The Impossible in Medicine

THE IMPOSSIBLE IN MEDICINE JEREMIAH A. BARONDESS* Medicine, someone has said, is the art of making adequate decisions based on inadequate evidence, which is another way of saying that medicine is, at its core, inexact. Certainty, the organizing ideal toward which medicine strives to move at both the clinical and basic scientific levels, is a luxury generally beyond the clinician's grasp and, in at least an absolute sense, beyond that of the basic biomedical scientist as well. It is, in effect, impossible to achieve. Given that certainty is fundamentally unattainable in clinical medicine, an examination of the systems which have evolved in diagnostic reasoning, in workup and management tactics, and in therapeutics will demonstrate that they are all techniques for managing (i.e., reducing or minimizing) uncertainty. The natural sciences in medicine are likewise bound ultimately by uncertainty. The theories of pure science are explanations of observed phenomena that satisfy the observations, do not clash with related explanations of related phenomena, and have predictive value. The strong inferences that can be drawn from such formulations permit new experiments to be designed, modifications of theories to be offered, and explanations of natural phenomena to be refined. Nevertheless, scientific theories retain a core of tentativeness that permits the possibility of revision or extension. The analogies to clinical medicine are rich. The present essay will concern itself with the clinical issues. If certainty is impossible and uncertainty is the rule, clinical medicine can be seen as based on systems of predictions, or probabilities. The domains of diagnostic reasoning, clinical tactics, and therapeutic management may be considered as expressions of processes that start with This article will appear in No Way. Essays on the Nature of the Impossible, edited by P. D. Davis and D. Park. New York: W. H. Freeman, in press. *Professor ofclinical medicine, Cornell University Medical College, 449 East 68th Street, New York, New York 10021.© 1986 by The University of Chicago. AU rights reserved. 003 1-5982/86/2904-0493$01 .00 Perspectives in Biology and Medicine, 29, 4 ¦ Summer 1986 | 52 1 maximal uncertainty and, through the application of systematic reductionistic reasoning, attempt to reduce that initial uncertainty to a minimum. Uncertainty, in this paradigm, may be considered to approach the zero level but never to reach it. In operating as a system based on probabilities, clinical medicine strives to deal effectively with the fact that derangements of physiological processes and alterations of anatomical arrangements can usually be appreciated only inferentially [I]. This can be understood further by examining the diagnostic process. Thus, the diagnosis "pneumonia" is based on clusters of data (chills, fever, cough, rales heard over the affected lung segment, abnormalities in the white blood cell count and on X-rays of the chest) rather than on direct observation of inflammation of the lung as one would see it under the microscope. Likewise, the diagnostic term "congestive heart failure" connotes a mix of symptoms (dyspnea , orthopnea), signs (cardiomegaly, gallop rhythm, rales in the lungs, hepatomegaly, edema) and physiologic aberrations (elevation of certain intravascular pressures) rather than reflecting direct evidence, at the level of the heart muscle itself, that it is "failing," that is, performing its hydraulic functions inadequately because of abnormalities in contractility . Clusters of data such as these, arranged so as to permit a diagnosis, amount, in effect, to clinical syllogisms; in the clinical situation, however, the syllogisms are rarely rigorous, because most signs, symptoms, and laboratory data are not disease specific but represent samplings of the rather limited repertoire of reactions to injury, recognizable by clinical techniques, which are possessed by each organ or organ system. Thus, chills, fever, rales, pleuritic chest pain, and cough are not specific to pneumonia but may be features of a large number of disorders. In attempting to understand the problem in a diagnostic sense, the clinician must generate an estimate of the frequency with which the phenomena observed in his patient are produced by various diseases, when grouped as they are in the particular patient at hand, at the point in the course of the disease which appears to obtain at the time; and he must also estimate the degree to which such initial estimates must be modified...