Karen Titus
November 2016—Progress is a complicated minuet. One popular adage talks of “one step forward, two steps back,” which is not only discouraging but, in an even less-gleaming light, happens to be the title of one of Vladimir Lenin’s books, published in 1904.
A more optimistic version (and one less centered on the crisis facing communists in turn-of-the-century Russia) suggests advances occur with two steps forward, mitigated by only one step back.
With PD-L1 testing, matters have moved several steps ahead in recent months. But given the complexities that chaperone this marker, these advances have already been followed by steps that could make pathologists feel like they’re moving sideways, or (in proper minuet fashion) even in exasperating circles.
Most recently, results from a phase three trial of pembrolizumab (anti-PD-1) suggest that the drug is a new option for first-line treatment for patients with advanced non-small cell lung cancer and high PD-L1 expression. The study was presented at the European Society for Medical Oncology 2016 Congress in Copenhagen in early October and published online in the New England Journal of Medicine on Oct. 9 (Reck M, et al).
Earlier in the year, the FDA approved atezolizumab (anti-PD-L1) to treat urothelial carcinoma, the first immune checkpoint inhibitor approved for this disease and a decades-spanning leap in bladder cancer treatment.
In both cases, testing for PD-L1 expression has brought excitement—and perhaps mild irritation—inside the laboratory and out as physicians consider all the options.
The approval of atezolizumab was huge news.
“It was the first time in 30 years the FDA has approved a new systemic therapy for bladder cancer,” says Donna E. Hansel, MD, PhD, professor, University of California, San Diego, Department of Pathology, and chief of the Division of Anatomic Pathology. “It’s been a very big deal.”
Now the real work begins.
Antibodies that target PD-L1 or PD-1 operate similarly to other immunotherapy approaches: The goal is to interfere with the pathways that allow cancer cells to hide from the immune system. In urothelial cancers, the blockade of PD-1 or PD-L1 has been shown to reactivate immune response and unmask tumor, says Dr. Hansel, with a number of pharmaceutical companies developing their own targeted inhibitors.
The approval of atezolizumab was based on a phase two trial (Rosenberg JE, et al. Lancet. 2016;387:1909–1920) that looked at use of the drug in 310 patients with locally advanced and metastatic urothelial carcinoma who progressed after treatment with platinum-based chemotherapy. Nearly 15 (14.8) percent of patients experienced at least a partial shrinking of their tumors, lasting from 2.1 to more than 13.8 months. In patients classified as positive for PD-L1 expression, 26 percent of patients had a tumor response, versus 9.5 percent who were classified as negative for PD-L1 expression.
“PD-L1 can be expressed on immune cells and cancer cells,” says Dr. Hansel. “And in this study, the investigators thought that there was an association between immune cell PD-L1 expression and the ability to respond to this antibody.”
(At the time of the drug’s approval, the FDA also approved the Ventana PD-L1 [SP142] assay [Roche] to detect PD-L1 protein expression on tumor-infiltrating immune cells.)
Other agents look promising as well and are being studied, including nivolumab (a PD-1 inhibitor) and durvalumab (a PD-L1 inhibitor). This includes multiple phase three trials underway in first-line, refractory, and adjuvant studies, with and without chemotherapy, as well as phase two single-agent and combination trials. Dr. Hansel says the data look promising so far.
Now might be a good place to cue the complications. Despite the excitement surrounding atezolizumab’s approval and the potential role for other agents that target PD-L1/PD-1, Dr. Hansel urges caution, noting that not all antibodies correspond specifically to each therapy, and that screening criteria vary by antibody. “Tissue testing will range in the cutoffs anywhere from one percent or more, to five percent or more, to 25 percent or more. Each antibody has different cutoffs.” Anti-PD-L1 antibodies include two from Ventana (the aforementioned SP142, as well as SP263), two from Dako (28-8 and 2230), and others.
There are considerable complications when using archived material, Dr. Hansel continues, since the epitope may not be stable. Sampling is another issue; pathologists need to ensure adequate material is available, depending on the cutoff. “It’s been very difficult to try to say it’s only one percent, versus two percent, versus five percent on a limited number of cells.”
