Alliance for Useful Evidence

Tea, Guinness and TB: the origins of the Randomised Control Trial

07.12.2012

There are many myths and misunderstandings surrounding randomized trials, an increasing part of international development and social policy.  Surprisingly, the best cuppa may have inspired the creation of RCTs, according to historian Rhodri Hayward at Queen Mary, University of London.

The waning of doctor's patrician power was also key to their growth, and the rise of a postwar democratic and bureaucratic spirit.

Tea-making has long been a point of contention in British culture.  In the early 1920s, R. A. Fisher (1890-1962), an agricultural statistician at Rothamstead Experimental Station, decided to test a colleague's claim that he could tell from the taste of a cup of tea whether the tea or milk had been poured first. The design of the experiment raised immediate problems. How should chance variations in the tea be controlled for? How many cups were required to demonstrate that correct answers were based on taste rather than luck? And what conclusions could there be if his colleague produced a perfect score, or made one or more errors?

Fisher's reflections on these problems led to a series of publications in statistical science that revolutionised the course of experimental design and, in turn, transformed the grounds of scientific truth. In particular, randomisation emerged as the key technique for controlling unconscious and unnoticed biases: both in the design of the experiment, and in the experimenter.

There had been attempts at the end of the nineteenth century, to use randomization techniques in experimental investigations into psychophysics and telepathic ability.  These trials made little impact.  They were bedeviled with organizational problems and were seen as over elaborate and time consuming. Fisher's protocol met similar opposition. His close friend and colleague, W. S. Gosset (master brewer for Guinness) argued that it was better to compensate for any accidental bias by using 'matched' or 'balanced arrangements', based on the experimenter's knowledge of the environmental conditions. Randomisation, it was claimed, undermined the store of practical insights that experimenters had developed in the field and was, moreover, likely to lead to errors in implementation.  But by the late 1930s, such objections had largely been marginalised - overcome not by experimental evidence, but by practical considerations in the burgeoning field of medical research. In epidemiological research and clinical trials, randomisation proved an effective way of dealing with the ethical problems involved in deciding which subjects received treatments and which were used as controls.

Medical use of randomisation techniques was led by Austin Bradford Hill (1897-1991), an MRC-funded epidemiologist at the London School of Hygiene and Tropical Medicine. Hill's 1937 publication, The Principles of Medical Statistics (based on a series of Lancet articles) became the cornerstone of future research.  He believed in the rhetorical power of randomisation, which lifted experimental design above the idiosyncratic biases associated with any particular researcher. But triumph over local experimental bias was also an institutional achievement, made possible by the coordination and standardisation of medical research laboratories during WWII, and by the National Health Service from 1948.

In 1946, Hill organised a team of physicians, radiologists, bacteriologists and statisticians to investigate the efficacy of the new antibiotic, streptomycin, in treating tuberculosis. Their 1948 publication on this trial became the template for future clinical research.

Anti-tuberculosis drugs were notoriously difficult to evaluate because consumptive patients often experienced spontaneous recovery: such remissions were behind the flawed results that led, in the late 1930s, to use of Sanocrysin, a gold compound, as a treatment. The streptomycin trial was further fraught by media reports of the drug's efficacy, combined with limited supply. Random allocation, decided by a central agency, protected the doctors involved from the knowledge that they might be withholding a lifesaving treatment from some patients. Fisher's logical analysis had evolved into a technique sustained by bureaucratic organisation, moral scruples and ethical concerns.

According to Nancy Mitford, tea-pouring was a defining feature of English social class: to add milk to the cup first marked one as "non-U"; outside the upper classes. This world of unspoken distinctions is one that Fisher's technique, originating in his tea-tasting trial, helped dismantle. The adoption of the randomised control trial moved hand in hand with the democratic and bureaucratic spirit that characterised post war Britain. It replaced the patrician assurance of medical authority with a relatively transparent technology rooted in an egalitarian system of distribution. Increasingly, it provides the basis for policy, offering the possibility of government grounded in social scientific knowledge rather than personal authority and belief.

 

Dr Rhodri Hayward is Wellcome Award Senior Lecturer in the History of Medicine at Queen Mary, University of London

(The views in this blog are the author's own and not necessarily those of the Alliance for Useful Evidence)