Recently, the US Food and Drug Administration (FDA) has started amending approvals for certain immunotherapy drugs, such as pembrolizumab and nivolumab (PD-1 checkpoint inhibitors), based on the level of programmed death-ligand 1 (PD-L1) in tumor cells. This is because new evidence emerged showing that, in some settings, the drugs’ efficacy is strongly linked to elevated PD-L1 levels and patients with low levels are unlikely to benefit. A couple of tests are available to measure PD-L1: the tumor proportion score (TPS), used in non-small cell lung cancer, and the combined positive score (CPS), used in head and neck, esophageal, cervical, and triple-negative breast cancers. Per the TPS, a score of 50% or greater indicates high amounts of PD-L1; per the CPS, a score of 1 or greater signals high levels while a score below 1 indicates low levels.

In a recent example, the FDA’s Oncologic Drug Advisory Committee voted against the use of checkpoint inhibitors in esophageal cancer with PD-L1 expression less than 1. While decisions like this help advance personalized medicine and save patients from unbeneficial treatment, the issue remains complicated. Below, several ECOG-ACRIN Cancer Research Group (ECOG-ACRIN) leaders share their perspectives.

Julie R. Brahmer, MD, Thoracic Cancer Committee Chair

Lung cancer, in particular, is a disease in which cell PD-L1 expression levels have helped shape the landscape of treatment. If tumor PD-L1 levels are used to restrict immune checkpoint inhibitor use, that will be challenging, and it will require us to design and conduct larger studies. Most of our current subset analyses are not powered to find a statistical difference. Additionally, treatment response and maintenance of response are not solely a matter of PD-L1 levels in the tumor. They involve multiple other factors that are difficult to measure, such as tumor-infiltrating lymphocytes and other cells or cytokines.

However, PD-L1 levels are important to guide clinical care in lung cancer. For example, patients with low tumor burden but high PD-L1 levels, greater than 50%, benefit from single agent immunotherapy. Others with low PD-L1 levels should receive immunotherapy combined with chemotherapy.

If treatments are restricted in the future based on CPS score, then we will need to start designing trials that take that into account. This will subdivide cancer further, which may help personalize treatment—but trials will also take longer to complete, which will slow the pace of improvements in therapies.

Naomi B. Haas, MD, Genitourinary Cancer Committee Chair

A few thoughts come to mind related to this issue. The first is that different companies use different assays to calculate PD-L1 levels. A patient’s CPS score could vary from assay to assay, so there is a lack of standardization. Another concern is that the score can differ based on what is biopsied. Sometimes the entire tumor is available, sometimes it is not. In urothelial cancer, for example, PD-L1 expression is more likely if the primary tumor is biopsied than if a lymph node is biopsied. Further, depending on tumor heterogeneity, PD-L1 expression could change based on what part of the tumor is biopsied. This means that patients who may benefit from checkpoint inhibitor therapy could be excluded simply based on the type of assay used or the location of their biopsy.

Variant histology is another consideration. In urothelial cancer, some variant histology is very sensitive to immune checkpoint inhibitor therapy. First-line treatment with enfortumab vedotin and pembrolizumab was recently proven superior to chemotherapy in metastatic disease. This combination appeared to have efficacy in variant histology too. If the PD-L1 signal is different, however, that could affect this type of approval. Finally, there are cancer types like renal cell carcinoma where treatment response is not well correlated with PD-L1 expression.

If strong data exist against the use of checkpoint inhibitors for particular tumors, then this practice may not affect our clinical trials much. However, if it becomes prevalent, it could affect trial eligibility, willingness of industry to provide study drugs, and inclination of insurance to pay for treatment.

Jedd D. Wolchok, MD, PhD, Melanoma Cancer Committee Chair

The challenge in melanoma is that data from previous trials show that PD-L1 levels do not distinguish responders from non-responders in a binary way. In some diseases it may enrich the population, but in melanoma it’s not a clear enough tool to inform treatment decisions.

There are two risks here: the first is that a patient is excluded from receiving a potentially life-prolonging therapy; the other is that a patient is exposed to the side effects of a treatment that has no chance of helping them. Both should be taken seriously. Another challenge, of course, is the availability of different tests. They are similar, but not identical, so there could be inconsistencies.

If there is strong evidence in certain diseases of zero benefit, then yes patients with low PD-L1 levels should be excluded. In those cases, however, perhaps there are other biomarkers that could help support the decision. For example, we know there is little to no benefit from immunotherapy for patients with non-small cell lung cancer and no history of tobacco use or those with an EGFR mutation. If we could assemble a suite of biomarkers, like PD-L1 in conjunction with a known driver mutation, that would create a strong case.

Biomarkers are important. Even if PD-L1 is not a perfect biomarker, utilizing them and refining them is a valuable goal. We want to be as precise as possible about which therapies we offer patients. The big picture is a positive one—but how we go about it, that is where the work is.