An early parallel group trial in cardiology

The 1948 report of the Committee for the Evaluation of Anticoagulants in the Treatment of Coronary Thrombosis with Myocardial Infarction describes a clinical trial sponsored jointly by the American Heart Association and the US Public Health Service. The idea that anticoagulation might be useful in coronary thrombosis had been around for a few years. It was based on the use of heparin in animal experiments, but it was thought that heparin would be too dangerous in humans. The development of dicoumarol seemed to present an opportunity to test the theory and to develop a useful new therapy. Patients were to be selected for anticoagulant or conventional therapy according to the date of their admission to hospital: patients admitted on odd dates were to receive anticoagulants (heparin for 48 h and then dicoumarol for 30 days), while those admitted on even dates were to receive (undefined) ‘conventional therapy’, without anticoagulants. The report does not include a clear definition of myocardial infarction, nor does it specify either an endpoint (the results being presented more as a series of observations) or an intended length of follow-up (results at six weeks are given). It had been planned to study about 1000 patients with acute myocardial infarction, but the results were reported after only 800 patients had been followed to completion, implying that the accumulating results had been monitored, for the authors judged that the addition of the last 200 cases was unlikely to alter the conclusions. The death rate among patients allocated to anticoagulation was 15% compared with 24% among those in the control group, a statistically significant difference (the authors explain in a footnote that this term meant that on the basis of chance alone, differences such as this would occur less than 1 in 100 occasions). Thromboembolic ‘complications’ (myocardial infarction, stroke and venous thrombosis) were also lower among those allocated to anticoagulants than among controls, albeit at the cost of an increased risk of bleeding (12% compared with 6%). To what extent were these differences in favour of anticoagulants a reflection of the effects of the drugs? The paper describes the two treatment groups as showing a ‘striking similarity’ in terms of age, history of previous infarction, and estimated severityof the admission event and tables suggest that thiswas so. The authors do not seem to have been concerned that the imbalance in the sizes of the anticoagulant and control groups (432 and 368 patients, respectively) may have reflected allocation bias, although they comment that 12% of the control patients received anticoagulants ‘because of pressure on the part of the family or a private physician’ (mentioning in a footnote that ‘this factor became intensified as the study progressed’). Six years after the report was published, reference was made to the study in one of the conferences on therapy organised by the Department of Pharmacology and Medicine of Cornell University Medical College and New York Hospital, one of the participating centres in the anticoagulant study. In this study allocation had been open, whereas in the multi-centre studies done under the aegis of the Medical Research Council allocation bias had been controlled by concealing the allocation schedules from the participants. The discussions on the other side of the Atlantic – led by Dr Harry Gold in particular – emphasised the value of using placebos to promote adherence to allocated treatment. One of the participants in the conference on ‘How to evaluate a new drug’ asked whether placebos are needed when comparing non-subjective outcomes like mortality rates in treated and untreated patients. In response, Dr William Foley commented as follows: