Methods for the Analysis of Event Processes Arising in Organ Transplantation

University Of Pennsylvania, Philadelphia PA

Investigators

Linked publications & trials

Abstract

PROJECT SUMMARY Importance: Currently, over 800,000 patients in the U.S. are on dialysis. In contrast, there were less than 28,000 kidney transplants carried out in 2023. Similar shortages have been observed for patients with end-stage liver disease (ESLD). Both ESKD and ESLD have substantial impacts on mortality, morbidity, health care cost and quality of life, and both have unmet healthcare needs. Problem: In order for patients, their health care providers and transplant surgeons to understand the relationships between risk factors and outcomes, methods must be developed which emit relevant and easily interpretable measures of a given risk predictor's importance. In time-to-event (survival) analysis, the area under the (survival) curve, also known as restricted mean survival time (RMST), is an increasingly popular method. An issue, however, is that the set of available methods for modeling RMST has not kept pace with what appears to be a substantially increasing demand. Overall Objective: To develop survival analysis methodology to support analyses that will produce a deeper understanding of morbidity and mortality patterns of ESKD and ESLD patients. This, in turn, should lead to improvements in therapy selection and, in turn, improved survival and quality of life. Target Audience: With respect to methodology, the target audience includes biostatisticians, particularly practitioners studying ESRD or ESLD or other chronic illnesses. Applications of the methods would be of interest to nephrologists, gastroenterologists transplant surgeons, and ESKD and ESLD patients. Products: Novel and innovative methods for the analysis of time-to-event data. Specific Aim 1: Flexible semiparametric direct modeling of restricted mean survival time (RMST) We will develop semiparametric methods for directly modeling the RMST that allow high-dimensional nuisance covariates to be handled non-parametrically. The methods will be applied to model post-kidney transplant graft survival, with center serving as the high-dimensional covariate. Specific Aim 2: Dynamic risk assessment via RMST regression We will develop methods for direct RMST modeling that accommodate time-varying predictors. We will apply such methods in order to model survival in the absence-of-liver-transplantation. Specific Aim 3: Effect of a time-dependent treatment on the RMST After establishing a pertinent causal inference framework, we will develop landmark methods to estimate the survival benefit of a time-varying treatment. The methods will be applied to estimate the survival benefit of liver transplantation as a function of pre-transplant death risk. For each Aim, the methods will be easily implementable since user-friendly software (SAS, R) will be developed and made available online.