AI Tool Identifies Life-Threatening Transplant Complications

Recent advancements at MUSC Hollings Cancer Center have unveiled a groundbreaking artificial intelligence (AI) tool designed to assist healthcare professionals in quickly identifying life-threatening complications following stem cell and bone marrow transplants. This innovation aims to enhance patient care, particularly during the precarious recovery phase.

For numerous patients, undergoing a stem cell or bone marrow transplant is crucial for survival. However, the journey to recovery extends beyond their hospital discharge. Some patients may face severe complications that arise months later, often catching them and their healthcare providers off guard.

One significant concern is chronic graft-versus-host disease (GVHD), where immune cells from the transplant mistakenly attack healthy tissues in the patient. This condition can impact various organs, such as the skin, eyes, mouth, joints, and lungs, leading to long-term health issues or even death.

Under the leadership of Sophie Paczesny, M.D., Ph.D., co-leader of the Cancer Biology and Immunology Research Program at Hollings, and in collaboration with Michael Martens, Ph.D., and Brent Logan, Ph.D. from the Medical College of Wisconsin, researchers have developed an AI tool called BIOPREVENT. This tool aims to help clinicians identify patients at an elevated risk for chronic GVHD, even before symptoms manifest, allowing for earlier and more effective monitoring.

By harnessing machine learning techniques in conjunction with immune-related proteins and verified clinical data, the team has formulated a risk assessment model that forecasts a patient’s likelihood of developing chronic GVHD and experiencing transplant-related mortality. Their findings, detailed in the Journal of Clinical Investigation, combine immune biomarkers, clinical information, and machine learning to provide real-world risk predictions.

A Window of Opportunity After Transplant

Despite significant progress in transplant medicine, chronic GVHD continues to be a leading cause of morbidity and mortality post-transplant. However, research indicates that the biological changes facilitating this disease initiate well before symptoms present themselves.

Particularly in the first few months following a transplant, patients may appear to be recovering well, yet underlying immune activity could be paving the way for serious complications.

By the time chronic GVHD is diagnosed, the disease process has often been unfolding for months, quietly hurting the body. We wanted to know whether we could detect warning signs earlier, before patients feel sick, and soon enough for clinicians to intervene, before the damage becomes irreversible.”

Sophie Paczesny, M.D., Ph.D., co-leader of the Cancer Biology and Immunology Research Program at Hollings Cancer Center

Transforming Blood Tests into Early Warnings

To achieve this, the researchers analyzed data from 1,310 patients who had undergone stem cell and bone marrow transplants, sourced from four extensive multicenter studies. Blood samples collected 90 to 100 days post-transplant were examined for seven immune proteins associated with inflammation, immune activation and regulation, and tissue injury and remodeling. The immune biomarkers utilized in BIOPREVENT had previously been identified and validated in an earlier study spearheaded by Paczesny.

These biomarkers were combined with nine clinical variables, including patient age, type of transplant, primary illness, and previous complications, all sourced from transplant registries. In the U.S., transplant centers are obligated to submit comprehensive, transplant-specific information to the Center for International Blood and Marrow Transplant Research, with added scrutiny for clinical trial participants. According to Paczesny, this uniform data collection was crucial in ensuring the model was built on consistent and high-quality clinical information.

The team experimented with various machine-learning techniques to determine their efficacy in predicting patient outcomes relative to traditional statistical approaches. The optimal model, based on a method known as Bayesian additive regression trees, formed the backbone of BIOPREVENT.

From Clinical Algorithm to Practical Tool

The results indicated that models leveraging blood biomarkers in conjunction with clinical data significantly outperformed those relying solely on clinical data, particularly in predicting mortality associated with transplants. The tool was further validated in an independent cohort of transplant recipients, confirming its reliability beyond the initial patient group used to create the model.

BIOPREVENT successfully categorized patients into low- and high-risk groups, demonstrating distinct differences in outcomes over an 18-month period. Furthermore, different biomarkers were found to correlate with various transplant outcomes, highlighting that chronic GVHD and transplant-related mortality are influenced by at least partially separate biological factors. For instance, one specific blood biomarker was strongly associated with mortality risk after the transplant, while others were more effective in indicating who might develop chronic GVHD later on.

To expand the practical application of their research, the team has developed BIOPREVENT into a complimentary, web-based application. Clinicians can input a patient’s clinical details and biomarker results to receive tailored risk estimates over time.

“It was vital for us that this tool is not confined to theoretical use or limited to a specific institution,” stated Paczesny. “By making BIOPREVENT freely accessible, we aim to ensure that clinicians and researchers can test it, learn from it, and ultimately enhance care for transplant patients.”

A Step Toward More Personalized Transplant Care

At this stage, BIOPREVENT serves to assist in risk assessment and clinical research rather than directing treatment decisions. According to Paczesny, the next steps involve conducting meticulously designed clinical trials to ascertain whether addressing these early risk indicators with closer monitoring or preventative therapies for high-risk individuals can lead to improved long-term outcomes.

This study signifies a broader movement toward precision medicine within transplant care, leveraging data to customize follow-up procedures to suit each patient’s specific risk profile.

“This is not about replacing clinical judgment,” emphasized Paczesny. “It’s about equipping clinicians with better information sooner, enabling more informed decision-making.”

While further validation is essential before this tool can be integrated into routine clinical practice, the researchers are optimistic that their approach represents a meaningful stride toward combating one of the most critical challenges in transplant medicine.

“For patients, the uncertainty that follows a transplant can be incredibly daunting,” concluded Paczesny. “We aspire for tools like BIOPREVENT to help us anticipate complications sooner, ultimately alleviating the burden of chronic GVHD.”