IN this issue of the Journal, Avram and Krejcie examine one of the conundrums that confront the design of target-controlled infusion (TCI) systems: The “standard” models used in pharmacokinetic and pharmacodynamic analyses are wrong. 1Specifically, such models assume that the plasma concentration peaks at the instant a bolus of drug is administered. Obviously, the concentration in the plasma is zero at the moment the drug is administered, because the drug must move through the veins, get mixed in the heart and great vessels, and ultimately flow through the aorta to the sampling site. All of this takes 30–45 s. Those of us who write software for TCI systems or study these devices have dismissed these 30–45 s of time delay as a minor nuisance, but Avram and Krejcie demonstrate that the way this error is handled by the model can measurably affect performance of TCI systems.
The international reader of Anesthesiology, accustomed to routine use of TCI systems, will doubtless find these results of interest. The North American reader, by contrast, will probably have no clue why these results are interesting, because exactly 0 of the estimated 13 million propofol anesthetics administered worldwide with TCI (written personal communication from James B. Glen, Ph.D., Glen Pharma, Knutsford, Cheshire, United Kingdom, June 2003) since the introduction of the Diprifusor (AstraZeneca, Macclesfield, Cheshire, United Kingdom) in Europe, Asia, the South Pacific, South America, and Africa have been performed in North America. The reason, at least in part, is that the U.S. Food and Drug Administration (FDA) has expressed a variety of concerns about computer-based drug delivery that have discouraged manufacturers from developing these systems, despite that the devices deliver approved drugs by approved routes at approved doses for approved indications. The specific concerns expressed by individuals within the FDA include “important health implications” that are not otherwise defined, “significant incremental risk” of anesthetic controllers (again undefined), concerns that “the use of high level languages, general-purpose computers, and complex operating systems results in products that are too elaborate for the product developer to verify entirely,” and a hesitation to accept the extensive literature supporting the clinical use of TCI on the basis that published reports “emphasize positive outcomes.”2
At the time these concerns were published (1995), AstraZeneca submitted regulatory documentation on the Diprifusor TCI system to FDA. Eight years later, there has been no discernible progress. In AstraZeneca's view, the primary problem has been the lack of regulatory precedent for a drug-device combination (written personal communication from James B. Glen, Ph.D., Glen Pharma, Knutsford, Cheshire, United Kingdom, August 2003). They have at various times been told that TCI would be regulated as a device (which it is), or as a drug. If regulated as a drug (the current FDA view), approval would require additional clinical studies and a revised package insert. The requirement for a change in the drug product labeling makes introduction of TCI drug delivery systems by device companies impossible, because device companies do not control the drug labeling.
Over the course of the 8-yr review, the FDA has demonstrated a poor understanding of the underlying scientific basis of TCI. Specifically, the FDA has not recognized that TCI devices can neither increase nor decrease underlying pharmacokinetic variability. As a result, the FDA has expressed unfounded concerns that the TCI mode of administration may lead to a greater frequency of adverse events. AstraZeneca performed a detailed review of sponsored TCI studies and the worldwide safety database on propofol, including propofol delivery by TCI, and found no evidence of increased risk of adverse events with TCI. This is consistent with the dozens of published manuscripts on the Diprifusor.
For the North American reader who is unfamiliar with these devices, we could perhaps explain them as the intravenous equivalent of a vaporizer, where one sets the desired concentration and a computer model, rather than physicochemical equilibration across the alveolus, aligns the plasma (and effect site) concentrations to the target concentration. 3However, we will instead explain the concept using a popular North American sport: surfing. The concentration versus response curves of anesthetic drugs are typically fairly steep, like a wave approaching the shore. Surfing the steep portion of the concentration–effect relationship makes it possible to produce the therapeutic drug effect while preserving the ability to decrease the level of effect rapidly at the end of the anesthetic with a small decrease in concentration. Figure 1shows the anesthesiologist surfing near the crest of a wave. Surfing beyond the crest (i.e. , the flat portion of the concentration–effect relationship) offers no clinical advantage; it only results in prolonged recovery and increased adverse effects with no measurable increase in therapeutic effect.
