Objective: To examine whether factors common to obesity prevention programmes in adolescence, namely exercise, sedentary activities, eating behaviours and psychological factors, predict change in BMI in 'free-living' adolescents followed into adulthood. Design: Longitudinal national birth cohort study. Subjects: 1970 British Birth Cohort: 4461 subjects with data on BMI at 16 years (1986) and 30 years (2000). Measurements and analysis: BMI z-score (zBMI) at 16 years (measured) and 30 years (self-reported). Obesity defined as BMI exceeding 95th British centile and 28.5 kg/m(2), respectively. Self-report data on exercise, eating behaviours, dieting and measures of psychological function (depression, psychological distress, self-esteem) at 16 years. Models were produced for the regression of zBMI at 30 years on each variable at 16 years, adjusted for socioeconomic status, sex and zBMI at 16 years. Results: In all, 467 (8.2%) were obese at 16 years and 730 (16.4%) were obese at 30 years. Of those obese at 16 years, 60.7% were also obese at 30 years. Loss of zBMI between 16 and 30 years was predicted by female sex (P 0.01), higher social class (P < 0.0001) and higher frequency of playing sport, although this was of borderline significance (P = 0.05). Increase in zBMI between 16 and 30 years was predicted by 4 or more hours per day of sedentary activities (P = 0.01), eating takeaway meals twice or more per week (P = 0.009), consuming two or more carbonated drinks per day (P = 0.04) and a history of dieting to lose weight (P 0.04). Conclusion: These data from 'free-living' adolescents followed into adult life support the importance of known risk factors such as reduction of sedentary behaviours and reduction of unhealthy eating patterns (consumption of carbonated drinks and takeaway foods, dieting) in the prevention of the persistence of obesity from childhood into adult life