Accounting for variability in control groups and reporting quality in systematic reviews of behavioural trials

Abstract

Background: Intervention and control groups in behavioural trials tend to be exposed to different levels of behavioural support. Systematic reviews ignoring variability in control group support may draw inaccurate conclusions about the effectiveness of intervention (components). Additionally, relying only on published intervention and control group descriptions may be problematic due to poor reporting. This presentation discusses these methodological problems and some solutions. Methods: The results from two (ongoing) systematic reviews (SR1: HIV treatment adherence (k=34), SR2: smoking cessation (k=148)) collecting additional intervention and control group descriptions from study authors, and coding these for behaviour change techniques (BCTs), are used to examine these issues using descriptive and correlation analyses. Findings: A minority of BCTs (about 40% for interventions and 20% for control group support in SR2) were described in publications; the remaining BCTs were only found in additional information obtained from study authors. The number of BCTs delivered to intervention and control groups varies widely between trials (1-60 and 1-45 BCTs respectively in SR2). The overlap in type of BCTs delivered to intervention and control groups in the same trial, varied between trials (SR1: 0-50%, SR2: 0-67%), and increased with the number of BCTs delivered to controls (SR1: r=.68, SR2: rho=.69, p’s<.01). Bivariate mixed-effects regression models can be used to account for variability in control group support (SR1). Discussion: Collecting additional information from study authors is essential for ensuring comprehensive intervention and control group descriptions and meaningful meta-analyses. Variability in control groups between trials strongly affects interventions’ ability to demonstrate superiority.