Anesthesiologists simultaneously use three techniques to stay on the crest (i.e. , the steep portion of the concentration–effect relationship). They start with pharmacokinetic guidance, the cookbook. In our view, most physicians dose commonly used drugs within a narrow range, reflecting a fundamental trust in pharmacokinetics to yield the desired target concentration and drug effect. For example, how far do your propofol infusions (combined with reasonable doses of opioid) differ from something like: 2–2.5 mg · kg−1bolus, then 100–150 μg · kg−1· min−1for 15 min, then 80–100 μg · kg−1· min−1for 30 min, then 70–90 μg · kg−1· min−1thereafter? Standard dosing guidelines such as these are based on the typical dose–concentration relationship (i.e. , pharmacokinetics). These standard dosing regimens represent a starting point in riding the wave's crest.
Inevitably, the initial attempt at riding the crest of the wave requires adjustment based on feedback from the patient. Perhaps the heart rate or blood pressure is higher than would be expected were the patient adequately anesthetized. Perhaps the Bispectral Index scale is 35, somewhat lower than clinically necessary. Pharmacodynamic guidance, the second technique used by anesthesiologists to stay on the crest of the wave, allows refining of the dose initially guided by pharmacokinetic knowledge to reflect the individual patient's unique pharmacologic characteristics.
The third guidance technique is pharmaceutical: choosing drugs with the right kinetic and dynamic properties to suit the patient and the duration of surgery, and to provide adequate safety margins between therapeutic and toxic doses. 4Currently, implementing the pharmaceutical technique to target the crest of the wave often means choosing drugs with responsive pharmacokinetic profiles (e.g. , propofol, remifentanil) so that if the initial pharmacokinetic guidance results in an overdose (or under-dose) as suggested by pharmacodynamic feedback, the levels can be quickly decreased (or increased) to an appropriate range.
In the context of this surfing analogy, TCI can be viewed as a tool to explore the wave while riding it. With a standard infusion pump, the “wave” that the anesthesiologist sees is not the concentration versus effect curve, shown in figure 1; rather, it is an infusion rate versus effect curve. Unfortunately, this wave changes constantly. A rate of 100 μg · kg−1· min−1of propofol translates to an effect site concentration of 0.5 μg · ml−1at 1.5 min, 1.0 μg · ml−1at 2.9 min, 2.0 μg · ml−1at 9.9 min, 3.0 μg · ml−1at 87 min, and 4 μg · ml−1at 747 min (fig. 2). 5,6The relationship between what is set on the device (the infusion rate) and what occurs in the patient changes every second. Thus, the wave the anesthesiologist is trying to surf constantly changes shape. If one suddenly needs to increase or decrease the concentration, the wave one was surfing has abruptly ceased to exist. So it becomes very difficult to characterize the wave, other than perhaps recognizing that “this patient needs more or less drug than average.”
With TCI, the wave is the concentration versus response relationship shown in figure 1. Admittedly, it is the predicted concentration, not the true concentration (which is unknowable), but the critical point is that the wave doesn't change shape during the ride to shore. When the anesthesiologist finds that a certain effect site concentration yields a given effect at 10 min into the anesthetic, that same predicted concentration should produce the same effect at 600 min. More than 220 peer-reviewed articles in MEDLINE on TCI (as of June 2003, including 40 articles on the Diprifusor alone) attest to the ability of TCI to preserve the shape of the wave and assist the clinician in exploring the wave and riding it to shore. Moreover, constant advances, such as those described by Avram and Krejcie in this issue of the Journal, continue to refine the technology.
Thirty-five years have elapsed since Kruger-Thiemer first proposed using computers to deliver drugs based on pharmacokinetic models, 7and more than 20 yr have elapsed since Helmut Schwilden first outlined the algorithm for anesthetic drugs. 8Although these developments began in Germany, American investigators added fundamental contributions as well. 9–11How ironic, therefore, that America, the country that brought the world surfing, 12continues to deny physicians access to the fundamental tools to surf the concentration response curves of intravenous anesthetic agents.