Even larger questions loom. PD-L1 is expressed on tumor as well as immune infiltrating cells; in general, expression by immunostaining on either type of cells appears to positively indicate response, but not all trials have shown clinical value, she says.
With each new study, more questions roll in like so many waves: How many antibodies should be tested on the sample? How variable are the results? Does impact differ for patients who have received chemotherapy versus those who have not? Does the therapy indicate a need for testing? How should labs quantify immune cells versus tumor cells? Do immune cells need to be in the center or in the periphery? What is the possible impact of molecular diversity on PD-L1 expression?
“It’s clear that there are many factors that are important,” Dr. Hansel continues.
Early data suggested that even a small amount of expression could be targeted by antibodies and instigate a better response to treatment. “We know now that’s not necessarily the case,” Dr. Hansel says, explaining that in some cases, good clinical response to immunotherapy has been seen even when PD-L1 or PD-1 expression has been shown to be negative by immunohistochemistry.
That has left laboratories with new questions. “Where are we now, for immunohistochemistry, with using these antibodies, which are proliferating at a remarkable rate?” Dr. Hansel asks. “Or do we find alternative methods to see if someone would be responsive to immunotherapy?”
Dr. Hansel points to a few possibilities. One is looking at mutational burden. Mutated proteins differ in appearance from normal proteins and, as neoantigens, can be recognized and targeted by the immune system. “The more you have in your cancer, the more likely your immune system will find something.” One study, the IMvigor-
210 trial, has linked better response to higher mutational burden in urothelial carcinoma.
Even in that scenario, however, the immune cells need access to the cancer. To address that challenge, says Dr. Hansel, researchers are looking at vaccines and trying immunotherapy post-chemotherapy. “You’ve disrupted the cancer cells, exposed these antigens, sensitized the immune system—in a way, you’ve primed the pump to then allow systemic antibodies to target the cancer.”
Past advances in bladder cancer provide at least a scrap of a road map. The disease has long been in the spectrum of cancers where modulating the immune response looked potentially useful, Dr. Hansel says. In an interesting historical aside, she notes that a long-ago study, from the 1920s, showed that patients with tuberculosis had a lower incidence of cancer at autopsy. Based on these findings, researchers proposed using mycobacterium as an anti-cancer therapy. In the 1970s bacillus Calmette-Guérin was used to treat bladder cancer, and ultimately, BCG became the gold standard for treating nonmuscle-invasive high-grade bladder cancer (and was the last major advance in bladder cancer therapy before atezolizumab’s approval, says Dr. Hansel).
At her own institution, which is a major regional referral center, a number of immunotherapy trials are underway. “We’re trying to better understand the antibody operating characteristics and how, in our hands, they play to response. In our CLIA lab we run two different antibodies, and we look at tumor cell expression and immune cell expression.” In the future, they’ll also likely start characterizing immune infiltrates to determine if CD8-positive T cells, FoxP3-positive T cells, or other features can be used to enhance whether a tumor is called positive or negative.
“As these companies come out with more antibodies, there is certainly a push to expand what we offer,” she says. Her laboratory is considering bringing on a third antibody.
What about patients who respond to immunotherapy yet are immunonegative? “We’re trying to figure out if there’s something we can do in lieu of or in addition to the immunostain that can better stratify these patients, so that we’re not overtreating,” Dr. Hansel says. Potential markers for response could also include microsatellite instability and genomic instability as well as the aforementioned mutational burden.
For laboratories trying to sort through PD-L1/PD-1 testing, Dr. Hansel recommends two time-honored steps: following the guidelines that accompany antibody use and performing solid validation. “Epitope stability may vary from lab to lab. Keep an eye on intrinsic factors that may be lab-specific that could alter the operating characteristics of these antibodies.”
Despite the unknowns, Dr. Hansel is upbeat. “From the bladder cancer and urinary tract cancer field, this has been an exciting development. Hopefully we expand on this and understand it better. We still have a lot of work to do to understand whether or not using these antibodies on tissues is really the best way to go about this.”
Anyone who thinks PD-L1 in lung cancer is a simpler tale should be ready for a burst bubble. (Easier, perhaps, to answer the question posed by Lenin, on the eve of the revolution, in his 1917 blockbuster, Will the Bolsheviks Maintain Power?)
