Previously published midazolam–alfentanil response surface model cannot predict patient response well in gastrointestinal endoscopy sedation : Journal of the Chinese Medical Association

Secondary Logo

Journal Logo

Original Article

Previously published midazolam–alfentanil response surface model cannot predict patient response well in gastrointestinal endoscopy sedation

Liou, Jing-Yanga; Ting, Chien-Kuna,b; Huang, Yu-Yingc; Tsou, Mei-Yunga,b,*

Author Information
Journal of the Chinese Medical Association: March 2016 - Volume 79 - Issue 3 - p 146-151
doi: 10.1016/j.jcma.2015.11.002


    1. Introduction

    Drug interactions have always been an important issue in daily anesthesia practice. Traditionally, isobolographic analysis is used to describe drug interactions, which can be characterized as additive, synergistic, or infra-additive (antagonistic).1 Isobologram is limited to presenting drug interactions at a specified response endpoint, for example, 50% chance of movement during laryngoscopy. The response surface model is a combination of the drug concentration–effect relation and the isobologram. It displays drug effects in a wide range of drug concentrations for two or more drugs.2,3 Various anesthetic combinations have already been evaluated, including hypnotic–hypnotic,4,5 opioid–hypnotic,6–8 and analgesic–analgesic9 pairs.

    The combination of midazolam and alfentanil can be used in some surgical procedures and examinations requiring moderate sedation and analgesia.10,11 Both drugs are still used very commonly. Few studies have investigated the response surface model for midazolam–alfentanil interaction.2,3 Minto et al3 used the volunteer data from Short et al12 and developed a response surface model for hypnosis without stimulus using midazolam and alfentanil. The aim of this study was to validate the accuracy of this published response surface model during the quiescent phases in gastrointestinal endoscopic procedure sedation.

    2. Methods

    2.1. Patient selection and anesthesia

    After approval from the Institutional Review Board (IRB) at Taipei Veterans General Hospital, Taipei, Taiwan (IRB 2014-12-001BC), 40 adults—aged < 65 years—scheduled for combined esophagogastroduodenoscopy (EGD) and colonoscopy were enrolled. All patients had documented written consent. Patients were assessed as being at a physical status of I or II, according to the American Society of Anesthesiologists classification system. Exclusion criteria included hearing impairment, neurologic or behavioral disorders, habitual sedative use, and allergy to midazolam or alfentanil. Strict fasting and colon preparation protocols were followed. A 22-gauge intravenous catheter was secured for drug administration. Each patient received standard anesthetic care monitoring comprising electrocardiography, pulse oximetry, and noninvasive blood pressure monitoring. Supplemental oxygen was given via a nasal cannula, with the SpO2 being maintained above 90%. Bolus intravenous doses of midazolam and alfentanil were administered by an experienced anesthesiologist. The patient was monitored with an auditory evoked potential monitor (AEP Monitor/2; Danmeter A/S, Odense, Denmark). Instrumentation began after a loss of response, as evaluated by the anesthesiologist, or an A-line auditory evoked potential index (AAI) of <60. The mean auditory evoked potential index values for various Observer's Assessment of Alertness/Sedation (OAA/S) scores were 81.2 at Score 5, 63.2 for Score 4, 48.8 for Score 3, 36.5 for Score 2, and 29 for Score 1 in patients undergoing gastrointestinal endoscopy sedation. According to the manufacturer of auditory evoked potential monitor monitors, an auditory evoked potential index value of >60 is indicative of the awake state.13 Intolerable desaturation was managed with mask ventilation or insertion of a nasal airway. Additional alfentanil boluses were given if the patient expressed pain or showed facial expressions of pain. Midazolam boluses were given if the patient had an OAA/S score of ≥4 with or without pain expressions. EGD was performed first, followed by colonoscopy. At the end of the procedure, the patient was observed until verbal arousal was possible. Sedation OAA/S (Table 1) scores were recorded by an independent observer. Patients with an OAA/S score of ≥4 were designated as “awake”. Each patient's response to a specific concentration of the midazolam and alfentanil pair was recorded during induction and emergence.

    Table 1:
    Observer's Assessment of Alertness/Sedation scale.a

    2.2. Response surface model

    Using SigmaPlot 12.5 (Systat Software, Inc., San Jose, CA, USA), patient response was calculated by a midazolam–alfentanil response surface model published by Minto et al3 [Eq. (1)].

