The Poverty of Epidemiology

Publisher Website
Google Scholar
Abstract
THE POVERTY OF EPIDEMIOLOGY PETR SKRABANEK* A perusal of the abstracts of papers presented at the twenty-third annual meeting of the Society for Epidemiological Research at Snowbird , Utah (June 12-15, 1990) [1], made me wonder whether epidemiology , in the absence of epidemics, is not a misnomer for scaremongering made respectable by the use of sophisticated statistical methods, and whether one of the reasons for this state of affairs is not a high prevalence of epidemiologists when the incidence of problems soluble by epidemiological methods is low. It would seem that any combination of "exposure" and disease, regardless of biological implausibility, or even without any underlying hypothesis , is fair game for calculating relative risks, odds ratios, or proportional hazards. The association game has three possible outcomes: positive association, negative association, or no association. As any of these three outcomes is generally thought to be "interesting," "controversial ," or just "in need of further research," they all get published. Combining these three possible outcomes with any two combinations of a potential risk factor (traditionally, sexual behaviour, alcohol drinking, and smoking, but now including any pleasurable activity, be it idleness, eating, or coffee drinking), the game has more possible combinations than Cluedo. The scope of epidemiological research has been enormously widened by including "passive" exposures to invisible electromagnetic waves, whether from home appliances, overhead wires, X-ray machines, or space, "passive" exposures to other people's smoke or other air pollutants, "passive" exposures to innumerable food additives, and other menaces of everyday life. Three groups of researchers looked at male breast cancer, a relatively uncommon disorder. According to one abstract, more than three medical X-rays increased the risk. Another study lent "support to the theory that exposure to electromagnetic fields may be related to [male] breast *Department of Community Health, Trinity College, University of Dublin, Ireland.© 1992 by The University of Chicago. AU rights reserved. 003 1 -5982/92/3502-0762$0 1 .00 182 Petr Skrabanek ¦ Poverty ofEpidemiology cancer." This was contradicted by a third study, in which no effect of occupational exposure to electromagnetic fields on breast cancer in males was observed, but the study was not altogether negative: widowed and never-married men had an increased risk, as shown by "unconditional logistic regression modelling." We are not one iota wiser, although the general feeling lingers that some radiation in some organisms may increase the risk ofsome cancers. As pointed out by Weinberg [2], transscientific questions are those that can be asked of science but cannot be answered by science, such as whether the amount of electromagnetic energy to which we are normally exposed increases the risk of cancer. In one of his examples, a study of the spontaneous mutation rate in mice exposed to 150 milirem of X-rays would require 8 billion mice in order to show an increase of one half percent (as extrapolated from much higher radiation doses, assuming a linear regression). Or perhaps 8 billion men, as extrapolation from mice to men is tricky. In other words, a potentially answerable question is unanswerable in real life. As so many scares have been disseminated by epidemiological research and avidly taken up by the media, who could hardly be blamed, further research is deemed necessary to confirm or deny previous reports . This time, one abstract provided reassurance to vasectomised men that they are not at an increased risk of dying of heart disease. This will hold until another report contradicts this. When too many such conflicting observations have accumulated a call is made for metaanalysis , that is, pooling studies with not quite the same methodologies and not quite the same populations, in hope that if none of them made much sense, the sum total will throw light on the matter. Women have been harassed more then men by epidemiologists, on account of having two sexual organs of interest: the uterus and the breast. Contraceptive pills "may have a substantial impact on incidence of invasive cervical cancer" was a conclusion of one abstract (though the verb "may" implies that the opposite may also be true), particularly if they do not eat tomatoes, since, according to another abstract, "increased odds of [cervical intraepithelial neoplasia] were associated with decreased...

This publication has 0 references indexed in Scilit: