A comparison of four tests, used in clinical trials to determine if a patient will respond to treatment with checkpoint inhibitors targeting PD-L1, found considerable disagreement as to their results — suggesting that patient misclassification could be a major problem in this type of cancer immunotherapy.
The findings are the result of an unusual collaboration involving pharmaceutical giants Merck, AstraZeneca, Genentech/Roche, and Bristol-Myers Squibb; the diagnostic companies Ventana and Dako; research organizations, such as the International Association for the Study of Lung Cancer (IASLC) and the American Association for Cancer Research; and academic research institutions, including the University of Colorado Cancer Center.
The study’s first part, “PD-L1 Immunohistochemistry Assays for Lung Cancer: Results from Phase 1 of the ‘Blueprint PD-L1 IHC Assay Comparison Project’,” was published in the Journal of Thoracic Oncology.
These four tests are used to determine if a tumor is dependent of PD-L1 for its growth and survival, as studies show that only tumors with high levels of PD-L1 respond to this type of immune checkpoint blockade therapy.
Researchers used 38 samples of human non-small cell lung cancer. All the tests signaled that half the tumors were PD-L1 positive and that five were negative. But when it came to 14 samples, making up 37 percent of the available samples, test results varied. While some tests signaled a tumor to be positive, others ruled it was negative for PD-L1.
In other words, in more than one-third of cases, the choice to offer a patient anti-PD-L1 immunotherapy may be flawed.
“Immunotherapy is evolving very fast and with very encouraging results in lung cancer as well as other cancers. However, a main issue is how to select patients for these therapies,” Fred R. Hirsch MD, PhD, an investigator at the University of Colorado Cancer Center and CEO of IASLC, and lead author of the study, said in a news release.
“Each company is pursuing their own predictive PD-L1 assay in order to select patients. However, the PD-L1 assays are all different in terms of antibody used and cut-off values for positive/negative results,” he added.
To get a clear picture of how the tests compare, Hirsch initiated the unique collaboration with support from the U.S. Food and Drug Administration.
In clinical practice, each test is developed and used together with a specific drug, so that each pharmaceutical company only uses its own test before treating patients with the company’s immunotherapy. All of the four drugs — Keytruda (pembrolizumab), Opdivo (nivolumab), Tecentriq (atezolizumab), and the not-yet FDA-approved durvalumab — work in the same way. That is, they block the interaction between the PD-1 molecule on immune T-cells and the PD-L1 protein on tumor cells.
So, in theory, all four tests — measuring the amount of PD-L1 found on the surfaces of tumor cells in patients — should do equally well in predicting the response to all four drugs.
“Rather than black and white, this can be a grey area. It is not that some tumors express PD-L1 and others do not, but rather that tumors express PD-L1 across a gradient and at some cutoff point in that gradient, the expression becomes clinically relevant,” Hirsch said.
While three of the four tests tended to show similar results, the study demonstrated that such a cutoff did not really exist, and the tests had to be evaluated by their own scales.
Research will now continue into the next phase, in which test results will also be compared to patient outcomes, further assessing the validity of the tests.