Use of Irwin's restricted mean as an index for comparing survival in different treatment groups--interpretation and power considerations.