Randomised Controlled Trials (RCTs) improve the robustness of evidence and the knowledge base but only if they are conducted well. A key methodological component of an RCT is allocation concealment. Inadequate allocation concealment can lead to selection bias. Selection bias needs to be prevented, and, where possible detected when undertaking quality assessments. This thesis makes an original contribution to the knowledge base by exploring a method to detect selection bias in systematic reviews and improve allocation concealment implementation and reporting in RCTs.
Paper 1 presents an exemplar of a novel technique to detect selection bias within systematic reviews. I found that systematic reviews exhibited baseline age heterogeneity, which can only be explained by the inclusion of RCTs with baseline selection bias. Paper 2 explores this technique further, demonstrating that more powerful baseline covariates in a baseline meta-analysis found greater heterogeneity.
In paper 3 I show this baseline heterogeneity is not a function of the reviews including ‘old’ trials because poor allocation concealment, which is the most likely explanation of selection bias, remains prevalent in recently published RCTs. In these recent RCTs, envelope and blocking implementation and reporting was suboptimal. In response, I focused my methodological research on improving the implementation and reporting of these methods.
In paper 4 I show envelopes are widely used as an allocation concealment method and their use will likely continue. Therefore, I created an original reference guide for researchers to utilise to aid the robust implementation and reporting of envelopes. In paper 5 I published the first report using empirical evidence, rather than modelling studies, that RCTs using small variable block sizes that include a block of two are at risk of subversion
Throughout my research I make recommendations to ensure allocation concealment is implemented and reported robustly to improve evidence generation and synthesis.
