More than half of transplant recipients develop chronic graft-versus-host disease (cGVHD), with 15% succumbing to causes unrelated to cancer relapse, according to a comprehensive analysis by a research team at the Medical University of South Carolina (MUSC) Hollings Cancer Center. This underscores the complex reality of allogeneic hematopoietic cell transplantation, where a potential cure may be accompanied by significant long-term health challenges.
cGVHD occurs when donor immune cells launch an attack on the recipient’s healthy tissues, leading to damage in the skin, eyes, mouth, joints, and lungs. It remains a leading cause of debilitating illness and nonrelapse mortality post-transplant. Dr. Sophie Paczesny, who spearheaded the research, aimed to identify earlier indicators of this condition. Alongside Dr. Michael Martens and Dr. Brent Logan, they developed a machine learning tool called BIOPREVENT, designed to estimate a patient’s risk for developing cGVHD and the likelihood of dying without a cancer relapse based on blood biomarkers and clinical factors.
The researchers assert that cGVHD often goes undiagnosed until overt symptoms appear across multiple organs, though they believe tissue disruptions may begin much earlier—soon after donor cells are infused. “By the time chronic GVHD is diagnosed, the disease process has often been unfolding for months, quietly hurting the body,” Paczesny explained. Their study focuses on the 90 to 100-day period post-transplant, aiming to identify biological signals before the onset of symptoms.
In a study published in the Journal of Clinical Investigation, the team analyzed data from 1,310 transplant recipients. This cohort included samples collected at Day 90 or 100 from patients across four multicenter groups: BMTCTN 0201 and 1202, the Dana-Farber Cancer Institute cohort, and a combined pediatric and adult cohort from trial NCT02194439. Of the participants, 52% developed any form of cGVHD, while 37% experienced moderate to severe cases, and 15% faced nonrelapse mortality.
The researchers measured seven plasma proteins linked to cGVHD risk, including CXCL9, MMP3, and DKK3, as well as the protein IL1RL1, which has demonstrated a strong correlation with nonrelapse mortality. Alongside these biomarkers, they combined nine clinical variables such as graft source and patient age to create the BIOPREVENT model. The objective was clear: to determine if machine learning could enhance the predictive power of existing clinical assessments.
Employing various analytical techniques—including penalized Cox regression and deep learning models—the investigators split their dataset into training and validation subsets. They found that models merging biomarkers with clinical data significantly outperformed those relying solely on clinical factors, particularly in predicting nonrelapse mortality. Notably, the Bayesian additive regression trees (BART) model emerged as a top performer, forming the basis for the BIOPREVENT algorithm, while deep learning methods did not provide a distinct advantage.
As the researchers drilled down into specific outcomes, they noted that different predictors emerged as particularly influential. For cGVHD risk, variables such as primary disease, graft source, and certain biomarkers were critical, while for nonrelapse mortality, IL1RL1 and patient age had the most significant impact.
With the aim of translating their findings into actionable clinical tools, BIOPREVENT also classifies patients into higher and lower risk categories based on predicted cumulative incidences. For example, a cutpoint of 0.45 for cGVHD at Day 360 indicated a cumulative incidence of 58.0% for those at or above it, compared to 32.7% for those below. For nonrelapse mortality, a threshold of 0.08 revealed a 13.9% incidence in the higher-risk group against a 3.5% incidence in the lower-risk group.
The study also discusses the practical implications of BIOPREVENT, now available as a public web application where clinicians can input clinical variables and biomarker values from Day 90 or 100 to obtain personalized risk estimates. The authors suggest that early risk assessments could facilitate closer monitoring and possibly influence future clinical trials aimed at preemptive interventions.
While the research offers promising insights, it does come with limitations. The model was built on a single biomarker timepoint, and the dataset did not encompass some potentially important clinical variables due to its retrospective nature. Despite these challenges, the authors remain optimistic that recognizing the biological signals of cGVHD earlier could enhance patient care and outcomes in the transplant setting.
See also
AI Study Reveals Generated Faces Indistinguishable from Real Photos, Erodes Trust in Visual Media
Gen AI Revolutionizes Market Research, Transforming $140B Industry Dynamics
Researchers Unlock Light-Based AI Operations for Significant Energy Efficiency Gains
Tempus AI Reports $334M Earnings Surge, Unveils Lymphoma Research Partnership
Iaroslav Argunov Reveals Big Data Methodology Boosting Construction Profits by Billions
