Article type
Year
Abstract
Background: Meta-analyses are rarely analyzed with trial sequential boundaries (TSB). TSB require calculation of OIS in order to determine when strong evidence is reached (Pogue ControlledClinTrials 1997). Allocation concealment of randomized clinical trials (RCT) are categorized as adequate or inadequate by The Cochrane Neonatal Review Group (CNRG). RCTs with inadequate allocation concealment may overestimate intervention effects (Kjaergard AnnInternMed 2001).
Objective: To examine meta-analyses for spurious p<0.05 values with Lan-DeMets discrete sequential boundaries based on optimal information size (OIS), calculated from empirical intervention effect of trials with adequate allocation concealment. Methods: We randomly selected six meta-analyses with at least five trials reporting a binary primary outcome from the 171 systematic CNRG reviews in The Cochrane Library (Issue 2, 2004). Relations between the cumulated z-curve, each cumulative z-value determined by fixed or random effects model, the traditional criterion z=1.96, and the TSB were analyzed using OIS. OIS was calculated based on intervention effect from trials with adequate allocation concealment.
Results: Three meta-analyses presented firm evidence for a beneficial intervention effect as the cumulated z-curve crossed both the z=1.96 and the TSB during the first trials. Two z-curves crossed the z=1.96 temporarily but returned to non-significant values. These z-curves never crossed the TSB. One z-curve never crossed the z=1.96 or the TSB. See figures on the following page.
Conclusions: Three meta-analyses with p<0.05 were supported by z-curves crossing the TSB. However, two meta-analyses showing temporary p<0.05 never crossed the TSB. Accordingly, trials sequential analyses based on trials with adequate allocation concealment may reduce the risk of type I errors without increasing the risk of type II errors.
Objective: To examine meta-analyses for spurious p<0.05 values with Lan-DeMets discrete sequential boundaries based on optimal information size (OIS), calculated from empirical intervention effect of trials with adequate allocation concealment. Methods: We randomly selected six meta-analyses with at least five trials reporting a binary primary outcome from the 171 systematic CNRG reviews in The Cochrane Library (Issue 2, 2004). Relations between the cumulated z-curve, each cumulative z-value determined by fixed or random effects model, the traditional criterion z=1.96, and the TSB were analyzed using OIS. OIS was calculated based on intervention effect from trials with adequate allocation concealment.
Results: Three meta-analyses presented firm evidence for a beneficial intervention effect as the cumulated z-curve crossed both the z=1.96 and the TSB during the first trials. Two z-curves crossed the z=1.96 temporarily but returned to non-significant values. These z-curves never crossed the TSB. One z-curve never crossed the z=1.96 or the TSB. See figures on the following page.
Conclusions: Three meta-analyses with p<0.05 were supported by z-curves crossing the TSB. However, two meta-analyses showing temporary p<0.05 never crossed the TSB. Accordingly, trials sequential analyses based on trials with adequate allocation concealment may reduce the risk of type I errors without increasing the risk of type II errors.