Pancreatic cancer breakthrough: AI-based blood test proves faster, simpler than biopsies

When it comes to pancreatic cancer, speed can mean survival. A new AI-powered method developed by Johns Hopkins Kimmel Cancer Center is giving doctors a faster, more accurate way to determine if a treatment is working — by analysing tiny DNA fragments floating in a patient’s blood. Known as ARTEMIS-DELFI, the technique could outpace traditional scans and even other advanced blood-based tests in identifying early responses to therapy, potentially transforming the treatment landscape for one of the deadliest cancers.

In two major clinical trials — CheckPAC and PACTO — ARTEMIS-DELFI and another method, WGMAF, were tested on blood samples from patients with pancreatic cancer. Both approaches outperformed standard imaging and molecular markers in predicting treatment outcomes within two months. However, ARTEMIS-DELFI proved to be more effective and practical.

Unlike WGMAF, which requires tumor tissue and compares it with cell-free DNA in the blood, ARTEMIS-DELFI relies solely on DNA fragments circulating in the bloodstream. Using machine learning, it identifies patterns in these fragments to detect treatment response — without needing a tumor biopsy.

Time is especially critical in pancreatic cancer cases, said senior author Dr Victor E Velculescu, co-director of the Cancer Genetics and Epigenetics Program at Johns Hopkins. “We want to know as quickly as we can if the therapy is helping the patient or not. If it is not working, we want to be able to switch to another therapy,” he explains.

Traditional imaging can be slow and is often unreliable for patients undergoing immunotherapy. ARTEMIS-DELFI, on the other hand, can show signs of response as early as four weeks into treatment, confirmed by its performance in the PACTO trial.

“The ‘fast-fail’ ARTEMIS-DELFI approach may be particularly useful in pancreatic cancer,” says lead author Carolyn Hruban, now a postdoctoral researcher at Dana-Farber Cancer Institute. “It's simpler, likely less expensive, and more broadly applicable than using tumor samples.”

Looking ahead, the team plans to conduct prospective studies to determine if the method can help clinicians more efficiently tailor treatments. A similar technique was also recently shown to be effective in tracking colon cancer therapies, suggesting the potential for even broader applications.