Author : Elizabeth Krakow
Publisher :
ISBN 13 :
Total Pages : pages
Book Rating : 4.:/5 (948 download)
Book Synopsis Can We Design Personalized Acute Graft-vs-host Disease Prevention and Treatment Strategies Using Registry Data and Sequential Multiple-assignment Randomized Trials to Improve Disease-free Survival of Blood & Marrow Transplant Patients? by : Elizabeth Krakow
Download or read book Can We Design Personalized Acute Graft-vs-host Disease Prevention and Treatment Strategies Using Registry Data and Sequential Multiple-assignment Randomized Trials to Improve Disease-free Survival of Blood & Marrow Transplant Patients? written by Elizabeth Krakow and published by . This book was released on 2015 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: "Objective: Graft-versus-host disease (GVHD) is a major complication of allogeneic hematopoietic cell transplantation (AHCT). We investigated whether the Center for International Blood and Marrow Transplant Research registry could be useful for (1) determining if the sequence of administering non-specific, highly T-lymphodepleting (NHTL) therapeutics in GVHD prophylaxis and in refractory GVHD impacts survival and (2) identifying donor- and patient-related factors that may guide individualized selection of classes of immunosuppressant agents over time, i.e. to develop adaptive treatment strategies (ATSs).Method: We employed a backwards induction method derived from reinforcement learning in a large cohort of patients who underwent AHCT for acute myeloid leukemia and myelodysplasia between 1995 and 2007. We devised logistic Q-models that first estimate the optimal treatment for each patient with refractory acute GVHD and then use a pseudo-outcome approach to estimate the optimal patient-specific GVHD prophylaxis, with the goal of maximizing 2-year disease-free survival (DFS).Results: In unadjusted analysis, NHTL prophylaxis and NHTL treatment of refractory acute GVHD were associated with inferior DFS compared to non-NHTL therapeutics. Yet, among the 9563 patients, the Q-model predicted that 4762 (50%) would have a higher probability of 2-year DFS with NHTL prophylaxis. For the 1411 patients with refractory acute GVHD, the Q-model predicted that 492 (35%) would have had a higher probability of 2-year DFS with NHTL salvage therapy. The magnitude of projected patient-specific benefit from choosing the optimal class of agent was modest. The models suggested that patient-specific combinations of characteristics could influence the choice of GVHD prophylaxis and treatment.Conclusions: Retrospective analysis with Q-learning can be used to propose personalized ATSs for GVHD prevention and treatment, which may then be tested in sequentially-randomized clinical trials. An important limitation which threatens the validity of the registry-derived strategies is that lack of detailed information about the indication for each immunosuppressant and other salient patient characteristics may lead to residual confounding." --