The evaluation of low-back disorder risk associated with materials handling tasks can be performed using a variety of assessment tools. Most of these tools vary greatly in their underlying logic, yet few have been assessed for their predictive ability. It is important to document how well an assessment tool realistically reflects the job's injury risk, since only valid and accurate tools can reliably determine whether a given ergonomic intervention will result in a future reduction in back injuries. The goal of this study was to evaluate how well a previously reported low-back disorder (LBD) risk assessment model (Marras et al. 1993) could predict changes in LBD injury rates as the physical conditions to which employees are exposed were changed. Thirty-six repetitive materials handling jobs from 16 different companies were included in this prospective cohort study. Of these 36 jobs, 32 underwent an ergonomic intervention during the observation period, and four jobs in which no intervention occurred served as a comparison group. The trunk motions and workplace features of 142 employees performing these jobs were observed both before and after workplace interventions were incorporated. In addition, the jobs' LBD rates were documented for these pre- and post-intervention periods. The results indicated that a statistically significant correlation existed between changes in the jobs' estimated LBD risk values and changes in their actual low-back incidence rates over the observation period. Linear and Poisson regression models also were developed to predict a change in a job's incidence rate and the number of LBD on ajob respectively, as a function of the job's risk change using this assessment model. Finally, this prospective study showed which ergonomic interventions consistently reduced the jobs' mean low-back incidence rates. These results support use of the LBD risk model to assess accurately a job's potential to lead to low-back injuries among its employees.