A 3-year prospective epidemiological study was conducted to investigate the relationship between musculoskeletal complaints (MS) and sensorineural complaints (SN) of the workers in the hand-wrist region. A group of 69 workers (G1) using vibrating tools in eight different working situations was compared to a group of 62 workers (G2) performing heavy work without vibration and 46 workers (G3) performing light work without vibration. Biomechanical constraints (force, postures, repetitiveness and movement velocities) were analysed for each working situation and the vibration exposure at the eight workplaces with the 69 workers. MS and SN data were collected using the nordic questionnaire, modified to collect information about the frequency intensity and duration of complaints. The prevalence of complaints at the start of the study was significantly greater for G1 (72.5%) than for G2 (56.5%), itself greater than for G3 (30.4%). The prevalence of SN was about 40% for G1 and 2.5 times smaller in the two other groups. During the two years follow-up, new cases of 'serious' MS and SN developed. The annual incidence was respectively 8.3 and 5.4% on average. The incidence of MS was slightly but not statistically significantly greater for G1, while the incidence of SN was statistically higher (P<0.01) for G1 (10.9%) than for the two other groups (4.1 and 2.1%). Forces and angular repetitiveness were the only biomechanical factors significantly greater for G1. The vibration exposure duration of the G1 workers varied, in average, from 10 to 70% of the work time and the weighted personal exposure amplitude (A(EPw)) varied from 0.5 to 25.4 ms(-2). The probabilities of complaints at the beginning of the study (cross-sectional study) were estimated using multiple logistic regression models. The prevalence odds ratio (POR) for MS was equal to 4 for G1 compared to G2 and equal to 9 compared to G3. Force and vibration exposure were the main constraint parameters associated with this likelihood. As far as the SN are concerned, G2 and G3 were not statistically different, but the POR for the G1 workers was 4.5 compared to both groups. The most significant constraint factor was the weighted personal exposure acceleration. The same procedure was used to estimate the likelihood of development of 'serious' complaints (longitudinal study). The three groups did not appear significantly different concerning the 'serious' MS, while the incidence odds ratio (IOR) of 'serious' SN was very high (28.5) and significantly greater for G1 than for the two other groups. The likelihood of development of 'serious' SN increased with A(EPw). According to this prediction model, the risk of 'serious' SN would be about 6% at the proposed European 'action' value (2.5 ms(-2)) and about 10% at the 'limit' value (5 ms(-2)).