Ranking patients on the kidney transplant waiting list based on fuzzy inference system

Background Kidney transplantation is the best treatment for people with End-Stage Renal Disease (ESRD). Kidney allocation is the most important challenge in kidney transplantation process. In this study, a Fuzzy Inference System (FIS) was developed to rank the patients based on kidney allocation factors. The main objective was to develop an expert system, which would mimic the expert intuitive thinking and decision-making process in the face of the complexity of kidney allocation. Methods In the first stage, kidney allocation factors were identified. Next, Intuitionistic Fuzzy Analytic Hierarchy Process (IF-AHP) has been used to weigh them. The purpose of this stage is to develop a point scoring system for kidney allocation. Fuzzy if-then rules were extracted from the United Network for Organ Sharing (UNOS) dataset by constructing the decision tree, in the second stage. Then, a Multi-Input Single-Output (MISO) Mamdani fuzzy inference system was developed for ranking the patients on the waiting list. Results To evaluate the performance of the developed Fuzzy Inference System for Kidney Allocation (FISKA), it was compared with a point scoring system and a filtering system as two common approaches for kidney allocation. The results indicated that FISKA is more acceptable to the experts than the mentioned common methods. Conclusion Given the scarcity of donated kidneys and the importance of optimal use of existing kidneys, FISKA can be very useful for improving kidney allocation systems. Countries that decide to change or improve the kidney allocation system can simply use the proposed model. Furthermore, this model is applicable to other organs, including lung, liver, and heart.

Computational Eurotransplant kidney allocation simulations demonstrate the feasibility and benefit of T-cell epitope matching

The EuroTransplant Kidney Allocation System (ETKAS) aims at allocating organs to patients on the waiting list fairly whilst optimizing HLA match grades. ETKAS currently considers the number of HLA-A, -B, -DR mismatches. Evidently, epitope matching is biologically and clinically more relevant. We here executed ETKAS-based computer simulations to evaluate the impact of epitope matching on allocation and compared the strategies. A virtual population of 400,000 individuals was generated using the National Marrow Donor Program (NMDP) haplotype frequency dataset of 2011. Using this population, a waiting list of 10,400 patients was constructed and maintained during simulation, matching the 2015 Eurotransplant Annual Report characteristics. Unacceptable antigens were assigned randomly relative to their frequency using HLAMatchmaker. Over 22,600 kidneys were allocated in 10 years in triplicate using Markov Chain Monte Carlo simulations on 32-CPU-core cloud-computing instances. T-cell epitopes were calculated using the www.pirche.com portal. Waiting list effects were evaluated against ETKAS for five epitope matching scenarios. Baseline simulations of ETKAS slightly overestimated reported average HLA match grades. The best balanced scenario maintained prioritisation of HLA A-B-DR fully matched donors while replacing the HLA match grade by PIRCHE-II score and exchanging the HLA mismatch probability (MMP) by epitope MMP. This setup showed no considerable impact on kidney exchange rates and waiting time. PIRCHE-II scores improved, whereas the average HLA match grade diminishes slightly, yet leading to an improved estimated graft survival. We conclude that epitope-based matching in deceased donor kidney allocation is feasible while maintaining equal balances on the waiting list.