BullFrog AI Platform Identifies Pancreatic Cancer Patient Subgroup with Threefold Survival Improvement
Key Insights
BullFrog AI's bfLEAP platform analyzed phase 3 clinical trial data and identified patient clusters in pancreatic adenocarcinoma (search), with one subgroup showing nearly threefold increase in mean survival when treated with glufosfamide versus best supportive care.
The AI-driven analysis of 134 patients from the TH-CR-302 study revealed four distinct clusters, with Cluster A characterized by lower baseline glucose and higher neutrophil (search)/monocyte (search) counts demonstrating superior overall survival outcomes.
This post-hoc analysis demonstrates the potential of machine learning approaches to identify biomarkers for patient stratification in one of oncology's most challenging cancers, where five-year survival rates remain below 10%.
BullFrog AI Holdings (search) has demonstrated the potential of artificial intelligence to transform treatment outcomes in pancreatic adenocarcinoma (search), one of oncology's most lethal malignancies. Using its proprietary bfLEAP platform, the company analyzed clinical trial data and identified patient subgroups with nearly threefold increases in overall survival rates when treated with the investigational drug glufosfamide.
The post-hoc analysis, conducted in collaboration with Eleison Pharmaceuticals (search) and the H. Lee Moffitt Cancer Center (search), examined data from the TH-CR-302 study, a randomized phase 3 clinical trial that evaluated glufosfamide against best supportive care in pancreatic adenocarcinoma (search) patients.
AI-Driven Patient Stratification Reveals Treatment Heterogeneity
The research team, led by Nikolas Naleid, M.D., Pharm.D., from Moffitt Cancer Center, utilized BullFrog AI's bfLEAP and bfPREP platforms to apply machine learning methodologies for analyzing multimodal biological data. The analysis constructed a patient-to-patient similarity network from baseline demographics, screening laboratories, and pre-randomization clinical features.
From 281 patients in the original study, the AI platform successfully clustered 134 patients (47%) into four distinct groups: Cluster A (n=12), Cluster B (n=26), Cluster C (n=24), and Cluster D (n=72). Patients without consistent groupings were not assigned to any cluster.
Biomarker Profile Predicts Superior Outcomes
Cluster A emerged as the most promising subgroup, characterized by lower baseline glucose values and higher baseline neutrophil (search) and monocyte (search) counts. This cluster demonstrated improved overall survival compared with best supportive care when treated with glufosfamide.
"The platform revealed biologically significant patient clusters, highlighting subgroups where mean survival increased from the control arm to nearly three times greater in the glufosfamide-treated cohort," according to the company's analysis.
Glufosfamide, originally developed as a glucose-conjugated derivative of ifosfamide designed to target glucose-dependent tumor cells, showed variable outcomes in the initial study but demonstrated potential for biomarker-driven applications through this AI analysis.
Addressing Critical Unmet Need in Pancreatic Cancer
Pancreatic adenocarcinoma (search) remains one of the most challenging cancers to treat, with a five-year survival rate of less than 10% and limited therapeutic options beyond conventional chemotherapy. The disease continues to represent a significant challenge in oncology due to its aggressive nature and poor prognosis.
"This glufosfamide case study in pancreatic cancer (search) effectively demonstrated the capability of our platform to furnish drug developers with a comprehensive analytical tool designed to address multimodal biological complexity at scale," said Vin Singh, founder and CEO of BullFrog AI.
Singh emphasized the inefficiencies in conventional drug development, noting that "excessive time and resources are squandered... resulting in patients lacking effective treatments." The AI technologies aim to address this by focusing efforts on viable pathways, reducing costs for payers, and accelerating access to targeted therapies.
Clinical Implications and Future Directions
The research team concluded that ensemble approaches like bfLEAP can successfully identify patient subgroups within existing clinical trial data. The analysis revealed treatment effect heterogeneity among clusters and identified possible early predictors of outcomes, highlighting the effectiveness of data-driven clustering approaches for refining patient stratification.
The study, titled "Data-driven subtyping and differential glufosfamide benefit in pancreatic adenocarcinoma (search)," was presented at the 2026 ASCO Gastrointestinal Cancers Symposium in San Francisco and published in a Journal of Clinical Oncology supplement.
This research represents part of a broader movement in precision medicine, where data analytics bridge the gap between trial results and real-world implementation. For managed care organizations, such developments may lead to more cost-effective formularies, with biomarker testing informing coverage decisions and improving population health outcomes.
The study was co-authored by Richard Kim, M.D., Service Chief of Medical Gastrointestinal Oncology at Moffitt Cancer Center, along with researchers from Eleison Pharmaceuticals (search) and BullFrog AI, demonstrating the collaborative approach needed to advance AI-driven precision oncology.