NUMBERS are part of medical education—everyone learns normal values for blood pressure, laboratory values, and so on; then, during training, one acquires numbers specific for one's specialty. One category of numbers—minimum alveolar concentration (MAC)—permeates the anesthesia community. For each inhaled anesthetic, that number is branded onto one's brain as a guide to dosing. However, there is a problem with MAC: If patients were dosed with 1 MAC, by definition, 50% of patients would move on incision. Surgeons would be unhappy, patients might be harmed, and anesthesiologists' professional standing would be compromised. Realistically, one of two things happens when inhaled anesthetics are given: Either larger doses are given routinely or movement is prevented by administration of other drugs.
One could argue that MAC, an ED50, is not an inherently useful number; instead, it makes more sense to report ED95 (i.e., a dose preventing movement in 95% of subjects). Yet, studies of every inhaled anesthetic in the modern era have focused on ED50 rather than any other measure of potency. This oddity can be explained through some history. The earliest studies of MAC, performed by Eger and associates, estimated MAC semiquantitatively and did not apply rigorous statistical evaluation (personal communication, Edmond I. Eger II, M.D., Professor, Department of Anesthesia, University of California, San Francisco, January 30, 2007). After the first several studies were reported, Eger learned of the “up–down” method in which the dose for a particular subject in a clinical trial is selected as a function of the response of the previous subject. This method was coupled with a variety of statistical techniques to estimate ED50. One advantage of these approaches was that studies could be quite small—MAC was estimated in as few as 10–20 subjects; in turn, MAC studies proliferated.
These MAC studies proved to be clinically useful, guiding numerous generations of anesthesiologists. Why would that be the case if the number that they provided could result in a high incidence of patient movement? The answer is simple. In Eger's early studies, it became apparent that doses minimally larger than ED50 typically prevented movement. For example, with halothane, in which ED50 was estimated to be 0.75%, none of the subjects with end-tidal concentrations larger than 0.78% moved.1 This observation suggested that the concentration–response relation for inhaled anesthetics was steep, i.e., ED95 differed minimally from ED50. In turn, knowledge of ED50 would have been sufficient to guide clinical dosing (and trials that formally estimated ED95 for inhaled anesthetics would not be needed).
Subsequently, de Jong and Eger2 formally evaluated the premise that, for inhaled anesthetics, the ratio of ED95 to ED50 is not markedly larger than 1. A study of nine inhaled anesthetics showed that ED95 was typically less than 20% larger than ED50; the ratio of ED95 to ED50 exceeded 25% only for ethylene and xenon (neither of which is used clinically). In that the safety margin of inhaled anesthetics is reasonably large, clinicians could dose inhaled anesthetics with concentrations sufficiently above ED50 to ensure lack of movement without compromising safety; in turn, knowledge of ED50 was sufficient to guide dosing (i.e., “give a little more than 1 MAC and things will be OK”).
Over ensuing decades, statistical techniques for the conduct/evaluation of MAC studies evolved minimally. Meanwhile, researchers in anesthesia extended traditional up–down methodologies to drugs other than inhaled anesthetics, e.g., local anesthetics. Unfortunately, published studies of these other drug classes have not formally evaluated steepness of the dose–response curve; in turn, relevance of ED50 to clinical dosing of anesthetic drugs other than inhaled anesthetics is understood less well. Perhaps the only evolution in statistical methods for estimation of potency of anesthetic drugs was the introduction of logistic regression to estimate ED50. Although I used logistic regression in one study,3 I understand now that it can yield flawed and biased estimates4 (logistic regression assumes that the independent variable, concentration, is selected randomly, whereas successive values are assigned systematically).
In contrast, statistical methods for estimation of dose–response have evolved significantly in other fields, driven by needs in therapeutic areas such as oncology (for which therapeutic margins are typically small and there is a need to minimize trial size). In the current issue of Anesthesiology, Pace and Stylianou4 review historic and new statistical methods for the determination of potency. Their important contribution to the anesthesia community is the description of methods to estimate EDg, potency at the gth percentile. The review is of particular importance to investigators who perform potency studies in which the outcome is quantal (e.g., movement vs. no movement). In addition, others interested in clinical research can learn much from the article.
During the review process for this article, two things became evident. First, as Paul and I5 reported several years ago in this journal, regardless of the methodology used for dose selection and analysis, studies with small numbers of subjects may yield biased and imprecise estimates (i.e., broad confidence limits), even for ED50. Second, as the potency estimate deviates further from 50% (e.g., ED95), problems of bias and imprecision become worse.
The anesthesia research community has benefited from studies that had relatively small sample sizes and reported ED50. If ED50 is useful, this practice can probably continue without harm: The long history with inhaled anesthetics suggests that this is the case. However, for other anesthetic drugs, e.g., local anesthetics, clinical utility of ED50 remains to be seen. Pace and Stylianou offer guidance on how to obtain other, possibly more useful, potency estimates.