OBJECTIVES: To obtain longitudinal information of the relation between smoking and musculoskeletal disorders in an industrial setting. METHODS: The associations of lifetime tobacco exposure (pack-years), current smoking status, and stopping smoking with back and limb disorders were studied in a cohort of white collar and blue collar employees in a metal industry. Measurements were made three times at 5 year intervals. Two thirds of an initial sample of 902 took part in both re-examinations. Musculoskeletal morbidity was measured as the abundance of symptoms during the past year and as clinical findings assessed by a physiotherapist (upper decile score/score difference = index category). Logistic regression and a generalised estimating equation were used, allowing for sociodemographic variables, physical workload, body mass index, exercise activity, and mental distress. RESULTS: By comparison with never smokers, exposure of 10-< 20 pack-years, the odds ratio (OR) (95% confidence interval (95% CI) of the 10 year change in neck-shoulder symptoms was 3.1 (1.4 to 6.8), in low back symptoms 2.4 (1.1 to 5.1), in upper limb symptoms 1.9 (NS), and in lower limb symptoms 3.4 (1.5 to 7.8). The highest exposure category of > or = 20 pack-years was associated with the change in upper limb findings 2.9 (1.4 to 6.2) and lower limb findings 2.9 (1.2 to 7.2). Those who continued to smoke through the follow up period had a higher increase in clinical findings 2.5 (1.1 to 5.9) than never smokers. There was a dose-response in the association of smoking intensity with future musculoskeletal symptoms. Also, those who stopped smoking during the follow up had a higher increase in symptoms 4.4 (2.0 to 9.9) and findings 3.5 (1.4 to 8.8) than never smokers. CONCLUSION: Smoking seems to predict the development in the occurrence of musculoskeletal symptoms and signs. Stopping smoking is associated with high morbidity.