Mario Sznol, MD, of Yale Cancer Center, New Haven, Connecticut, was formal discussant of the MK-3475 presentations at the American Association for Cancer Research (AACR) Annual Meeting. He alluded to all the factors that complicate assays of PD-L1: the disparate nature of the assays themselves, the type of biopsy, the interval between biopsy and treatment, primary tumor vs metastatic tumor, antibody and staining conditions, automated vs manual readings, and the definition of a positive result with a cutoff point.

“Several prior trials of PD-1/PD-L1 pathway antibodies have shown correlations between PD-L1 expression and response in melanoma and also in combined cohorts of patients with various malignancies, mostly including renal cancer and non–small cell lung cancer, and several studies show that PD-L1 expression is also prognostic,” he continued.

Key Results

The findings in melanoma, reported by Daud et al, showed a strong correlation between PD-L1 expression and response or disease control rate, and PD-L1 expression with improved progression-free survival but no correlation with survival.

“Survival was still better than expected in the PD-L1–negative melanoma patients. These data suggest that the drug may have an effect on survival in PD-L1–negative patients despite the apparent lack of impact on progression-free survival and response,” Dr. Sznol commented.

Because of the drug’s effects on PD-L1–negative as well as –positive patients, “this study suggests that we should not be using [PD-L1] as a predictive assay in the clinic in melanoma,” Dr. Sznol stated.

Regarding the data in lung cancer reported by Gandhi et al, Dr. Sznol said that using the 50% cutoff for PD-L1 “loses 40% of potential responders.” As in the melanoma cohort, no major difference in overall survival was observed between PD-L1–positive and PD-L1–negative patients.

Better Biomarkers Needed

“Multiple assays are in development for PD-L1 by drug companies. It is confusing to know which assay to use. There can be heterogeneity of PD-L1 expression within a single patient. Most of these patients don’t have good alternative therapies. MK-3475 has clear treatment activity in a subset of low PD-L1 expressers, and this is relevant in patients with no other alternatives,” he continued.

An assay for PD-L1 expression might be useful in drug development, and it may be a way to stratify patients in clinical trials to target different response rates and expectations, Dr. Sznol said. It also may be useful in studying combinations of therapies.

“Moving forward, we need better biomarkers. We don’t yet have the technology or capacity to decide whether an immune intervention is going to work or not at present. There are too many unanswered questions…. At this point, I would not use these assays to predict for response in the clinic for individual patients. We will also need phase III trials to determine benefit of anti–PD-1 therapy in thelow [PD-L1–]–negative group,” he concluded. ■

Disclosure: Dr. Sznol served as an advisor or consultant for Bristol-Myers Squibb, Genesis Biopharma, Celgene, and Prometheus Labs.  He received grants for clinical research from Bristol-Myers Squibb Company and owns stock, stock options, or bonds from Genesis Biopharma. Has research support from GlaxoSmithKline, LLC, and Merck.