Dataset Information

Machine learning reveals chronic graft-versus-host disease phenotypes and stratifies survival after stem cell transplant for hematologic malignancies.

ABSTRACT: The application of machine learning in medicine has been productive in multiple fields, but has not previously been applied to analyze the complexity of organ involvement by chronic graft-versus-host disease. Chronic graft-versus-host disease is classified by an overall composite score as mild, moderate or severe, which may overlook clinically relevant patterns in organ involvement. Here we applied a novel computational approach to chronic graft-versus-host disease with the goal of identifying phenotypic groups based on the subcomponents of the National Institutes of Health Consensus Criteria. Computational analysis revealed seven distinct groups of patients with contrasting clinical risks. The high-risk group had an inferior overall survival compared to the low-risk group (hazard ratio 2.24; 95% confidence interval: 1.36-3.68), an effect that was independent of graft-versus-host disease severity as measured by the National Institutes of Health criteria. To test clinical applicability, knowledge was translated into a simplified clinical prognostic decision tree. Groups identified by the decision tree also stratified outcomes and closely matched those from the original analysis. Patients in the high- and intermediate-risk decision-tree groups had significantly shorter overall survival than those in the low-risk group (hazard ratio 2.79; 95% confidence interval: 1.58-4.91 and hazard ratio 1.78; 95% confidence interval: 1.06-3.01, respectively). Machine learning and other computational analyses may better reveal biomarkers and stratify risk than the current approach based on cumulative severity. This approach could now be explored in other disease models with complex clinical phenotypes. External validation must be completed prior to clinical application. Ultimately, this approach has the potential to reveal distinct pathophysiological mechanisms that may underlie clusters. Clinicaltrials.gov identifier: NCT00637689.

SUBMITTER: Gandelman JS

PROVIDER: S-EPMC6312024 | biostudies-literature | 2019 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine learning reveals chronic graft-versus-host disease phenotypes and stratifies survival after stem cell transplant for hematologic malignancies.

Gandelman Jocelyn S JS Byrne Michael T MT Mistry Akshitkumar M AM Polikowsky Hannah G HG Diggins Kirsten E KE Chen Heidi H Lee Stephanie J SJ Arora Mukta M Cutler Corey C Flowers Mary M Pidala Joseph J Irish Jonathan M JM Jagasia Madan H MH

Haematologica 20180920 1

The application of machine learning in medicine has been productive in multiple fields, but has not previously been applied to analyze the complexity of organ involvement by chronic graft-versus-host disease. Chronic graft-versus-host disease is classified by an overall composite score as mild, moderate or severe, which may overlook clinically relevant patterns in organ involvement. Here we applied a novel computational approach to chronic graft-versus-host disease with the ...[more]

PMID: 30237265

Similar Datasets

Project description:Severe graft-versus-host disease is a major barrier for non-T-cell-depleted haploidentical stem cell transplantation. There is no consensus on the optimal graft-versus-host disease prophylaxis. This study compared the two most commonly used graft-versus-host disease prophylaxis regimens (post-transplant cyclophosphamide-based vs. the anti-thymocyte globulin-based) in adults with acute myeloid leukemia reported to the European Society for Blood and Bone Marrow Transplantation. A total of 308 patients were analyzed; 193 received post-transplant cyclophosphamide-based regimen and 115 anti-thymocyte globulin-based regimen as anti-graft-versus-host disease prophylaxis. The post-transplant cyclophosphamide-based regimen was more likely to be associated to bone marrow as graft source (60% vs. 40%; P=0.01). Patients in the post-transplant cyclophosphamide-based regimen group had significantly less grade 3-4 acute graft-versus-host disease than those in the anti-thymocyte globulin-based group (5% vs. 12%, respectively; P=0.01), comparable to chronic graft-versus-host disease. Multivariate analysis showed that non-relapse mortality was lower in the post-transplant cyclophosphamide-based regimen group [22% vs. 30%, Hazard ratio (HR) 1.77(95%CI: 1.09-2.86); P=0.02] with no difference in relapse incidence. Patients receiving post-transplant cyclophosphamide-based regimen had better graft-versus-host disease-free, relapse-free survival [HR 1.45 (95%CI: 1.04-2.02); P=0.03] and leukemia-free survival [HR 1.48 (95%CI: 1.03-2.12); P=0.03] than those in the anti-thymocyte globulin-based group. In the multivariate analysis, there was also a trend for a higher overall survival [HR 1.43 (95%CI: 0.98-2.09); P=0.06] for post-transplant cyclophosphamide-based regimen versus the anti-thymocyte globulin-based group. Notably, center experience was also associated with non-relapse mortality and graft-versus-host disease-free, relapse-free survival. Haplo-SCT using a post-transplant cyclophosphamide-based regimen can achieve better leukemia-free survival and graft-versus-host disease-free, relapse-free survival, lower incidence of graft-versus-host disease and non-relapse mortality as compared to anti-thymocyte globulin-based graft-versus-host disease prophylaxis in patients with acute myeloid leukemia.

Project description:BackgroundAdvances in allogeneic hematopoietic stem cell transplant (HSCT) have increased patient survival, although substantial treatment-related toxicity remains, including chronic kidney disease (CKD). We assessed the association between CKD and survival and transplant-specific outcomes in HSCT recipients.MethodsWe conducted a retrospective study of all 408 adult patients with allogenic HSCT at Princess Margaret Cancer Centre (Toronto, Canada, 2015-18). We used logistic regression to identify risk factors for CKD at 1 year post-transplant. Associations between CKD at 1 year and overall survival, relapse-free survival, graft-versus-host-disease (GVHD)-free/relapse-free survival, relapse and transplant-related mortality were examined using extended time-varying Cox models. In a sensitivity analysis, we restricted the cohort to survivors at 1 year, using standard Cox proportional hazard models to examine associations between CKD and overall survival, relapse-free survival and GVHD-free/relapse-free survival, and Fine and Gray's competing risk models to determine associations between CKD and relapse/transplant-related mortality.ResultsThe prevalence of CKD at 1 year was 19% (46 patients) with median follow-up of 23 months. Multivariable regression identified age at transplant [adjusted OR (aOR) 1.09, 95% confidence interval (95% CI) = 1.05-1.14; P < 0.0001), female gender (aOR 2.83, 95% CI = 1.34-5.97; P = 0.006) and acute kidney injury during the first 100 days (aOR 3.86, 95% CI = 1.70-8.73; P = 0.001) as risk factors for CKD at 1 year. Patients with CKD at 1 year had significantly poorer overall survival than those without CKD, when adjusted for relevant covariates [adjusted HR (aHR) 1.93, 95% CI = 1.02-3.66; P = 0.04 in the time-varying Cox model, and aHR 2.06, 95% CI = 1.04-4.07; P = 0.04 using the standard Cox model]. CKD at 1 year was also associated with worse GVHD-free/relapse-free survival (aHR 1.65, 95% CI = 1.04-2.61; P = 0.03).ConclusionsCKD adversely affects the long-term prognosis for allogeneic HSCT recipients, with increased mortality risk and worse GVHD-free/relapse-free survival.

Dataset Information

Machine learning reveals chronic graft-versus-host disease phenotypes and stratifies survival after stem cell transplant for hematologic malignancies.

Publications

Machine learning reveals chronic graft-<i>versus</i>-host disease phenotypes and stratifies survival after stem cell transplant for hematologic malignancies.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets