Biomedical Engineering Reference
In-Depth Information
the observed time of death and no later than the observed time of death
itself.
3. The participant may have been lost to follow-up for a variety of reasons,
or the study may have been terminated before virological failure. Loss
to follow-up may have resulted, for example, from withdrawal of the
participant from treatment due to drug toxicity. In this case, again,
which of virological failure or death occurred first would not be known,
but the composite outcome would be right-censored by the last study
monitoring time before loss to follow-up.
It follows then that the outcome of interest is subject to a combination
of interval-censoring and right-censoring, determined by which of the above
scenarios applies to a given participant. This, in addition to the wealth of
time-varying covariates recorded during the study period, poses significant
methodological challenges. The scientific question of focus, here, consists of
determining the causal effect of various antiretroviral therapies on time until
virological failure or death using data collected as part of the Tshepo study.
8.1.2
Overview
As indicated above, in contrast to classical survival analysis, where interest lies
in estimating the distribution of a time-to-event outcome, in this chapter we
wish to develop estimators of a causal effect of treatment on a time-to-event
outcome, using survival data subject to a combination of interval-censoring
and right-censoring, and incorporating data from longitudinally collected time-
dependent covariates. The motivating example presented above is but one
setting in which the need for such methodology arises. Yet, because in most
studies rather than being followed in continuous time, participants are mon-
itored at study visits occurring at random times, it is possible to argue that
most longitudinal studies where individuals are followed for a specific event
will in fact give rise to time-to-event data subject to a combination of interval-
 
Search WWH ::




Custom Search