ESTIMATING THE EFFECT OF ZIDOVUDINE ON KAPOSIS-SARCOMA FROM OBSERVATIONAL DATA USING A RANK PRESERVING STRUCTURAL FAILURE-TIME MODEL

Researchers commonly express scepticism about using observational data to estimate the effect of a treatment on an outcome the treatment is intended to affect. In this paper, we consider using data from the Mul ticenter AIDS Cohort Study (MACS) to determine whether zidovudine prev ents the development of Kaposi's sarcoma among HIV-positive gay men. S everal methodologic issues common to observational data characterized the study: information on potentially important confounders was missin g at some study visits; investigators did not always know the time of changes in treatment level, nor the value of confounders at that time, and the censoring process depended strongly on time-varying covariate s related to outcome. We describe application to our data of Robins' p aradigm for defining, modelling and estimating the effect of a time-va rying treatment and show how to modify his approach to deal with the m ethodologic issues we have mentioned. Further, we demonstrate that rel ative risk regression is less well equipped to deal with these issues, We compare our results to the findings from randomized trials, and co nclude that observational studies may sometimes be useful in evaluatin g the effect of treatment on an intended outcome. (C) 1998 John Wiley & Sons, Ltd.