<p>The response variable in clinical studies is frequently ordinal. In the comparison of treatment efficacies, the latent Weibull model has been shown to be a powerful tool that is appropriate for testing even when the data are highly skewed, a condition under which the traditional proportional odds model may be unsuitable owing to the possible inflation of the type-I error rate. A response-adaptive treatment randomization scheme based on the latent Weibull model has recently been proposed to allow more patients to receive the better treatment while maintaining competitive test power. Nevertheless, the proposed procedure has been restricted to the comparison of two treatments. This paper generalizes the aforementioned idea to include trials with multiple treatments. Statistical procedures for pairwise comparisons and multiple comparisons with a control are discussed. Furthermore, a simulation study is performed to illustrate the superiority of the proposed method. The practical benefits of the suggested procedures are demonstrated through the redesign of two clinical trials.</p>