Event history analysis and inference from observational epidemiology

Systematic inclusion of time in observational epidemiological studies may h elp strengthen the inference to be drawn, but new epidemiological challenge s arise, such as time-dependent confounders - covariates which may change f rom being confounders to being intermediate variables. The focus of this pr esentation concerns two sets of tools: event history analysis and structura l nested failure time models, both applied to a particularly intricate prob lem in observational epidemiology, of empirically assessing the graft-versu s-leukaemia effect after bone marrow transplantation. Copyright (C) 1999 Jo hn Wiley & Sons, Ltd.