The Sealed Envelopes

Everybody has met the phrase randomised controlled trial by now. It is what people mean when they ask whether something has actually been tested. The first one that was properly done ran in British hospitals in 1947, and the reason it was allowed to happen is that the country could not afford the drug.

Streptomycin had come out of Selman Waksman's laboratory, where his assistant Albert Schatz isolated the organisms that produced it in 1944. It killed tubercle bacilli in the test tube, then in guinea pigs, and early results in patients looked similar. Tuberculosis was the most important cause of death among young adults in Europe and North America. The publicity was enormous.

The drug was also expensive and American, and Britain in 1946 had very few dollars. The government would buy only a limited quantity for testing.

Why withholding it was allowed

That scarcity is what made the trial possible, and the argument was made explicitly. Austin Bradford Hill, Professor of Medical Statistics at the London School of Hygiene and Tropical Medicine, put it to the Medical Research Council's committee that given how little of the drug there was, and given that the outcome of pulmonary tuberculosis was genuinely uncertain, it would be unethical not to find out what advantage streptomycin offered over the standard treatment, which was bed rest. The committee accepted it.

The committee was chaired by Sir Geoffrey Marshall, a senior consultant who had been one of Churchill's doctors, which mattered, because his standing was part of what persuaded other consultants to enter their patients. Philip D'Arcy Hart, who ran the Council's tuberculosis unit, was secretary. Marc Daniels coordinated the trial across the centres, with a trial manager, Charlene Agnew.

There was a second argument that is easy to miss. There were more tuberculosis patients in Britain than there were sanatorium beds. Patients allocated to bed rest alone were given priority for admission, so being in the control group bought you a bed you might otherwise have waited for, and if the drug worked you would get it later when supplies improved.

The envelopes

Eligibility was narrow: patients aged between 15 and 30 with acute progressive bilateral pulmonary tuberculosis of presumably recent origin, bacteriologically proved, and unsuitable for collapse therapy, which was the surgical alternative of the day.

When a consultant found somebody who might fit, the details went to Daniels at the coordinating centre. If the patient qualified, a bed was arranged at the nearest participating hospital.

Then the mechanism. Each hospital held a numbered series of sealed envelopes, one series for men and one for women, marked with nothing but the name of the hospital. Inside each was a card reading S for streptomycin or C for control. The order of the cards had been set from a table of random numbers, and when a patient was approved the next envelope in that hospital's series was opened.

That is the whole invention, and it is worth being precise about why it matters. Randomising is not the hard part. The hard part is that the doctor entering the patient could not know what the envelope said until after the patient was in, so he could not steer the sicker cases towards the drug out of kindness, or the healthier ones towards it to make the drug look good. The allocation was concealed from the person making the decision. That is what the 1948 report described in detail, and it is why the report became a landmark rather than merely a result.

Patients on streptomycin and patients on bed rest alone were usually put in different wards and otherwise treated identically.

The part that is uncomfortable

Neither group was told they were in a trial. It stayed confidential for its whole fifteen months. By the standards of 1947 this was unremarkable and nobody involved appears to have thought it required defending. By the standards that now govern medical research it would be impossible, and the rules that make it impossible were written in large part because of what people learned from the generation of studies this one belongs to.

What they measured

Every month each patient had a chest X-ray. The films were graded by three specialists who did not know which group any patient was in, with disagreements, usually slight, settled by discussion. Sputum was smeared and cultured monthly by bacteriologists who were also kept ignorant of the allocation. Temperature, weight and sedimentation rate were recorded throughout.

What it showed

In the first six months, four of the 55 patients allocated streptomycin died. Fifteen of the 52 allocated bed rest alone died. The X-rays told the same story.

Then it stopped being a triumph. Over the following six months there were eight further deaths in the streptomycin group and nine in the control group, which is to say almost no difference at all. At six months, tubercle bacilli could still be grown from the sputum of 47 of the 55 patients who had been given the drug, against 50 of the 52 who had not.

The reason was resistance. The bacilli in treated patients became resistant to streptomycin, particularly after the fourth month, and patients who had improved began to deteriorate. The original plan had been six months of daily injections. The Council cut it to four.

So the first modern randomised trial in medicine, conducted on the wonder drug of its moment, returned a verdict that the wonder drug could not be relied on to cure the disease. That was the finding, and it was published in the British Medical Journal on 30 October 1948, running to fourteen pages, rather than quietly filed.

It was also the finding that mattered. Because the trial had shown resistance developing under observation, the next question was obvious, and the answer turned out to be combination: streptomycin together with para-aminosalicylic acid worked far better than either alone and slowed the emergence of resistance. Every tuberculosis regimen since has been a combination, for that reason.

Streptomycin itself is now a second-line drug, held back for the cases where the first-line combinations have already failed. The envelopes are the part that lasted.

Sources

• Medical Research Council, "Streptomycin Treatment of Pulmonary Tuberculosis: A Medical Research Council Investigation", British Medical Journal, 30 October 1948 (the trial report itself, with the design, the allocation procedure and the results).
• The James Lind Library, "Medical Research Council (1948) Streptomycin treatment of pulmonary tuberculosis" (the use of concealed random allocation to generate the comparison groups, the concealment of the allocation schedule from those entering patients into the trial, and the report's status as a methodological landmark for describing those steps in detail).
• John Crofton, "The MRC randomized trial of streptomycin and its legacy: a view from the clinical front line", James Lind Library Bulletin, 2004 (a first hand account by the clinician who ran the Brompton Hospital component: Schatz isolating the streptomycin producing organisms in Waksman's laboratory in 1944; the drug's expense and postwar Britain's shortage of dollars; Bradford Hill's argument that it would be unethical not to compare it against bed rest; the committee under Sir Geoffrey Marshall with D'Arcy Hart as secretary, Marc Daniels coordinating and Charlene Agnew as trial manager; the eligibility criteria of ages 15 to 30 with bacteriologically proved acute progressive bilateral disease unsuitable for collapse therapy; the priority given to control patients for scarce beds; the numbered envelopes by hospital and by sex containing cards marked S or C in an order set by random numbers; the trial being kept confidential from patients for its fifteen months; the blinded monthly radiographs graded by three specialists and blinded monthly sputum culture; four deaths among 55 against fifteen among 52 in the first six months and eight against nine in the second; bacilli still cultured from 47 of 55 against 50 of 52 at six months; the emergence of resistance after the fourth month and the decision to stop treatment at four months; and the subsequent finding that streptomycin combined with para-aminosalicylic acid was far more effective than either alone).
• The James Lind Library, "Records of fair tests of treatments" (the wider collection documenting how controlled comparison developed, of which the 1948 trial is the pivotal twentieth century entry).
• National Library of Medicine, "Streptomycin treatment of pulmonary tuberculosis, PubMed record" (the bibliographic record of the 1948 report in the British Medical Journal).