controlled trials- the 1948 watershed

7/23/2019 Controlled Trials- The 1948 Watershed

1/4

Education and debate

Controlled trials: the 1948 watershedRichard Doll

Clinical trials before 1946

Important scientific advances seldom occur out of theblue but can be seen in retrospect to have been the cul-mination of processes that have built up over the years.This was certainly true of the introduction of the newmethod of conducting clinical trials, first reported in1948, that has played such a major part in the progressof clinical medicine in the last half century. When I

qualified in medicine in 1937, new treatments werealmost always introduced on the grounds that in thehands of professor A or in the hands of a consultant atone of the leading teaching hospitals, the results in asmall series of patients (seldom more than 50) hadbeen superior to those recorded by professor B (orsome other consultant) or by the same investigatorpreviously. Under these conditions variability ofoutcome, chance, and the unconscious (leave alone theconscious) in the selection of patients brought aboutapparently important differences in the resultsobtained; consequently, there were many competingnew treatments. The treatment of peptic ulcer was, per-haps, more susceptible to claims of benefit than mostother chronic diseases;so that in 1948, when I began toinvestigate it, I was soon able to prepare a list of treat-ments beginning with each letter of the alphabet.Standard treatments, for their part, tended to be passedfrom one textbook to another without ever beingadequately evaluated.

A few clinicians had realised that this wasunsatisfactory and had called for the use of concurrentcontrols, usually suggesting that alternate patientsshould be treated by one or other of the two methodsbeing compared. Fibiger is recorded as having usedthis method in a trial of sensitised serum for the treat-ment of diphtheria in 1898,1 but use of the methodspread slowly. By the 1930s it was used regularly by theMedical Research Council, presumably on the advice

of its statistical committee which was under thechairmanship of Major Greenwood. It was used toevaluate the serum treatment of lobar pneumonia in1934,2 and by DArcy Hart in a trial of patulin for treat-ment of the common cold in 1944.3This was expandedby Wilson et al in 1946 to a factorial design to enabletwo comparisons to be made in the same group ofpatients. This necessitated four treatment groups inorder to test both the value of a low fat diet and dietarysupplementation with di-cysteine in the treatment ofacute hepatitis.4 Patulin, which had been discovered bya chemist who had failed to isolate penicillin, had noapparent effect on the common cold, and a low fat diet

had no apparent effect on acute hepatitis, butdi-cysteine seemed to reduce the duration of acutehepatitis by a few days.

The introduction of randomisation

The technique of alternate allocation had one majordisadvantage: the investigator knew which treatmentthe next patient was going to receive and couldbeand indeed often wasbiased by knowing what thenext treatment wouldbe when deciding whether or nota patient was suitable for inclusion in the trial. Even

blinding the investigator to the nature of the giventreatment, which was often possible, by presenting thetreatments in similar forms labelled A and B did notget over the difficulty completely; the investigator couldquickly get the impression that one treatment wassuperior to the other and subsequently be biased indeciding on the next patients eligibility.

One way of avoiding such biases would be to dividethe patients into two similar groups before it wasknown which group would get which treatment andthen, at the last minute, allocate one whole group toone treatment and the other to another by tossing acoin. This method was proposed by van Helmont, a

Summary points

Claims of therapeutic benefit are often misleadingunless there are concurrent control groups

When treatments are allocated to alternatepatients there is a risk of bias in the selection ofpatients

Randomly allocating individuals after entry into atrial eliminates bias and provides a properestimate of random error

Randomisation can be done within strata of thelikely response to treatment, if clinicians candefine the strata sufficiently clearly

Ethical standards should be the same fortherapeutic trials and routine care

Large numbers of participants may be needed ina trial if moderate yet important effects are to bedetected

Clinical TrialService Unit andEpidemiologicalStudies Unit,University ofOxford, RadcliffeInfirmary, OxfordOX2 6HE

