AI Predicts Life-Threatening Complications After Stem Cell Transplants

A powerful artificial intelligence (AI) tool may offer clinicians a significant advantage in identifying potentially life-threatening complications following stem cell and bone marrow transplants, according to new research from the MUSC Hollings Cancer Center in the United States. The findings, published in the ‘Journal of Clinical Investigation’, detail a method combining immune biomarkers, clinical data, and machine learning to predict risk in real-world scenarios.

Stem cell or bone marrow transplantation can be life-saving for many patients. However, recovery extends far beyond hospital discharge. Serious complications can emerge months later, often without warning. One of the most challenging of these is chronic graft-versus-host disease (GVHD), a condition where immune cells from the transplanted tissue attack the patient’s healthy tissues. This can affect multiple organs – skin, eyes, mouth, joints, and lungs – leading to long-term disability or even death.

Researchers, led by Sophie Paczesny, MD, PhD, co-leader of the Cancer Biology and Immunology Research Program at Hollings, along with Michael Martens, PhD, and Brent Logan, PhD, from the Center for International Blood and Marrow Transplant Research at the Medical College of Wisconsin, have developed an AI-based tool designed to identify patients at higher risk for chronic GVHD *before* symptoms appear, potentially enabling earlier monitoring, and intervention.

The team applied machine learning to immune-related proteins and validated clinical information, resulting in a tool called BIOPREVENT. This tool estimates a patient’s future risk of developing chronic GVHD and dying from transplant-related causes. According to Paczesny, the team made BIOPREVENT freely available “to help ensure that researchers and physicians can test it, learn from it, and ultimately improve care for transplant patients.”

Despite significant advances in transplant care, chronic GVHD remains a leading cause of morbidity and mortality post-transplant. However, the disease process doesn’t begin with the onset of symptoms; the underlying biological changes that trigger it start much earlier. The first few months after transplant are particularly critical, with patients potentially feeling well while underlying immune activity quietly sets the stage for future complications.

“By the time chronic GVHD is diagnosed, the disease has often been developing for months, silently damaging the body,” explains Paczesny. “We wanted to know if we could detect warning signals earlier, before patients feel unwell, and with enough lead time for physicians to intervene before the damage becomes irreversible.”

To achieve this, the researchers analyzed data from 1,310 recipients of stem cell and bone marrow transplants across four large, multi-center studies. Blood samples collected between 90 and 100 days post-transplant were analyzed for seven immune proteins associated with inflammation, immune activation and regulation, and tissue damage and remodeling. The immune biomarkers used in BIOPREVENT were previously identified and validated in a prior study led by Paczesny.

These biomarkers were combined with nine clinical factors, including patient age, transplant type, primary disease, and prior complications, extracted from transplant registries. The standardized data collection practices of transplant centers, which submit detailed transplant-specific data to the Center for International Blood and Marrow Transplant Research (with additional review for patients in clinical trials), were crucial in ensuring the model was based on consistent, high-quality clinical data, according to Paczesny.

The team tested various machine learning approaches to determine if they could predict patient outcomes more accurately than traditional statistical methods. The best-performing model, based on a statistical technique called Bayesian additive regression trees, became the foundation of BIOPREVENT.

Results demonstrated that models combining blood biomarkers with clinical data consistently outperformed those based solely on clinical data, particularly in predicting transplant-related mortality. The tool was validated in an independent group of transplant recipients, confirming its reliable prediction of risk beyond the patients used to build the model.

BIOPREVENT also effectively categorized patients into low- and high-risk groups, with clear differences in outcomes observed up to 18 months post-transplant. Notably, different biomarkers predicted different transplant outcomes, highlighting that chronic GVHD and transplant-related death are, at least in part, driven by distinct biological factors. For example, one blood biomarker was strongly associated with the risk of death after transplant, while others were more indicative of who would later develop chronic GVHD.

To maximize the research’s impact, the team developed BIOPREVENT as a freely accessible web-based application. Clinicians can input a patient’s clinical information and biomarker values to receive personalized risk estimates over time.

“It was important to us that this didn’t remain a theoretical model or a tool limited to a single institution,” Paczesny stated. “Free access to BIOPREVENT helps ensure that researchers and physicians can test it, learn from it, and ultimately improve care for transplanted patients.”

Currently, BIOPREVENT is intended to support risk assessment and clinical research, not to directly guide treatment decisions. The next step, according to Paczesny, will be to conduct carefully designed clinical trials to evaluate whether acting on these early warning signals of risk – such as closer monitoring or preventative therapies for high-risk patients – can improve long-term outcomes.

More broadly, the study reflects a shift towards precision medicine in transplant care, utilizing data to tailor monitoring to each patient’s individual risk profile. “This isn’t about replacing clinical judgment,” Paczesny emphasizes. “It’s about providing physicians with better information, earlier, so they can make more informed decisions.”

While further validation is needed before the tool can become part of routine care, researchers believe this approach represents a significant step towards preventing one of the most serious complications in transplant medicine.