A study has been made of the way in which the number of events available for analysis in a clinical trial was dependent on the recruitment period, the maximum follow-up time on individual patients and the length of time between the start of the trial and its analysis. The events considered were deaths, local recurrences and late radiation effects on normal tissue in patients treated for cancer of the laryngo-pharynx by two different fractionation regimes. The relationship is demonstrated between the number of events and the 95% confidence intervals that can be placed on differences between results in the two arms of the trial. It was found, in this particular trial, that no significant improvement in precision was gained by following up patients beyond 5 years or carrying out the analysis later than 2 years after the end of recruitment. The results are discussed in the context of the initial design of clinical trials, particularly those in which the aim is to test therapeutic equivalence.