Surgery (Austin & Northern Health) - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemOptimising allocation of deceased donor liver grafts for transplantation( 2020)Liver transplantation is a life-saving procedure for many patients. However, the number of liver transplants that can be performed is limited by donor organ availability. As with all scarce health resources, the overall objective is to ensure fair allocation of donor grafts. An ideal liver allocation model would consider the key ethical considerations of equity and utility, as well as fulfilling other requirements such as efficiency and transparency. Liver allocation has evolved to meet the growing problem of graft scarcity. An evidence-based approach to liver allocation is widely accepted, with an emphasis on objective measures to predict patient mortality. Need-based allocation (according to Model for End-stage Liver Disease score) is commonly practiced worldwide. However, utility, the expected gain from liver transplantation, is also an important consideration in liver allocation. The concept of transplant benefit balances both equity (need) and utility in allocation decisions. Transplant benefit is measured by the difference between a patient’s expected survival with and without transplantation. In the literature, transplant benefit is utilised both as an outcome (i.e. a metric of the gain from liver transplantation for an individual or group of patients with similar characteristics), and as a basis for liver allocation models. Transplant benefit-based allocation meets many requirements of an ideal liver allocation model, although there are challenges in creating accurate survival prediction models. The development of transplant benefit models and expected outcomes of transplant benefit-based allocation have not been investigated in the Australian and New Zealand (ANZ) context. Transplant benefit allocation models are not universally relevant due to differences in transplant populations and context. Two nationally-developed transplant benefit allocation models from US and UK populations demonstrated modest prediction performance in an Australian transplant population. An ANZ transplant benefit allocation model, the ANZ transplant benefit score (TBS), was created from waitlist and transplant survival prediction models. The methods used to create the ANZ TBS address statistical challenges that arise from the characteristics of transplant data. The ANZ TBS demonstrated good prediction performance on internal validation. In order to determine whether transplant benefit allocation would improve survival outcomes in ANZ, it is necessary to simulate allocation to compare models. There was no existing simulation program available for use in the ANZ population. An ANZ Liver Simulated Allocation program (ALSA) was thus created to simulate liver allocation processes and outcomes in ANZ. Using ALSA, transplant benefit-based allocation was shown to have the best expected survival outcomes overall in a single ANZ transplant centre, over 5 years of simulated allocation. Allocation according to the ANZ TBS was predicted to result in a reduction in deaths and additional life-years when compared to need and utility-based allocation and actual outcomes over the same time period. Novel modelling methods were explored to attempt to improve the prediction performance of the ANZ TBS. Random survival forests and artificial neural networks failed to improve the performance of the ANZ TBS, and predicted worse overall survival outcomes from simulated allocation. This thesis concludes that a transplant benefit-based allocation model could potentially improve survival outcomes in ANZ, although further developments and refinements are required before considering adoption into practice. These results provide insights to help guide the development and evolution of evidence-based liver allocation policy in ANZ.