Speaking shortly before the European Society for Medical Oncology announcement, Lynette M. Sholl, MD, noted that matters already were complicated. For example, when is PD-L1 testing appropriate for nonsquamous NSCLC? “That’s where things are, at the moment, most contentious,” said Dr. Sholl, associate pathologist and associate director, Center for Advanced Molecular Diagnostics, Brigham and Women’s Hospital, Boston. Efficacy has been shown in both squamous cell carcinoma and in adenocarcinoma. But the squamous cell trials of checkpoint inhibitors as second-line therapy have not, to date, shown a strong enough correlation between benefits of therapy and extensive PD-L1 expression to justify the addition of a companion diagnostic. Adenocarcinomas, on the other hand, have shown a stronger and statistically significant correlation between extent of PD-L1 expression and response.
Until now, the focus had been on second-line treatment. Anticipating the new study, Dr. Sholl, who is also an assistant professor, Harvard Medical School, said, “So it becomes only more complicated in a couple of days!”
Quite right. The study showed the efficacy of pembrolizumab versus chemotherapy in the first-line setting for patients with advanced NSCLC expressing PD-L1 in more than 50 percent of tumor cells (with median progression-free survival of 10.3 months for patients receiving pembrolizumab versus six months for chemotherapy). “In that scenario, we will probably need to get PD-L1 testing for both squamous and nonsquamous.”
The study, Dr. Sholl says, will likely create a fourth category of patients who receive a type of targeted first-line therapy. Oncologists will, up front, look at EGFR, ALK, ROS-1, and PD-L1 status in patients with nonsquamous, non-small cell lung carcinoma.
Dr. Sholl calls this a “clean and easy” way to think about such patients, since subgroup analysis from the original pembrolizumab lung cancer trials showed that patients with EGFR and ALK alterations did not benefit as much from immunotherapy. Another study by a Massachusetts General Hospital colleague, Justin Gainor, MD, published earlier this year, showed similar results, she says. (In fact, the study published Oct. 9 in the New England Journal excluded patients with EGFR and ALK from enrollment. “A number of immunotherapy trials are doing that now,” says Dr. Sholl.)
At her institution, “We’ve already instituted reflexive testing of PD-L1 in all our advanced-stage lung cancer patients,” including those awaiting first-line therapy and those who’ve relapsed after progression. “We’re trying to catch as many as we can. And we sort of get help from our oncology colleagues. If there’s somebody we didn’t perform PD-L1 testing on up front, they’ll gently remind us,” she says with a laugh.
Even without the news from the European Society for Medical Oncology, laboratories can be forgiven for thinking PD-L1 testing might benefit from a KonMari decluttering effort. At Brigham and Women’s, Dr. Sholl and her colleagues, in considering testing for second-line therapies, looked at the different platforms that have been approved along with the associated companion diagnostics.
“We decided it was too cumbersome and too expensive for us to bring on the range of assays that have been approved,” she says, calling it a business decision. Her laboratory doesn’t operate as a reference lab, which made it difficult to justify the cost of the companion diagnostic, especially “when we know we have a biologically robust and validated laboratory-developed test.” They use the E1L3N antibody, which has been optimized to a variety of platforms.
The laboratory uses Hodgkin’s lymphoma samples for its controls, since they’re known to have a very high rate of PD-L1 overexpression. “That’s a function of amplification of the gene that encodes for PD-L1,” Dr. Sholl explains; moreover, the vast majority of those patients respond quite dramatically to anti-PD-1 or anti-PD-L1 therapies. The lab has also done cross-comparisons with other laboratories to confirm readouts are being scored in a similar fashion. “We’ve done studies of [tissue microarrays] looking across two hospitals, run with the same antibody on different platforms, read out by different pathologists, and have shown very good concordance.”
In short, Dr. Sholl and her colleagues are confident they’re delivering reproducible and biologically validated results.
That leaves other issues, naturally.
“The challenge with what we have to offer now is that it’s not validated in the clinical trial setting,” Dr. Sholl says. Consequently, the test result is treated as an additional piece of information. “Often the clinicians are looking at a variety of different data points at once,” she says, noting that the lab offers a comprehensive genomic sequencing panel that looks at over 400 cancer-related genes. Clinicians may be deciding between a targeted therapy involving less common targets BRAF or ERBB2 (HER2), for example—both of which are implicated in lung cancer—versus immunotherapy. If PD-L1 expression is strong, but information about a particular gene variant is less clear, “that may push them toward the immunotherapy route.
