For time-to-event data, the number of events (and associated number of patients) required to provide the desired power should have been specified up front, and when this is the case it provides a simple guide as to the 'right time' to publish. This will depend both on the event rate and the accrual rate; the slower the rate of accrual, the longer the median follow-up at the closure of the trial and the shorter the time required after closure for the required number of events to accrue. It is necessary to bear in mind that those running trials have an obligation to the patients who agree to enter trials that they should contribute to the final results. Therefore presentation of results should not generally take place until all patients have at least completed their treatment. Exceptions to this may occur, firstly when the trial treatment is very long-term - for example ten years of hormone therapy for prostate cancer - and secondly when a trial stops early because of unequivocal evidence of superiority of one treatment over another. In this latter situation it is important to publish the results, as they were when the decision was made to close the trial, as soon as possible. In this situation, careful thought must be given as to how these results are communicated to patients, particularly those who may be still receiving treatment and those treated with the inferior therapy. If the results change with further follow-up, then subsequent publication maybe appropriate, subject to the guidance given in the next section.
Was this article helpful?