The study demonstrates the performance of Caris AI Insights™ to predict first-line treatment benefit using an AI–based molecular signature. In the testing cohort, standard molecular risk patients predicted to benefit from FOLFIRINOX achieved substantially longer median overall survival when treated with FOLFIRINOX rather than gemcitabine plus nab-paclitaxel (gem/nab-p) (16.0 vs 9.9 months). Approximately half of the patients in the study received a different first-line therapy than the model would have recommended, highlighting the potential opportunity for more biologically informed treatment selection to improve survival and reduce toxicity.

The Caris AI Insights signature for pancreatic cancer is included in the Caris Molecular Tumor Board Report. This report, built using whole exome sequencing (WES) and whole transcriptome sequencing (WTS) data, is available upon request at no additional cost when ordering MI Cancer Seek®. Caris AI Insights deliver clinically relevant findings across all tumor types, with disease-specific algorithms to support treatment decision-making in colon, breast, ovarian, pancreatic and lung cancer.

For patients with advanced pancreatic ductal adenocarcinoma (PDAC), first-line therapy options include FOLFIRINOX, gem/nab-p, and NALIRIFOX. The study focused on the first two of these regimens, both of which can extend survival but can also exact a significant physical toll on the patient. Despite differences in toxicity and intensity, clinical judgments about treatment selection are still not sufficiently guided by tumor biology. Currently, clinicians must make the call without a widely adopted biomarker to guide them. As a result, some patients endure unnecessary side effects for limited benefit, while others may never receive the intensity of therapy their disease demands.

"For too long, pancreatic cancer treatment decisions have forced clinicians to choose between toxicity and uncertainty," said David Spetzler, MS, PhD, MBA, President of Caris Life Sciences. "Pancreatic AI shows that tumor biology can help guide that decision, identifying patients who may not need the most aggressive therapy, while also flagging those who may derive greater benefit from aggressive treatment. This is about using data and AI to move beyond trial-and-error and toward more precise care from the very first line of treatment."

Rather than relying on individual biomarkers, the pancreatic signature applies machine learning techniques to identify complex molecular patterns associated with real-world treatment benefit. The validation study leveraged Caris' large-scale clinico-genomic datasets, linking comprehensive molecular data with treatment outcomes across thousands of patients. Clinicians are provided with two results: risk stratification to categorize patients as standard or high molecular risk and guidance to help inform selection between FOLFIRINOX and gem/nab-p. The model identifies a meaningful subset of patients who may achieve similar or greater benefit from gem/nab-p, while also flagging patients who are more likely to require the intensity of FOLFIRINOX, potentially allowing greater flexibility in treatment selection.

Caris received FDA approval in November 2024 for MI Cancer Seek. This tissue-based assay is the first and only simultaneous WES and WTS-based assay with FDA-approved companion diagnostic (CDx) indications for molecular profiling of solid tumors.