A multi-treatment two-stage adaptive design is developed for phase III clinical trials with ordinal categorical responses. Starting with an equal randomization phase, asymptotic p-values for testing equality of treatment effects in an interim analysis are used to derive allocation probabilities for second-stage patients. Performance of the proposed design is studied both theoretically and empirically and compared with those of the suitable designs. The practical applicability of the proposed design is further envisaged through redesigning of a real clinical trial. The results of the thorough statistical study show that the proposed design is able to assign most of the patients to the best treatment while maintaining the other ethical concerns like preserving substantial power of a statistical test and maintaining type-I error rates.
AMS Subject Classification: 62P10
