Background The evidence base for effective interventions to manage childhood overweight and obesity is growing. However, results from research may not generalise to service delivery settings, and scaled-up interventions may not reduce health inequalities. We examine a scaled–up childhood weight management intervention, examining variations in body mass index by person, family, place and programme.

Methods MEND 7–13 (Mind, Exercise, Nutrition, Do It!) is a multi-component family-based community weight management intervention shown to be effective in a randomised controlled trial (RCT). We used MEND service data from 10,080 children attending MEND programmes from 2007–10. Outcomes were: Change in Body Mass Index (BMI) change over the programme (10 weeks). Anthropometry is measured by trained staff following standardised procedures. We estimated associations between BMI change and participant, family, neighbourhood and programme factors – including pre-specified interactions and random slopes. Multilevel multivariate regression models were used with multiple imputation for missing values. We described participants by BMI at programme start (pre-BMI), age, sex and ethnicity; families by parent-reported employment, lone parent and housing tenure status; places by residential neighbourhood income deprivation, urbanicity, food and built environments; and programmes by the percentage of sessions attended and variables describing programme composition (e.g. group size).

Results BMI was reduced by similar amounts in the RCT and service delivery (RCT BMI change = –0.91 [95%CI: –1.13 to –0.68]), service BMI change = –0.75 [–0.78 to –0.73]. Service BMI reductions were clinically significant in all socio-demographic groups analysed. However, in multilevel models, pre-BMI, age, ethnicity, unemployment status and programme attendance were independently associated with BMI change. For example, in comparison to white children, reductions in BMI were statistically significantly smaller for black and minority ethnic group children (Asian b=0.29, p<0.001, Black b=0.20, p<0.001, (positive coefficients indicate smaller BMI reduction)). Similarly, BMI fell less in children whose parents were unemployed (b=0.14, p<0.001, baseline employed parent). There were no significant differences by sex or neighbourhood factors. Tests of interactions and random slopes were non-significant.

Conclusion Clinically significant BMI changes, similar to those achieved under research conditions, may be replicable in service delivery settings for children of all socio-demographic groups analysed. However, at the population level, scaled up programmes may work better for some groups than others. Public health implications of these results for health inequalities will be discussed.