SE), including the burst suppression ratio (BSR), were measured every 5 s with a M-Entropy module of the S/5 Anesthesia Monitor (GE Healthcare). SpE indices with all other measured parameters were collected into a laptop computer (Acer; Acer Group, New Taipei City, Taiwan) using the Rugloop software (Demed, Temse, Belgium). Baseline measurements were obtained before proceeding to anesthesia induction, and the data collection continued until the extubation of the patient.

We relied on the M-Entropy module on the detection of burst suppression (BS). The occurrence of BS was defined as the BSR of the M-Entropy exceeding 1%, and the disappearance of BS was interpreted as the moment when the BSR was again back at the value of 0%.

To detect possible lighter levels of hypnosis (indicated by higher SpE values) during maintenance and emergence (= the last 15 min of surgery), occurrences of SpE index values, between 50–60 and above 60, were registered and analyzed as percentage of corresponding overall maintenance duration of the individual patients.

The pharmacokinetic data (C^p) of the propofol TCI-pump, noninvasive arterial blood pressure, and heart rate (HR) of the S/5 monitor were collected in parallel with the SpE data. Blood pressure measurements were taken at baseline and every 5 min during anesthesia maintenance but not during induction. In addition to study monitoring, routine anesthesia monitoring was carried out including: end-tidal oxygen concentration, pulse oximetry saturation, capnography, and nasopharyngeal temperature.

This study was divided into two phases: 1) the induction phase (lasting 24 min), from the start of induction until the increase of C^pto 6 μg/ml just before intubation, and 2) the maintenance phase, from the start of surgery to the extubation of the patient.

A member of the research team who was blinded to the SpE monitor's output estimated the level of consciousness by using the UMSS21(table 1). UMSS scores were taken at baseline before induction and then during the induction phase at 1-min intervals. The UMSS scores were stored as annotations with a timestamp accuracy of 1 s in a computer file. The annotations were subsequently synchronized with the data files obtained from the S/5 monitor and the drug-related data from the Rugloop system.

The estimated TCI-propofol C^pwas considered to be in pseudo-steady state with the effect site concentration (C^eff) when the C^phad been stable for 7 min. Seven minutes was approximately four times that of12,23,24the plasma effect site equilibration rate constant (k^e0) half-life (1.7 min),25which allowed enough time for C^effto equiliberate with C^p.

The moment that UMSS 2 score changed to UMSS 3 or 4 was determined as the moment of loss of consciousness. Loss of consciousness were determined by means of the timestamps of “UMSS 2” and “UMSS 3 or 4” annotations. The moment of loss of responsiveness (LOR) was defined as the moment when a patient state for the first time was annotated as “does not open eyes or squeeze the hand of the UMSS evaluator when asked.” This moment was determined as being the mean of the timestamps of the annotations “obeys commands” and “does not obey commands.”

The use of extra remifentanil boluses and vasoactive drugs were recorded.

The patients were stratified into three age groups based on the recommendations of the European Medicines Evaluations Agency26: 3–6 yr (youngest), 7–11 yr (middle), and 12–16 yr (oldest). Comparisons between the results for different age groups were carried as outlined below. The normality of the distributions of variables was assessed using the Lilliefors27test. Depending on its outcome we performed one-way ANOVA or Kruskal-Wallis tests for comparison between age groups. A significance level P < 0.05 was used in all tests to accept or reject hypotheses.

Prediction probabilities (P^k)28were calculated during induction between: RE, SE versus  UMSS; RE, SE versus  C^p, C^eff; and RE, SE versus  HR to assess the relationships between these variables. During maintenance the P^kvalues between RE, SE versus  C^eff, and also RE, SE versus  HR, and arterial blood pressure were analyzed. A P^kvalue of 1.0 would indicate that a parameter such as RE was perfectly following changes in the sedation level, whereas a P^k-value of 0.5 would indicate that the prediction was no better than chance alone. In some cases, variables changed inversely to one another such as SpE values increasing with decreasing UMSS values. In such cases a P^kclose to 0 indicated a strong but opposite directional relationship. Consequently, those results were presented as 1-P^kin order to allow easy comparison with other variables' associations.

