An important aspect of the drug/device evaluation process is to have an integrated benefit-risk assessment to determine, using some quantitative measures, whether the benefit outweighs the risk for the target population. The subject-level benefit-risk response is a five-category random variable with cell counts following a multinomial distribution. Assuming that the cell probabilities follow a Dirichlet distribution, we develop a Bayesian approach for the longitudinal assessment of benefit-risk using the global measures proposed by Chuang-Stein et al. In a more generalized approach, a power prior is used through the likelihood function to discount the information from previous visits. For the subject-level benefit-risk assessment, the cell-probability of the subject, with respect to a reference category, is modeled, on the logarithmic scale, as a generalized linear model using a Dirichlet process as a prior. The model is applied to drug/device clinical trial datasets.