Richard Doll,honorary member

BMJ 1998;317:121720

1217BMJ VOLUME 317 31 OCTOBER 1998 www.bmj.com


2/4

medicinal chemist, in 1662 when he challenged the

academics of the day to compare their treatmentsbased on theory with his based on experience. Let ustake out of the hospitals, out of the Camps, or fromelsewhere, 200, or 500 poor People, that have Fevers,Pleurisies, etc. Let us divide them into half, let us castlots, that one half of them may fall to my share, and theother to yours . . . We shall see how many funerals bothof us shall have. But let the reward of the contention orwager, be 300 florens, deposited on both sides.5 Sadly,the challenge was not accepted. The technique was,however, actually put into practice by Amberson et al260 years later to assess the value of sodium gold thio-sulphate in the treatment of pulmonary tuberculosis.6

Amberson et al divided 24 patients into two groups of12, the members of each group being individually

matched in pairs. They then tossed a coin to decidewhich treatment each group should get.

This technique suffers, as Armitage has pointedout,5 not only from the virtual impossibility of trulybeing able to match cases, but also because it providesno means of measuring the relevant random error.Both difficulties are overcome within quantifiable limitsby the randomisation of individuals. This techniquehad been used in agricultural experiments describedby Fisher in 1926 (when plots of land were individuallyrandomised)7; Bradford Hill had recognised the desir-ability of using it in clinical medicine when hepublished a series of articles on the principles of medi-cal statistics in 1937.8 He did not recommend the ran-

domisation of individuals then, preferring that the twotreatments for comparison be allocated to alternatepatients because, as he wrote in 1990, by referring tothe randomisation of treatments he might have scareddoctors off any use of concurrent controls. 9 In 1946,when he judged the time was right, he recommendedthe randomisation of individual patients and thisrapidly gained acceptance among medical scien-tists.10 11 He advocated it not so much because itprovided a proper estimate of randomerror,whichwasthe principal reason it was advocated by Fisher, but onthe practical grounds that it eliminated bias inselection.

The trial in which treatments first began to be allo-

cated randomly to individuals was one designed to testthe efficacy of immunisation against whoopingcough,10 not the trial of streptomycin for treatingpulmonary tuberculosis.11The latter trial, organised byDArcy Hart and Daniels, started in September 1946, afew months after the whooping cough trial; but it wasreported in 1948, three years earlier than the results ofthe whooping cough trial. Consequently, although itwas certainly the first to be reported, it undeservedlyearned the reputation of being the first trulyrandomised trial. In both cases efforts were made toblind the assessor to the participants treatment and,when practicable as in the first trial, to blind theparticipants. In both cases ethical considerationsplayed a major part.

Ethical considerations

At that time there were not any ethical committees toconsult nor were there any ethical criteria laid down bythe Medical Research Council or any other responsiblebody. Medical ethics were primarily defined by theHippocratic oath, which all newly qualified doctorswere required to swear. Bradford Hill was not a doctor,but he had such a deep understanding of the nature ofmedical practice that the lecture that he gave at theRoyal College of Physicians on medical ethics andcontrolled trials was listened to with respect and waswidely acclaimed.11

In the trial of streptomycin, the first issue that had

to be faced was whether it was ethical to withhold fromany patient a drug that had been effective in animalexperiments and had had encouraging clinical resultsin the few published reports. There was, however, onlya small amount of the drug in Britain and it was notpossible to buy more from abroad. It was agreed to usethe limited supplies to treat patients with twoconditions that had previously been invariably fatal:miliary tuberculosis and tuberculous meningitis. Theamount of streptomycin left over was insufficient totreat more than a tiny proportion of the peopledesperately ill with other types of tuberculosis. TheMedical Research Councils Streptomycin in Tubercu-

NICK

SINCLAIR

Richard Doll, Philip DArcy Hart, and Francis Avery Jones. Portraits by Nick Sinclair (published in With Head and Heart and Hand: Photographs of 20th Century BritishDoctors, BMJ Books, 1997: see www.bmj.com/collections/century/data/exhibit.htm)


1218 BMJ VOLUME 317 31 OCTOBER 1998 www.bmj.com


3/4

losis Trials Committee agreed that it would have beenunethical not to have seized the opportunity to designa strictly controlled trial which could speedily andeffectively reveal the value of the treatment.12 Thequestion of whether it was ethically justifiable towithhold the drug from any patient was, therefore,answered with an unhesitating Yes.