The mean of the two SpE values recorded in the 10 s immediately preceding each UMSS annotation were used to generate (SpE, UMSS) pair values for calculation of the P^kbetween SpE and UMSS. The same approach was used to calculate (HR, UMSS) and (C^p, UMSS) and (C^eff, UMSS) pairs. We identified the steady state and nonsteady-state periods during the induction for the three increasing steps of propofol. The mean SpE and C^p/C^effvalues for each propofol step for each patient were calculated to compare SpE values and HR values with C^pand C^eff. Thus, for each patient we had three (SpE, C^p/C^eff) pairs for the P^kcalculation. All the observation pairs from all patients were then pooled into one dataset to obtain a mean P^kand its corresponding SE.

All annotations available for each patient were pooled into one large dataset to calculate the prediction probability in order to obtain the associations between SpE and the change of “obeys commands” to “does not obey commands.” The relationships between SpE and UMSS 2 versus  UMSS 3 annotations were obtained by calculating mean values of SpE at each level for each patient then pooling all data.

We detected those periods in the maintenance phase during which C^phad not changed more than 0.01 μg/ml for 7 min or more for the maintenance analysis. The mean of the variables over those stable periods was then calculated to create (SpE, C^eff, etc. ) pairs for each patient. Again all data pairs for all patients were pooled within the appropriate age group to calculate the final P^kvalues.

The pharmacodynamic relationship between SpE values and C^pwas evaluated by employing the E^maxmodel.29 

where E is the recorded entropy value and C the propofol concentration. E⁰is the value at a drug concentration of 0, whereas E^maxis the maximum (suppressive effect) value for E. EC⁵⁰is the drug concentration that corresponds to the half-maximal effect of entropy. γ (γ) is a measure that describes the steepness of the drug concentration-effect relationship (the Hill coefficient).

All C^pand SpE data obtained from all patients during the maintenance phase were pooled and a nonlinear least squares fit using the E^maxmodel as reference function was calculated using Matlab's function nlinfit(). This was done separately for the three specified age groups.

E⁰and E^maxwere specified as 100 for RE and as 91 for SE17The results of the fit were the estimated values of EC⁵⁰and γ together with their 95% CIs and plotted as curves of the equations calculated from these data, in addition to the 95% CIs for predicted values of SE and RE, were calculated when new observations of drug concentrations were obtained in a simulated data range of 0–10 μg/ml with increasing steps of 0.01 μg/ml.

Our main aim was the validation of SpE during TIVA; consequently we did not perform a pre hoc power analysis for detecting SpE differences between age groups at equal sedation or equal C^plevels. The means and standard deviations of entropy at different UMSS levels were calculated. The Kruskal–Wallis test was used to compare the distributions of SpE values among age groups at each UMSS level. For these tests, for each patient one SpE value per UMSS level was used, calculated as the average SpE value observed at that UMSS level. To investigate a possible relationship between EC⁵⁰and age, additional individual fits were performed for the data of each subject separately, again using Matlab's nlinfit() function. Linear correlation and regression analysis was performed to assess possible presence of a relationship.

All patients were included in the analyses. During induction and the maintenance phases all patients had missing data for very short periods. The mean lengths of these missing data were 95, 80, and 75 s per patient for the youngest, middle, and oldest age groups, respectively. The corresponding mean percentages of missing data as a proportion of total recording per age group were 3.2, 2.7, and 2.2%.

The average total doses of propofol in mg/kg/h per age group were (SD; ranges): the youngest 15.9 (1.8; 12.0–18.9), the middle 15.9 (2.2; 13.2–21.3), and the oldest 15.1 (2.0; 11.9–18.0). These doses include both the induction and maintenance doses, and should be interpreted together with the duration of surgery (table 2).

A representative patient case with relevant data are shown in figure 1. Patient demographics are shown in table 2.

The Lilliefors test indicated that the distribution of all the hemodynamic variables could not be assumed to be normal, and thus nonparametric tests were used for further analysis. The length of surgery was not statistically different between age groups analyzed (Kruskal–Wallis test, P = 0.09).

