OBJECTIVES: A surveillance program for upper-limb work-related musculoskeletal disorders (UWMSD) based on assessment of health and risk factors was implemented between 1996 and 2000 in a large shoe factory with overall high levels for biomechanical exposure. The study aimed to identify workers with an increased risk of UWMSD incidence. METHODS: In 1996, 1997 and 2000, 166 workers filled out a questionnaire and underwent a standardized physical examination. Factors from the 1996 questionnaire (general, personal and occupational factors) associated with UWMSD incidence in 1997 were selected. The predictive role of these variables was studied with a logistic model, taking into account also gender and age. The performance of a risk score based on this model was studied in 2000, using the Wilcoxon test and ROC curves. RESULTS: In 1997, 28 incident cases of UWMSD were observed (N = 107, 26.2%). Work pace and prior history of UWMSD were the only factors significantly associated with UWMSD incidence in 1997 (respectively 33% versus 13%, P = 0.02 and 58% versus 22%, P = 0.01). Psychological distress (36% versus 21%, P = 0.10), physical fatigue (35% versus 22%, P = 0.14), repetitiveness (30% versus 18%, P = 0.17) and task precision (33% versus 21%, P = 0.16) were also included in the logistic model for 1997 UWMSD incidence. Controlling for these variables, prior history of UWMSD remained associated with incidence in 1997 (OR = 5.5, 95% CI = 1.4-21.8). In the period from 1997 to 2000, 24 incident cases were observed (N = 102, 23.5%). The risk score, based on variables from the 1997 model, was significantly higher for incident cases (median = 6 in incident cases versus 4.5 for healthy subjects, P = 0.02). ROC curves indicated that the highest agreement reached 67% for sensitivity and 59% for specificity. Among subjects who did not change their task (N = 71, 18 incident cases), performance reached 66% for specificity with the same sensitivity. CONCLUSION: These results suggest that surveillance programs of UWMSD at a company level are possible even with overall high levels for biomechanical exposure and should take into account occupational and personal factors, including prior history of UWMSD.