    E represents the drug effect, which is the probability of a loss of response. It ranges from 0 to 1, with 0 indicating no drug effect and the patient having 100% probability of response, and 1 indicating no response to stimuli. Emax(Θ) is defined as the maximal drug effect (effect to achieve an OAA/S score of <4), whereas E0 is the baseline effect when no drug is present. Their values are designated as 1 and 0 for Emax and E0, respectively, to simplify the equation. C50 stands for the effective drug concentration that is required to achieve 50% maximal effect. U is the unitless normalized potency of the drug relative to a plasma concentration of 50% drug effect [Eq. (2)].

    The model introduces a central concept, Θ, to represent a new drug as a ratio of the drugs under investigation [Eq. (3)]. The term Θ should not be confused with an actual measurable drug concentration; it is a concept developed for the model parameters. The range of Θ varies from 0 (only midazolam present) to 1 (only alfentanil present).

    In our research, the drugs under investigation were midazolam (Symbol) and alfentanil (Symbol); γ is the sigmoidicity factor, a function of Θ, that determines the steepness of the effect. U50(Θ) is the potency of the new drug, at ratio Θ, which yields half the maximal response. It can be calculated according to Eq. (4):


    The parameter Symbol is an interaction parameter that originated from a fourth-order polynomial function, as described by Minto et al.3 When the value is 0, it denotes an additive effect. It is synergistic or antagonistic when >0 or <0, respectively.


    Pharmacokinetic profiles, including plasma and effector-site concentrations, were estimated using the TIVA trainer simulation program (version 8, Build5, Guttabv, EuroSIVA, Netherlands). We used the Maitre et al's14 model for alfentanil and the Zomorodi et al's15 model for midazolam.

    The C50 values given by Minto et al3 were 0.144 mg/kg for midazolam and 0.0936 mg/kg for alfentanil. The reported response surface model was built on the raw data given by Short et al,12 and the volunteers were all females. By inputting average height, weight, and age into the simulation software, the estimated maximum plasma concentration at such a dosage yielded 311 ng/mL for midazolam and 705 ng/mL for alfentanil. A modified response surface model was then constructed using the above approximations. The investigation endpoint was the patient's arousal status as binary data, E = 1 for OAA/S score <4 or E = 0 for OAA/S score ≥4. The primary aim was to validate the model in terms of accuracy. Accuracy of prediction is assessed by calculating the difference between the true response (OAA/S scores < 4 and ≥ 4) and the model-predicted probability. The model is considered “accurate” if the difference is < 0.5. The total percentage of accurate predictions was obtained for both the original (mg/kg) and the modified (ng/mL) response surface models.

    3. Results

    Forty patients were initially enrolled; however, five were excluded, two because of previous known neurologic diseases and three because of inadequate records. The demographic data are summarized in Table 2. EGD and colonoscopy were completed in all our patients. Forty-one observations were eligible for pooling among the 35 patients studied. Six additional observations were available from the period between EGD and colonoscopy when the patient became arousable and the OAA/S score was ≥4. Each observation produces two concentration pairs, one is at the maximal model-predicted chance of an OAA/S score of <4 (E = 1, which all occurred after initial boluses during induction) and the other is at true patient arousal at the end of the procedure (OAA/S score ≥ 4 or E = 0). A total of 82 concentration pairs for midazolam and alfentanil were available. The average body mass index was 22.1 ± 2.3 kg/m2. The mean cumulative dose for midazolam was 0.047 ± 0.015 (standard deviation) mg/kg and that for alfentanil was 0.012 ± 0.004 mg/kg. The plasma concentration ranged from 27 ng/mL to 112 ng/mL and from 12 ng/mL to 106 ng/mL for midazolam and alfentanil, respectively. Accuracy of prediction was 0% for the OAA/S score <4 group, since the difference between the true response and prediction was >0.5 in all 41 observations (Fig. 1). Drug synergy for producing hypnosis is clearly delineated as bowing of the isoboles toward the origin. The estimated probability of loss of response ranged from 0.04% to 2.94% at the time of arousal (OAA/S score ≥ 4) and from 0.24% to 15.55% when the patient was asleep at maximal model-predicted probability (OAA/S score < 4). All the observations were below the 50% probability isobole.

    Table 2:
    Patient demographic data.
    Fig. 1:
    Contour graph of the modified response surface model versus plasma concentration. Solid lines represent the 95%, 50%, and 5% chances of patients having an OAA/S score of <4. All the observations lie below the 50% line, regardless of the patient's actual response. OAA/S = Observer's Assessment of Alertness and Sedation scale.

    The modified response surface model is shown in Fig. 2. Significant synergy is seen between midazolam and alfentanil. At a commonly used dosage, alfentanil alone is insufficient to produce an OAA/S score of <4. Maximum drug concentrations at an OAA/S score of <4 are plotted, which show a large discrepancy between the predicted and true patient responses (Figs. 2 and 3).

