Time-dependent prediction of mortality and cytomegalovirus reactivation after allogeneic hematopoietic cell transplantation using machine learning

Abstract

Allogeneic hematopoietic cell transplantation (HCT) treats high-risk hematologic diseases effectively but can entail HCT-specific complications, which may be minimized by appropriate patient management and accurate, individual risk estimation. Existing clinical scores typically provide a single risk assessment before HCT and do not incorporate additional data as it becomes available. We developed machine learning models which integrate both baseline patient data and time-dependent laboratory measurements to individually predict mortality and cytomegalovirus (CMV) reactivation after HCT at multiple time points per patient. These models provide well-calibrated time-dependent risk predictions and achieved areas under the receiver-operating characteristic of 0.92 and 0.83 and areas under the precision-recall curve of 0.58 and 0.62 for prediction of mortality and CMV reactivation, respectively, in a 21-day time window. Both were successfully validated in a non-interventional, prospective study and performed on par with expert hematologists in a pilot comparison.

Implication of machine learning for relapse prediction after allogeneic stem cell transplantation in adults with Ph-positive acute lymphoblastic leukemia

“…In some of them the AUC reaches acceptable for practical application levels: in the study Pan L. et al the developed model demonstrated high AUC level of 0.904 [34]. Interestingly, the number of studies based on posttransplant variables is even more limited: in one of them by Eisenberg L. et al machine models provided wellcalibrated, time-dependent risk predictions and achieved appropriate level of AUC of 0.92 and 0.83 for prediction of mortality and CMV reactivation, respectively, in a 21-day time window after allo-HSCT [35]. However, to the best of our knowledge, there are no reliable machine learning approaches to predict relapse in adult Ph-positive ALL patients after allo-HSCT.…”

A pilot study of implication of machine learning for relapse prediction after allogeneic stem cell transplantation in adults with Ph-positive acute lymphoblastic leukemia

“…In some of them, the AUC reaches acceptable for practical application levels: in the study by Pan et al, the model demonstrated a high AUC level of 0.904 34 . Interestingly, the number of studies based on posttransplant variables is even more limited: in one of them by Eisenberg et al, machine models provided well-calibrated, time-dependent risk predictions and achieved appropriate levels of AUC of 0.92 and 0.83 for prediction of mortality and CMV reactivation, respectively, in a 21-day time window after allo-HSCT 35 . However, to our knowledge, there are no reliable machine-learning approaches to predict relapse in adult Ph-positive ALL patients after allo-HSCT.…”