The baseline SpE data collection from all patients was successful. The median baseline values of RE and SE with ranges per age group were: the youngest 97 (89–98) and 88 (83–90), the middle 95 (87–99) and 88 (83–90), and the oldest 97 (92–99) and 88 (85–91), respectively. There were no significant differences in SpE baseline measurements between any age groups (P = 0.70 for SE and P = 0.34 for RE).

The box plots of RE and SE versus  UMSS in the induction phase are shown in figure 2. The P^kvalues during the induction and the maintenance phase in each age group are shown in tables 3and 4, respectively, for: RE, SE, HR, C^p, C^effversus  UMSS; RE, SE, HR versus  C^p; RE, SE, HR versus  C^eff; RE, SE, HR, arterial blood pressure versus  C^eff. Tables 3and 4also include the relationship with hemodynamic variables expressed as prediction probabilities.

The pharmacodynamic relationship between SpE indices and C^pconcentrations data for each age group during maintenance are shown in figure 3. The estimations of TCI propofol EC⁵⁰and γ with 95% CIs are shown in table 5. Based on the relative paucity of C^effdata as compared to C^pdata, we used C^pdata to perform the fits.

Figure 4shows the age dependency of EC⁵⁰values. An associated linear relationship using least mean squares regression is drawn for RE: EC⁵⁰= −0.0826 * age(yr) + 3.17, and for SE. EC⁵⁰= −0.0788 * age(yr) + 3.19. For RE the correlation coefficient between age and EC⁵⁰is −0.363 (P = 0.004), and for SE the correlation coefficient is −0.347 (P = 0.008). The results show that for both RE and SE there were significant differences between age groups at UMSS levels 2, 3, and 4, but not at levels 0 and 1. The figure 5shows the age dependency of the SpE by the plots of SpE values versus  age of the patient at equal UMSS levels.

The P^kanalyses of the specificity of SpE to differentiate between moderate (UMSS 2) and deep sedation (UMSS 3 and 4) and to measure LOR are shown in table 3. Moreover, 55/60 (83%) of patients had BS as defined by the BSR being greater than 1%. The median C^pat appearance and disappearance of BS are shown in table 6, and there is a significant difference (using the ANOVA test) between age groups for appearance (P = 4 × 10⁻¹³) as well as disappearance (P = 5 × 10⁻²⁰) of BS. Table 7shows the mean C^pwith SD and 95% CIs at loss of consciousness and LOR (the significance levels using ANOVA: at loss of consciousness, P = 0.04 and at LOR, P = 0.08). The mean RE and SE with their SD and 95% CIs at LOR are shown in table 8. The means of SpE indices at LOR between age groups are significantly different (for RE, P = 3 × 10⁻⁶and for SE, P = 2 × 10⁻⁵). The incidence of higher SpE values for RE and SE, between 50–60 or above 60, during maintenance are shown in table 9.

The numbers of patients that received rescue analgesic remifentanil boluses during the maintenance period were: 3–6 yr: 9 pts 1 dose and 3 pts 2–3 doses; 7–11 yr: 8 pts 1 dose and 1 pt 3 doses; 12–16 yr: 1 pt 1 dose. Four patients received phenylephrine: one patient three times, one patient twice, and two patients once. One patient received atropine twice. The vasoactive dosing did not happen more commonly during burst suppression.

This study shows that SpE adequately measures the level of hypnosis and also the hypnotic drug effect in 3 to 16 yr old children during propofol induction and propofol-remifentanil TIVA. Although the P^kvalues of other published studies are not strictly comparable,15our P^kcorrelations were generally high, especially during the induction and in the oldest age group.

During the induction with propofol as the only anesthestic, the P^kvalues of SpE indices versus  UMSS were approximately 0.87 in the two younger age groups and 0.93 in the oldest age group. These correlations are high, but as shown by the box plot graphs there were large overlaps of SpE indices with UMSS values at each level of hypnosis.

