South Korean researchers have developed technology to predict high-risk patients for a fatal liver complication that occurs during pediatric hematopoietic stem cell transplantation.
Seoul National University Hospital announced on the 13th that a joint research team led by Professors Hong Kyung-taek and Kang Hyoung-jin of the Department of Pediatrics, along with Professor Han Do-hyun of the Department of Transdisciplinary Medicine, identified key biomarkers for early detection of high-risk patients for hepatic veno-occlusive disease (VOD) among pediatric stem cell transplant candidates and developed an artificial intelligence prediction model based on these findings.
Hematopoietic stem cell transplantation is a treatment for patients with blood cancers such as leukemia and lymphoma, involving the removal of cancer cells and the patient's own hematopoietic stem cells followed by transplantation of new stem cells. The procedure requires high-intensity chemotherapy to eliminate diseased bone marrow. The problem is that highly toxic chemotherapy agents such as busulfan administered during this process can damage liver microvasculature and trigger veno-occlusive disease. The condition occurs in 15-30% of pediatric patients receiving stem cell transplants and can cause liver enlargement and deterioration of liver and kidney function, with mortality rates reaching up to 80% in severe cases.
The research team conducted detailed analysis of 720 blood proteins before and after chemotherapy in 51 pediatric patients who received busulfan-based high-intensity conditioning prior to haploidentical allogeneic stem cell transplantation. Of the 51 patients, 26 developed severe veno-occlusive disease while the remaining 25 did not.
Analysis revealed that patients who did not develop complications had higher levels of a liver detoxification enzyme (GCLC) even before chemotherapy—essentially equipped with sufficient "cleaning tools" to flush out the highly toxic chemotherapy agents. In contrast, patients who developed complications not only lacked this detoxification enzyme but also showed significantly lower expression of a specific protein (FBP1) responsible for the liver's "baseline resilience." This indicates they were vulnerable to toxic stimuli even before chemotherapy began.
Based on these findings, the research team identified 15 early biomarkers capable of distinguishing high-risk patients. When five key proteins with high predictive power (HRNR, FBP1, DCD, GCLC, LSAMP) were selected and applied to the team's proprietary machine learning model, high-risk patients could be identified with 92% accuracy.
"The newly identified protein patterns will serve as an important turning point in enabling effective prevention and safe transplant treatment for high-risk patients," said Professor Hong Kyung-taek.
The research findings were published in the latest online edition of Transplantation and Cellular Therapy, the official journal of the American Society for Transplantation and Cellular Therapy (ASTCT).