SEGMENTED DOSE-RESPONSE MODELS FOR REPEATED MEASURES DATA

Abstract

In a dose-response analysis, logit-transformed responses are modelled as a function of log-transformed doses. The linear trend is commonly observed. The comparison among treatment groups can be made based on the linear trend. An example in this paper came from a study to estimate the effect of aminophylline on dose-response curve of atracurium. Unlike the usual dose-response curve, this example has repeated measures and seems to have two slopes to which the usual dose-response model is not adequate to fit. We propose segmented regression models that allow two different slopes. The proposed model is an extension of the segmented regression model with a univariate response per subject. We illustrate the proposed model fits data better than the usual dose-response model.

Keywords:

• Dose-response curve
• Interaction
• Repeated measures
• Segmented model

ACKNOWLEDGMENT

This work was supported by Korea Research Foundation Grant through the Research Institute of Basic Sciences, Seoul National University (KRF-2000-015-DP0059) and by Brain Korea 21 Project.