Skip to main content
Premium Trial:

Request an Annual Quote

AstraZeneca Study Finds High Concordance in Comparison of Three PD-L1 Assays


NEW ORLEANS (GenomeWeb) – In a head-to-head comparison funded by AstraZeneca, three commercially available PD-L1 tests showed high concordance in their measurement of protein expression in non-small cell lung cancer samples when researchers were able to align their different cut points, supporting potential future interchangeability of these assays to guide therapy with different immune checkpoint inhibitors.

The results, presented in a poster at this week's annual meeting of the American Association for Cancer Research, contrast with those of some other studies, which have highlighted discordance between different methods for assessing PD-L1. Among those evaluations was a study published in JAMA Oncology in late 2015, in which researchers led by David Rimm of the Yale Cancer Center measured PD-L1 protein distribution in non-small cell lung cancer samples using both conventional immunohistochemistry tests and a novel quantitative immunofluorescence method his team had previously developed.

Overall, their results showed significant heterogeneity of PD-L1 expression within NSCLC tumors as well as clear discordance between the different antibodies used, the authors reported.

Researchers in the new study, led by Marianne Ratcliffe, a diagnostic associate director at AstraZeneca, compared three assays: the Ventana SP263 assay, which was developed for use alongside AstraZeneca's durvalumab; Dako's 22C3 assay, approved by the US Food and Drug Administration as a companion diagnostic for Merck's Keytruda (pembrolizumab); and the Dako 28-8 assay, approved by the FDA as a complementary diagnostic for Bristol-Myers Squibb's Opdivo (nivolumab).

According to Ratcliffe, all three tests assess the percentage of tumor cells whose membranes stain positive for the PD-L1 protein. Each test has its own cutoff point, above which, a tumor is determined to be more likely to respond to the corresponding therapy.

"PD-L1-directed antibodies are emerging as effective therapeutics in monotherapy and in combination in multiple oncology settings," Ratcliffe said in a statement.

"Before our study, we did not know whether the different assays identified the same patients," Ratcliffe said. "Clearly, for the oncology community, this presents a number of issues, including a lack of confidence in being able to identify appropriate patients for treatment with these targeted therapies."

In the study, Ratcliffe and other AstraZeneca colleagues tested approximately 500 tumor biopsy samples from patients with NSCLC.

Comparing results of the three different assays they focused on, the investigators determined that if the different assay cutoff points could be properly aligned, the three tests were actually highly concordant in regard to patient response. In other words, the patient population defined by Ventana SP263 test at the 25 percent cutoff point was similar to the group identified by the Dako 28-8 test at the 10 percent cutoff.  Though the two cutoff points are different, the tests appeared to be measuring the same thing if accordingly aligned.

The results also indicated that there was a high degree of concordance between the SP263 and 22C3 assays if a 50 percent cutoff point was applied in both cases. Overall, Ratcliffe reported that the three tests achieved agreement of more than 90 percent.

Though Ratcliffe said more research will be needed to confirm the findings, the new cutoff point comparison data could help clinicians and researchers compare existing results from clinical studies that have used different tests, or use information from one test to guide treatment decision-making for a different drug approved with a different test. In other words, they provide a rationale for responsibly extrapolating the results from one test to another.

Another ongoing effort aimed at immunotherapy companion diagnostic harmonization is the collaborative Blueprint assay harmonization effort, involving AstraZeneca, as well as several other pharma companies and diagnostic developers.

Ratcliffe told GenomeWeb this week that her study was independent of, but complementaty to that larger harmonization initiative.

"Some of what we are doing is the same … [in that] we are both using the fully validated versions of these diagnostic tests with trained pathologists," she said. "What's different is that we've done a large-scale study looking at 500 samples. We wanted to really address robustly the assay agreement across the whole range, and you need a lot of samples to do that."

In contrast, Ratcliffe said, Blueprint has focused so far on analysis of a smaller cohort of samples — about 40.

One thing Blueprint is expected to be able to inform on that Ratcliffe and her colleagues did not, is a fourth assay, the SP142 antibody test developed by Roche's Spring Bioscience to identify patients more likely to respond to Roche/Genentech's investigational cancer immunotherapy atezolizumab.

Since an SP142 assay is not yet commercially available, Ratcliffe said she and her team could not include it in their study, but they hope to do so in the future. She and her AstraZeneca colleagues did not include any other research-based assays in their comparison either, of which there are a growing number in the cancer community.

However, she said, in taking a first step toward standardizing, or harmonizing the few clinically validated, commercially available assays, she and her team hope to be able to spearhead a benchmark against which newer tests can compare themselves in order to maintain harmonization. "In the future, as people develop more tests it will be up to them to show how they compare with these [more established] assays."

Overall, Ratcliffe said, she and her colleagues hope their results can help contribute to moving the field away from a binary definition of PD-L1 positive and negative.

"It's highly likely we are going to see multiple clinically validated cut points coming to the fore, so at least if we can show the assays are doing the same thing, that reduces one area of question or confusion [for clinicians]," she said.

Overall, the creation of standards that can help researchers to understand how to directly compare the results of different immunotherapeutic studies, or help clinicians to understand how the results of one assay might apply across different drugs, should be a benefit to the cancer diagnostics community, Ratcliffe argued.

With the rise of immunotherapies, the competitive landscape of molecular diagnostics and companion diagnostics has heated up, but it also has become more complicated.

In October last year, the FDA approved its first 'complementary diagnostic,' Dako's PD-L1 IHC 28-8 pharmDx, as a tool to identify non-squamous, non-small cell lung cancer patients who might benefit most from Bristol-Myers Squibb's Opdivo (nivolumab). In January this year, the agency then expanded the intended use for the complementary test to advanced melanoma patients.

Only a few months earlier, the administration had approved the first lung cancer immunotherapy for a molecularly defined subset of patients, Merck's anti-PD-1 drug Keytruda for advanced non-small cell lung cancer patients. Alongside Keytruda the agency approved a different Dako IHC test using the 22C3 antibody.

Although both tests were developed by Dako using the same IHC platform, there are key differences between the assays, the most important being how the two tests define the cutoff point for PD-L1-positive tumors differently, which in turn differentially characterizes the best responder populations for these two drugs.

According to Rimm, the Yale researcher and author of the JAMA study highlighting discordance between PD-L1 assays published last year, these type of discordances illustrate a pressing clinical issue wrought by the cloistered development of immunotherapy companion tests — that an assay for one drug cannot be expected to effectively predict response to others.

Clinicians and pathologists simply don't want to have to order or perform four different assays to ask a single question. And without a system for transposing the results of one assay to the context of a drug developed with a completely different assay, it's not hard to imagine a scenario where they might have to do that very thing.

In addition to Ratcliffe's work and the larger BluePrint effort, Rimm is also co-chairing a separate NCCN-supported study with Bristol-Myers Squibb to better understand how different assays measuring PD-L1 expression correspond to one another.