Open-Label Extension Studies

Abstract

Open-label extension studies (OLEs) create a number of challenges for data analysis and interpretation. These challenges include bias in outcomes and adverse clinical events that may exist because of subject selection and participation. OLEs are usually performed following the successful completion of a randomised, controlled clinical trial (RCT) and, thus, the characteristics of the study participants who chose to continue to be enrolled are likely to differ significantly from the individuals who dropped out of the RCT. Bias may also exist for outcome and adverse clinical events because of unmasking. The effect that treatment knowledge may exert on participant behaviour may be profound and, in addition, leads to an inability to distinguish temporal trends from differences due to administration of the treatment. Statistical issues that must be considered in OLEs include the introduction of both type I and type II errors into the analysis. There may be an increased type I error rate for outcome assessment because of multiple looks at the data set and a lack of rigour in the selection of variables for analysis of outcomes. There may be type II errors in the detection of uncommon adverse clinical outcomes and type I errors for those adverse events that do appear to occur. Finally, statistical procedures to describe effect sizes for outcomes offer challenges for interpretation. This paper will describe the issues associated with interpretation of data from OLEs and the potential sources of bias associated with these types of studies.