NEW YORK – Autologous CAR T-cell therapies can lead to deep and durable responses, almost akin to a cure in some patients with lymphoma, and researchers have been trying to better identify these best responders.
In a series of presentations at the American Society of Hematology annual meeting on Saturday, researchers shared data on the ability of different biomarker approaches to home in on who is most likely to benefit from CD19-directed cell therapies.
Inflammation-based signatures
Sandeep Raj, a hematology oncology fellow at Memorial Sloan Kettering Cancer Center, presented an approach to predicting CAR T-cell treatment failure in large B-cell lymphoma patients, dubbed InflaMix. The signature is based on biomarkers of inflammation, he explained.
"We wanted to understand the role of inflammation in treatment failure and identify a reproducible metric, which could best identify which patients are at high risk of relapse," he said.
To develop InflaMix, Raj and colleagues measured 14 key lab values and cytokines in large B-cell lymphoma patients prior to their CAR T-cell therapy infusions. They then modeled the clusters associated with clinical outcomes and homed in on two unique signatures they could assess with a single blood draw before CAR T-cell infusion.
The inflammatory signature, enriched for biomarkers such as interleukin-6, ferritin, and TNF-alpha, among other markers of high inflammation, was associated with lower rates of complete responses to CAR T-cell therapy, as well as shorter progression-free and overall survival.
"Our interpretation is that InflaMix clustering is a modifiable risk factor, and that inflammation immediately prior to CAR T-cell infusion is actually what drives poor outcomes," Raj said.
In his presentation on Saturday, Raj shared data from his team's work validating the signature across 352 large B-cell lymphoma, follicular lymphoma, and marginal zone lymphoma patients treated with Gilead Sciences' Yescarta (axicabtagene ciloleucel) across the ZUMA-1, ZUMA-7, and ZUMA-5 registrational studies that led to the CAR T-cell therapy's approval in various lines of therapy. Thanks to the availability of "rich correlatives already available for this dataset," Raj said his team was able to access the biomarker data they needed to conduct their validation studies in the ZUMA cohorts.
After adjusting for other variables, Raj said the InflaMix signature inflammation clustering was "significantly associated with diminished responses" in these studies.
In the ZUMA-1 cohort, InflaMix assigned 46 percent of patients to the inflammatory cluster, and the same was true for 30 percent of patients in ZUMA-7. After 100 days post-infusion, Raj shared that the patients stratified to the inflammatory cluster were less likely to have a complete response to Yescarta than those without the high inflammatory signature. In both ZUMA-1 and ZUMA-7, patients with high inflammatory signatures were less likely to experience a complete response to Yescarta.
The signature showed a consistent association with overall survival outcomes in ZUMA-1 and ZUMA-7 and with progression-free survival in ZUMA-1, he added.
Given these findings, Raj said he and his team wanted to understand what might be driving this inflammatory signal. Here, they came up with two hypotheses. The first, he said, was that the inflammatory signature could reflect an immune-suppressing tumor environment, and the second was that it could impact in vivo CAR T-cell expansion after infusion.
To investigate these theories, the researchers first investigated RNA-sequencing data from pretreatment samples taken at the time patients had their immune cells harvested. They also looked at multiplex immunohistochemistry data from tumor biopsies to try to better explain the tumor microenvironment. Patients with high-inflammation signatures according to InflaMix had lower activated CD8-positive phenotypes as well as lower scores on Immunosign-21, which is a metric of activated immune contexture derived from RNA sequencing.
To probe the second hypothesis about CAR T-cell expansion, Raj and colleagues measured peak expansion of Yescarta cells in patients' blood within the first month of their infusions using flow cytometry. Here, they found that the inflammatory signature at infusion was associated with diminished peak CAR T-cell expansion.
Given that the InflaMix signature could reproducibly predict outcomes and that the inflammatory biomarkers played a part in both patients' tumor microenvironments and their CAR T-cell therapy expansion post-infusion, Raj suggested that the signature could help identify patients who could benefit from some sort of anti-inflammatory intervention.
"We think InflaMix could be used to identify high-risk patient populations in clinical trials to study either prophylactic anti-inflammatory therapies or pursue consolidating therapies shortly after infusion," he said. "It requires only a blood draw and has good performance even when unconventional labs and cytokines are not available."
As for whether patients who have a high inflammatory signature might be treated with alternative approaches instead of CAR T-cell therapy, such as bispecific antibodies, Raj said future prospective studies will need to answer that. This is especially true since the ZUMA-based InflaMix validation studies were retrospective, and as such, many of the patients didn't have the option to receive newer bispecific antibodies at the time when they were treated with Yescarta.
Either way, looking ahead, Raj said he hopes the InflaMix signature will play a role in trial designs and high-risk patient stratification in the CAR T-cell therapy space.
Circulating tumor DNA
In a separate presentation Saturday, Ash Alizadeh, medical oncologist at Stanford Health Care, shared data suggesting that large B-cell lymphoma patients' circulating tumor DNA (ctDNA) could predict early outcomes with Bristol Myers Squibb's CAR T-cell therapy Breyanzi (lisocabtagene maraleucel).
