Abstract
BACKGROUND: In clinical research, the most appropriate way to assess the effect of an intervention is to conduct a randomized controlled trial (RCT). In the field of rare diseases, conducting an RCT is challenging, resulting in a low rate of clinical trials, with a high frequency of early termination and unpublished trials. The aim of the EPISTOP trial was to compare outcomes in infants with tuberous sclerosis (TSC) who received vigabatrin preventively before the seizures onset with those who received it conventionally after. The study was designed as a prospective, multicentre, randomized clinical trial. However, ethics committees at four centres did not approve this RCT design, resulting in an open-label trial (OLT) in these four centres and an RCT in the other six centres. In this paper, we re-analyse the data from the EPISTOP trial using methods to investigate the influence of allocation bias on the results of the EPISTOP trial. METHOD: A bias-corrected analysis is used to support and strengthen the published results. We included a term representing the effect of selection bias as an influencing factor on the corresponding endpoint in the statistical model. Thus, the treatment effect estimates for the primary endpoint of time to first seizure and additional secondary endpoints are adjusted for the bias effect. RESULT: The bias-corrected analyses for the primary endpoint show that the estimated hazard ratio and associated confidence intervals are in a very similar range (original analysis: HR 2.91, 95%-CI [1.11 to 7.67], p-value 0.0306; bias-corrected analysis: HR 2.89, 95%-CI [1.10 to 7.58], p-value 0.0316). This was also the case for the secondary endpoints. CONCLUSION: The statistical re-analysis of the raw trial data therefore supports the published results and confirms that there is no additional bias introduced by randomization, thereby increasing the value of the results. However, this highlights that this aspect needs to be considered in future trials, especially in rare diseases, to avoid additional biases in an already small sample size where it may be difficult to reach significance.