As the immune checkpoint inhibitors progress into more tumor types, earlier lines of therapy and combination regimens, cost concerns are pushing demand for biomarkers to help guide patient selection.

At the American Society of Clinical Oncology annual meeting, held in Chicago May 29-June 2, there was continued debate over the use of expression of the PD-L1 ligand on patients’ tumor cells – the leading marker for the PD-1/L1 inhibitors – as well as focus on emerging biomarkers.

One of the predominant themes of the conference was that although high levels of PD-L1 expression tend to correlate with better response rates across multiple clinical trials, it has significant limitations as a biomarker.

There is a natural continuum in the level of expression – the biomarker is not merely present or absent – and those levels can vary over time and change in response to cancer treatment. Expression differs between the primary tumor and metastases, and it can be limited by immunosuppressive effects of other cells in the tumor microenvironment. Furthermore, there is no standardized testing (it is done with an off-the-shelf assay and different companion diagnostics are in development) and different companies are looking at different cutoffs.

Stratification Strategies

Imperfect as it may be, PD-L1 expression has been the only biomarker game in town and has been part of immune checkpoint inhibitor development programs. The major sponsors have enrolled their clinical trials with stratification by levels of PD-L1 expression as a way to assess effect both in patients with low/no levels and to see if higher expression correlates with better response.

Rocheand Merck & Co. Inc. have made PD-L1 expression a significant focus and are developing companion diagnostics.

The first look at how FDA will handle PD-L1 expression in labeling should come with the expected approval of Merck’s Keytruda in lung cancer. The agency recently accepted the sBLA for NSCLC, setting a priority review deadline of Oct. 2, though approval will likely come sooner. The pivotal data for that submission showed twice the response in patients with high expression as the response rate in the total population in KEYNOTE-001 ([A#00150427007]).

In March, Bristol-Myers Squibb Co.’s Opdivo became the first PD-1 inhibitor approved in lung cancer, for squamous cell lung cancer, but it was cleared for all-comers. Data presented at ASCO from the CheckMate 017 trial in squamous disease showed the biomarker didn’t matter: the survival benefit and overall response rate were independent of PD-L1 expression level, lead investigator David Spigel, Sarah Cannon Research Institute, reported.

The firm’s submission of the non-squamous NSCLC study, CheckMate 057, poses more of a challenge, because there the stratification showed better response rates at higher levels of PD-L1 expression.

With all the data favoring nivolumab regardless of status in CheckMate 017, there is “no place for a biomarker in squamous cell carcinoma,” Keith Kerr, Aberdeen University, stated in a talk assessing the role of biomarkers for immune checkpoint inhibitors in NSCLC.

“On the other hand in non-squamous NSCLC, in the CheckMate 057 study … nivolumab over docetaxel is favored in PD-L1-positive cases in non-squamous cell carcinoma,” Kerr said. “So quite a difference between these two studies using the same drug and the same biomarker, and we have no reason to believe the biomarker assessment was any different, though I’m sure those investigations will be ongoing.”

Still, BioMedTracker’s Robert Jeng expects approval for the broader non-squamous population. “The biomarker results are intriguing but still too far from being a meaningful tool for patient selection in their current form,” he said. “Also, the low-biomarker expressing patients in the study still did perform very similarly to docetaxel [in terms of overall survival], the current standard of care, so it’s not like there was no activity in that population. I think the label will reflect that with an approval for simply second-line non-squamous patients.”

Opdivo did provide an advantage over docetaxel in non-expressers in terms of duration of response, and showed a markedly better tolerability profile (see related story, [A#00150608005]).

Kerr noted that in Roche’s POPLAR study, which included both histologies, there was “a rather similar effect as [CheckMate] 057 of a differential effect depending on level of PD-L1 expression.” He said, “PD-L1 did appear to be a predictive biomarker for effect for atezolizumab.”

The presentation of an interim analysis of POPLAR on May 31 gave the first look at Roche’s lung cancer program for atezolizumab (MPDL3280A). The risk of death was reduced by 23% in the total population. But stratified by level of PD-L1 expression, the risk was reduced 54%, then 44% and then 37% in the lowest level group, “virtually the same” as the control arm, lead investigator Alexander Spira, Virginia Cancer Specialists Research Institute, said.

The firm’s Phase III trial, OAK, is in all-comers but is stratified for PD-L1 expression. Data are expected in 2016.

The PD-L1 testing in POPLAR was “significantly different from all the other trials,” Kerr pointed out. Roche’s diagnostic measures PD-L1 expression on the tumor cells as well as immune cells, and uses four levels of expression (more than the two or three used by competitors).

Is Specificity The Solution?

Roche believes that the greater level of specificity in its proprietary assay, developed with its subsidiary Ventana Medical Systems Inc., will be a better guide to patient selection, not just for single-agent PD-1/L1 use, but to have enough information to potentially identify patients who could respond if given the PD-L1 agent as part of a combination. “This, we think, is very important to determine the future of how we're going to see these agents used in immunotherapy,” Roche Pharmaceuticals Chief Operating Officer Daniel O’Day said at the firm’s June 1 ASCO analyst briefing.

“Our feeling very strongly has been, from the very beginning, that we need to understand the biomarkers of response and lack of response, both to allow us to enrich those patients … that are most likely to respond to single agents, not to prevent other patients from being treated, but to rather understand how best to treat the patients who don't have the positive biomarker,” Genentech VP-Immunology Ira Mellman added. The best option for those patients is likely combination therapy ([A#14150601001]).