The concordance of SpE with the hypnotic drug concentration (C^p, C^eff) was better with the level of hypnosis as assessed by UMSS. This may be explained by inter- and intraobserver variation of the UMSS estimation.21Surprisingly, the correlations between SpE versus  C^pwere higher than those between SpE versus  C^eff. It was expected that the (pseudo) effect site concentration would reflect the hypnosis better than the plasma concentration data. This finding may be because of statistical reasons (number of correlation pair observations) and interindividual pharmacokinetic or pharmacodynamic variations. The correlations of C^pand C^effversus  UMSS were higher than SpE versus  UMSS for the youngest age group, but lower for the two older age groups. This may support the speculation that the younger the child, the poorer the performance of the DoH monitor.20 

During the anesthesia maintenance and surgery, only the correlation between SpE indices and C^effwere calculated for simplicity and statistical reasons. Theoretically the C^eff, as a brain effect site concentration, should reflect better the hypnosis as measured by the SpE, even in the pseudo-stable state. The P^kcorrelations between SpE indices and C^effwere much lower than during the induction, 0.86, 0.75, and 0.81 for the youngest, middle and oldest age groups respectively. This is not surprising, as there are several possible reasons for this: the varying surgical stimuli, the combining of a stable remifentanil infusion not titrated to surgical stimuli, the possible pharmacodynamic interactions, the less standardized anesthetic regimen, and methodology compared with the induction. However, the steering of anesthesia cannot solely rely on SpE values because of the overlapping of SpE indices over a range of different UMSS scores.

The nonlinear regression analysis (fig. 3) demonstrated an obvious pharmacodynamic relationship between SpE indices and C^p. The fitting of the classic E^maxmodel is justified and in practice the validity of the fitted curve is good (table 5). Our estimates of propofol EC⁵⁰for SpE were 3.2, 2.38, and 2.05 μg/ml for the youngest, middle, and oldest age groups, respectively. Figure 4indicates the clear dependency of EC⁵⁰on the patient age. In three subjects the obtained EC⁵⁰is very low (smaller than 0.5). This is because of the fact that the maintenance data of these cases contain SE and RE values that are below 15, thus leading to an almost flat C^p-Entropy curve that made estimation of the equation parameters very difficult.

We used the C^pdata instead of the C^effdata for pharmacodynamic modeling because in this study there were only relatively few C^effvalues at different SpE levels. In contrast, the much higher number of C^pvalues made the use of C^pmore suitable for pharmacodynamic modeling, although the C^effvalues would probably reflect the real pharmacodynamic effect better, and take into consideration the possible hysteresis.

In two pediatric Bispectral Index® (BIS) studies, Munoz et al.  30found TCI propofol (Paedfusor pharmacokinetic model) EC^{50(effect site)}of 3.65 μg/ml for children aged over 3–11 yr. In contrast, Rigouzzo et al.  12infused TCI propofol (Kataria pharmacokinetic model) and obtained an estimate and a measurement of propofol of EC⁵⁰of 2.9 and 4.0 μg/ml for children aged 3–10 yr. The variation between these studies can partly be explained by differences regarding: the electroencephalogram monitor, the pharmacokinetic model, anesthetic regimen, and the differences in age ranges between the studies. Our study demonstrated none of the paradoxal increases of SpE indices as the hypnosis deepens that has been reported for BIS during sevoflurane inhalation anesthesia in children.31 

Detecting the difference between moderate and deep levels of sedation is challenging. We studied the effectiveness of SpE to discriminate between moderate (UMSS 2) and deep sedation/anesthesia (UMSS 3 and 4), and to detect the LOR. The ability of SpE to differentiate between moderate and deep sedation/general anesthesia was high for the oldest age group and moderate in the middle age group. Recently, Malviya et al.  32concluded that BIS poorly differentiated between moderate and deep levels of sedation in all age groups in a combined secondary analysis of four independent studies evaluating age- and sedative agent-related differences in BIS in a large sample of children younger than 18 yr. However, the SpE values in the youngest age group tended to be higher at the UMSS 1–3 sedation levels and then fell more steeply with the change from UMSS 3 to 4 when compared with the two older age groups. This indicates a poor discrimination between the light to deep sedation levels (UMSS 1 to 3) in the youngest age group.

