Premium
Non‐inferiority trial design for recurrent events
Author(s) -
Cook Richard J.,
Lee KerAi,
Li Hongdan
Publication year - 2007
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.2885
Subject(s) - sample size determination , statistics , poisson distribution , robustness (evolution) , null hypothesis , econometrics , computer science , event (particle physics) , marginal distribution , random effects model , mathematics , medicine , random variable , meta analysis , biochemistry , chemistry , physics , quantum mechanics , gene
We describe methods for designing non‐inferiority trials with recurrent event responses arising from mixed‐Poisson models. Sample size formulae are derived for trials in which treatment effects are expressed as relative rates and as absolute differences in cumulative mean functions at a particular time. Simulation studies are conducted to provide empirical validation of the frequency properties of the design and testing procedures under the null and alternative hypotheses using both mixed‐Poisson models and robust marginal methods. The robustness of the design to misspecification of the random effect distribution is also studied empirically. Sample size requirements based on the proposed method are contrasted with those from a design based on the time to the first event for a motivating study of patients with bone metastases at risk of skeletal complications. When the between‐patient heterogeneity in the event rate is small, there may be a considerable reduction in sample size with recurrent event outcomes. Copyright © 2007 John Wiley & Sons, Ltd.