Background: Control groups are used in clinical trials to increase confidence that any improvements in patient outcomes are due the therapy under investigation and not to other factors. The reported effect size of any intervention is estimated from differences in outcomes achieved by intervention participants in comparison to control participants. Clinical heterogeneity in control groups across different studies can make the pooling of data from these studies in one meta-analysis questionable or reduce certainty in their results.
Aim: The aim of this study was to evaluate: 1) the variability in the types of control groups used in studies that have been pooled in meta-analyses in Cochrane reviews on neurorehabilitation interventions; and 2) how authors of Cochrane reviews on neurorehabilitation interventions have taken information about control groups into consideration when making decisions to undertake meta-analyses and interpreting their results.
Methods: We searched the Cochrane library for reviews on neurorehabilitation interventions published between 2012 and 2016 that included at least one meta-analysis involving a control group. We extracted data from included reviews on the review characteristics, the characteristics of the included meta-analyses, and any information on how the review authors managed control groups in the conduct and interpretation of meta-analyses.
Results: The 43 included reviews pooled data from 358 clinical trials, with an average of 5±5 clinical trials (range: 2-45) contributing to each meta-analysis. The majority of clinical trials involved a control group containing active treatments (61.7%; 221 of 358), often "treatment as usual" controls without any additional placebo or sham intervention. Over half (58.1%; 25 of 43) of the included meta-analyses involved pooling of data from studies with a mix of different types of control groups, with an additional 25.6% pooling data from studies where control participants had received a range of different active treatments. The influence of different control groups on the summary results from meta-analyses was not analyzed in 21 (48.8%) of the included reviews.
Conclusions: Further work is needed to develop: standardized ways to categorize control conditions in rehabilitation trials; more guidance on reporting criteria for control groups in rehabilitation trials; and agreed methods for managing different control types in one meta-analysis.