NEW YORK (GenomeWeb) – Researchers from Novartis subsidiary Genopitx recently presented early data showing that a novel quantitative multiplexed PD-1/PD-L1 immunohistochemistry test could better predict immunotherapy treatment response for patients with melanoma and potentially other cancers.
The study, presented last month at the American Association for Cancer Research meeting by Genoptix senior scientist Jennifer Bordeaux, used the approach to measure not only PD-L1 expression in cancer cells, but also the interaction between these cells and PD-1-positive immune cells in formalin-fixed, paraffin-embedded clinical specimens.
Bordeaux and her colleagues also performed an initial clinical validation of the combined assay in a small group of melanoma patients, and found that it was more predictive of response and associated outcomes than measurement of PD-L1 expression alone.
With the rise of an exciting generation of new immune-modulating cancer drugs, researchers are pursuing diverse strategies to develop better ways to identify responders, not only testing for single biomarkers like PD-L1, but also using whole-genome profiling to count the overall landscape of cancer mutations, measuring T-cell diversity, and assaying other markers of immune-cancer interaction.
PD-L1 expression tests, however, have been the mainstay for a number of pharmaceutical companies as they advance these drugs into the clinic, with two so far garnering US Food and Drug Administration approval alongside specific therapies — Dako's PD-L1 IHC 22C3 pharmDx as a companion diagnostic to Merck's Keytruda (pembrolizumab), and Dako's PD-L1 IHC 28-8 pharmDx as the first FDA-approved "complementary" diagnostic alongside Bristol-Myers Squibb's Opdivo (nivolumab).
Spurred by data highlighting a lack of concordance between various IHC-based PD-L1 tests, multiple groups have begun working to try to harmonize the different assays on the market or in development.
However, based on the research presented by Bordeaux at last month's AACR meeting, there could be ways to better predict which patients are more likely to respond to immunotherapies than with tests that measures PD-L1 alone, regardless of how well harmonized they are.
She and her colleagues' new combined PD-1/PD-L1 assay relies on quantifying the amount of interaction between PD-L1-positive cancer cells and nearby PD-1-positive immune cells in a tissue sample to yield a predictive "interaction score."
"The goal of our study was to [develop] a single-test that could provide reliable guidance for selecting of anti-PD1 or anti-PD-L1 therapies across multiple indications," Bordeaux said at the meeting. "Our hypothesis was that in looking only at PD-L1 you are only looking at half the story," she said.
The resulting combined PD-1/L1 interaction score test involves fluorescently staining samples with a combination of anti-PD-1 and PD-L1 antibodies and analyzing them using PerkinElmer's Vectra imaging platform along with algorithms to translate the imaging data into an interaction score.
The group also tested anti-CD8 antibodies as part of a combined approach as they developed the method, Bordeaux said, but CD8-positivity did not appear predictive of therapy response either alone or in combination.
In developing their approach, Bordeaux and her colleagues first established analytical sensitivity, precision, and accuracy using standardized PD-L1 and PD-1 tissue control arrays composed of cell lines and lymphoid organs.
They then went on to test the method on a small group of archived melanoma clinical specimens in collaboration with Vanderbilt University, and have since also tested another independent set from Yale and the University of California, San Francisco.
According to Bordeaux, when the team applied the assay to a cohort of 21 advanced melanoma patients treated with nivolumab or pembrolizumab, the PD-1/L1 interaction score could reliably distinguish responders from non-responders. In contrast, PD-L1 expression alone was not predictive of response in the same samples.
Furthermore, patients with higher PD-1/L1 interaction scores had superior response rates, Bordeaux and her coauthors reported — 78 percent versus only 17 percent in those with lower scores. Responders with higher PD-1/L1 interaction scores also experienced significantly longer median progression-free survival — 177 versus 85 days — and fewer deaths — 22 percent versus 58 percent — compared with patients who had lower PD-1/L1 interaction scores,
In terms of diagnostic utility, Bordeaux and her colleagues also found that their the PD-1/L1 interaction test showed superior predictive power, with a 78 percent positive predictive value and an 83 negative PV compared with PD-L1 expression alone.
According to Bordeaux, she and her team have moved on to additional studies to fully establish diagnostic utility of their test, including following patients whose samples they have already tested to track the relationship between their PD-1/L1 interaction score and their overall survival.
"We are moving forward with a full clinical validation," she said at the AACR meeting. In addition to continuing with melanoma patients, the group also plans to extend what she said are "promising proof-of-concept studies" in additional tumor types.
Genoptix did not respond by press time to a request for details on its commercial plans for a PD-1/L1 test.