Roche sees the universe of data presented at ASCO as supporting its diagnostic-driven strategy. “This year, both from our data as well as from competitor data, it's really starting to appear that that strategy is starting to be supported by clinical results pretty broadly,” Mellman said.

The firm had its VP of biomarker development and companion diagnostics, Garret Hampton, explain why its assay is “different from other folks' assays.” The company set out to see expression in all cell types within the tumor, which allows it to probe deeper and to try to understand “differences from tumor to tumor and maybe some of the differentiations between tumors and why immune cells might be important, why tumor cells might be important.” The results show that expression in tumor cells, as opposed to immune cells, is different between indications and could better correlate with benefit. In some indications tumor cells are important and in others they aren’t, so Roche’s assay should help it better understand response as atezolizumab moves into other cancers.

Should We Stick With PD-L1?

“Is this the correct biomarker? Well, maybe, maybe not,” Kerr concluded. “It is the one we know most about and preclinical studies support us looking at this marker. But there are many other biomarkers on the surface of T-cells regulating their function, and we also know the type and the microanatomical location of immune cells of different types matters for prognosis in certain types of cancer.”

As to why PD-L1 expression was predictive in CheckMate 057 and wasn’t in 017, Kerr suggested that the histology might be behind the differences. It is “not unreasonable to speculate” that in squamous cell lung cancer, where there are more smokers, there might be a greater mutational load. It could be that the immunomodulatory mechanisms are different in these tumors, and that immune status or microenvironment could be affected by differences in prior treatment.

Kerr also suggested that Bristol’s cutoff values “could be too low.” The firm used 1%, 5% and 10%, “but … in [Merck’s pivotal lung study] KEYNOTE 001 the cutoff value that seemed to show a differential effect was actually 50%,” he pointed out.

He concluded that expectations of the biomarker in lung cancer are unreasonable. The cutoff levels are “artificial lines in the sand,” and immune checkpoint inhibition is “not like hitting and knocking out oncogenes,” Kerr said. “I think we have to tailor our expectations of the biomarker accordingly.”

Disparities in testing and cutoff levels also will be problematic. “We have the potential for four drugs and four different tests all purporting to test the same thing,” he pointed out. Merck and AstraZeneca PLC confirmed that there is an industry team working together on standardizing testing, which Bristol, Roche, Ventana and Dako AS are also part of.

Kerr said immunohistochemistry remains challenging, and likened the situation to the early difficulties with HER2 testing. “But with standardization, very good results are possible,” he added.

Exploring New Markers

Still, different markers will be needed for different cancers, and there is research advancing various other markers for immunotherapy.

Thomas Gajewski, University of Chicago, described an “inflamed phenotype” of tumors that have “rich chemokine levels and variable T-cell infiltration, but also contain increased levels of immunosuppressive T-regs and immune checkpoint modulators (and therefore might benefit from therapies designed to inhibit T-regs or immune checkpoint molecules).”

During a session at ASCO about immunotherapy’s potential in genitourinary cancers, David McDermott, Harvard Medical School, suggested the inflamed phenotype could help select patients who would be appropriate for single-agent therapy in renal cell cancer, but “it will be hard to get single-agent to the first-line [setting] without biomarkers.” He thought tumor grade and mutational load, which seems to be correlated to PD-1 outcomes, “may add value” in selecting patients. “The bigger challenge for the field [is] the non-inflamed tumor type, which is the vast majority." Combination therapy might be best there, he said.

As to whether immunotherapy is ready for prime time in bladder cancer, Piyush Agarwal, National Cancer Institute, noted that while “we’re ready with single-agent checkpoint inhibitors,” because of the high mutational burden, researchers are "not yet" ready "to look at combinations and more customized” options.

In lung cancer, Kerr said that a broader immune signature might be of value, and mutational burden would also make sense in this setting.

Merck appears to be taking the lead with these alternative biomarkers approaches. At ASCO, the firm released results of a Phase II Keytruda (pembrolizumab) trial in colorectal cancer stratified by DNA mismatch repair deficiency, a marker of mutational burden, to good effect ([A#00150608002]). “We believe that when you have a higher mutation burden, you’re more likely to be producing new antigens,” Merck Research Labs Senior VP Global Clinical Development Roy Baynes explained in an interview at ASCO.

“We’ve also looked at inflammatory signatures,” he added. “The notion here would be, if you have a lot of inflammation and a lot of active lymphocytes, it’s more likely that’s a milieu that you can do something with.” At ASCO, Merck “rolled out some really nice data around a platform called NanoString, where we’ve been focused on a number of different signatures – gamma interferon, some T-cell receptors, some activation signatures,” Baynes said. The program is yielding “really quite good science looking at evidence of inflammation.” The firm expanded its work with NanoString Technologies Inc. on the eve of ASCO ([A#14150528003]).

Baynes also stressed that there was “tremendous convergence” across the PD-L1 stratified findings in clinical trials presented at ASCO, which will reinforce the use of PD-L1 expression for guiding PD-1/L1 monotherapy decisions in the lung cancer setting.

The newer biomarkers will gain more traction as the agents move into other cancer types and also, as biomarkers identify patients who are unlikely to respond to monotherapy, into combination therapy.

So, while biomarker-guided selection of appropriate patients usually offers a cost containment strategy, with immunotherapy it could have the opposite effect. Patients who aren’t candidates for monotherapy could be candidates for combination therapy, resulting in double the cost – but also likely with increased benefit that will increase the value of treatment.

Building A Better Biomarker: PD-L1 Expression Under Spotlight At ASCO

More from Clinical Trials

More from R&D