Predicting risk of bias in clinical trials at study design

Risk of bias is central to the appraisal of evidence and refers to the internal validity of a study, that is the extent to which a study is impacted by systematic errors. There are various tools for assessing risk of bias. These include the Cochrane Risk of Bias (RoB) 2.0 tool for randomised trials, Physiotherapy Evidence Database (PEDro) scale, Effective practice and Organisation of Care (EPOC) RoB Tool, Critical Appraisal Skills Programme (CASP) checklist, Scottish Intercollegiate Guidelines Network (SIGN) methodology checklist and the Joanna Briggs Institute (JBI) critical appraisal checklist. However, most risks of bias occur due to limitations of the methods in the study design. There are no known models that play a role to predict an RCTs risk of bias during the design stage, and it is unclear whether an existing risk of bias assessment tool can be modified for this purpose. RCTs are prospective studies which aim to measure the effectiveness of interventions (e.g., pharmaceutical agents, surgical interventions or even therapeutic or preventative therapies). The design involves assigning individuals at random to one or more clinical interventions, which are then followed prospectively, with outcomes of interest and comparisons between the control and intervention groups documented. The Cochrane Risk of Bias 2.0 tool is an updated version of the original Cochrane Risk of Bias tool which was one of the most widely utilised tools for assessing risk of bias in randomised controlled trials (RCTs). The RoB 2.0 tool has five domains, which consider both aspects of empirical evidence and theoretical considerations. These domains are: (i) Bias arising from the randomisation process, applying to the whole study; (ii) Bias due to the deviations from intended interventions, applying to the outcome being measured; (iii) Bias due to missing outcome data; (iv) Bias in measurement of the outcome; and, (v) Bias in selection of the reported result, applying to the specific result. RCTs are regarded as the ‘gold standard’ for proving evidence in healthcare interventions. The aim of a RCT is to reduce sources of bias during the study process. The primary goal of random assignment of individuals to interventions is to minimise selection bias by randomly distributing patient variables that may impact outcome between groups. The process of randomisation increases the chances that any variation in observed outcomes across groups is most likely related to the intervention rather than any other cause. Randomisation is critical to reducing bias as it provides an effective measure of the cause-effect relationship between interventions and outcomes by ensuring that only one factor is tested at a time and that no other variables are present that could introduce bias and influence results. Examples of other ways bias can increase could be due to poor blinding or concealment. High risk of bias is considered when the study’s design or methodology could play a significant role in influencing results further leading to inaccurate conclusions. Assessing risk of bias typically occurs during or after study completion. The assessment is usually informally conducted by peer reviewers, who provide their critical evaluation of the study, or formally by systematic reviewers and/or guideline developers, who assess the strength of evidence across multiple studies. During the design stage of an RCT, it is important that researchers are mindful of potential sources of bias that can arise so that adequate strategies can be implemented.