    Fig. 2:
    The modified response surface model for midazolam–alfentanil interaction showing significant synergy between the two drugs. As shown by the model surface, an OAA/S score of <4 cannot be reliably achieved with alfentanil alone, even at high doses. OAA/S = Observer's Assessment of Alertness and Sedation scale.
    Fig. 3:
    The modified response surface model is rotated to a different perspective. Black circles are maximum drug concentrations. All patients have attained an OAA/S score of <4 despite the model predicting that they would not, even at plasma drug concentrations with maximal model-predicted probability of an OAA/S score of <4. OAA/S = Observer's Assessment of Alertness and Sedation scale.

    The original response surface model in units of mg/kg is plotted as a contour graph. The cumulative patient drug dose is marked as black dots (Fig. 4). Two out of 35 patients (6%) reached > 50% predicted probability of an OAA/S score of <4 using the model. One patient had difficult EGD and colonoscopy, and the examination time was much longer than average. The second patient was not remarkably different from others, but was simply manifested as more resistant to midazolam and alfentanil and required more boluses.

    Fig. 4:
    This is the contour graph derived from the original response surface model. Dashed lines represent the 95%, 50%, and 5% probabilities of an OAA/S score of <4. The cumulative drug doses are mostly scattered below the 50% isobole. Only two out of 41 observations were considered accurate (difference between true and predicted probability <0.5). OAA/S = Observer's Assessment of Alertness and Sedation scale.

    4. Discussion

    In this study, we assessed the ability of a previously reported response surface model of midazolam–alfentanil to predict patient response during gastrointestinal endoscopy sedation and wake-up time. The model reported by Minto et al3 is in a class by itself and is the only available response surface model for the midazolam–alfentanil interaction to date. The model, which predicts hypnosis, was built based on the raw data collected by Short et al12 from 400 Chinese female patients with an average weight of 50.3–54.1 kg and an average age of 30.6–32.6 years. In the literature, hypnosis was defined as failure to open the eyes on verbal command. This was equivalent to an OAA/S score of <4 in our study. Our population's average body weight and age were 60.54 kg and 49.1 years, respectively. The derived plasma concentration differed by < 10% for the age range of 29–60 years and body weight range of 40–60 kg under simulation for both drugs.

    A midazolam and opioid combination has been used extensively for moderate conscious sedation during various examinations11,16–20 by both anesthesiologists and nonanesthesiologists. The previously reported C50 values for midazolam and alfentanil were 0.144 mg/kg and 0.0936 mg/kg, respectively.3 For a 60-kg patient, that would translate into 8.64 mg midazolam and 5616 μg alfentanil. The single-bolus dosage of alfentanil is extremely high and rarely used in modern anesthesia practice, especially for such a short procedure. Using the model, the predicted probability for hypnosis was inaccurate for all doses. All our observed patients have a <20% chance of being asleep (OAA/S score < 4). This implied that patients should be awake throughout the entire procedure, although actually most of the patients fell asleep in our 41 observations.

    Pharmacokinetic studies on alfentanil reported a large discrepancy in C50 concentration. Jhaveri et al21 reported a loss of response C50 of 1012 ng/mL for loss of consciousness in healthy volunteers when alfentanil was the sole agent used. Loss of consciousness was defined by failure to respond after three consecutive commands to take a deep breath. Plasma concentration for suppressing response to skin incision in combination with 70% N2O was 200–300 ng/mL.22 Ausems et al23 also reported a similar plasma C50 when supplemented with 66% N2O. The C50 for alfentanil was 475 ± 28 ng/mL for tracheal intubation, 279 ± 20 ng/mL for skin incision, and 150 ± 23 ng/mL for skin closure. Vuyk et al24 studied the interaction between propofol and alfentanil during lower abdominal surgery in women. At low propofol plasma concentrations (2 μg/mL), the C50 for alfentanil was 170 ng/mL for laryngoscopy, 280 ng/mL for intubation, 259 ng/mL for opening of the peritoneum, and 209 ng/mL for intra-abdominal surgical stimuli. Albeit under noxious stimuli, the reported alfentanil C50 values with concomitant propofol or N2O were below 500 ng/mL. The wide range of the pharmacokinetic profile may reflect a great interindividual pharmacodynamic response and poor ability to reliably produce a loss of response when opioids are used alone.14

