Study Of Trial Policy Of Interim Trucation-2 (STOP-IT-2)
Lead Research Organisation:
University of Oxford
Department Name: Primary Care Health Sciences
Abstract
A recent review 143 randomized trials that stopped earlier than
planned because of a larger than expected treatment effect showed
that such trials are increasingly common but often show implausibly
large treatment effects. This study (STOP IT-2) aims to understand
how biased such early stopping is, and how we might fix the problem.
To do this we will find and summarise all other research trials that
answer the same questions as those that stopped early and compare the
results. The results will have considerable importance to the
interpretation of trials that stop early, and will help clinicians,
policy makers and consumers in interpreting such trials.
planned because of a larger than expected treatment effect showed
that such trials are increasingly common but often show implausibly
large treatment effects. This study (STOP IT-2) aims to understand
how biased such early stopping is, and how we might fix the problem.
To do this we will find and summarise all other research trials that
answer the same questions as those that stopped early and compare the
results. The results will have considerable importance to the
interpretation of trials that stop early, and will help clinicians,
policy makers and consumers in interpreting such trials.
Technical Summary
Recently, Montori et al. published STOPIT-1 , a systematic review
(SR) of 143 randomized trials that stopped earlier than planned
because of a larger than expected treatment effect (JAMA
2005;294:2203-2209). Such truncated randomised trials (tRCT) are
becoming increasingly common but often show implausibly large
treatment effects. STOP IT-2 seeks to determine the magnitude and
determinants of bias that truncation introduces. We will search for
SRs addressing the same question as the 143 tRCTs in STOPIT-1. For
each of the SRs we will conduct meta-analyses addressing the outcome
that led to the early termination of the tRCT(s) and compare the
relative risk (RR) generated by the tRCT with the RR from all other
studies. Second we will use multivariable regression to determine the
independent contribution of the decision to stop early on the
magnitude of the treatment effect and modifiers of that contribution
(such as, for instance, the number of events that had occurred at the
time of truncation). Finally, we will compare possible methods for
debiasing the estimates from tRCTs, in particular the use of Bayesian
methods using conservative informative priors to regress to the
mean overoptimistic small studies.
(SR) of 143 randomized trials that stopped earlier than planned
because of a larger than expected treatment effect (JAMA
2005;294:2203-2209). Such truncated randomised trials (tRCT) are
becoming increasingly common but often show implausibly large
treatment effects. STOP IT-2 seeks to determine the magnitude and
determinants of bias that truncation introduces. We will search for
SRs addressing the same question as the 143 tRCTs in STOPIT-1. For
each of the SRs we will conduct meta-analyses addressing the outcome
that led to the early termination of the tRCT(s) and compare the
relative risk (RR) generated by the tRCT with the RR from all other
studies. Second we will use multivariable regression to determine the
independent contribution of the decision to stop early on the
magnitude of the treatment effect and modifiers of that contribution
(such as, for instance, the number of events that had occurred at the
time of truncation). Finally, we will compare possible methods for
debiasing the estimates from tRCTs, in particular the use of Bayesian
methods using conservative informative priors to regress to the
mean overoptimistic small studies.