Geographic variation in drug response could arise from genetics and a variety of environmental factors that often are difficult to define or quantify. Despite evidence that propofol requirements and recovery times vary between patients living in different geographic locations,1,2 regional variation is still not acknowledged as a factor that could influence drug effect. This has been a relatively neglected area of investigation. Despite the fact that authorities in some countries, e.g., Japan, Korea, and People’s Republic of China, require information on local populations before drug registration, drugs remain licensed under the same recommended dose for patients in different geographic locations. The local licensing authorities accept data from a limited number of European or North American Caucasian subjects and do not demand specific information on precise dosage efficacy and toxicity from patients in their geographic location.

The bispectral index (BIS) is a processed electroencephalographic (EEG) parameter derived from an empirical database using a complex proprietary algorithm that combines three subparameters into a single metric.3 It quantifies EEG response to anesthetics. The DiprifusorTM (AstraZeneca Pharmaceuticals, Macclesfield, UK) target-controlled infusion (TCI) pump uses a pharmacokinetic model to achieve and maintain a set propofol plasma concentration4; however, plasma is not the site of drug effect. Thus, the Diprifusor displays both the estimated plasma concentration (Cp) and the predicted effect-site concentration (Ce).4

There are studies wherein estimated concentrations of propofol have been examined at peak BIS effect using a TCI technique in Chinese populations.5,6 Other studies using logistic regression sigmoid Emax curves individually assessed the propofol estimated effective concentration (EC50) at which 50% of Caucasian7 or Chinese patients8 did not respond to skin incision. We are not aware of previous studies directly comparing the performance of the widely used Diprifusor system, which incorporates the commonly used Marsh pharmacokinetic model,9 between Chinese patients in China and Caucasian patients in Europe. There are data in the literature showing that estimated concentrations according to the Marsh pharmacokinetic model did not accurately reflect measured blood concentrations in the Chinese population,10 or in other populations,11 but no model leads to accurate predictions. Thus, we assumed that the Marsh pharmacokinetic model might not apply to either the Chinese or the Austrian group, even if it was developed in a European population. The propofol effect was measured using the widely available BIS monitor, which was developed primarily in a North American population.3 Thus, we sought to examine whether the administration of propofol using a TCI technique based on the popular Marsh pharmacokinetic model,9 and thereby providing the same estimated concentration to both groups, resulted in the same dynamic end points, such as loss of consciousness (LOC), in patients with different geographic origins, i.e., Chinese patients in China and Caucasian patients in Austria. The purpose of our study was to compare the estimated plasma concentration-response (Cp-BIS) and the predicted effect site concentration-response (Ce-BIS) in a group of Caucasian patients in Austria vs a group of Han Chinese patients in China.

Methods

Our study was registered at the European Clinical Trials Database, EudraCT (trial registration number: 2009-015582-31). Our report was prepared in conformity with the guidelines of the Consolidated Standards of Reporting Trials (CONSORT) statement.12 A prospective controlled study was conducted at the Medical University of Graz, Austria and the Zhe Jiang University hospitals, People’s Republic of China from September 2008 to March 2010. After approval by the ethics committees, all patients who agreed to participate in the study gave written informed consent. As health differences were a major concern, we delineated the health status between groups by including only those patients who were American Society of Anesthesiologists (ASA) classification I and II with body mass index 18-26 kg·m−2 and aged 30-50 yr. We excluded patients with alcohol, tobacco, or substance abuse, or medical conditions that might affect their level of consciousness, such as stroke, stupor, or encephalopathies. Thirty consecutive (15 male/15 female) second-generation Han Chinese patients, who were living in China for at least ten years and scheduled to undergo general anesthesia for orthopedic or general surgery, were recruited at the Zhe Jiang University hospitals. The Chinese patients were matched with 30 second-generation Caucasian patients who were living in Austria for at least ten years, met the same sex, ASA classification, and approximate age and weight criteria, and were recruited at the Medical University of Graz, Austria. The Caucasian patients were classified as members of the Indo-European white race from supposed origin in the Caucasus.

