ARE LIVING-DONOR LIVER TRANSPLANTS TIMED TO MAXIMIZE LIFE EXPECTANCY?

Purpose. Living-donor liver transplantation (LDLT) is a treatment for end-stage liver disease (ESLD) in which the major clinical decision is the appropriate time to transplant. The purpose of this study is to empirically compare the timing decision suggested by a Markov decision process (MDP) model, which maximizes patient’s total life expectancy, to the actual transplant decision of the patient.
Methods. The optimal timing of the LDLT problem is modeled as a nested MDP composed of a series of finite-horizon MDPs (to represent the different natural history surrounding diagnosis and listing) followed by an infinite-horizon optimal stopping problem (to represent chronic progression). The state of the process is described by the current patient health as measured by the MELD score (the sicker the patient the higher her MELD score). The quality of the living-donor liver is assumed to be fixed and is described by a combination of characteristics of the donor and the recipient. Transition probabilities of the process are estimated using the natural history of liver disease. The objective is to maximize pre- and post-transplant life expectancy. The parameters of the model are estimated using data from UNOS and a large urban transplant hospital. The optimal decision rules are found using standard solution techniques for MDPs. These decisions are compared to the actual decisions made by 387 ESLD patients diagnosed with a cirrhotic disease (such as primary biliary cirrhosis, primary sclerosing cholangitis, alcoholic liver disease, and autoimmune disorders) from the UNOS database.
Results. The nested MDP model is tested for a variety of parameter values. Overall, about 85% of all patients are found to be transplanted at a MELD score lower than what is suggested by the model. As a result of these differences in patients’ decisions compared to their model suggested decisions, about 75% of all patients are predicted to have shorter life expectancy. The average life expectancy is shortened by about 1310 days with a standard deviation of 473 days.
Conclusions. The empirical analysis shows that a significant number of ESLD patients are transplanting earlier than would be expected to maximize overall life expectancy. Although it is possible that maximizing quality of life would eliminate some of this disparity, this work provides evidence that many patients with living-donors are transplanted too early.