Premium
Accounting for leadtime in cohort studies: evaluating when to initiate HIV therapies
Author(s) -
Cole Stephen R.,
Li Rui,
Anastos Kathryn,
Detels Roger,
Young Mary,
Chmiel Joan S.,
Muñoz Alvaro
Publication year - 2004
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.1579
Subject(s) - hazard ratio , cohort , observational study , context (archaeology) , medicine , confidence interval , hazard , cohort study , proportional hazards model , human immunodeficiency virus (hiv) , family medicine , history , chemistry , archaeology , organic chemistry
Commonly reported comparisons of differences in disease progression according to disease staging at therapy initiation may be subject to bias if they do not account for the time it took the deferred group to reach the latter stage (that is, leadtime) and for previous events in those who initiate therapy at late stage (that is, unseen fast progressors). To estimate the impact of deferring initiation of highly active antiretroviral therapies (HAART) on time to clinical AIDS in the context of data from observational cohort studies, we describe a method that capitalizes on data from a pre‐HAART period to multiply impute estimated leadtimes and the unseen events among fast progressors. After accounting for leadtime and the unseen events, data from two large cohort studies ( N =739) indicate that deferring HAART initiation until CD4 is below 200 cells/mm 3 was detrimental compared to initiating between 201 and 350 (hazard ratio=1.97; 95 percent confidence interval [CI] 1.09, 3.54), and that failure to account for leadtime resulted in a 38 per cent higher hazard ratio. In contrast, initiating HAART between 201 and 350 did not increase the hazard of AIDS compared to initiating with CD4 between 351 and 500cells/mm 3 (hazard ratio=0.70; 95 per cent CI 0.35, 1.42). Methods presented here offer an approach to analysing prevalent cohort studies and provide procedures to maximize the usefulness of observational data. Copyright © 2004 John Wiley & Sons, Ltd.