Skip to main content
Premium Trial:

Request an Annual Quote

Cofactor Genomics RNA-seq Assay Outperforms PD-L1 IHC to Predict Immunotherapy Responders


This article has been updated to clarify that OncoPrism combines immune cell quantification and T-cell subtyping. 

NEW YORK – Cofactor Genomics' OncoPrism platform, which combines immune cell quantification and T-cell subtyping, has outperformed PD-L1 testing as a method to identify cancer patients who will respond to immunotherapy.

The San Francisco-based company's platform has shown in the PREDAPT trial that it was twice as accurate in identifying which patients would respond to immunotherapy compared to PD-L1 by immunohistochemistry. At a scientific meeting earlier this month, Cofactor reported data from the trial evaluating OncoPrism against standard PD-L1 IHC testing in patients with head and neck cancer, non-small cell lung cancer, and melanoma.

"We heard from physician interviews time and time again that people were not very happy with what was being used, which was IHC-based PD-L1 testing," Cofactor CEO Jarret Glasscock said. "Those physicians said that they would not actually show the results of the PD-L1 assay to their patients because they felt that they were very inaccurate and did not set realistic expectations." Based on that feedback, Cofactor knew there was a need for better diagnostics to guide immunotherapy decisions.

Recent trial readouts and approvals of checkpoint inhibitors have added to the debate that PD-L1 expression is an unreliable biomarker to predict response to these drugs. For example, Bristol Myers Squibb's immunotherapy combination Opdualag (nivolumab and relatlimab) was recently approved for an all-comer melanoma population in the US but restricted to the PD-L1-low melanoma population in Europe. The European Commission relied on exploratory biomarker analysis in the pivotal study in the PD-L1 low group to make its decision, but the FDA used the primary analysis to approve the broader indication.

"Five years or so ago, the immunotherapy space was on fire, but what we were seeing was that the diagnostics to identify the patients who were going to respond to those therapies were really lagging," Glasscock said. "We were using 100-year-old technology, and [the diagnostics space] was going to have to play catch up in order for the area of immunotherapy to really be practicing precision medicine."

To that end, in 2021, Cofactor began the PREDAPT (Predicting Immunotherapy Efficacy From Analysis of Pre-treatment Tumor Biopsies) trial. The study aims to enroll 1,650 participants with a range of solid tumors to evaluate OncoPrism against PD-L1 testing, though it has initially focused on patients with head and neck cancer.

Patients in the study were screened with both OncoPrism and PD-L1 IHC. The OncoPrism assay analyzes T-cell subtypes and gives patients a score from 0 to 100; patients scoring 46 or above were considered potential responders. The study also stratified patients by PD-L1 expression. Eligible patients were then treated with an anti-PD-1 inhibitor, such as Merck's Keytruda (pembrolizumab), and followed to assess clinical outcomes.

In the initial data presented at the AGBT Precision Health Conference in September, OncoPrism had greater accuracy than PD-L1 IHC testing in predicting which patients would respond. Patients who were predicted to respond to immunotherapy by OncoPrism had a median overall survival of 374 days, compared to 197 days for patients who OncoPrism predicted would not respond. By comparison, patients predicted to respond using a PD-L1 test didn't show a significant difference in overall survival. Those with a PD-L1 combined positive score (CPS) of 20 or greater, a potential indicator of better response, had a median overall survival of 222 days, while those with a PD-L1 CPS of less than 20 had a median survival of 289 days.

The OncoPrism test characterizes the abundance of five T-cell subtypes from tumor tissue samples using RNA-seq models. The assay measures gene expression associated with certain T-cell subtypes: naïve, activated, exhausted, effector memory, and central memory. The amount of gene expression associated with each T-cell subtype is measured and developed into a score from 0 to 100.

In a previous study, published in Nature Scientific Reports in January, Cofactor researchers validated the test's ability to predict response across several tumor types. In head and neck cancer, for example, they found that effector memory T cells were more abundant in tumors of immunotherapy responders and that overall T-cell infiltrate was higher in responders. Based on that finding, they built a multianalyte biomarker using machine learning that could predict the probability of response to immunotherapy. They then analyzed samples from NSCLC and melanoma to develop similar biomarkers.

"We set out on this path to connect the RNA expression to the immune profile of the tumor," Glasscock said. "If we could become more comprehensive in reporting on that immune component and more quantitative, our thesis was that we could do a better job of predicting would-be responders."

The firm hoped to create a test with better specificity for identifying responders to immunotherapy than PD-L1 testing, Glasscock said. In the data presented at the AGBT conference, OncoPrism had a specificity of 72 percent versus 59 percent for PD-L1 IHC testing. The positive predictive value for OncoPrism was 59 percent compared to 46 percent for PD-L1.

OncoPrism demonstrated 75 percent accuracy in predicting would-be responders to checkpoint inhibitors in head and neck cancer, compared to 42 percent accuracy for PD-L1 IHC, based on the PREDAPT results. It also had a lower false-positive rate than PD-L1 IHC, 28 percent and 41 percent, respectively. The two assays had a similar false-negative rate and negative predictive value.

The PREDAPT trial is expected to continue through 2023. The study is currently enrolling patients with 11 different tumor types across 20 sites, Glasscock said. Cofactor's initial focus is on head and neck cancer, and Glasscock expects the firm to make OncoPrism available as a laboratory-developed test in 2023, assuming the data continues to show improvement over PD-L1 testing. The next LDT indication Cofactor will pursue, he expects, will be NSCLC.

Beyond these two tumor types, Cofactor is also studying OncoPrism as a predictive immunotherapy response test for small cell lung, urothelial, gastric, cervical, esophageal, triple-negative breast, liver, kidney, and colorectal cancers in the PREDAPT study.

"We put these pieces together to define what does that immune profile look like for the cohort of patient samples who respond versus the cohort of patients which are non-responders?" he said. "The hope is those distributions don't overlap, that there's a distinct difference, and that's what's playing out in our work. When we say we have 75 percent accuracy, that's showing that there is a distinct difference between those two cohorts in the context of our technology."

Based on the positive data for OncoPrism, the firm has also begun pilot studies with drugmakers that have immunotherapies in clinical trials, Glasscock said, but declined to disclose which companies Cofactor is working with. "This makes sense given the increasing importance of predictive diagnostics to identify the subset of the population that will benefit from the immunotherapies. Numerous trials have failed as a result of missing this component," he added.

As Cofactor moves forward with commercialization of OncoPrism, it has to contend with the easy accessibility of IHC in many hospital labs and its relatively low cost. In its studies so far, OncoPrism has a turnaround time of about 10 days, which Glasscock noted was "on par with other high-complexity LDTs like those from Foundation Medicine, Genomic Health, and Exact Sciences." By comparison, a 2019 real-world data study found that IHC test results are returned to patients, on average, in around three days.

As for cost, a 2016 paper found that IHC tests cost an average of $90 per sample in the US. While RNA-seq assays, like OncoPrism, are more expensive, Glasscock acknowledged, the firm is working with health insurers to better understand how its test could be reimbursed.

While the company is still working out reimbursement details and shoring up support from pharma in clinical trials, the company sees clear demand for its test among oncologists. "We work closely with physicians across the US, both in the 20 hospital systems that have partnered with Cofactor, and more broadly to understand their needs," Glasscock said. "The message we hear consistently is that the turnaround time and cost for OncoPrism fits well within their standards for patient care and will enable them to make vastly improved treatment choices for their patients."