Two other questions that the committee considered

were whether the doctors involved could modify thetreatment schedule and whether the control patientsshould be given apparently similar placebos. It wasagreed that the doctor must always do for his patientwhatever he really believes to be essential for thatpatient to return him to health.12 This meant that ifany patient seemed likely to benefit from an inducedpneumothoraxthe only specific treatment availablefor pulmonary tuberculosis before the introduction ofstreptomycinthe treatment must be given irrespec-tive of whether it upset the balance of the streptomycinand the control groups, as in fact it proved to do. Theuse of a placebo was ruled out in the interest of thepatients because it would have required an intramuscu-lar injection four times a day for four months. The

response to treatment could be assessed objectivelywithout it: psychological factors would have littleimpact on such a serious disease, and there was noneed in the search for precision to throw commonsense out of the window.12

The committee did not discuss the need to obtaininformed consent; the overriding issue was the welfareof the patient. Bradford Hill inveighed strongly againstthe compulsion to obtain formal consent if thisrequired giving a frightening account of the risks asso-ciated with the patients condition. Does the doctorinvariably seek the patients consent before using a newdrug alleged to be efficacious and safe? If the answer isNo, then what process, one may ask, makes it needful

for him to do so if he chooses to test the drug in such away that he can compare its effects with those of theprevious orthodox treatment?12

The question he asked might be answereddifferently today, but the principle with which he wasconcernedthat there should not be one standard forthe ethics of therapeutic trials and another for routinemedical careis still valid and ought to be a majordetermining factor in our approach to patients.13 Toseek informed consent when it went against thepatients interest was, in Bradford Hills opinion,unethical and should be dispensed with subject, aswould now generally be agreed, to approval by anappropriate independent committee.

The situation in the whooping cough preventiontrial was different. Parents of children aged 6-18months were asked to volunteer to have their childrenentered into the trial; they were given a pamphletdescribing the study, which included the informationthat half the inoculations would not be against whoop-ing cough but would be anti-catarrhal. No child wasentered until a consent form had been received; thiscondition was emphasised in the report by theacknowledgment of the many parents who, in the fullknowledge that their children would not necessarilyreceive pertussis vaccination, consented to take part inthe investigation.9

Clinical opposition

The spread of randomisation, until it became an essen-tial element of trials submitted to licensing authoritiesfor the approval of new drugs, was initially slow andnot without opposition. This was most commonlyexpressed along the lines of Lewiss criticism of whathe called the statistical method of testing treatment.14

Lewis was the doctor in charge of the department of

clinical research at University College Hospital,London, and the doyen of clinical research in theUnited Kingdom in the 1930s. He died in 1945 beforerandomised trials were introduced but I can imaginewhat his reaction would have been. Lewis thought thatwhen testing treatments for acute diseases two groupsof patients that were as similar as possible should betreated in exactly the same way and concurrently,except that one group should receive the remedy andthe other should not. However, he added that it is tobe recognised that the statistical method of testingtreatment is never more than a temporary expedient,and that but little progress can come of it directly: for ininvestigating cases collectively, it does not discriminatebetween cases that benefit and those that do not, and so

fails to determine criteria by which we may knowbeforehand in any given case that treatment will besuccessful.14

Lewiss objection was repeated many times in thefirst few years after randomisation was introduced.Bradford Hill would reply: Tell me the criteria to dis-tinguish patients who will respond from those whodont and we will build this into the trial(A BradfordHill, personal communication). There was never anyserious response to Bradford Hills challenge, and itcame to be accepted that randomisation was appropri-ate within strata defined by the clinicianif theclinician could define them sufficiently clearly for prac-tical use.

