A panel of statisticians will present statistical tools and approaches on causal inference to help researchers working through enormous databases to better isolate treatment effects in hematopoietic cell transplantation (HCT) and CAR T-cell therapies.

Kwang Woo Ahn, PhD, chief statistical director for CIBMTR®, said these techniques are becoming increasingly important to consider and apply as researchers expand from focusing on small clinical trials and make increasing use of large registry databases, such as the CIBMTR data repository.

“In clinical trials, investigators control important patient characteristics and can ensure that the biggest difference between two groups is the treatment. In those cases, causation can be obvious,” Ahn said. “But when we work with real-world data from observational data, we don’t have any controls and we cannot design the studied groups, so it becomes more difficult to identify causation.”

He will chair the Feb. 14 session Concurrent: Causal Inference and Machine Learning for Right-Censored Data Analysis and Bioinformatics/Genetics Data Analysis at 10:30 a.m. in Ballroom B.

Ruta Brazauskas, PhD, associate professor in the Division of Biostatistics, Medical College of Wisconsin (MCW), will cover propensity scores and their applicability as a tool for matched cohort studies with multiple variables. Michael Martens, PhD, assistant professor in the Division of Biostatistics, MCW, will examine machine learning for biomarkers of chronic graft-versus-host disease (GVHD) and non-relapse mortality risks in allogeneic HCT. Paul Auer, PhD, professor of biostatistics and data science, MCW, and director of the Biostatistics Shared Resource for the MCW Cancer Center, will discuss Mendelian randomization, a tool increasingly applied to work with genetics data.

Ahn said although the three experts will provide insights from various perspectives, each presentation focuses on leveraging statistical tools to improve a database study’s chance to identify causes, rather than associations, of an effect.

“There are always questions regarding the accuracy about conclusions based on registry data or observational data analysis because they are not from clinical trials, and so it is not straightforward to make causal inferences,” Ahn said. “Modern statistical techniques can be used to address such limitations.”

The statistical models that will be discussed at the panel session are general tools with applications for any medical study, but they are particularly adapted for HCT and CAR T-cell treatment studies.

“Many studies in other medical fields will compare two drugs, study the outcome, and look at the association of whether the patients fared better or not,” Ahn said. “But in HCT and CAR T-cell therapy, the biggest fundamental questions are how to extend life expectancy and improve quality of life. It is not just a yes or no answer, it is a question of how long.”

The presentations and follow-up discussions are designed to provide investigators, clinicians, and researchers insights into statistical approaches and encourage researchers to involve statisticians at the design stage of new studies.

“This is a statistical session, but I try to make this session as applied as possible,” Ahn said. “Researchers should discuss their questions with statisticians at the start of a study, to collaborate and design the study together from the very beginning.”

This and other sessions at the 2025 Tandem Meetings | Transplantation & Cellular Therapy Meetings of ASTCT® and CIBMTR® will be available for on-demand viewing for registered attendees following the live presentation.