Data-driven risk stratification guides childhood brain tumor treatment

Medulloblastoma is a highly variable pediatric brain tumor with several distinct molecular groups. It is often treated with radiation and chemotherapy, but in children, these therapies can have toxicities that impact long-term health. The potentially adverse effects associated with treatment underscore the need to understand individual patients’ risk better. However, it has been difficult to determine which factors are crucial for understanding risk and stratifying patients to different therapies.

To address this need, a team of St. Jude scientists performed an integrated analysis of three clinical trials. The study of 898 patients, published in Neuro-Oncology, examined treatment outcomes and compared them between the trials, accounting for the treatment differences and each tumor’s molecular features, such as DNA sequences and methylation profiles.

“We found data to support that 40% of patients with medulloblastoma can receive lower doses of craniospinal radiation therapy and almost all can receive less chemotherapy than they received in the clinical trials we analyzed and maintain the same or better survival,” said corresponding author Giles Robinson, MD, Department of Oncology, Division of Neuro-Oncology director.

These findings led the researchers to identify new risk factors and predictors of therapy effectiveness. For example, patients whose tumor falls into one of the two most common molecular groups of the disease, G3 or G4, could be subdivided based on their metastatic status, the loss and gain of certain chromosomes, the methylation subgroup, and amplifications of the known oncogene MYC. The result is the ability to group patients into one of four actionable treatment recommendations from low to high intensity. Classifying patients in this way ultimately leads to a greater number of lower-intensity treatments and fewer high-intensity treatments.

“Our results provide a blueprint to better risk-stratify therapy so that not everyone receives such toxic treatment and only the most aggressive tumors receive the most aggressive therapy,” said Robinson. 

The work was facilitated by the creation of a medulloblastoma data portal stemming from a collaboration between co-corresponding authors Robinson and Xin Zhou, PhD, Department of Computational Biology, Data Visualization director, which was published in Cancer Research. The portal incorporates patient data and outcomes, genomic alterations, and methylomic profiling results in an easy-to-use interface. Zhou, Robinson, and their teams used the portal to explore medulloblastoma-driving mutations in KBTBD4, which encodes a protein in the CULLIN-RING-E3 ubiquitin ligase complex. They identified two different novel subgroups in KBTBD4, suggesting that there may be a better way to classify these patients. When they associated some KDTBD4 mutations with other molecular features, it also hinted at previously unknown disease biology, providing a new area of investigation into how these mutations drive disease, which may eventually lead to novel treatments.

“We’ve made these data available at anybody’s fingertips,” Zhou said. “We hope that they can lead to more insights, which we can one day use to fuel new clinical trials to better treat medulloblastoma.”