Statistical methods for testing effects on "maximum lifespan"

It has been noted that certain interventions such as caloric restriction may increase maximum lifespan, whereas other interventions may increase mean or median lifespan but not maximum lifespan. Here the term "maximum lifespan" is used to refer to the upper percentiles of the distribution of lifespan. This is of great interest because increasing maximum lifespan may be an indicator that an intervention is slowing the general process of aging and not merely retarding the development of specific diseases. However, formal methods for testing maximum lifespan have not been elucidated. Herein, we show via simulation that conditional t-test (CTT), a method that is sometimes used, is invalid. We then offer a new method based on quantile regression and we show that this method is, at worst, conservative and remains powerful and valid. (C) 2004 Elsevier Ireland Ltd. All rights reserved.