BackgroundViolence in and around premises licensed for the on-site sale and consumption of alcohol continues to burden the NHS with assault-related injuries.Trial designA randomised controlled trial with licensed premises as the unit of allocation, with additional process and cost-effectiveness evaluations.MethodsPremises were eligible (n = 837) if they were licensed for on-site sale and consumption of alcohol, were within 1 of the 22 local authorities (LAs) in Wales and had previously experienced violence. Data were analysed using Andersen–Gill recurrent event models in an intention-to-treat analysis. An embedded process evaluation examined intervention implementation, reach, fidelity, dose and receipt. An economic evaluation compared costs of the intervention with benefits.InterventionPremises were randomised to receive a violence-reduction intervention, Safety Management in Licensed Environments (SMILE), which was delivered by an environmental health practitioner (EHP; the agent). SMILE consisted of an initial risk audit to identify known risks of violence, a follow-up audit scheduled to enforce change for premises in which serious risks had been identified, structured advice from EHPs on how risks could be addressed in premises and online materials that provided educational videos and related material.ObjectiveTo develop intervention materials that are acceptable and consistent with EHPs’ statutory remit; to determine the effectiveness of the SMILE intervention in reducing violence; to determine reach, fidelity, dose and receipt of the intervention; and to consider intervention cost-effectiveness.OutcomeDifference in police-recorded violence between intervention and control premises over a 455-day follow-up period.RandomisationA minimum sample size of 274 licensed premises per arm was required, rounded up to 300 and randomly selected from the eligible population. Licensed premises were randomly assigned by computer to intervention and control arms in a 1 : 1 ratio. Optimal allocation was used, stratified by LA. Premises opening hours, volume of previous violence and LA EHP capacity were used to balance the randomisation. Premises were dropped from the study if they were closed at the time of audit.ResultsSMILE was delivered with high levels of reach and fidelity but similar levels of dose to all premises, regardless of risk level. Intervention premises (n = 208) showed an increase in police-recorded violence compared with control premises (n = 245), although results are underpowered. An initial risk audit was less effective than normal practice (hazard ratio = 1.34, 95% confidence interval 1.20 to 1.51) and not cost-effective. Almost all eligible intervention premises (98.6%) received the initial risk audit; nearly 40% of intervention practices should have received follow-up visits but fewer than 10% received one. The intervention was acceptable to EHPs and to some premises staff, but less so for smaller independent premises.ConclusionsSMILE was associated with an increase in police-recorded violence in intervention premises, compared with control premises. A lack of follow-up enforcement visits suggests implementation failure for what was seen as a key mechanism of action. There are also concerns as to the robustness of police data for targeting and assessing outcome effectiveness, while intervention premises may have received greater attention from statutory agencies and, therefore, the identification of more violence than control premises. Although SMILE had high reach and was feasible and acceptable to EHPs, it was found to be ineffective and associated with increased levels of violence, compared with normal practice and it requires additional work to promote the implementation of follow-up enforcement visits. Future work will aim to better understand the role of intervention dose on outcomes and seek more objective measures of violence for use in similar trials.Trial registrationCurrent Controlled Trials ISRCTN78924818.FundingThis project was funded by the NIHR Public Health Research programme and will be published in full inPublic Health Research; Vol. 3, No. 10. See the NIHR Journals Library website for further project information.