Analyzing transplant survival data.

Stratified proportional hazards regression is described as a method of estimating multifactorially preoperative factor effects on graft survival--and, at the same time, making due allowances for unknown transplant-center-specific influences. The multifactorial aspect of the method overcomes the biases inherent in analyzing transplant survival data one factor at a time, and stratification allows the data from many centers to be used simultaneously without the dangers associated with simple pooling of data from many sources. The proportionality and regression assumptions implicit in the method enable the data to be used efficiently, but must be validated on the data. Methods by which these assumptions may be relaxed are described--in particular, the stratified piece-wise proportional hazards regression method.