    One study, which involved 54 intensive care unit patients who had undergone coronary artery bypass graft surgery, reported a midazolam C50 of 171 ng/mL at a Ramsay score of 3 and 260 ng/mL at a Ramsay score of 5.25 Another study conducted by Vinik et al26 also investigated the hypnotic synergism between propofol, midazolam, and alfentanil in their binary and triple combinations. In the midazolam–alfentanil group, the C50 values were 0.04 mg/kg and 0.03 mg/kg for midazolam and alfentanil, respectively. The pharmacokinetic profile was magnitudes smaller than that adopted in this study. The unexpectedly high alfentanil and midazolam C50 values in Minto et al's3 model may partly explain the poor predictive ability of the response surface model. We have plotted the cumulative doses against the contour graph derived from the original response surface model. It may seem slightly more accurate (6% and 0%), but prediction is still poor and should be interpreted carefully. Drugs are given in multiple boluses during the examinations. The cumulative dose increases with examination time and does not correlate well with plasma concentrations. It will most likely overestimate the predicted loss of response probability.

    Endoscopic procedures are associated with discomfort and pain, both during and after the procedure.27 Postprocedural pain can occur in up to 36% of the patients. Our study did not assess the response to instrumentation, but the presence of postprocedural pain may imply that drug concentrations lower than those in our study may be sufficient to produce an OAA/S score of <4 for such procedures.

    Instead of volunteers, patients were used in this study. This is a strength but also a weakness. The patient population was limited and may underestimate the model's capabilities. We derived the C50 value from Minto et al's3 model based on the mean body weight and height of the study population. The original C50 unit was mg/kg, and the derived plasma concentrations given by TIVA trainer were in ng/mL. Age difference had little effect on the three-compartment pharmacokinetic model used in the program, and the change was very small. Although assumptions were made to simplify the model, the variation in derived C50 plasma concentration is unlikely to alter the results. Plasma concentration equilibrated with effect site relatively quickly for alfentanil,28 but that was not the case with midazolam.15 Effect-site concentrations peaked minutes after bolus administration for both studied drugs, and cannot be reliably assumed and fitted for model modification.

    In conclusion, the reported response surface model poorly predicted patient response during gastrointestinal endoscopy sedation. Minto et al's3 model was constructed with volunteer data and a single bolus of drugs, which makes it difficult to be implemented into clinical practice. Analysis of the noxious stimulations during gastrointestinal endoscopy with respect to plasma or effect-site concentrations will make the model more accessible to clinicians. Further research is required in order to adjust the response surface parameters to provide a good prediction, as well as to develop new models for different circumstances.


