To Dr. Cooper, some of the most exciting tools that have come out of digital pathology recently are prognostic assays with heavy input from oncologists. “They’re consumers of this information and I think those projects are successful because of their involvement,” he says. “Pathologist input is important because they know the details of the lab and the tissue processing. All of those things can be very important in the success and failure of these projects.”
Pathology and oncology have always worked hand in hand, Dr. Lu says, “but now with AI, there’s another interesting layer.” Decisions made in clinical practice can now be supported by tens of thousands of data pieces and experience from multiple modalities.
“The future is not pathology AI versus oncology AI. It is integrated, team-based clinical intelligence.”
In patients with melanoma, a positive ctDNA is associated with poor outcomes.
“It’s a very specific test,” says April K. S. Salama, MD, medical oncologist and associate professor of medicine at Duke University School of Medicine. “If your ctDNA is positive postoperatively, or it changes from negative to positive at any point in surveillance, it’s a pretty poor prognostic factor. That’s been borne out now over a series of small prospective cohorts, as well as correlative analyses to some of our larger adjuvant trials.”
Yet that doesn’t mean ctDNA is “ready for prime time” in melanoma, Dr. Salama says, because while a positive test clearly indicates a higher risk of recurrence, a negative ctDNA result does not conclusively exclude patients with a recurrence, which have been missed with ctDNA. “The question that remains is, how do we use this information?”
For the majority of patients, a positive ctDNA can’t be used in real time to inform adjuvant decision-making, because “most are negative at the first postoperative timepoint—exactly when adjuvant decisions are being made,” she says. And it doesn’t reveal which therapy might work best for the patient or if the patient will respond to therapy. “We have to be careful we’re not conflating a prognostic marker with a predictive marker of benefit from therapy,” she says. It also isn’t a one-time test. “A negative test today doesn’t mean you’re truly negative. It means you’re still at risk for recurrence based on stage and other factors.” The test’s value is in longitudinal monitoring.
Dr. Salama will review the recent research on ctDNA for disease monitoring and recurrence prediction in “Beyond Routine Surveillance Imaging: Emerging Technologies for the Early Detection of Melanoma Recurrence.” Her co-presenters are Cicero Luiz Cunha De Sousa Martins, MD, of the Latin American Cooperative Oncology Group, who will discuss the global perspective on melanoma surveillance, and Michael Farwell, MD, of the University of Pennsylvania Health System, who will address the role of CD8-targeted PET imaging to predict early response to immunotherapy in melanoma.
A study published last year investigated whether ctDNA can predict survival outcomes during adjuvant targeted therapy or placebo treatment in stage III melanoma (Syeda MM, et al. Lancet Oncol. 2025;26[5]:641–653). The authors compared associations between survival outcomes and baseline (post-resection) ctDNA copies per milliliter. In a subset of patients, ctDNA quantities during follow-up or at recurrence were measured. Patients with adverse longitudinal ctDNA kinetics (molecular relapse or persistently positive) had markedly shorter median recurrence-free survival compared with patients with favorable kinetics.
“ctDNA positivity at any time point was a poor prognostic factor,” Dr. Salama says, “and conversion from negative to positive did predict relapse.” The patients who received adjuvant therapy had better outcomes, but even within the treatment group, the patients who were ctDNA-positive had poorer outcomes than those who remained ctDNA-negative.
One limitation of ctDNA studies is the difficulty of sample collection. “We need to remember how important these samples are,” she says. In the Lancet Oncology study, baseline plasma samples were available for 597 of 870 patients. Samples for assessing the ctDNA positivity rate at three months, six months, nine months, and 12 months after treatment initiation were available for only 94 of 870 patients.