Clinical Pharmacology on Display

In this issue of Anesthesia & Analgesia are 3 articles that most subscribers will never read.1–3 Not even the abstract will garner their interest. Nevertheless, these articles and related work help form the foundation underlying 2 recently introduced commercial displays, Navigator Applications Suite (GE Healthcare, Helsinki, Finland) and SmartPilot View (Dräger Medical, Lübeck, Germany), intended for everyday use in the operating room. These sophisticated displays show the concentrations and predicted effects of combined anesthetic drugs to facilitate more precise titration and outcome. The accuracy of these displays depends on the accuracy of the models they use. This editorial, and the following two,4,5 attempt to explain how complex research in pharmacokinetic/ pharmacodynamic (PK/PD) modeling is being translated to everyday clinical practice. We need to titrate anesthetic drugs because of the individual variability in response, and the rapidly changing level of stimulation in the operating room. We adjust our dose according to the effect that we want to achieve. However, because of our incomplete understanding of the complex anesthetic state and consequent technical limitations, we often have very little direct feedback to precisely guide this titration. It would be helpful if we could “see” the effect of our drugs in real time, as well as forward projections of the current trajectory. For example, cardiovascular drugs are typically easy to titrate because we have monitors that display heart rate and arterial blood pressure in real time. Can we develop a real-time display showing the effects of all our anesthetic drugs? For many years, we have measured the effect of neuromuscular blocking drugs using peripheral nerve stimulators. Although mechanomyography was a research tool, electromyography was available for routine clinical practice more than 20 years ago using the Relaxograph (Datex, Helsinki, Finland). The graphical trend of twitch height depression and train-of-four ratio allowed one to appreciate the time course of drug effect (Fig. 1). There has been a resurgence of interest in neuromuscular monitoring, with integration into anesthesia workstations permitting centralized displays of neuromuscular block. With this ready display of drug effect, anesthesiologists have not generally been interested in knowing the actual plasma concentration of neuromuscular blocking drugs.Figure 1: Time course of neuromuscular block. The black bar shows the height of the first twitch (T1) as a percentage of control and the blue bar shows the train-of-four ratio as percentage. x-axis = time stamp; y-axis = scale from 0% to 100%.For the volatile anesthetics, the picture is more complex because they cause more than one desirable effect. Is the drug being titrated to cause hypnosis, suppress somatic responses, or suppress autonomic responses? The effect of volatile anesthetics on the suppression of somatic response has long been characterized by the minimum alveolar concentration (MAC). This value (sometimes uncorrected) is displayed on many anesthetic monitors, permitting anesthesiologists to adjust inhaled anesthetic delivery to obtain a desired fraction or multiple of MAC. MAC is an effective dose (ED)50 on the quantal response/no response curve, meaning that it is the concentration at which 50% of subjects do not move. Who wants to give a concentration at which half the patients move? Who knows at what concentration or multiple of MAC nearly all patients (say 95%) would not move? Instead of displaying the volatile concentration as a fraction or multiple of MAC, it is possible to display the volatile concentration as its effect on the quantile dose-response curve. For example, 2.07% sevoflurane is 1.21 MAC, but this is also the ED95 for no movement (Fig. 2).6 Similarly, if our effect of interest was hypnosis, the concentration could be displayed as an EDx where x% of the population would be expected to not respond to verbal command. This may be more useful than a fraction of MAC (or multiple of MAC-awake). In contrast to the neuromuscular blocking drugs, some knowledge of the volatile agent concentration seems important, but at least we can readily measure it.Figure 2: Dose-response curve for sevoflurane and the probability of no movement. The AD95 is the concentration at which 95% of patients do not move in response to skin incision. MAC = minimum alveolar concentration; S.E. = standard error. (From Katoh and Ikeda,6 with permission from Lippincott Williams & Wilkins.)The population values of MAC and MAC-awake were derived from patients that may be very different from any particular patient. Is it possible to have a measure of anesthetic drug effect in the patient currently undergoing anesthesia, similar to that for neuromuscular block? Individual monitoring of motor spinal reflexes is possible, but rarely used. There is much more interest in monitoring hypnosis using derived measures of the electroencephalogram (EEG) such as the Bispectral Index (BIS) and entropy. Although the raw EEG reflects the current state of the patient, the derived indices are also based on models from previous studies. The BIS is supposed to reflect