Contamination within trials of community-based public health interventions: lessons from the HENRY feasibility study.

Contamination occurs when participants allocated to trial control arms receive elements of the active intervention. Randomisation at cluster level, rather than individual level, may reduce or eliminate contamination, avoiding the dilution of intervention effectiveness that it may cause. However, cluster randomisation can result in selection bias and may not be feasible to deliver. We explored the extent of contamination in a qualitative study nested within a feasibility study of HENRY (Health, Exercise and Nutrition for the Really Young); a UK community-based child obesity prevention programme. We aimed to determine the nature and impact of contamination to inform a larger planned trial and other trials in community based public health settings. We invited participants to take part in the nested qualitative study who were already involved in the HENRY feasibility study. Semi-structured interviews/focus groups were conducted with children's centre managers (n=7), children's centre staff (n=15), and parents (n=29). Data were transcribed and analysed using an integrative approach. First, deductively organised using a framework guided by the topic guide and then organised using inductive thematic analysis. Potential for contamination between treatment arms was recognised by all stakeholder groups. Staff within the intervention centres presented the greatest risk of contamination, predominantly because they were often asked to work in other children centre's (including control group centres). 'Sharing of best practice' by staff was reported to be a common and desirable phenomenon within community based settings. Parental sharing of HENRY messages was reported inconsistently; though some parents indicated a high degree of knowledge transfer within their immediate circles. The extent of contamination identified has influenced the design of a future effectiveness trial of HENRY which will be clustered at the centre level (with geographically distinct clusters). The common practice of knowledge sharing amongst community teams means that this clustering approach is also likely to be most suitable for other trials based within these settings. We provide recommendations (e.g. cluster randomisation, training intervention facilitators on implications of contamination) to help reduce the impact of contamination in public health intervention trials with or without clustering, whilst enabling transfer of knowledge where appropriate. ClinicalTrials.gov Identifier NCT03333733 registered 6th November 2017.