Abstract
The use of two or more primary correlated endpoints is becoming increasingly common. A mandatory approach when analyzing data from such clinical trials is to control the family-wise error rate (FWER). In this context, we provide formulas for computation of sample size and for data analysis. Two approaches are discussed: an individual method based on a union–intersection procedure and a global procedure, based on a multivariate model that can take into account adjustment variables. These methods are illustrated with simulation studies and applications. An R package known as rPowerSampleSize is also available.
ACKNOWLEDGEMENT
We thank the Danone Research Clinical Study Platform for making the data available.
Notes
Note. Global: Global method based on multivariate model.
Indiv: Individual procedure for known covariance matrix.
*: Significant association.