Controlling for diversity in control groups: assessing control group support and its impact on outcomes in smoking cessation evidence synthesis

Presentation video:




Oral session: Understanding and using evidence (4)


Thursday 24 October 2019 - 11:00 to 12:30


All authors in correct order:

Johnston M1, Black N2, Viechtbauer W3, West R4, Hartman-Boyce J5, Michie S4, Eisma M6, deBruin M1
1 University of Aberdeen, UK
2 University of New South Wales, Australia
3 Maastricht University, The Netherlands
4 University College London, UK
5 University of Oxford, UK
6 University of Groningen, The Netherlands
Presenting author and contact person

Presenting author:

Marie Johnston

Contact person:

Abstract text
Background: it is widely recognized that support received by control or comparator groups in trials of interventions varies greatly, can influence outcomes and may therefore bias results in evidence synthesis. However there is no agreed method of assessing and allowing for control group variability in behavioural trials.

We investigated the feasibility of obtaining sufficient information on the potential active content of support delivered to control groups included in a systematic review of trials of behavioural smoking cessation interventions, and whether variability in support predicted heterogeneity in smoking cessation.

Objectives: to examine:
- feasibility of obtaining information on the active content of control group support;
- variability in smoking cessation support between trials;
- relationship between control group content and control group outcomes and trial effect sizes
in smoking cessation interventions.

Methods: in a systematic review of randomized controlled trials (RCTs) of behaviorual smoking cessation interventions for adults, with biochemically-verified outcomes at six months or more, we identified studies by searching the Cochrane Tobacco Addiction Group Specialised Register (PROSPERO: CRD42015025251). Additionally, we asked study authors to complete a checklist of information about their control groups. Data, including active content assessed as behaviour change techniques (BCTs), were reliably extracted from published and unpublished (i.e. obtained from study authors) materials by independent coders.

Results: of 142 eligible control groups, it was possible to collect (near-) complete data on smoking support content and smoking cessation outcomes for 104 studies (74%, N = 23,706 participants). Number of BCTs delivered varied considerably between control groups (M = 15.42, SD = 13.45, range: 0 to 45). Delivery of more smoking cessation BCTs was associated with higher smoking cessation rates in control groups (B = 0.020, P < .001), and smaller trial effect sizes (B = −0.012, P < 0.05); primarily in person-delivered interventions.

Conclusions: it was feasible to obtain usable information on the support provided to control groups in three-quarters of studies, but contacting study authors was resource intensive. Control group support varied in active (BCT) content and this predicted outcomes. When we reanalyzed trial results allowing for control group support, effect sizes were reduced, confirming the importance of variability in control groups.
It is recommended that if control group support is not reported, it can be obtained from study authors, and allowed for in analyzing trials.

Patient or healthcare consumer involvement: smokers who had attended smoking cessation services and had successful or unsuccessful outcomes were consulted individually on the design of the study and on the content of the control group intervention checklist resulting in an amended protocol.