    1. Dexter F. Statistical analysis of drug interactions in anesthesia. J Theor Biol. 1995;172:305-314.
    2. Liou JY, Tsou MY, Ting CK. Response surface models in the field of anesthesia: a crash course. Acta Anaesthesiol Taiwan. 2015 Dec;53(4):139-145.
    3. Minto CF, Schnider TW, Short TG, Gregg KM, Gentilini A, Shafer SL. Response surface model for anesthetic drug interactions. Anesthesiology. 2000;92:1603-1616.
    4. Diz JC, Del Rio R, Lamas A, Mendoza M, Duran M, Ferreira LM. Analysis of pharmacodynamic interaction of sevoflurane and propofol on Bispectral Index during general anaesthesia using a response surface model. Br J Anaesth. 2010;104:733-739.
    5. Schumacher PM, Dossche J, Mortier EP, Luginbuehl M, Bouillon TW, Struys MM. Response surface modeling of the interaction between propofol and sevoflurane. Anesthesiology. 2009;111:790-804.
    6. Manyam SC, Gupta DK, Johnson KB, White JL, Pace NL, Westenskow DR, et al. Opioid-volatile anesthetic synergy: a response surface model with remifentanil and sevoflurane as prototypes. Anesthesiology. 2006;105:267-278.
    7. Ting CK, Johnson KB, Teng WN, Synoid ND, Lapierre C, Yu L, et al. Response surface model predictions of wake-up time during scoliosis surgery. Anesth Analg. 2014;118:546-553.
    8. Kern SE, Xie G, White JL, Egan TD. A response surface analysis of propofol–remifentanil pharmacodynamic interaction in volunteers. Anesthesiology. 2004;100:1373-1381.
    9. Hannam J, Anderson BJ. Explaining the acetaminophen–ibuprofen analgesic interaction using a response surface model. Paediatr Anaesth. 2011;21:1234-1240.
    10. Baudet JS, Borque P, Borja E, Alarcon-Fernandez O, Sanchez-del-Rio A, Campo R, et al. Use of sedation in gastrointestinal endoscopy: a nationwide survey in Spain. Eur J Gastroenterol Hepatol. 2009;21:882-888.
    11. Donelly M, Scott WA, Daly DS. A comparison of sedation for upper GI endoscopy using diazepam with demerol or midazolam with alfentanil. Can J Anaesth. 1990;37:S21.
    12. Short TG, Plummer JL, Chui PT. Hypnotic and anaesthetic interactions between midazolam, propofol and alfentanil. Br J Anaesth. 1992;69:162-167.
    13. Huang YY, Chu YC, Chang KY, Wang YC, Chan KH, Tsou MY. Performance of AEP Monitor/2-derived composite index as an indicator for depth of sedation with midazolam and alfentanil during gastrointestinal endoscopy. Eur J Anaesthesiol. 2007;24:252-257.
    14. Maitre PO, Vozeh S, Heykants J, Thomson DA, Stanski DR. Population pharmacokinetics of alfentanil: the average dose-plasma concentration relationship and interindividual variability in patients. Anesthesiology. 1987;66:3-12.
    15. Zomorodi K, Donner A, Somma J, Barr J, Sladen R, Ramsay J, et al. Population pharmacokinetics of midazolam administered by target controlled infusion for sedation following coronary artery bypass grafting. Anesthesiology. 1998;89:1418-1429.
    16. Chan WH, Chang SL, Lin CS, Chen MJ, Fan SZ. Target-controlled infusion of propofol versus intermittent bolus of a sedative cocktail regimen in deep sedation for gastrointestinal endoscopy: comparison of cardiovascular and respiratory parameters. J Dig Dis. 2014;15:18-26.
    17. Kongkam P, Rerknimitr R, Punyathavorn S, Sitthi-Amorn C, Ponauthai Y, Prempracha N, et al. Propofol infusion versus intermittent meperidine and midazolam injection for conscious sedation in ERCP. J Gastrointestin Liver Dis. 2008;17:291-297.
    18. Garewal D, Powell S, Milan SJ, Nordmeyer J, Waikar P. Sedative techniques for endoscopic retrograde cholangiopancreatography. Cochrane Database Syst Rev. 2012;6:CD007274.
    19. Lera dos Santos ME, Maluf-Filho F, Chaves DM, Matuguma SE, Ide E, Luz Gde O, et al. Deep sedation during gastrointestinal endoscopy: propofol–fentanyl and midazolam–fentanyl regimens. World J Gastroenterol. 2013;19:3439-3446.
    20. Milligan KR, Howe JP, McLoughlin J, Holmes W, Dundee JW. Midazolam sedation for outpatient fibreoptic endoscopy: evaluation of alfentanil supplementation. Ann R Coll Surg Engl. 1988;70:304-306.
    21. Jhaveri R, Joshi P, Batenhorst R, Baughman V, Glass PS. Dose comparison of remifentanil and alfentanil for loss of consciousness. Anesthesiology. 1997;87:253-259.
    22. Scott JC, Stanski DR. Decreased fentanyl and alfentanil dose requirements with age. A simultaneous pharmacokinetic and pharmacodynamic evaluation. J Pharmacol Exp Ther. 1987;240:159-166.
    23. Ausems ME, Hug CC Jr, Stanski DR, Burm AG. Plasma concentrations of alfentanil required to supplement nitrous oxide anesthesia for general surgery. Anesthesiology. 1986;65:362-373.
    24. Vuyk J, Lim T, Engbers FH, Burm AG, Vletter AA, Bovill JG. The pharmacodynamic interaction of propofol and alfentanil during lower abdominal surgery in women. Anesthesiology. 1995;83:8-22.
    25. Somma J, Donner A, Zomorodi K, Sladen R, Ramsay J, Geller E, et al. Population pharmacodynamics of midazolam administered by target controlled infusion in SICU patients after CABG surgery. Anesthesiology. 1998;89:1430-1443.
    26. Vinik HR, Bradley EL Jr, Kissin I. Triple anesthetic combination: propofol–midazolam–alfentanil. Anesth Analg. 1994;78:354-358.
    27. Zubarik R, Ganguly E, Benway D, Ferrentino N, Moses P, Vecchio J. Procedure-related abdominal discomfort in patients undergoing colorectal cancer screening: a comparison of colonoscopy and flexible sigmoidoscopy. Am J Gastroenterol. 2002;97:3056-3061.
    28. Shafer SL, Varvel JR. Pharmacokinetics, pharmacodynamics, and rational opioid selection. Anesthesiology. 1991;74:53-63.

    alfentanil; gastrointestinal endoscopy; midazolam; pharmacodynamic; response surface model

    © 2016 by Lippincott Williams & Wilkins, Inc.