Skip to main content
Premium Trial:

Request an Annual Quote

PD-L1 Cutoff for TNBC Keytruda Benefit Confirmed in KEYNOTE Analyses, But Questions Linger

NEW YORK – Combined with chemotherapy, the PD-1 immune checkpoint inhibitor pembrolizumab (Merck's Keytruda) improves overall survival versus chemotherapy alone for advanced PD-L1-high triple-negative breast cancer patients.

Now, a new subgroup analysis from the Phase III KEYNOTE-355 trial has confirmed that PD-L1 with a combined positive score, or CPS, of at least 10 is the best cutoff to predict this benefit.

Javier Cortes, head of the International Breast Cancer Centre in Barcelona, presented updated survival data by PD-L1 expression status during the 2021 San Antonio Breast Cancer Symposium on Tuesday.

It's been slightly over a year since the US Food and Drug Administration granted accelerated approval to pembrolizumab plus chemotherapy for patients expressing PD-L1 with a CPS of at least 10 based on earlier KEYNOTE-355 data. The agency converted the accelerated approval to a full approval this past summer following additional results showing overall survival benefit. Regulators in Japan and the European Union have also followed suit with approvals.

Previous readouts from the trial showed that the immunotherapy combination improved progression-free survival for patients with PD-L1 expression with a CPS of at least 10, but that there wasn't a benefit for patients whose tumors expressed lower levels of PD-L1. It was still slightly unclear, however, whether the CPS of 10 was the best cutoff to ensure patient benefit, in part because the investigators had not analyzed results among narrower groups like patients with PD-L1 CPS of less than 1 or higher than 20.

According to the new data that Cortes presented Tuesday, the median overall survival among patients with a PD-L1 CPS of 10-19 treated with the pembrolizumab combo was 20.3 months versus 17.6 months with chemo and placebo. In patients with PD-L1 CPS of 20 or higher, these median overall survival times were 24 months versus 15.6 months. Because there was a clear benefit in both of these PD-L1-high cutoff groups, the data showed that the overall observed benefit was not driven solely by the patients with extremely high PD-L1 expression above CPS 20.

Meanwhile, in patients whose tumors expressed PD-L1 both with a CPS below 1 and with a CPS between 1 and 9, there was no statistically significant overall survival benefit with the addition of pembrolizumab. Progression-free survival results more or less mirrored that of overall survival in all of the PD-L1 expression subgroups assessed, doubling down on the choice of CPS 10 as the ideal cutoff for pembrolizumab benefit among these advanced TNBC patients.

"CPS of 10 or more is a reasonable cutoff to define the population of patients with metastatic triple-negative breast cancer expected to derive treatment benefit from pembrolizumab plus chemotherapy," Cortes said. "In my opinion, these results provide further support for pembrolizumab in combination with chemotherapy as a new standard-of-care treatment regimen for [this patient population]."

Discordance with early-stage TNBC

In a discussion of the data following Cortes' presentation, Hope Rugo, director of breast oncology clinical trials at the University of California, San Francisco, pointed out a lingering question that became all the more vexing following these new data supporting the PD-L1 cutoff: Why is PD-L1 an important biomarker for immunotherapy response in metastatic triple-negative breast cancer, but not in earlier stages of the disease?

The question was top of mind during the symposium on Tuesday, given the placement of the KEYNOTE-355 presentation right after a separate presentation of the Phase III KEYNOTE-522 trial confirming early-stage TNBC patients' event-free survival benefit with neoadjuvant pembrolizumab plus chemo followed by adjuvant pembrolizumab versus neoadjuvant chemo and placebo followed by adjuvant placebo. In that trial, early-stage TNBC patients benefited from the immunotherapy addition regardless of their PD-L1 expression status, and additional data presented Tuesday confirmed this benefit was consistent in multiple sensitivity analyses.

In her discussion, Rugo suggested that immune differences between primary and metastatic disease may account for this discordance; prior studies of paired primary and metastatic tumor samples, for example, have shown that primary tumors have higher PD-L1 expression generally than metastases, as well as higher interferon signaling and more tumor-infiltrating lymphocytes, among other characteristics that decline as the tumor metastasizes, leading to immune escape.

"This is clearly complicated by mutational complexity under the pressure of treatment," Rugo added, giving a nod to the fact that cancers tend to develop more complex mutational landscapes as they adapt to resist different treatments.

Addressing a similar question following his KEYNOTE-355 presentation, Cortes acknowledged this discordance of PD-L1 utility in the early-stage and metastatic settings. "Of course, these are two completely different populations [and] we still have so many unanswered questions," he said. "We have to explore other biomarkers in detail."

Indeed, although the data from both KEYNOTE trials presented Tuesday were clarifying — one showing that (neo)adjuvant pembrolizumab benefits early-stage TNBC patients regardless of characteristics including PD-L1 and the other showing that PD-L1 expression of at least CPS 10 is necessary for advanced metastatic TNBC patients to benefit from the checkpoint inhibitor — some researchers in the field are hesitant to rely on PD-L1 as a foolproof immunotherapy biomarker for TNBC. As Rugo noted in her discussion, PD-L1 is an "imperfect marker, but the only one we have at present." She also pointed out that the different tests that have been widely used to define PD-L1 positivity in breast cancer, for example, don't overlap perfectly.

Previously, moreover, researchers, such as Justin Balko of the Vanderbilt University Medical Center, have highlighted that PD-L1 expression can play a role in response to chemotherapy as well as immunotherapy, complicating its value in guiding immune checkpoint inhibitors specifically. The key to homing in more exactly on the population of TNBC patients most likely to benefit from immune checkpoint inhibition will be to validate a biomarker with unique sensitivity to the immunotherapy portion of treatment. Efforts are ongoing to identify a biomarker like this. Balko, for instance, published research this past summer suggesting that MHC-II expression could be promising.

Even though pembrolizumab plus chemo is now an approved option for both early-stage and metastatic TNBC patients, the search to identify a better biomarker than PD-L1 must continue, researchers have noted, because there is significant toxicity — both financial and physical — that come with treating patients with checkpoint inhibitors, which should be avoided in patients unlikely to benefit.