A study of the use of the probability-of-being-in-response function as a summary of tumor response data.

The probability-of-being-in-response function (PBRF), proposed by Temkin (1978, Biometrics 34, 571-588) as a vehicle for analyzing transient response data, is examined using the simplified parametric model in which each of the failure processes is assumed to be exponential. Properties of the PBRF under these assumptions are discussed. It is demonstrated that the PBRF is a very complete summary of the data and has attractive properties for its intended use as a visual display. The model is extended to incorporate covariates, and it is demonstrated that the desired properties can be generalized under specific conditions for the covariates.