#107 Epidemiologic methods and the tobacco debate.

摘要

PURPOSE: Ronald Aylmer Fisher and Joseph Berkson remain infamous today for their stand in the debate over the emerging evidence on the health effects of tobacco during the 1950s and 1960s. However, throughout this period there were a substantial number of senior statisticians and epidemiologists, including Caroll Palmer, Jacob Yerushalmy, Donald Mainland, Herbert Lombard, KA Brownlee, Edwin Wilson, George Saiger, Alexander Gilliam, and others, who were critical of claims that cigarette smoking had been demonstrated to be a cause of lung cancer. We sought to characterize the scientific arguments of these skeptics and to understand the historical context that contributed to their concerns.METHODS: We identified research articles, commentaries, and reviews on the epidemiology of smoking and lung cancer between 1950 and 1968. We also reviewed textbooks from this period on epidemiology, survey design, and experimental statistics. We searched for additional documents, conference transcripts, and unpublished papers by key names in electronic tobacco industry document repositories.RESULTS: Epidemiologists and biostatisticians expressed concerns about widespread abuses of statistical methods in the literature, particularly an excessive emphasis on statistical significance tests to find differences between groups of patients. Epidemiologists and statisticians on both sides of the tobacco debate described themselves as methodological reformers at a time when statistical thinking was still working its way into medicine.CONCLUSION: The methodological arguments expressed by Berkson and Fisher were shared by many of their colleagues, including those who argued that the case against cigarettes was sufficient for taking action. However, the adversarial nature of the public debate and the industry's aggressive public relations campaign exaggerated methodological differences that were often quite subtle. This case provides useful lessons for epidemiologists today involved in methodological debates with implications for policy and financial interests.