A BIS Quatro Sensor (Aspect Medical Systems, Newton, MA, USA) was placed on each patient’s forehead according to manufacturer’s guidelines and connected to a BIS-XP monitor. The raw EEG signal was band-pass filtered to 2-70 Hz and processed in real time using version 3.4 of the BIS algorithm. In addition, the BIS monitor displayed the electromyographic (EMG) activity calculated in decibels (dB) in the 70-110 Hz frequency band. The BIS recordings were started after verifying a signal quality index > 95%, electrode impedance < 5 kΩ, and EMG values < 35 dB.

The patients received no premedication prior to the study. Propofol was obtained from the same AstraZeneca Pharmaceuticals European production line, and the same study protocol was strictly implemented in both countries. A propofol DiprifusorTM (software version 2) was used that incorporated the Marsh pharmacokinetic model.9 A target-controlled infusion was started after the patients’ anthropometric data were entered. In all patients, the Diprifusor Cp was set to reach a target of 5 μg·mL−1 gradually over five minutes by initially setting the target at 1 μg·mL−1 and increasing the target by 1 μg·mL−1 increments every minute. Thus, the estimated Cp variation with time was the same for all patients. The exit rate constant out of the effect site (keo) was set at 0.26 min−1 to obtain effect-site concentrations (Ce). After 5 min, the estimated Cp was maintained at 5 μg·mL−1 with no further changes. To eliminate differences in the observers’ approaches to LOC, we used a consistent set of identical verbal commands in the respective languages. With a gentle tap on the shoulder every ten seconds, patients were asked whether they were still awake until a score of one was reached on the Observer’s Assessment of Alertness and Sedation (OAA/S) Scale, i.e., no eyelash reflex and no response to verbal commands, which defined LOC.5 The patients’ lungs were ventilated via a facemask. Digitized BIS values, estimated Cp, and predicted Ce were collected continuously and stored on a laptop computer for the duration of the study. Stable BIS values that showed no further decline and remained within ± 5 of the previous BIS value were considered an indicator of a steady state propofol concentration. A remifentanil infusion of 0.1-0.3 μg·kg−1·min−1 was then started, and rocuronium 600 μg·kg−1 was administered for tracheal intubation, which marked the end of our study recording period. The BIS values showing artifacts, which are very clearly displayed as such in the off-line format, were identified and eliminated in the off-line analysis. Patients were warmed using a forced hot-air blanket, and mean arterial pressure and heart rate were recorded continuously.

Statistical analysis

Using the NONMEM Scientist® program (MicroMath Scientific Software Inc., Saint Louis, MO, USA), propofol estimated Cp-BIS and predicted Ce-BIS relationships were modelled using a non-linear regression analysis (NLR) via a non-linear mixed effects modelling (NONMEM) sigmoid Emax pharmacokinetic-pharmacodynamic model.13 The goodness of fit of each patient was determined using Akaike’s information criterion.14 Each patient’s Cp-BIS and Ce-BIS data points were fitted to a sigmoid Emax mathematical model of the nature:

$$ {\text{E}} = \left( {{\text{E}}_{\text{baseline}} -\,{\text{E}}_{ \max } } \right){\frac{{{\text{C}}^{\gamma } }}{{{\text{C}}_{50}^{\gamma } + \,{\text{C}}^{\gamma } }}} $$

where Ebaseline is the BIS value before any propofol administration (BIS = 100); Emax is the BIS value at maximal propofol effect, i.e., when steady state Cp = 5 μg·mL−1 is reached; C50 is the estimated concentration when 50% of Emax is obtained; and γ is a Hill coefficient sigmoidicity factor, i.e., the slope of the sigmoid curve. In all patients over the five-minute period, we obtained a consistent set of BIS vs estimated Cp data points on the sigmoid curves, which were used to create the sigmoid curves of the two geographical locations.

Our primary outcome measurement was Emax (lowest BIS value at estimated Cp = 5 μg·mL−1). We performed an interim power analysis Student’s t test (α = 0.05) based on the first ten Austrian pilot patients who were included in the final analysis, which indicated a mean BIS Emax of 54.8 ± 6.4 compared with 43.2 ± 2.6 for ten Chinese patients. This result showed that a group size of 25 patients would be required to reveal a statistically significant difference between the two groups with > 90% power.

We used repeated measures analysis of variance (ANOVA) to compare parameter differences over time (group and time factors), and we used Dunnett’s two-sided multiple comparison post hoc test to compare BIS values at different propofol estimated plasma concentrations. Data were expressed as means ± SD, with P < 0.05 considered statistically significant.

Results

Patients’ demographic data are presented in Table 1. Compared with the Austrian group, the Cp-BIS (Fig. 1) and Ce-BIS (Fig. 2) sigmoid Emax curves were shifted significantly to the left in the Chinese group. The lowest BIS value at the propofol estimated Cp of 5 μg·mL−1 was less in the Chinese group than in the Austrian group (47.2 ± 3.6 vs 63.6 ± 5.4, respectively; P = 0.0006).

Table 1 Austrian and Chinese patients’ demographics
Full size table
Fig. 1
figure 1

Bispectral index (BIS) vs propofol estimated plasma concentration (Cp) sigmoid Emax curve model (95% confidence intervals) in the Austrian (n = 30) and Chinese groups (n = 30)

Full size image
Fig. 2
figure 2

Bispectral index (BIS) vs propofol predicted effect-site concentration (Ce) sigmoid Emax curve model (95% confidence intervals) in the Austrian (n = 30) and Chinese groups (n = 30)

Full size image

The propofol estimated Cp at LOC, the predicted Ce at LOC, and time to LOC were significantly less in the Chinese group than in the Austrian group; however patients of both groups lost consciousness at similar BIS values (Table 2). Two-way ANOVA (group x time) revealed differences (P < 0.0001) in BIS values over time. Dunnett’s two-sided multiple comparison post hoc test revealed differences between the two groups (P < 0.001) in BIS values starting from the propofol estimated Cp = 2.0 μg mL−1.

Table 2 Time to loss of consciousness, propofol estimated plasma concentration, predicted effect site concentration, and bispectral index at loss of consciousness in the Austrian and Chinese patients
Full size table

Before induction, the mean arterial pressure was greater in the Chinese patients than in the Austrian patients (99 ± 15 mmHg vs 90 ± 14 mmHg); however, after induction, there was a similar decrease with time in both groups to 90 ± 15 mmHg vs 80 ± 15 mmHg, respectively (two-way ANOVA). Heart rate was almost identical in the Chinese and Austrian groups both before and after induction (77 ± 16 beats·min−1 vs 76 ± 12 beats·min−1, respectively, and 74 ± 14 beats·min−1 vs 75 ± 10 beats·min−1, respectively).

Discussion

Using a similar estimated Cp based on the Marsh pharmacokinetic model, which incorporates patients’ age, weight, and height, we demonstrated that Chinese patients lost consciousness 1.1 min earlier than their Austrian counterparts. This result corresponds with an estimated propofol Cp that was 0.8 μg·mL−1 lower. Interestingly, patients of both groups lost consciousness at similar BIS values. This could be interpreted either as Chinese patients responding more to the same estimated propofol concentration as their Austrian counterparts (a pharmacodynamic explanation), or as Chinese individuals achieving a greater propofol plasma concentration following the same Diprifusor infusion (a pharmacokinetic explanation), or a combination of both factors. This issue could be resolved only by measurement of propofol blood concentrations. Irrespective of the true explanation, the Diprifusor concentration settings should be reduced in Chinese patients. The pharmacokinetic model used to drive the Diprifusor was probably erroneous in both populations, because it missed the target BIS value of 40 in both groups (BIS 47 in Chinese patients and BIS 64 in Austrian patients).

