An early-biomarker algorithm predicts lethal graft-versus-host disease and survival
Matthew J. Hartwell, Umut Özbek, Ernst Holler, Anne S. Renteria, Hannah Major-Monfried, Pavan Reddy, Mina Aziz, William J. Hogan, Francis Ayuk, Yvonne A. Efebera, Elizabeth O. Hexner, Udomsak Bunworasate, Muna Qayed, Rainer Ordemann, Matthias Wölfl, Stephan Mielke, Attaphol Pawarode, Yi-Bin Chen, Steven Devine, Andrew C. Harris, Madan Jagasia, Carrie L. Kitko, Mark R. Litzow, Nicolaus Kröger, Franco Locatelli, George Morales, Ryotaro Nakamura, Ran Reshef, Wolf Rösler, Daniela Weber, Kitsada Wudhikarn, Gregory A. Yanik, John E. Levine, James L.M. Ferrara
Matthew J. Hartwell, Umut Özbek, Ernst Holler, Anne S. Renteria, Hannah Major-Monfried, Pavan Reddy, Mina Aziz, William J. Hogan, Francis Ayuk, Yvonne A. Efebera, Elizabeth O. Hexner, Udomsak Bunworasate, Muna Qayed, Rainer Ordemann, Matthias Wölfl, Stephan Mielke, Attaphol Pawarode, Yi-Bin Chen, Steven Devine, Andrew C. Harris, Madan Jagasia, Carrie L. Kitko, Mark R. Litzow, Nicolaus Kröger, Franco Locatelli, George Morales, Ryotaro Nakamura, Ran Reshef, Wolf Rösler, Daniela Weber, Kitsada Wudhikarn, Gregory A. Yanik, John E. Levine, James L.M. Ferrara
View: Text | PDF | Corrigendum
Clinical Research and Public Health Oncology Transplantation

An early-biomarker algorithm predicts lethal graft-versus-host disease and survival

Abstract

BACKGROUND. No laboratory test can predict the risk of nonrelapse mortality (NRM) or severe graft-versus-host disease (GVHD) after hematopoietic cellular transplantation (HCT) prior to the onset of GVHD symptoms. METHODS. Patient blood samples on day 7 after HCT were obtained from a multicenter set of 1,287 patients, and 620 samples were assigned to a training set. We measured the concentrations of 4 GVHD biomarkers (ST2, REG3α, TNFR1, and IL-2Rα) and used them to model 6-month NRM using rigorous cross-validation strategies to identify the best algorithm that defined 2 distinct risk groups. We then applied the final algorithm in an independent test set (n = 309) and validation set (n = 358). RESULTS. A 2-biomarker model using ST2 and REG3α concentrations identified patients with a cumulative incidence of 6-month NRM of 28% in the high-risk group and 7% in the low-risk group (P < 0.001). The algorithm performed equally well in the test set (33% vs. 7%, P < 0.001) and the multicenter validation set (26% vs. 10%, P < 0.001). Sixteen percent, 17%, and 20% of patients were at high risk in the training, test, and validation sets, respectively. GVHD-related mortality was greater in high-risk patients (18% vs. 4%, P < 0.001), as was severe gastrointestinal GVHD (17% vs. 8%, P < 0.001). The same algorithm can be successfully adapted to define 3 distinct risk groups at GVHD onset. CONCLUSION. A biomarker algorithm based on a blood sample taken 7 days after HCT can consistently identify a group of patients at high risk for lethal GVHD and NRM. FUNDING. The National Cancer Institute, American Cancer Society, and the Doris Duke Charitable Foundation.

Authors

Matthew J. Hartwell, Umut Özbek, Ernst Holler, Anne S. Renteria, Hannah Major-Monfried, Pavan Reddy, Mina Aziz, William J. Hogan, Francis Ayuk, Yvonne A. Efebera, Elizabeth O. Hexner, Udomsak Bunworasate, Muna Qayed, Rainer Ordemann, Matthias Wölfl, Stephan Mielke, Attaphol Pawarode, Yi-Bin Chen, Steven Devine, Andrew C. Harris, Madan Jagasia, Carrie L. Kitko, Mark R. Litzow, Nicolaus Kröger, Franco Locatelli, George Morales, Ryotaro Nakamura, Ran Reshef, Wolf Rösler, Daniela Weber, Kitsada Wudhikarn, Gregory A. Yanik, John E. Levine, James L.M. Ferrara

×

Figure 4

Study scheme of algorithm development and validation.

Options: View larger image (or click on image) Download as PowerPoint
Study scheme of algorithm development and validation. Clinical data and ...
Clinical data and plasma samples from day 7 after hematopoietic cellular transplantation were available from 1,287 patients transplanted at 11 MAGIC centers. Patient samples from the 2 largest centers, the University of Michigan and the University of Regensburg, were randomly assigned to the training and test sets in a 2:1 proportion. The remaining 358 patients were assigned to the independent multicenter validation set. The training set alone (n = 620) was used to develop the algorithm. All possible combinations of 1 to 4 biomarkers were used to model 6-month nonrelapse mortality (NRM) by competing-risks regression. Rigorous comparison of models through a Monte Carlo cross validation of 75 different, randomly created training sets confirmed that the models using ST2 and REG3α were superior to all other biomarker combinations. We used this model to predict the probability of 6-month NRM in the patients from the training set, rank ordered them from lowest to highest, and chose a threshold to separate risk groups for the final algorithm (see Methods). We then applied the algorithm to the test set in a first validation and to the multicenter validation set in a second validation.