Clinical Trials

❘❙❚■ Chapter 31 | Interim Monitoring and Stopping Rulesplaces almost all of the Type I error on the final analysis. The Lan–DeMetsmethod is flexible enough to incorporate either of these approaches and alsoallow for compromise between the two, depending on recruitment rates.It is possible to use group sequential methods not only for treatment-differencetrials but also for equivalence trials, one-sided tests, and asymmetric boundaries.An example of an asymmetric boundary might be to use an O’Brien–Flemingmethod to monitor the benefit of the new intervention and a Pocock method formonitoring detrimental effects of the new intervention. Different methods couldbe used in the same trial to monitor different endpoints. For example, anO’Brien–Fleming method might be used to monitor efficacy at the same time asusing the Pocock method to monitor safety.What situations can drive early trial termination?There are four main outcomes from interim monitoring:• trial continues as planned• trial continues with protocol amendments (more patients or centers, orrequests about details of adverse events for closer scrutiny or amendmentto inclusion/exclusion criteria due to unexpected observations)• trial is halted temporarily for further safety data or for protocolamendment and further training in the new protocol• trial is stopped prematurely and permanently due to early divergenceof treatment effects (either in favor of or against the new treatment)Interim monitoring allows the DSMB to evaluate whether a divergence of eventrates between groups is occurring. This divergence usually has to be present fortwo consecutive interim analyses. It will then result in early trial termination toavoid further disadvantage to patients in the higher event rate group. If the DSMBis blinded then the higher event rate may or may not be in favor of the newtreatment. If the DSMB is unblinded then they will know if the new treatment hasgreater benefit or not. For example, the BHAT (Beta-blocker Heart Attack Trial)was terminated 9 months early, mainly because the observed treatment benefitwith beta blockers was judged significant and likely to remain so during the9 months remaining in the trial [6].Similarly, if a new treatment is unlikely to be superior, ie, there is no curvedivergence some way through the trial, the trial may be stopped under futilityrules since it is unlikely to have future benefit if the drug effect is supposed to bedemonstrated in the first few months. CARET (Beta-Carotene and Retinol360