Unique statistical approach lends promise to therapies for rare childhood brain cancer

Promising treatments for a rare, aggressive form of childhood brain cancer may move closer to clinical adoption thanks to a new analysis method that strengthens evidence from small clinical trials. Researchers at the University of Louisville used a modern statistical method to reanalyze the results of children with a special type of high-risk medulloblastoma called Group 3, who participated in two major national clinical trials of therapies. The analysis method, known as Bayesian Dynamic Borrowing, revealed that although they were somewhat promising, the trial results may have underrepresented the effectiveness of the therapies due to low numbers of participants.

By carefully combining information from previous studies with the results from these two recent clinical trials, the researchers showed that the therapies tested in the recent trials now show stronger evidence of meaningfully improving outcomes in children with high-risk Group 3 medulloblastoma.

High-risk Group 3 medulloblastoma is a fast-growing and notoriously hard-to-treat childhood brain cancer. Because so few children are diagnosed each year, even large national studies can enroll only a handful of patients. In one of the trials, only 10 children with this tumor type received the new therapy. In another, only 43 children were treated, despite the fact that the trial was open in dozens of cancer centers in the United States.

Unfortunately, these numbers are too low to thoroughly evaluate a therapy's effectiveness using traditional analysis methods.

"These small numbers make it extremely difficult for traditional statistical methods to show with certainty whether the therapies truly work," said Akshitkumar Mistry, a neurosurgeon and scientist at UofL and UofL Health - Brain & Spine Institute who led the reanalysis study. "As a result, promising treatments for these children can remain in limbo - not because they fail, but because the evidence isn't strong enough using traditional approaches."

To overcome this challenge, the UofL team used a novel statistical approach called dynamic borrowing via Bayesian models, which carefully "borrows" information from previous studies to strengthen the results of new trials. The idea is to let the model learn how similar the past and present data are, and to borrow more past data that match and less when they differ. The researchers ran 10,000 computer simulations using this process, ensuring that the findings remained both reliable and not artificially inflated.

Using this method, they reanalyzed data from two recent national trials and found a greater than 90% probability that the therapies tested in the clinical trials truly do provide benefit for children with high-risk Group 3 medulloblastoma. The therapies that had limited statistical power under traditional analyses now appear strongly promising under the new approach and as a result, may warrant renewed consideration as effective treatment options.

For children and families facing the devastating diagnosis of this aggressive cancer, these findings bring renewed hope that these treatments are not only worth trying but also are likely to be effective.

The research team published their study in the journal Neuro-Oncology in September.

"This work is part of a larger effort at UofL to modernize how we design, conduct and analyze clinical trial data, helping scientists and physicians learn as much as possible from the small, precious data that take years to collect in rare diseases," Mistry said. "Our goal is to make the most of every patient's experience - past and present - to improve the care of future patients. It is our way of honoring every child and adult who participates in clinical trials by ensuring their contributions continue to shape the treatments of tomorrow."

Mistry, who was profiled in the Dec. 2025 issue of MD-UPDATE, also led a team that recently published the world's largest database of neurological tumors, showing the composition of tumors at the genomic level, combined with clinical information such as patient age, tumor location and survival outcomes. This Atlas of Nervous System Tumors is a free, publicly available tool that promises to speed up the discovery of treatments for brain and nerve tumors, especially rare ones that have had limited research attention, like Group 3 medulloblastomas.

This project was supported by the Kentucky Pediatric Cancer Research Trust Fund and the Kentucky Department for Public Health. Mistry's work also is supported by the Louisville Clinical and Translational Research Center at UofL and by a UofL Presidential Scholars award.