The relationships between HR versus  UMSS, C^pand C^effduring induction, and those between HR and arterial blood pressure versus  C^effduring maintenance assessed by P^k, were weak. These data suggest that hemodynamic variables do not reflect the level of hypnosis or the hypnotic drug effect on children during TIVA. This has also been previously shown for children undergoing inhalation anesthesia.20,33 

The incidence of BS was demonstrably and accurately detected by the BSR in another study.34Without SpE guidance, 83% of the patients in our study reached BS that corresponded to a very deep level of hypnosis. There are preliminary data that suggest deep anesthesia in adults to be associated with a long-term negative outcome.35,–,38Moreover, laboratory neonatal animal data indicate that toxicity of anesthetics for the developing brain might warrant minimizing the anesthetics dosage. Higher levels of SpE values were not registered during anesthesia maintenance and emergence (last 15 min of surgery) period except for the youngest age group. Our BSR occurrence and disappearance in addition to the higher SpE data suggest that clinically one should not probably aim for propofol C^plevels higher than 3.5–5 μg/ml. It should be noted these values depend on the patient's age and also assumes that adequate analgesia has been ascertained. This is in concordance with pediatric pharmacodynamic data12,30and also with the optimal propofol and opioid concentrations in adults proposed by Vuyk et al.  39 

On the other hand, our C^pvalues at the loss of consciousness or LOR (without surgical stimulus in this study), may hint at the minimum dosing of propofol. Our data suggest a lowest C^plevel of approximately 2.9 μg/ml, and again this assumes that adequate analgesia by adjuvant methods is used.

Measuring precisely the anesthetic drug concentrations per se  may prevent awareness as reported by Avidan et al.  40However, those authors measured end-tidal concentrations of inhalation anesthetics in adults, and their data should not be extrapolated to TIVA and to pediatric patients. Our study data suggest that the TCI is a valuable tool in guiding the level of hypnosis, as the associations of C^pwith UMSS were good and equal to the correlations of SpE against UMSS.

The need for rescue boluses of remifentanil on a clinical basis was quite common in the two youngest age groups, but rare in the oldest age group. This finding may indicate a different pharmacokinetic and/or pharmacodynamic action of the remifentanil-propofol combination between the age groups.

Our results show significant differences in SpE values between the age groups at the corresponding UMSS levels 2–4. The patients aged 3–6 yr were deeply sedated (= UMSS 3) at SpE values of 80, and half of them were under general anesthesia (= UMSS 4) at SpE indices over 60. So the higher-than-recommended SpE values for surgical anesthestic depth (for adults) by the manufacturer do not indicate an inappropriate anesthesia in this age group. The younger the patients in the 3–16 yr age group, the higher were their SpE indices at clinically similar sedation scores (fig. 5). This is a clinically important finding and is consistent with previous studies on the age dependency of patients for DoH-monitors.31,41,–,44As no a priori  power calculations were conducted and the age dependent differences in SpE values over the total UMSS scale were not large, statistical power of our study may have been an issue. The current evidence strongly suggests that a calibration of DoH-monitors for patient age is warranted.

This was the first study to assess SpE during TCI and TIVA in children; thus there are no previous results for comparison. SpE has been studied in children during inhalation anesthesia18,–,20and demonstrated reasonable performance in children but not in infants. A comparative accuracy between SpE and BIS was also found. In a more detailed comparison between our present and previous20studies, SpE seems to measure the depth of hypnosis better during TIVA than during inhalation anesthesia. However, this disparity may be caused by differences in patient age selection, as older patients were investigated in the present study, and other methodological issues, such as end-tidal sevoflurane being used instead of effect site sevoflurane in the previous study.

In contrast to studies on SpE in children during TIVA, BIS12,30,43,45,46and Narcotrend index44have been investigated during TCI or TIVA anesthesia in children.

