Changes to design and analysis elements of research plans during randomised controlled trials in Australia.

Abstract

To investigate the frequency and legitimacy of substantive changes to the research plans of published randomised controlled trials (RCTs) undertaken in Australia. Comparison of methodology and analysis plans for RCTs specified in protocol documents (full protocols, published protocol articles, statistical analysis plans, Australian New Zealand Clinical Trials Registry [ANZCTR] registration entries) and described in publications of primary results. 181 RCTs registered with the ANZCTR, 1 September 2007 - 31 December 2013, for which primary results had been published. Changes made to research plan, both overall and by specific item (primary outcome, analysis set, eligibility criteria, sample size, primary analysis method, and treatment arms included in the primary comparison in multi-arm trials); trial characteristics associated with changes. Protocol documents were available for 124 of 181 eligible RCTs (69%; 46 publicly available, 78 provided by trial groups on request). Full audit of RCTs with protocols found clear or probable changes in 111 trials (90%), for 101 of which (91%) it was unclear whether changes had been made blinded to treatment outcomes. After seeking clarification from investigators, changes to 78 trials were confirmed (63%), for 61 of which (78%) changes were made blinded to treatment outcomes. Any change was less likely for trials with publicly available protocols than for trials for which we needed to request protocols (odds ratio, 0.22; 95% CI, 0.06-0.77). Limited reviews of trials without protocols identified that changes had been made to 42 of 57 trials (74%). Changes to RCT study plans in Australia are both frequent and usually made appropriately blinded to treatment outcomes. However, the documentation of changes made to RCT protocols should be formalised to improve transparency.