Fisher and Bradford Hill: theory and pragmatism?

provided the theoretical underpinning for tests of statistical significance.1 Because of this, it is often assumed that Fisher must have played a key role in the evolution of randomized trials in medicine during the 1930s and 1940s. Randomization was adopted by Austin Bradford Hill for Medical Research Council (MRC) trials for a more pragmatic reason, however. Bradford Hill was aware that an alternate allocation scheme had not been strictly observed in a MRC trial done in the early 1930s2 and that selection bias had thus probably undermined the validity of the comparisons made in the study. In designing MRC trials a decade later, he therefore used allocation schedules based on random numbers. If properly concealed, these made it more difficult for those recruiting participants to know which allocations were next in line, and, thus less likely to introduce bias in assembling therapeutic comparison groups.3 The papers in this issue of the International Journal of Epidemiology—by Peter Armitage and Richard Doll (formerly colleagues of Bradford Hill), Walter Bodmer (formerly a student of Fisher), and Harry Marks (author of a history of clinical trials4)—were commissioned to reflect on this characterization of ‘Fisher the theoretician’ and ‘Bradford Hill the pragmatist’. This is obviously a simplification—Fisher was interested in practical problems, and Bradford Hill was aware of statistical theory—but it does help to emphasize the very different contributions made by the two men. Fisher left a wealth of personal papers, which have been archived very helpfully at the University of Adelaide, in Australia, and an excellent biography exists.5 By contrast, very few of Bradford Hill’s personal papers have survived (his daughter has told me that he ‘threw everything away’) and no one has yet taken up the challenge of writing a biography. I obtained copies of correspondence between Bradford Hill and Fisher from the Fisher Archive, and the authors of the four papers that follow were able to draw on this. However, none of them has used any material from an unpublished hand-written ‘memoir’ that Bradford Hill prepared in 1988 for the Librarian at the London School of Hygiene and Tropical Medicine.6 I want to take this opportunity to bring some of this material to a wider readership. Passages in this memoir make clear that Bradford Hill recognized his limitations as a mathematical statistician, declaring that, ‘having no mathematical knowledge’ he needed first class statisticians to work with him (ref. 6, p. 5). This acknowledgement did not prevent him criticizing statisticians who were out of touch with practical realities, however. Commenting on some American statisticians who had criticized a IJE vol.32 no.6 © International Epidemiological Association 2003; all rights reserved. International Journal of Epidemiology 2003;32:922–924 DOI: 10.1093/ije/dyg295