The purpose of this project is to develop and refine new quantitative techniques that will enhance our ability to predict the optimal timing of liver transplantation in end stage liver disease (ESLD). We intend to accomplish this goal by conducting methodologic research in the application of Monte Carlo methods to the evaluation of complex Markov process-based simulation models. The clinical issue being studied is an important one. Substantial resources are consumed each year in the treatment of ESLD, which ranks seventh as a cause of disease-related death, and produces substantial morbidity and decreases quality of life when patients develop ascites, encephalopathy or bleeding varices. Transplantation is an effective but expensive therapy for ESLD, which the NIH consensus Conference on Liver Transplantation recognized as providing """"""""prolongation of life of good quality"""""""" for patients with several types of liver disease. More recently, the AHCPR evaluation of liver transplantation recommends transplantation for ESLD from primary biliary cirrhosis or sclerosing cholangitis. However, the timing of liver transplantation is an important clinical question that has received little attention in the literature. Transplanting too late increases perioperative risks and decreases survival; transplanting too early exposes many patients who may have a benign course to unnecessary risks. Both timing errors waste a scarce resource. The timing decision provides fertile ground for methodologic research in the generic problem of technologic intervention in chronic disease. For ethical and pragmatic reasons it is unlikely that a randomized controlled trial of the optimal timing of transplantation would ever take place. Therefore, observational and decision analytic techniques provide appropriate tools for examining this problem. However, Markov processes, the most common decision analytic technique used to model diseases for which events are spread out in time, are unable to effectively model complex natural histories. Technically, this is because transition probabilities in standard processes may not be path dependent. This necessitates the creation of models of unmanageable size, often involving thousands of separate states. To solve the path dependence problem, this work will continue research on Monte Carlo evaluation techniques that incorporate """"""""memory"""""""" into decision analysis models based on Markov processes. It will construct the quantitative descriptions of the natural history of ESLD necessary to calibrate such models, and it will evaluate these models in the context of the problem of the optimal timing of transplantation. Finally, because decision analysis models easily incorporate costs, this work will investigate the economic impact of various transplantation selection and timing criteria. Survival and costs are directly linked, and the timing decision impacts both survival and resources spent on ESLD.
