The purpose of the present study was to investigate the long-term effect of the Active Back School programme on minimizing recurrences of episodes of low back pain. Forty-three subjects were randomly allocated to the Active Back School group and 38 to the control group. There were no significant differences between the groups with regard to baseline characteristics. The Active Back School programme comprised 20 lessons each divided into a 20-min theoretical and a 40-min exercise part during a 13-week period. Nine participants (11%) dropped out during the study period. Recurrence of new low back pain episodes was significantly less (p = 0.04), and the time from inclusion to the first new low back pain episode was significantly on the side of the Active Back School group (p < 0.01). The duration of sick leave was found to be significantly shorter (p < 0.01) in the Active Back School group compared to the control group. The Active Back School reduced the recurrence and severity of new low back pain episodes at 36 months' follow-up.