A model based analysis for two period, two treatment change-over clinical trials

Abstract

This dissertation is concerned with developing a method of analysis for data arising from a series of clinical trials which involve the testing of a new drug and a standard drug in patients with Rheumatoid Arthritis. Data on response variables are collected at biweekly intervals from two consecutive sex week treatment periods. The trials are change-over trials in that half the patient population receives one drug during treatment period one and the other half receives the second drug. At the end of the treatment period the first group (denoted sequence group 1) receives the second drug and the second group (denoted sequence group 2) receives the first drug. The category of change-over designs is summarized. The current analysis for data such as described here is to use the method proposed by Grizzle [1965]. The observations at week six of each treatment period are usually the only data analyzed. Thus the information collected at two and four weeks is wasted. A non-linear model, E[y] = C + Ae [superscript kt], arising as the solution of a 'rate equation' is fitted to the data from the first treatment period for each sequence group separately in a trial in order to use all the data available during that period. The Modified Gauss-Newton procedure proposed by Lawton and Sylvestre [1971] are both used to fit the data to the model and the results are compared.