Skip to main content
Premium Trial:

Request an Annual Quote

UPenn Team Develops Test to Predict Treatment Response for Castleman Disease Patients


NEW YORK – A team of researchers led by scientists from the University of Pennsylvania have developed a protein-based test to identify patients with a rare blood disorder who will respond to treatment.

In a study published earlier this month in Blood Advances, the team, which included researchers from the Castleman Disease Collaborative Network, Janssen Pharmaceuticals, and hospitals across the globe, analyzed 190 patient samples to narrow down which analytes would best determine treatment responses for Castleman disease and validate the test.

Patients with idiopathic multicentric Castleman disease, or iMCD, a hematologic disorder involving an overgrowth of cells in the body's lymph nodes, experience systemic inflammation and are at risk for fatal multiorgan failure. The cause of the disease is unknown, although previous research has shown that interleukin-6 is a disease driver in about one-third of patients, the researchers said.

Currently, only one treatment — anti-IL-6 therapy siltuximab (EUSA Pharma's Sylvant) — is approved by the US Food and Drug Administration for the disease, and there are no tests to predict the likelihood of a patient's response to treatment, according to the authors of the Blood Advances study. Based on the drug's Phase II registrational study, 66 percent of iMCD patients did not respond to the drug, the authors noted.

EUSA Pharma did not respond to requests for comment for this story.

The senior author of the study, David Fajgenbaum, director of the Center for Cytokine Storm Treatment & Laboratory at UPenn's Perelman School of Medicine, noted that if siltuximab doesn't work for a patient, the next option is aggressive chemotherapy, which should be started as soon as possible.

He added that although researchers don't know why IL-6 is elevated in so many patients with iMCD, "we do know that it is the key/driver cytokine in patients who improve on/respond to anti-IL-6 therapy. … In patients that don't improve on/respond to anti-IL-6 therapy, we assume that there must be another cytokine or pathway driving the disease process."

Fajgenbaum, who has Castleman disease and "nearly died five times" as a result, said the team originally looked at 1,305 analytes and asked, "Among those 1,300 proteins we measured … are there certain signatures that were found in the people who responded [to the drug] and certain signatures in people who didn't respond" to the drug.

The team wanted "to get a sense for what's going on in the blood of responders and non-responders," which would allow them "to start to dig into what is it that actually differentiates it," he said.

Using SomaLogic's SomaScan instrument, they measured the 1,305 analytes, of which 1,178 passed quality control measures.

By testing the serum of 88 iMCD patients, 60 patients with clinico-pathologically overlapping diseases — including human herpesvirus associated MCD, Hodgkin's lymphoma, and rheumatoid arthritis — and 42 healthy controls, Fajgenbaum and his team found that iMCD patients have serum proteomes that differentiate them from the patients with overlapping diseases.

Further, they found that within the iMCD patients, there was a subgroup with a superior response to siltuximab.

To first determine each patient's disease activity, the team used a score that relies on C-reactive protein, hemoglobin, and albumin developed by a Japanese research team and published in Modern Rheumatology in 2018.

The first result, based on the disease activity, found that iMCD proteomic profiles are broadly distributed compared to other comparable diseases, which "suggests that iMCD is heterogeneous" and "that multiple etiologies are able to elicit iMCD," meaning there's no one cause of the disease, the researchers wrote.

Using an algorithm they developed on the iMCD samples, six proteomically defined clusters of patients were differentiated. The researchers found that the first cluster demonstrated significantly higher disease activity, higher baseline IL-6 levels, and significantly higher response to siltuximab — with 65 percent of patients responding, compared to 19 percent of patients from the other clusters.

The team then utilized an elastic net algorithm to create a "cluster 1 score" for that first group of patients with higher disease activity and better treatment response to siltuximab in order to determine which specific proteins predicted cluster 1 membership in the discovery cohort.

In other words, the higher the cluster 1 score was, the more likely the patients were to respond to siltuximab, and the team looked for the proteins those samples had in common, Fajgenbaum said.

The researchers were then able to narrow the list of 1,178 analytes down to "the seven most important" proteins, Fajgenbaum said — apolipoprotein E, amphiregulin, serum amyloid P-component, inactivated complement C3b, immunoglobulin E, IL-6, and erythropoietin.

The team assessed the score in a 23-sample validation cohort, finding a P value of .0757. They wrote that there was "a trend towards a positive association between cluster 1 score and siltuximab response … and cluster 1 score was significantly associated with increased disease activity" and IL-6 levels.

Although the result technically indicates a trend toward positive association and not a confirmed positive association, Fajgenbaum noted that because of the smaller sample size of the validation cohort, the researchers felt it was a strong signal that the test worked for this cohort.

"The seven proteins included in our validated algorithm could form the basis for development of a clinical predictive signature," the researchers wrote.

The team also analyzed each protein individually to see if a specific one would be significantly associated with treatment response in both the discovery and validation cohorts, but none were discovered. Fajgenbaum said that "as of now, it seems that the seven proteins together outperform any other combination of fewer proteins," so a future test would include all seven.

For future studies, the research team recommended assessing the predictive signature "in association with outcomes of siltuximab treatment other than lymph node response" and potentially in patients treated with anti-IL-6 receptor blocker tocilizumab (Roche's Actemra).

Fajgenbaum said the team plans to continue assessing the algorithm's applicability in independent cohorts.

Sunita Nasta, an associate professor of clinical medicine at UPenn not involved with the study, said that this standardized evaluation "will permit faster elimination of certain diagnoses to the benefit of the patient," although she noted the test needs to be validated in a larger group of patients.

She added that the method is "time-consuming" and requires specialized equipment, "so small labs will not be able to implement [the method] easily." As a result, "centralized evaluation in tertiary care centers will still be needed" for patients with the disease.

According to Fajgenbaum, the researchers are "eager to find a partner that would want to measure these proteins" and commercialize the algorithm. They are working on identifying potential partners, although he said they haven't had any conversations with partners yet.

Once the team identifies one, Fajgenbaum said the team is in "a good place" to offer the test to the commercializing partner, since they filed for a provisional patent from the US Patent and Trademark Office last year.

Regulatory plans, meantime, he said, were out of his area of expertise, although those would likely be determined by the commercial partner.