Book Review

Quantification and the Quest for Medical Certainty

N Engl J Med 1996; 334:670March 7, 1996

Article

Quantification and the Quest for Medical Certainty
By J. Rosser Matthews. 195 pp. Princeton, N.J., Princeton University Press, 1995. $39.50. ISBN: 0-691-03794-9

Succinctly put, Matthews's book chronicles some of the pivotal events in the evolution of the art of medicine into the science of medicine. Starting with the debates surrounding Pierre Louis's numerical method in the 1800s, we traverse a series of intellectual, sociological, and political confrontations leading up to the introduction of the modern randomized clinical trial.

The first portion of the book focuses on the resistance to the aggregation that was inherent in Louis's numerical method. Louis stressed the need for carefully observed and recorded data. He championed the use of summary statistics as a means of investigating therapeutic efficacy. His approach was vigorously opposed by those such as Double, who thought that the aggregation of individual data into averages carried an unacceptable loss of clinical relevancy. Contemporaneously, the campaign for quantification and aggregation was being carried forward by Quetelet, with his focus on l'homme moyen, and Poisson. The latter was responsible for development of the law of large numbers, which served to quantify the behavior of averages of large numbers of measurements. Those who find the adoption of 95 percent confidence intervals and 5 percent tests of significance arbitrary may be intrigued to learn that Poisson proposed a criterion that was much more stringent. He argued for the use of a criterion of 99.5 percent for calculating the limits of oscillation to distinguish random from systematic occurrences.

Although Louis's approach was vigorously attacked in the French Academy of Medicine, quantification was not rejected out of hand. Information gained from the use of more precise instruments of measurement, such as the thermometer, was viewed by many as an appropriate means of advancing medical knowledge. Many critics had opposed Louis's method, in large part because his focus on the population was viewed as diverting attention from the study of underlying mechanisms. With the availability of more precise means of quantifying phenomena at the individual level, critics such as Lister and Bernard were able to advocate laboratory experimentation as the appropriate route to medical knowledge. This approach represented a compromise: it explicitly recognized the value of quantification and embraced instrumentation as a means of obtaining relevant information about physiologic function while rejecting the need to focus at any level other than that of the individual.

Continuing developments and struggles along similar axes are traced through debates in the German physiologic literature and the establishment of the British school of biometry. The acceptance of the clinical trial, which came into its own with the pioneering work of Sir Austin Bradford Hill almost 50 years ago, was a milestone that marked a change in the relationship between the clinician and the epidemiologist or medical statistician. The change reflected a transition from a more combative to a more collaborative relationship between those whose primary focus was directed to the individual and those whose primary focus was directed to the population. Matthews notes that although the reaction to Hill's streptomycin trial was extremely positive, it was primarily the U.S. legislative requirement for evidence of drug efficacy arising from the Kefauver congressional hearings of the early 1960s, along with the emergence of a host of industrially produced drugs, that provided the catalyst to “bring statistical methods into the open.”

During the course of the previous century and a half, the struggle to advance medical knowledge was marked by both skirmishes and pitched battles between those whose preferred response to biologic variability was to narrow their focus to the specific patient and those whose response was to extend it to a population of patients. Matthews provides us with a chronicle of some of the key developments and debates concerning that struggle. Although he does an admirable job of guiding us on this journey, there are some intriguing side trips that it is a shame to have missed. In particular, given that the book is oriented toward showing the evolution of statistical reasoning in medicine, the absence of any mention of the Bayesian approach is striking. There are some obvious opportunities to have done so. In particular the discussion on page 32 repeats a common but mistaken classic inferential interpretation of the situation in which a value falls outside specified confidence limits.

Despite the occasional technical lapse, such as a reference to a statistically significant population, the book is careful in its approach and usually will not lead readers astray. One of its strengths is a general richness of background detail, although there are exceptions. For example, after the author has spent many pages describing the difficulties and frustrations experienced by pioneers Pearson and Greenwood in their attempts to establish biometry as a recognized discipline with viable career paths, readers are rather abruptly informed that in the early 1920s the Rockefeller Foundation suddenly began to fund those who wished to learn statistical methods. Given the nature of the book, the absence of a discussion of the factors that led to this decision is notable. However, it is always easy to fault someone for what has not been done rather than concentrate on what has been done. In general, what Matthews does he does quite well, and interested readers will find the time invested in reading his book to be well spent.

The book also allows us to see the origins of some of the paths that we occasionally find ourselves stumbling along. In particular, the concern of some about the incompatibility of the dual roles of physician-scientists in clinical trials that has been expressed in the Journal (S. Hellman and D.S. Hellman. “Of Mice but Not Men: Problems of the Randomized Clinical Trial.” 1991;324:1585-9) bears a striking similarity to arguments put forward almost 150 years ago by opponents of Louis's numerical method.

Fittingly, this study of the evolution of medical epistemology ends with the French proverb Plus ça change, plus c'est la même chose. For those who want to know how change has come about, Matthews's book provides a thoughtful and comprehensive analysis. It is also a striking illustration of the fact that the past is often prologue.

Stanley H. Shapiro, Ph.D.
McGill University, Montreal, QC H3A 1A2, Canada