Why First-Line Nivolumab Is No Better than Chemo

Researchers expected positive results from a trial studying first-line treatment of advanced non–small cell lung cancer (NSCLC) with nivolumab (Opdivo; Bristol-Myers Squibb [BMS]), a PD-1 inhibitor (N Engl J Med 2017;376:2415–26). After all, a phase III trial showed that another PD-1 inhibitor, pembrolizumab (Keytruda; Merck), outperformed chemotherapy as a first-line treatment for the disease (N Engl J Med 2016;375:1823–33). However, nivolumab did not significantly extend progression-free survival (PFS) or overall survival (OS) relative to platinum-based chemotherapy, leaving researchers to explain why the trials came to different conclusions.

The nivolumab findings came as a surprise even to the investigators running the trial. Two phase III trials in patients with advanced NSCLC whose disease had progressed on or after platinum-based chemotherapy had shown that nivolumab increased survival relative to docetaxel chemotherapy. Plus, some patients in the current trial responded well to nivolumab. “We have multiple patients continuing on nivolumab with a continued excellent response, though most of them were originally randomized to chemotherapy and have since crossed over to nivolumab,” says David Carbone, MD, PhD, of The Ohio State University in Columbus, who led the trial.

Nevertheless, among 423 patients with stage IV or recurrent NSCLC who had not previously received chemotherapy, median PFS was 4.2 months for nivolumab vs. 5.9 months for chemotherapy, and median OS was 14.4 months vs. 13.2 months, respectively. Neither of these differences was statistically significant.

The discrepant trial results may lie in the selection of patients based on PD-L1 expression levels. Investigators in the earlier pembrolizumab trial enrolled patients with an expression level of 50% or higher. However, BMS set their cutoff at 5%, based on promising results from an earlier phase I trial, as well as findings from their larger research program using nivolumab to treat various cancers, explains Audrey Abernathy, the company's immuno-oncology communications lead.

This low cutoff may have let too many patients into the study who weren't likely to respond. “BMS was trying to be overly inclusive. They broadened the study population excessively,” says Corey Langer, MD, of the University of Pennsylvania in Philadelphia, noting that marketing concerns may have motivated the decision.

Repeating the trial with a higher PD-L1 cutoff, as in the pembrolizumab trial, is unlikely to happen, explains Edward Garon, MD, of the University of California, Los Angeles: “Now the standard of care for patients with 50% PD-L1 expression or greater would be pembrolizumab rather than chemotherapy. Although a trial to establish the non-inferiority of nivolumab to pembrolizumab is possible, the number of patients required would be large.”

However, in the CheckMate 227 trial, BMS will broaden the array of nivolumab-based treatment options studied, comparing platinum-doublet chemotherapy with nivolumab monotherapy, nivolumab plus chemotherapy, and nivolumab paired with the CTLA4 inhibitor ipilimumab (Yervoy; BMS). “Everyone realizes it is time to move the bar higher, to increase responses over single-agent therapy,” says Elizabeth Jaffee, MD, of the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University in Baltimore, MD.

Going forward, the nivolumab results are likely to affect future PD-1/PD-L1 inhibitor trials by encouraging investigators to stick with higher PD-L1 cutoffs—such as the 50% cutoff used in the earlier pembrolizumab trial—says Langer. Although the pool of patients who could participate in trials would be smaller, and the resulting FDA approval would thus be narrower, the proportion of responders should be higher.

These results have also turned the field's attention to biomarkers of response. First, in an editorial, Garon noted that multiple assays for PD-L1 expression are currently used, impeding efforts to draw comparisons across trials (N Engl J Med 2017;376:2483–5). For example, given a cutoff of 50% PD-L1 expression and the same pool of patients, two assays might not identify the same individuals. A consortium of pharmaceutical and diagnostic companies called the Blueprint Project is currently comparing various PD-L1 diagnostic assays. But “right now, there are financial and intellectual property reasons why companies will not share each other's assays,” says Jaffee.

Second, many researchers believe that patient selection must move beyond PD-L1 expression. PD-L1 expression levels can be difficult to interpret, as they reflect the number of cells present in a sample, how much PD-L1 each cell expresses, and the type of cells present. Jaffee suggests also measuring T-cell infiltration, which is crucial to the therapy's success. “The bottom line is that if T cells are not infiltrating the tumor, then PD-L1 expression does not matter,” she says.

Carbone notes that tumor mutational burden, which can be measured with widely available genomic profiling assays, and expression of PD-L2, another PD-1 ligand, are also promising biomarkers. Further, more sophisticated approaches to patient selection, such as profiling dozens of immune regulatory molecules in a tumor to identify the dominant immunosuppression pathway, could be implemented. “Until we have a really good biomarker or panel of biomarkers to select these patients, I think we will see surprisingly variable outcomes from these trials,” he concludes. –Kristin Harper