[Roaming through the methodology. XIV. The premature ending of a randomized trial]

A randomized controlled trial that is stopped prematurely because of a striking benefit or a strong untoward effect is most probably suffering from a 'random high'. At premature stopping the conclusions will often either be too optimistic or too pessimistic. In the early phases of an investigation the intermediate results show wider fluctuations around the hypothetical 'truth' than in the later phases. Subjective arguments involved in the stopping of a trial may be eliminated by a so-called triple blind design: the data monitoring and safety committee receives the intermediate results but is not told which data are from the experimental group and which ones from the placebo group.