Multi-state models for the Analysis of Quality of Life and Clinical Data from Psoriatic Arthritis Patients (Handling time-dependent explanatory variables and the choice of random effects distributions)

Speaker: Vern Farewell, Professor, Medical Research Council Biostatistics Unit, Cambridge.

Speaker

Vern Farewell, Professor, Medical Research Council Biostatistics Unit, Cambridge.

Abstract

There is considerable interest in the relationship between a patient’s quality of life (QoL) and clinical manifestations of disease. A particular question with respect to psoriatic arthritis patients is the link between QoL and active (painful and/or swollen) joints. Both QoL and active joint status are observed intermittently and this is a particular problem for active joint status which fluctuates rapidly over time. This relationship will be investigated by expanding a multi-state model for QoL to handle such an intermittently observed explanatory variable.

A long-standing question concerning psoriatic arthritis is whether there exists a sub-population of patients who, while having manifestations of active joint disease, will not progress to the development of permanently damaged joints. In the multi-state model context, so called mover-stayer models are often used to investigate this but inference from such models can be problematic. Recently the mixture aspect of these models has been incorporated in patient level random effects distributions. It will be shown that the choice of parametric form for these distributions can have a major effect on the apparent conclusions. The evidence for separate populations of psoriatic arthritis patients will be assessed in light of the analyses presented.

 

Published Aug. 13, 2012 3:30 PM - Last modified Sep. 6, 2012 4:52 PM