Modern developments

Early randomised trials can properly be criticised onthe grounds that they were often too small to have anychance of detecting moderate effects. Small trials canbe successful when the effect is large but this seldomoccurs. They can also be successful when the effect ismoderate and the outcome is measured quantitatively,as in the series of trials of treatment for gastric ulcerconducted by Avery Jones and me in the 1950s and60s, which recorded the percentage reduction in thesize of the ulcer over a fixed time.15 However, when theoutcome is qualitative rather than quantitative, moder-ate but important effects will often be missed unless the

number of patients treated is large. For example, manyearly trials of the treatment of myocardial infarction,stroke, and cancer, which were not large by modernstandards, consequently led to the misleading conclu-sion that there was no benefit. Bradford Hill and hisstudents, such as myself, were primarily concernedwith getting the principle adopted, and to have pressedfor trials on thousands of patients would have been selfdefeating. There were too few physicians, leave alonesurgeons, who were willing to expose their theories tocold scientific investigation. Multicentre trials wereorganised from the beginning (the streptomycin trialinvolved six centres), several centres were successfully


1219BMJ VOLUME 317 31 OCTOBER 1998 www.bmj.com


4/4

involved in trials like those of adrenocorticotrophichormone and cortisone for the treatment of ulcerativecolitis, and more centres became involved in the Medi-cal Research Councils trials of treatment for differentforms of leukaemia.16 It was many years beforerandomisation was accepted as such a normalprocedure. Only then did it become possible to organ-ise the groundbreaking international study of infarctsurvival (ISIS) trials for the treatment of myocardial

infarction, which involved hundreds of centres andrandomly allocated tens of thousands of patients, andthereby showed the value of moderate improvementsin the treatment of common diseases.17

Without Bradford Hill, randomisation would havecome about sooner or later, perhaps introduced byRutstein in the United States. Rutstein collaboratedwith Bradford Hill in the design of an Anglo-Americantrial of adrenocorticotrophic hormone, cortisone, andaspirin in the treatment of acute rheumatic fever.18

Randomisation would have been adopted much moreslowly, however, without Bradford Hills understandingof medical susceptibility and medical ethics andwithout his concern for simplicity of design and clarityof presentation. Modern authors please note.

Competing interests: None declared.

1 Fibiger J. Om serumbehandlung af Difteri.Hospitalstidende 1898;6:309-25.

2 Medical Research Council Therapeutic Trials Committee. The serumtreatment of lobar pneumonia.BMJ1934;i:241-5.

3 Medical Research Council Patulin Trials Committee. Clinical trial ofpatulin in the common cold.Lancet1944;ii:373-4.

4 Wilson C, Pollock MR, Harris AD. Diet in the treatment of infectioushepatitis.Lancet1946;i:881-3.

5 Armitage P. The role of randomisation in clinical trials. Stat Med1982;i:345-52.

6 Amberson JB Jr,McMahon BL, Pinner M. A clinical trial of sanocrysin inpulmonary tuberculosis.Amer Rev Tuberc1931;24:401-35.

7 Fisher RA. The arrangement of field experiments. Journal of the Ministryof Agriculture1926;33:503-13.

8 Hill AB.Principles of medical statistics. London: Lancet, 1937.9 Hill AB. Memories of the British streptomycin trial in tuberculosis.

Control Clin Trials1990;11:77-9.10 Medical Research Council Whooping-Cough Immunization Committee.The prevention of whooping-cough by vaccination.BMJ1951;i:1463-71.

