Statistical Analysis of Survival Experiments2

Final incidences of late-occurring diseases are seriously affected by mortality rates preceding their onset. Meaningful analysis of survival experiments can be obtained by estimations of the net probability of death due to the cause of interest. A comparison of some interval techniques and a noninterval technique developed by Kaplan and Meier suggests that the latter is preferable when the actual ages-at-death are known. This representation can be analyzed for a significant difference between a treated and a control population with a nonparametric test for randomly censored data. The mean age-at-death, a summarized description of the cumulative mortality function, must also be adjusted for competing risks (diseases). Different techniques for correction of incidences of nonlethal diseases must be used, however, since changes in the mortality paHerns for lethal diseases will change the "net probabilities" for nonlethal diseases. Nonlethal diseases can be analyzed by an interval method or a noninterval method similar to the Kaplan-Meier estimate used with lethal diseases.