Sequential Analysis of Quality-of-Life Measurements Using Mixed Rasch Models

Abstract

Early stopping of clinical trials in the case of either beneficial or deleterious effect of a treatment on quality of life (QoL) is an important issue. QoL is usually evaluated using self-assessment questionnaires and responses to the items are usually combined into QoL scores assumed to be normally distributed. However, these QoL scores are rarely normally distributed and usually do not satisfy a number of basic measurement properties. An alternative is to use item response theory (IRT) models such as the Rasch model for binary items which takes into account the categorical nature of the items. In this framework, the probability of response of a patient on an item depends upon different kinds of parameters: the “ability level” of the person (which reflects his or her current QoL) and a set of parameters characterizing each item. Sequential analysis and mixed Rasch models were combined in the context of phase II or III comparative clinical trials. The statistical properties of the triangular test (TT) were compared using mixed Rasch models and the traditional method based on QoL scores by means of simulations. The type I error of the TT was correctly maintained for the methods based on QoL scores and the Rasch model assuming known item parameter values, but was higher than expected when item parameters were assumed to be unknown. The power of the TT was satisfactorily maintained when Rasch models were used but the test was underpowered with the QoL scores method. All methods allowed substantial reductions in average sample numbers as compared with fixed sample designs, especially the method based on Rasch models. The use of IRT models in sequential analysis of QoL endpoints seems to provide a more powerful method to detect therapeutic effects than the traditional QoL scores method and to allow for reaching a conclusion with fewer patients.