Fixing the number of events in large comparative trials with low event rates: a binomial approach

In designing large clinical trials where the event rates are low, capabilities can be improved if one fixes the number of events rather than the number of patients. The dependent variable, the number of events in the treatment group, has essentially an unconditional binomial distribution. Simple tables giving the minimum detectable difference are provided for common type I error values and power. Implications for sequential monitoring are also presented. The methods can be applied to large randomized, stratified trials utilizing censored survival data, planned under proportional hazards and equal target population competing losses, provided that event rates are low.