Abstracts

Rationale: Open-label extension studies or follow-on randomised controlled trials (FORCTs) are increasingly conducted in healthcare research. They may provide information about long-term efficacy and tolerability of newer therapies. Concerns have been raised regarding their validity and the potential for patient selection bias which may inflate effect estimates. We compare the common reporting practice of the inferential Intension To Treat (ITT) analysis, with a worst case scenario ITT (WCS-ITT)analysis. The inferential ITT analysis implicitly assumes that patients who drop out of the trial are representative of those who remain in the trial. The WCS-ITT analysis treats patients who drop out of the trial as non-responders. Methods: A systematic review of FORCTs and RCTs of topiramate, levetiracetam and gabapentin as adjuvant therapy in refractory adult epilepsy was conducted. General study attributes, sample sizes, number of responders, numbers dropping out, and reported outcomes for the RCT and FORCT studies were extracted. Reported outcomes and outcomes based on a WCS-ITT analysis were compared for FORCTs and parent RCTs. Results: Ten FORCTs were identified; seven of which were abstracts. All studies reported outcomes based on the sample participating within the FORCT that is ignoring patients who dropped out of the RCT. Less than two thirds of FORCTs reported the original RCT sample size and only half reported on dropouts between the RCT and FORCT. The majority of dropouts occurring during the RCT were for reasons of adverse events or inefficacy. Recalculated fifty percent responder rates in six studies based on a WCS-ITT analysis, led to smaller effect estimates than those reported. For example, a FORCT of levetiracetam reported a responder rate of 43%, which when recalculated based on a WCS-ITT analysis reduced to 28%, comparable to a WCS-ITT analysis outcome of 26% for the parent RCT. Similarly, a FORCT of gabapentin reported responder rates of 35-71%, which when based on a WCS-ITT analysis reduced to 23% comparable to a WCS-ITT analysis outcome of 19% for the parent RCT. Conclusions: FORCTs investigating the longer-term efficacy of anti-epileptic drugs are poorly reported, and as such are very likely to be misleading.RCTs usually report results assuming any dropouts are non-responders. Results for FORCTs, at the other extreme, assume that those dropping out of the parent RCT and FORCT, are representative of those remaining in the extension study.