Modelling the liver function data in diabetes.

Abstract

We have enrolled diabetic patients into a long-term follow-up study that entails visits to the eye clinic every six months at which time use of a classification scheme determines whether or not the patient is normal regarding liver function. Since the study is still in progress, not all patients have complete follow-up data. We fit models of complete symmetry to the multi-visit binary data with supplemental margins and also investigate the data as stationary Markov chains with testing for stationarity and determination of the order of chain. The final model, a synthesis of both approaches, has potential use in future clinical trials of new drugs for diabetic patients.