We demonstrated a significant difference in both the propofol estimated Cp-BIS as well as the propofol predicted Ce-BIS response relationships between the Austrian and Chinese patients. The Ce-BIS is probably the more clinically relevant of these, because the site of action of propofol on BIS, an EEG-derived parameter, is the brain, not the plasma. Our results seem to concur with the results of a previous study in which recovery time from anesthesia that was maintained at similar propofol infusion rates was shorter in Caucasian than in Chinese patients.1 Our group also demonstrated a geographic difference concerning a neuromuscular blocking agent, as the duration of action of rocuronium was significantly longer in Han Chinese patients in China than in Caucasians in Austria.15

On the other hand, our study results do not concur with other individual studies that failed to reveal or precisely quantify differences between Caucasian and Chinese populations.7,8 Two separate studies conducted by the same team implementing the same study design reported almost identical results for Caucasians in the United Kingdom7 and Chinese patients in Hong Kong.8 The estimated EC50 at LOC was 2.8 μg·mL−1 in Caucasian patients in the United Kingdom7 and 2.7 μg·mL−1 in Chinese patients in Hong Kong.8 However, this could also indicate that a pharmacokinetic explanation seems to be a more plausible explanation rather than a pharmacodynamic one.

Despite the fact that patients in a single geographic region share certain traits, both genetic and environmental, in today’s multicultural world, populations from different or even mixed ethnic origins could be living in distant geographic locations. Thus, we tried to identify two ethnic populations that share the same factors that could affect propofol distribution, such as diet, pollution, physical exercise, body composition, fat distribution, and skeletal muscle fibre type proportion. Our study was an attempt to evaluate the “collective effect” of these combined ethnic/environmental variables. It was not the aim of our study to determine which variables might have contributed to our findings. That being said, the differences between the Chinese and Caucasian patients we encountered in our study are likely to be multifactorial.

A possible explanation of our results is the degree of plasma protein binding in both populations. Only the unbound free fraction can exert a pharmacological effect. However changes in plasma protein binding were shown to have very little clinical relevance,16–18 except for a small group of drugs, such as propofol, that are extensively protein bound (>96%) and have a very large volume of distribution.19,20 This was clearly demonstrated when sudden cardiopulmonary bypass hemodilution resulted in a twofold increase in the unbound propofol with an immediate decrease in BIS values, despite propofol total plasma concentration remaining unchanged.21 Compared with the Caucasians, the Chinese subjects were shown to have different plasma protein binding to drugs, as they possess lower α-1-acid glycoprotein plasma concentrations, a glycoprotein shown to accelerate recovery from drugs by increasing protein binding.22

The effect of propofol, with its high hepatic extraction ratio23 is expected to depend on hemodynamic changes during induction. In our study, hemodynamic parameters did not seem to play a significant role, as there were no significant differences between the two groups. However, cardiac output was not measured, and this factor cannot be ruled out to account for faster response in Chinese patients.24

Our study has limitations, as we did not confirm the actual propofol concentrations through blood samples. Furthermore, as we only assessed the Marsh pharmacokinetic model,9 it would have been interesting to see how other TCI models, such as the Schnider model,25 would describe both populations.

Finally, the significant differences in propofol BIS responses to similar infusion regimens would necessitate either a TCI pharmacokinetic model designed especially for a Chinese population or a change in target concentrations using the same model. Given that there is probably already considerable and often unpredictable variability even among individuals of the same ethnic origin, TCI propofol should be titrated to clinical effect.

In conclusion, we demonstrated significant differences between Han Chinese and Austrian Caucasians in BIS responses to similar propofol infusion regimens. Given our observed results of clear differences between the two populations, we recommend larger studies in different geographic locations to prepare a map of differential propofol TCI infusion models, in addition to those already established predominantly for the Caucasian population. Larger studies of propofol clinical pharmacology are warranted to explore all potential covariates that may contribute to our understanding of these differences, along with a thorough analysis of why these differences occur and in which ethnic groups these covariates are relevant.