Background:Randomisation is used to allocate participants and generate similar baseline characteristics/prognostic factors between groups. Baseline comparability has been argued by some to be an important methodological indicator for evaluating the risk of bias of randomized controlled trials (RCTs). However, to what extent baseline comparability is associated with distortion of treatment effects in RCTs has not been exhaustively explored in the literature.
Objectives:To determine the influence of baseline comparability on treatment effects in physical therapy (PT) trials.
Methods:Meta-epidemiological study. Data extraction was conducted independently by 2 reviewers. Trials were considered to have appropriate baseline comparability when authors stated and tested that groups were comparable at baseline on prognostic factors. A trial was considered not to have appropriate baseline comparability when the authors reported that baseline comparability was not accomplished, and there was no indication of any type of analysis to take into account this imbalance. Trials' assessment and categorisation were performed by two independent assessors and verified by a third assessor. We calculated effect sizes for each trial, and we used a two-level analysis using a meta-meta-analytic approach with a random-effects model to perform these analyses.
Results:We analyzed 43 meta-analyses in the PT area, including 393 trials and 44,622 patients. From 393 trials, 288 trials (73.28%) had appropriate baseline comparability, and 105 trials (26.71%) were considered to have inappropriate/unclear baseline comparability. There was no statistically significant difference between treatment effect estimates of trials with appropriate or inappropriate baseline comparability (ES= -0.02; 95% CI -0.14-0.11). When stratifying the analysis by meta-analysis characteristics, it was found that trials in the musculoskeletal area with appropriate baseline comparability seem to have a more beneficial effect than trials with inappropriate baseline comparability [ES: -0.16 ( CI 95% -0.34 , 0.02)].
Conclusions:These results could shed light into the importance of evaluating baseline comparability in RCTs. More research is necessary to gather more evidence in this field.
Patient, public and/or healthcare consumer involvement:The results of this project will inform about the importance of evaluating baseline comparability in individual RCT when included in a systematic review in the context of knowledge synthesis.