Tirel et al.  43assessed the relationship between BIS and C^pof TCI propofol (Kataria pharmacokinetic model), combined with remifentanil infusion during surgery in 50 children aged 3–15 yr. Their multiple correspondence analysis demonstrated a difference in BIS values between C^p2 and 4 μg/ml, but not between 4 and 6 μg/ml. An explanation for the difference between the BS data reported by Tirel et al.  43and our data might be that their patients probably had already reached BS at the C^p4 μg/ml level, which would have attenuated the difference in BIS between the two higher propofol levels.

The study by Rigouzzo et al.  12deserves special attention. They compared the estimated TCI propofol plasma pseudo-steady state concentrations (C^t) and measured propofol concentrations of blood samples (C^m) in 45 children aged 6–13 yr using Kataria pharmacokinetic model and the corresponding parameters of 45 adults aged 14–32 yr using Schnider pharmacokinetic model. They found that the pharmacokinetic models of the TCI pumps systematically underestimated the real C^mlevels. This bias increased as the concentration of propofol increased. Their pharmacodynamic E^maxmodel showed EC⁵⁰(estimated/measured) 2.9/4.0 μg/ml in children and 2.6/3.3 μg/ml in adults. They speculated that the higher need for propofol in children compared to adults to reach the same BIS level might be caused by both pharmacokinetic and pharmacodynamic factors. Rigouzzo et al.  used BIS in their study, whereas SpE was determined in our study. If their results on the real propofol concentrations were to be extrapolated to our study, our E^maxmodel curve would be shifted to the right and become more shallow, and the BS would be reached at higher concentrations.

Various pharmacokinetic and pharmacodynamic factors may cause bias. The main limitation of our study was the lack of measured propofol concentrations of blood samples. The use of estimated plasma and effect site concentrations instead of measured plasma concentrations may have impact especially on the results regarding the pharmacodynamic modeling and the observed age dependency of propofol EC⁵⁰. The effect of remifentanil on SpE values is not known, and the possible pharmacokinetic and pharmacodynamic interactions between propofol and remifentanil cannot be ruled out. Both before mentioned issues may have different effects depending on the age of the patient. Moreover, the data regarding these issues in adults are conflicting and in children even lacking. We assume that interactions between propofol, lidocaine, and midazolam should not have any major impact on our results. The growth and maturation of the central nervous system and their effects on the electroencephalogram may have impacts on the results between different age groups, although these factors are probably not as important as they are in infants and toddlers.

In addition to known problems in pediatric depth of hypnosis research,47we rely on many surrogate measures. There is no exact definition for anesthesia or level of hypnosis. The UMSS probably represents the best validated sedation scale in pediatrics. As the hypnotic drug effect cannot be precisely defined and the effect site concentrations of hypnotics cannot be measured, we have to estimate those concentrations based on pharmacokinetic modeling. We used the equilibrium time of 7 min, which approximately equals four times the only known estimation of plasma-effect site equilibration half-life in children25assessed with A-Line ARX-index® (Danmeter A/S, Odense, Denmark).

Probably the most important aim for future studies and for the manufacturers of DoH monitors, initially designed for adults, is the calibration of these monitors for pediatric patients of various ages. Our study among numerous other pediatric studies emphasizes the age dependency aspect of the calculated DoH indices. The ability to steer anesthesia on the basis of a pediatricly validated effect site concentration would be of great scientific and clinical significance. Research attention should be especially paid to the very young age groups for which the benefits of hypnosis monitoring may be of the highest value.

This study found a correlation of SpE indices with the level of hypnosis as assessed by UMSS sedation scale in addition to estimated propofol concentrations during TCI propofol induction and TCI propofol-remifentanil maintenance in children aged 3–16 yr. There was an age dependency in respect of SpE values. Pediatric anesthesia cannot be steered solely on the basis of SpE values.

The authors thank Seppo Ranta, M.D., Ph.D., Datawell Ltd, Espoo, Finland; the Instru Foundation, Helsinki, Finland; and the Department of Anesthesiology and Intensive Care Medicine, Helsinki University Central Hospital, Helsinki, Finland.