Rapid Responses to:

EDUCATION AND DEBATE: Alan Yoshioka Use of randomisation in the Medical Research Council's clinical trial of streptomycin in pulmonary tuberculosis in the 1940s BMJ 1998; 317: 1220-1223 [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

The first clinical trial with true randomisation of individuals?

S F Olsen   (11 November 1998)

The first clinical trial with true randomisation of individuals? 11 November 1998
S F Olsen Send response to journal: Re: The first clinical trial with true randomisation of individuals?	Editor - In 1936 G.W.Theobald, an obstetrician at St. Mary Abbots Hospital in London, studied the effect of calcium, vitamin A and vitamin C on toxaemia. According to his report1, "an equal number of blue and white beads (presumably 50 plus 50) were placed in a box. Each woman accepted for the experiment was asked to draw a bead from the box. Those who drew blue beads were placed in Group A and those who drew white beads were placed in Group B. The beads drawn out were placed in a separate container. The patients in group A were requested to take daily ... while those in group B served as controls." True randomisation of individuals was thus apparently used. We are informed that the two randomisation groups were similar with respect to numbers of women older than 30, primigravidae below 20, as well as primigravidae older than 30 (9 v. 11, 2 v. 2, and 2 v. 2, respectively). Further, "symptoms were recorded by independent antenatal officers who had no knowledge as to which patients were receiving the additional substances", although Theobald himself "saw every patient … and confirmed the findings". He also considered that incorporating methylene blue in the preparation could be expedient to assess patient compliance. Women with symptoms of toxaemia tended to be less frequent in the treatment group. E.S. Pearson did the statistical analysis and "expressed the opinion that the difference in incidence of complications between the two groups is very unlikely to have arisen by chance" (no statistics were provided, however, and reconstructing the tests would require a post hoc selection among several different outcome measures provided in the tables). Theobald himself stated that "it therefore seems logical to assume that the difference ... must, if not due to chance, be attributed to the substances given". We cannot know how rigorously he performed the randomisation. The method was vulnerable to pressures from patients or others to take a new bead until the patient ended in the preferable group, which would be difficult to check if each attempt were not numbered (this problem was avoided in the streptomycin trial which used numbered, sealed envelopes and where the patients who ended in the control group were not informed that they were part of the trial2,3). The most serious potential flaw, however, follows from Theobald's statement that the small number of participants was "due to the fact that only a small proportion of the patients booked sufficiently early and attended the hospital antenatal clinic throughout their pregnancies"1. Does this mean that a substantial proportion of the women were enrolled and later excluded? If he truly had understood the underlying ideas of randomisation, however, he would try to avoid these problems. London was certainly focus of controlled trials during those years. From March 1938 to the end of 1939 the People's League of Health (PLH) undertook their trial4,5,6,7, which is still likely to be the largest properly controlled trial of a dietary supplement in pregnancy ever published. Apart from the size, the PLH trial had many similarities to Theobald's trial. More than 5000 pregnant women from 10 London hospitals were alternately allocated to receive no supplement or a supplement which provided calcium and vitamins A and D, but at other doses than in Theobald's trial and with other vitamins and minerals added as well. It is interesting that, after the publication of the preliminary report4,5, the PLH trial was criticised in very general terms by R.B. Fisher in these columns8; this may have contributed to the PLH trial's bad reputation7 which in many ways was undeserved, as the trial was an enormous achievement and, as far as we can judge afterwards, had strong methodological features7. Theobald foreshadowed the PLH trial by stating that "experiments conducted on these lines would show to what degree, if any, the different protective substances are associated with toxaemia symptoms" and that he published the results "in the hope that further experiments on a larger scale will be conducted elsewhere"1. Of particular note is the fact that W.C.W.Nixon of St Mary Abbots Hospital was medical secretary to the PHL committee1. The PHL investigators team must certainly have been aware of Theobald's trial, and in retrospect it is a pity that they did not use his idea of implementing true randomisation; one can only speculate why this was so. Despite the fact that it was ahead of its time, Theobald's trial is remarkably little cited. Might it be because of the potential flaws mentioned above? Or were there other serious problems which could justify its dismissal from various reviews? Or has it simply been forgotten? Sjúrður F Olsen Senior Epidemiologist Danish Epidemiology Science Centre, Statens Serum Institut 5 Artillerivej, DK-2300 Copenhagen C, DENMARK sfo@ssi.dk 1 Theobald GW. Effect of calcium and vitamin A and D on incidence of pregnancy toxaemia. Lancet 1937; ii: 1397-1399. 2 Medical Research Council. Streptomycin treatment of pulmonary tuberculosis. BMJ 1948; ii: 769-82. 3 Yoshioka A. Use of randomisation in the Medical Research Council's clinical trial of streptomycin in pulmonary tuberculosis in the 1940s. BMJ 317: 1220-3. 4 People's League of Health. Nutrition of expectant and nursing mothers. British Medical Journal 1942; ii: 77-78. 5 People's League of Health. Nutrition of expectant and nursing mothers. Interim report. Lancet 1942; ii: 10-11. 6 People's League of Health. The nutrition of expectant and nursing mothers in relation to maternal and infant mortality and morbidity. Journal of Obstetrics and Gynaecology of the British Empire 1946; 53: 498- 509. 7 Olsen SF, Secher NJ. A possible preventive effect of low-dose fish oil on early delivery and pre-eclampsia: indications from a 50-year old controlled trial. Br J Nutrittion 1990; 64: 599-609. 8 Fisher RB. Dietetics in pregnancy. British Medical Journal 1943; i: 144-5.