Health promotion interventions cannot work if people do not engage with them. The aim of this study was to examine whether disengagement from an adolescent smoking prevention and cessation intervention was an independent risk factor for regular smoking 1 and 2 years later. The data were taken from a cluster randomised controlled trial, in the West Midlands, UK, based on the transtheoretical or stages of change model. In this trial, 8,352 13-14-year old school pupils enrolled, and the data in this report were based on the 7,413 and 6,782 pupils present at 1 and 2 years follow-ups, respectively. The intervention group undertook three sessions using an interactive computer programme. At the end of the programme, pupils recorded their responses to it. Pupils were classed as engaged if they thought the intervention was both useful and interesting; all others were classed as disengaged. Random effects logistic regression related the number of times engaged to regular smoking at 1 and 2 years follow-up, adjusted for school absences and 11 potential confounders. The majority of pupils were engaged by the intervention. For participants using the intervention three times but not engaging once, the odds ratios (95% confidence intervals) for smoking at 1 and 2 years relative to the controls were 1.83 (1.41-2.39) and 1.70 (1.38-2.11). For those engaging three times, they were 0.79 (0.60-1.03) and 0.96 (0.75-1.21). There was no interaction with baseline intention to smoke, classified by stage of change, but there was a borderline significant interaction with baseline smoking status, with disengagement acting as a stronger risk factor among baseline never-smokers. We conclude that disengagement from interventions is a risk factor for smoking independently of experimentation with cigarettes. The best explanation is that disengagement from school, an established risk factor for smoking, generalises to disengagement from didactic school-based health promotion programmes.