the hypnotic component of anesthesia and help guide drug dosage with the presumption that we do not want the patient to be awake during surgery or receive more drug than necessary.7 The BIS is presented as a number from 0 to 100, and the user is expected to understand what the numbers indicate, as a guide to dosage. It is possible to relate the BIS to the probability of the patient responding to verbal command (Fig. 3).8 However, the BIS versus probability of consciousness curve is different for different drugs and drug combinations. Comparing the BIS and the predicted hypnotic effects (from sometimes predicted concentrations) may be a useful check for the models (and the patient)!Figure 3: The relationship between Bispectral Index and probability of response to verbal command for propofol alone and in combination with alfentanil. (From Iselin-Chaves et al.,8 with permission from Lippincott Williams & Wilkins.)For IV anesthetics such as propofol, the effect of interest is also hypnosis, but unlike the volatile anesthetics, the concentration cannot be measured in real time. One can use a target-controlled infusion pump that can calculate and deliver a target propofol concentration. However, we need to know what concentrations are required for the effect that we want, and consider the limitations of the PK model being used.9 As for inhaled anesthetics, one can use EEG measures such as the BIS to guide propofol titration. For opioids, a useful measure of effect has been more difficult to derive. In clinical practice, we usually use a combination of experience and cardiovascular signs to titrate dose. One can also use software programs to predict the opioid concentration given the dosing history and PK model. One can then relate predicted concentrations to opioid concentration versus response relationships for end points such as no autonomic response to surgical incision.10 This would seem to be a rational (but seldom used) method for adjusting opioid dose, although variability in response is large. Using this information, we could aim for an opioid concentration at the end of anesthesia with a high probability of analgesia but a low probability of apnea. As if the effects of single drugs were not difficult enough to display, an additional complication is that there are drug interactions among our anesthetics affecting all our end points. Indeed, we take advantages of the synergies between our drugs to produce the anesthetic state with the least amount of drug, minimizing side effects. Some propofol infusion pumps display a predicted time to wake up after the infusion is stopped. This is necessarily inaccurate because it only considers propofol and does not adjust for the effects of other drugs. Still, we do have a good understanding of drug interactions that have been characterized using response surface methodology, as explained in the subsequent editorial.4 Programs have been written to predict the effect of drug combinations, and suggestions were made more than 10 years ago for appropriate target concentrations of combinations of drugs.10 We titrate anesthetic drugs empirically to effect, occasionally influenced by measured or predicted concentrations. Software has long been available to provide more information about predicted drug concentrations and effects, but only few researchers and enthusiasts have used these diverse programs. They are not in a format easy to use by most anesthesiologists who do not even perceive a great need for them. However, there is now (controversial) evidence that anesthetic drugs may cause long-term adverse effects, and morbidity, when excessive doses are used.11,12 Effective graphic displays of complex information have improved decision-making in high-risk environments.13 The novel displays from GE Healthcare and Dräger try to translate the wealth of PK/PD research to create usable tools for everyday clinical practice. These displays show past and present estimates of drug concentrations and predicted effects of our combined anesthetic drugs, as well as predicted future paths if one continues (or changes) drug dosing. We can see more clearly where we might be going! What pilot or driver or anesthesiologist would not appreciate knowing where the potential dangers might be ahead of them? There are many expectations, limitations, and unanswered questions about this new technology. The 2 displays were developed in collaboration with 2 different PK/PD groups using slightly different models. The predicted effects are based on population models of only some of the drugs that we use, and modified by only a few patient covariates. Are the predictions accurate (or at least accurate enough)? Can we individualize the predictions? What training and expense will be required to make this technology worthwhile? Which interface has more style?14 I hope that anesthesiologists will be interested in this technology, and that it will improve the accuracy with which we titrate anesthetic drugs, benefiting patient care. It is the translation to everyday clinical practice of the nearly unreadable articles, published by the anesthetic PK/PD community over the past 20 years, that mathematically characterize the behavior of our anesthetic drugs.