Blood test may reveal which lymphoma patients will respond to immunotherapy  

Cornell engineers and Dana-Farber oncologists found that RNA floating in the plasma carries a signature of lymphoma tumor’s surroundings, one that tracked with whether patients were still in remission a year later.  

Of 91 lymphoma patients who received a powerful cell therapy at Dana-Farber Cancer Institute, those whose blood carried a particular gene expression signature before treatment were far less likely to see their cancer progress. Patients with high scores on that signature had an 80% lower risk of progression than those with low scores, according to a new study from Cornell and Dana-Farber. The research published Nov. 18 in the journal Cell Reports Medicine.  

The genetic signature comes from cell-free RNA, fragments of genetic material shed by cells into the bloodstream. The researchers say it offers a glimpse of the tumor’s local environment without the need for a biopsy.  

The research addresses a deficit in a current therapy for an aggressive blood cancer known as large B-cell lymphoma. Anti-CD19 CAR T-cell therapy, in which a patient’s own immune cells are engineered to hunt cancer, has transformed care for people with large B-cell lymphoma. But nearly half of patients either fail to respond or relapse within a year. Until now, doctors have few reliable ways to predict who will do well.  

“CAR T-cell therapy is one of the major breakthroughs of the past couple of decades, but not every patient responds,” said Iwijn De Vlaminck, professor in the Meinig School of Biomedical Engineering in the Cornell Duffield College of Engineering, who led the research.  

Glory Feyisayo Agun, a doctoral student in De Vlaminck’s lab and one of the study’s first authors, said the treatment’s complexity makes outcomes hard to anticipate. “CAR T-cell therapy has a strong overall response rate that drops over time, but the factors driving differences in patient response are not fully understood,” she said.  

Reading the neighborhood  

Recent research has pointed to the tumor microenvironment, the mix of immune, structural and blood vessel cells surrounding a tumor, as a key factor. A 2025 study in the journal Cancer Cell sorted lymphomas into three archetypes based on biopsy samples. Tumors with a “lymph node-like” environment fared better with CAR T therapy. Two others, one dominated by exhausted T cells and one by follicular macrophages and accessory cells, were linked to treatment failure.  

“If you analyze the bone marrow tissue of the patients, you can pick up on a signature of response,” De Vlaminck said. “You can collect tissue and biopsy, but it’s quite invasive.”  

His team wondered whether a blood draw could do the same job. “RNA that is circulating in blood plasma might provide the signal, which gives you a window not just to the immune system but also potentially to tissues,” he said. “The advantage is that you can sample it noninvasively.”  

The researchers analyzed plasma collected before treatment from patients receiving a particular CAR T product therapy across three separate groups. The median patient age was 65.  

The lymph node-like signature showed up clearly. Patients who responded had higher scores than non-responders in the plasma RNA, mirroring what the earlier biopsy study found in tumor tissue. Patients with above-median scores had significantly longer survival in all three groups.  

The signal was absent from immune cells circulating in the blood, which the team examined in matched samples. That suggests the plasma RNA reflects activity in the tissue itself rather than a change in the blood’s cellular makeup.  

“We don’t see that with circulating cells,” Agun said. “This is one of the first times we’re seeing this with plasma being a proxy for the tumor environment.”  

“Being able to access it from the blood – and not tissue – is a major advance, in our opinion, because it’s minimally invasive,” De Vlaminck said.  

One obvious worry is that the signature simply measures how much cancer a patient has. The researchers checked against a standard proxy for tumor burden, lactate dehydrogenase, and found no significant correlation before chemotherapy preparation. In statistical models, the score also predicted response better than clinical factors. They also ran several other tests suggesting the signature is specific to CAR T response.  

Next steps  

The authors say the model needs validation in larger, multi-institution groups. The findings have not been shown to change patient outcomes.  

“The study was focused on a specific patient group,” Agun said. “The next step would be to expand that to other patients who are receiving CAR T therapy. Next would be validating across a cohort and seeing if this is reproducible across different contexts.”  

If the findings hold, the authors envision two uses. Doctors could identify patients with unfavorable tumor environments before infusion and perhaps offer them additional therapies aimed at reshaping that environment, such as targeted radiation or drugs aimed at macrophages. A blood test could also confirm that such an intervention had worked before CAR T cells are given.  

“This approach has potential beyond lymphoma,” said co-first author Conor Loy, Ph.D. ’24 and co-founder of Romix Biosciences, a biotechnology startup combining cell-free RNA and AI to advance precision. “This technology could help guide treatment decisions, matching patients to treatments that are right for them, and helping doctors to choose treatment to pair with immunotherapies.” 

Media Contact

Kaitlyn Serrao