Alizadeh and his team used a ctDNA testing approach developed at his and Maximilian Diehn's labs, called Phased Variant Enrichment and Detection Sequencing, or PhasED-Seq, on LBCL patients enrolled in BMS's Phase III TRANSFORM clinical trial of second-line Breyanzi. A minimal residual disease test that uses PhasED-Seq is marketed by Foresight Diagnostics, which Alizadeh cofounded.
"PhasED-Seq, a new assay leveraging phased variants that are multiple single-nucleotide variants on individual DNA strands, has been shown to improve sensitivity and specificity," Alizadeh said.
All patients enrolled in TRANSFORM had blood drawn for ctDNA assessment at baseline. Variants were identified in plasma after censoring for germline variants and clonal hematopoiesis. The researchers identified tumor-derived phased variants, and these were then used to longitudinally measure ctDNA among patients randomized to the Breyanzi arm.
The ctDNA assessments post-treatment, which were limited to 63 patients in the Breyanzi arm of the trial, took place on the day of the CAR T-cell therapy infusion, then again at day 15 post-infusion, and again one month, three months, and a year after infusion. Alizadeh noted that the longitudinal analysis of ctDNA in TRANSFORM's standard of care arm is still in progress.
Alizadeh said that a comparison of patients' ctDNA levels after treatment against baseline showed "strikingly" that patients who achieved complete responses experienced rapid reductions in ctDNA levels with some complete responses being measurable as early as 15 days after infusion. This was not the case for patients with partial responses or with stable or progressive disease, he said.
Fifteen days after infusion, around half of the patients who achieved complete responses had completely cleared their ctDNA, whereas 89 percent of the patients who had progressive disease after Breyanzi still had detectable ctDNA.
Among patients who achieved complete radiographic responses, those with detectable ctDNA levels were still at high risk of relapse or death, Alizadeh added. "The numbers are small, but all patients with detectable ctDNA levels, even at month 12, went on to have progressive disease or death," he said. "[This] demonstrates the value of minimal residual disease for disease surveillance and early prediction of a durable clinical benefit with … Breyanzi."
The high risk of persistent ctDNA after day 15 signaling radiographic progression provides an opportunity to try to mitigate that risk in future clinical trials, Alizadeh proposed, with reinfusion strategies and other therapies.
Prognostic score for overall survival
In a third study presented Saturday, Gloria Iacoboni, a hematologist at Vall d'Hebron Hospital in Barcelona, presented data on a new prognostic tool for predicting overall survival among LBCL patients after CAR T-cell therapy progression.
"CAR T-cell therapy fails to achieve durable responses in approximately 60 percent of relapsed or refractory LBCL patients [but] data regarding prognostic factors at the time of progression are scarce," Iacoboni said, explaining that she and her colleagues set out to develop a prognostic tool that could predict overall survival after CAR T-cell therapy progression using easily available markers from routine clinical practice.
The researchers derived their score, called the post-CAR prognostic index, or PC-PI, from what began as 15 total variables, including patient demographics and age; the number and type of prior treatments they received; lab test results from the time of progression, such as blood cell counts; and the time from infusion to disease progression.
To train their PC-PI model, Iacoboni and colleagues collected data from 216 relapsed or refractory LBCL patients who experienced disease progression after CAR T-cell therapy at 12 centers in Spain. To validate the score, they used data from 204 patients treated in three European countries, including the UK, France, and Germany.
They aimed to determine patients' overall survival from the date of their disease progression after CAR T-cell therapy in the context of their PC-PI scores, which ultimately included five of the original 15 markers: ECOG performance status, hemoglobin levels, LDH levels, number of extra nodal disease sites, and the time from CAR T-cell infusion to disease progression. These factors helped researchers group patients into four buckets: low-risk, intermediate-low risk, intermediate-high risk, or high-risk.
In the training cohort, median overall survival was not reached in the low-risk group and was 7.3 months in the intermediate-low risk group, 4.9 months in the intermediate-high risk group, and 1.8 months in the high-risk group. In the validation cohort, Iacoboni noted that the distribution of patients across the four risk groups was similar to the training cohort, and the median overall survivals were 15.2 months, 5.3 months, 2.9 months, and 0.9 months, respectively.
When Iacoboni and colleagues compared the performance of their PC-PI score to the international prognostic index, or IPI, as well as the revised international prognostic index, or R-IPI, they found that the PC-PI outperformed both of these commonly used indexes in predicting overall survival.
"The PC-PI was more discriminative than the IPI and the R-IPI in the post-CAR T-cell therapy setting," she said, referring to the score as "a novel, clinically useful tool for overall survival prediction and risk-adaptive treatment planning in large B-cell lymphoma patients progressing after CAR T-cell therapy."
Iacoboni underscored as a caveat that "the aim of the score was to predict overall survival after CAR T-cell therapy progression, not to predict which treatment would be better for that patient." Even so, she and her colleagues did take a look at the treatments patients received — be they palliative care, chemo-radiation, or immunotherapy — after they progressed following CAR T-cell therapy and observed several correlations between the PC-PI risk scores and these subsequent interventions. They found, for instance, that the palliative care group, as expected, was enriched for patients who were high- or intermediate-high risk, for instance.
As for how this tool could immediately inform clinical practice, Iacoboni suggested that anyone designing a trial that includes patients who have progressed after treatment with CAR T-cell therapy could use this score to risk-stratify patients in the context of the study.