Large simple trials and trial overviews (Peto and colleagues)

The Medical Research Council had backed collaborative trials in leukemia and other cancers for years, and a group of statisticians wanted to pass on the expertise those trials had built up. Richard Peto led ten British and American statisticians, among them Malcolm Pike, Peter Armitage and David Cox, in writing a two-part guide for doctors with no statistical training. Part I, in the British Journal of Cancer in December 1976, covered design. Part II, in January 1977, gave full instructions for analysis with life tables and the logrank test, including a worked example. The authors called it the first simple account of such analysis for non-statistical physicians; the test itself was older, and the guide cited Peto and Pike's 1973 paper on it.

The case for very large trials came in 1984, in a paper by Salim Yusuf, Rory Collins and Peto titled 'Why do we need some large, simple randomized trials?' A large simple trial would collect only the information necessary to answer the primary question: randomize patients, record the outcome of interest, and avoid the elaborate sub-protocols and extensive data collection that made conventional trials slow and expensive to run. In 1987 the group wrote that moderate effects on serious outcomes needed trials much larger than was then standard.

The ISIS trials, which began recruiting in mid-1981, tested this approach in acute myocardial infarction. ISIS-1 randomized 16,027 patients in 245 coronary care units to intravenous atenolol or none and found 15 percent fewer vascular deaths in the first week, with wide confidence limits. ISIS-2, published in the Lancet in 1988, enrolled 17,187 patients in 417 hospitals; aspirin alone reduced the odds of vascular death at five weeks by 23 percent, streptokinase alone by 25 percent and the two together by 42 percent. The two drugs' effects appeared to be additive.

The group also championed the systematic overview: in 1985 Yusuf, Peto and colleagues pooled some 65 randomized trials of beta-blockers in myocardial infarction, and in 1987 Collins, Gray, Godwin and Peto argued that conclusions should rest on all relevant randomized trials, not one or a few. The 1985 overview concluded that about a year of beta-blockade after discharge could cut mortality by about 25 percent. The Cochrane Collaboration, founded in 1993, extended systematic reviews across medicine.

Later ISIS trials were larger still: ISIS-3 compared three clot-dissolving drugs, and aspirin with or without heparin, in 41,299 patients, and ISIS-4 tested captopril, oral mononitrate and intravenous magnesium in 58,050. ISIS-3 pooled its heparin results with those of the Italian GISSI-2 trial, which had tested the same question.