Extracranial-intracranial bypass, one; clinical trials, nil.

A large prospective trial published last year seemed to show that extracranial-intracranial bypass was not effective for symptomatic but conventionally inoperable extracranial vascular disease.' The trial, which was purportedly ran-domised but like all surgical trials was not blind, recruited 71 centres and 1377 randomised patients, and cost $9m. The null hypothesis was that the procedure would not make a difference to the incidence of cerebral events over the ensuing five years in patients already symptomatic from conventionally inoperable extracranial vascular disease. The outcome confirmed this negative view and at the time was widely regarded as having put the ultimate scientific boot into what seemed an elegant and rational procedure.2 Nobody doubts that the researchers made a valid deduction from the data received at the study centre, but whether this result can be generalised to the whole population is now open to doubt. Dr T Sundt at the Mayo Clinic in a strenuous round of telephoning, writing, and interviewing contacted most of the international participants in the trial and has shown that the researchers randomised only a small proportion of the eligible patients who came under their care.3 The exact figures are open to dispute even after exhaustive inquiry and reanalysis4 because potential entrants who were medically ineligible were lumped together with those eligible but from whom consent for randomisation from either the referring doctor or the patient could not be obtained. Dr Sundt concluded from his survey of 57 of the centres that 1695 medically eligible patients were operated on outside the trial at a time when 1077 patients were randomised (560 to medical and 517 to surgical treatment). Extrapolating these figures to the whole trial may not be justified, but on their own they imply a high degree of selection. I find it difficult to penetrate the formal reanalysis undertaken by a committee of the American Association of Neurological Surgeons,4 but of all the patients in the study who were truly eligible 570 seem not to have been randomised at a time when 1377 were. Because of the method of collecting data the first figure may be too low; in addition, what actually happened to those excluded is not clear, although I expect that most had an operation. From all this incomplete and confused evidence we may conclude that 50-70% of eligible patients were not included in the trial. To exclude this proportion is serious cause for concern and …