11 Medical Research Council Streptomycin in Tuberculosis Trials Com-mittee. Streptomycin treatment for pulmonary tuberculosis. BMJ1948;ii:769-82.

12 Hill AB.Medical ethics and controlled trials.BMJ1963;i:1043.13 Collins R,Doll R, Peto R. Ethics of clinical trials.In: CJ Williams,ed.Intro-

ducing new treatments for cancer:practical,ethical, and legal problems.Chiches-ter:John Wiley,1992.

14 Lewis T. Clinical science illustrated by personal experiences. London: Shawand Sons, 1934:178-83.

15 Doll R. Medical treatment of gastric ulcer.Scott Med J1964;9:183-96.16 Medical Research Council Working Party on Leukaemia in Childhood.

Improvement in treatment for children with acute lymphoblastic leukae-mia.Lancet1986;i:408-11.

17 Second International Study of Infarct Survival Collaborative Group.Randomised trial of intravenous streptokinase, oral aspirin, both, or nei-ther among 17,187 cases of suspected acute myocardial infarction:ISIS-2.Lancet1988;ii:349-60.

18 Rheumatic Fever Working Party of the Medical Research Council of

Great Britain, Subcommittee of Principal Investigators of the AmericanCouncil on Rheumatic Fever and Congenital Heart Disease, AmericanHeart Association. Treatment of acute rheumatic fever in children: acooperative clinical trial of ACTH, cortisone, and aspirin. BMJ1955;i:555-74.

(Accepted 6 October 1998)

Use of randomisation in the Medical Research Councilsclinical trial of streptomycin in pulmonary tuberculosis inthe 1940s

Alan Yoshioka

A slight mystery surrounds the clinical trial that thisspecial issue of the BMJcommemorates.1 It is nowwidely recalled that pulmonary tuberculosis patients inthe Medical Research Council (MRC) trial wereallocated to a streptomycin treatment group and acontrol group by a process using random samplingnumbers and sealed envelopes (box). The editorial thatintroduced the MRCs report to BMJ readers on 30October 1948 called attention to this new scheme, dis-tinguishing it from the older practice of takingalternate cases in order of admission to hospital as themethod of creating a control group.2 The statisticianinvolved, Professor (later Sir) Austin Bradford Hill, had

been promoting the use of random allotment sincebefore the second world war.3 Remarkably, however,the word random appeared nowhere in the MRCsfiles on streptomycin for 1946. During that year, thenow famous scheme was explicitly mentioned in asingle letter. So why did the MRC use randomisation inthis clinical trial?

Any answer is tentative as contemporary evidenceabout the reasoning of the members of the MRCscommittee is thin. We can make some inferences fromthe arguments they used after the fact to justify the ran-dom allocation scheme to the medical profession. Theirdesign, and how it was presented, should be understood

The MRC trialreport isreproduced onour website,together with itstwo accompanyingeditorials

Summary points

Randomised allocation of patients is rarelymentioned in the Medical Research Councilsdocuments on streptomycin clinical trials

The meaning of the term randomisation hasshifted over time; justifications for usingrandomisation have also changed

Streptomycin was isolated in 1943 at RutgersUniversity in New Jersey; the MRC began

planning clinical trials in 1946; the first patientswith tuberculosis entered the trials in 1947

The British government initially purchased 50 kgof American streptomycin for the MRC; most ofthe supply went to the clinical trial in pulmonarytuberculosis

Public demand for streptomycin was far in excessof supplies in Britain. Randomisation relieved theMRCs clinicians of responsibility for decidingwho would be treated


Centre for Historyof Science,Technology andMedicine, ImperialCollege, LondonSW7 2AZ

Alan Yoshioka,historian

Correspondence to:Mr Yoshioka,PO Box 166, Pictou,NS B0K [email protected]

BMJ 1998;317:12203

1220 BMJ VOLUME 317 31 OCTOBER 1998 www.bmj.com

controlled trials- the 1948 watershed

Documents