“Then it’s in the oncologist’s court,” Dr. Sholl continues. “Honestly, they may or may not ask us to even perform that [PD-L1] stain.” She traces the oncologist thought process: I don’t see a great biomarker here for response to another targeted therapy. This patient can get nivolumab commercially. I don’t need to know the PD-L1 status. I may not even be able to get my hands on more tissue to get a stain.
In these scenarios, oncologists might order a PD-L1 stain after the fact, “to see if that can help explain why the patient had such a great response to the drug,” says Dr. Sholl. “Again, it’s a point of information.” And while some providers will send out the test to another reference lab or commercial lab prior to administering therapy—sometimes they’ll do that in the context of a clinical trial, she says, or if they have a particular interest in putting a patient on pembrolizumab versus nivolumab—in her experience, “A lot of the oncologists take the path of least resistance, given that they already have another commercially available agent that seems to have similar efficacy in lung cancer patients.”
Dr. Sholl is quick to point out that at her institution the circumstances may be a bit unusual, given the comprehensive sequencing available for all cancer patients. At the same time, most practitioners are likely operating with knowledge of EGFR and ALK status, and often with ROS-1 status as well, in addition to PD-L1.
Is PD-L1 testing difficult? “The literature has been a little bit back and forth on this,” says Dr. Sholl, noting, for example, “Some studies show that there is poor interobserver agreement on calling tumor percentages.”
A couple of factors come into play. Some antibodies target the extracellular domain; others target the intracellular domain, and these may be associated with different patterns of expression. Plenty of studies have looked across the antibodies, Dr. Sholl says, including comparisons of extracellular—versus intracellular—targeting antibodies. “Once you normalize for the staining platform and standardize the way you’re treating the specimens, you minimize a lot of the variability” at the antibody level. Even though other variables remain—amount of nonspecific cytoplasm expression, intensity of expression—many of the antibodies perform in a similar fashion once the platform and detection system are standardized, she says.
And what about that interobserver variability? “These can be very challenging to read, right?”
Dr. Sholl says. When she and her colleagues perform PD-L1 testing, they report out the percentage of tumor cells expressing PD-L1 as well as the percentage of inflammatory cells that are present, either within the tumor or at the tumor-stromal interface.
“So for the most part we can discriminate between an inflammatory cell that is positive versus a tumor cell that’s positive. But it’s not always easy,” Dr. Sholl says. “And putting a number on it—in particular, putting a number on the immune cells—is really hard.” Given that, the correlation studies between pathologists looking at tumor cells versus looking at immune cells “are pretty good when it comes to quantifying PD-L1 on tumor.”
“But they’re pretty lousy when it comes to quantifying PD-L1 on immune cells,” she adds.
At this point, her clinical colleagues aren’t asking for much more than what the laboratory already offers. But from the research perspective, she says, plenty of people are pushing the PD-L1 field forward.
There is, for instance, a great deal of interest in doing routine immunoprofiling in tumor specimens. Several institutions have embarked on multiplex immunofluorescence to understand the relationship between tumor cells and immune cells, and their PD-L1 and PD-1 expression, respectively.
Given the sequencing option offered by her laboratory, physicians can look at mutational burden in a prospective way. If that becomes the norm at other institutions, that could also influence the role PD-L1 testing will eventually play. That points to another, less-recognized aspect of PD-L1 testing: “A lot of folks are just going on immunotherapy because it’s got a low side-effect profile, and it’s effective in a subset of patients without a lot of preselection,” says Dr. Sholl. If a clinician sees that a patient has an unusually high number of mutations, however, that could influence therapy decisions. “In first-line therapy, we’re going to need to be able to discriminate in a much more robust fashion between the likely responders and the nonresponders,” she says. She recalls the heady, early days of EGFR mutations—or, as she puts it, “There was a lot of excitement about putting everybody and their mother on EGFR TKIs initially.” Subsequent trials showed that patients who lacked an EGFR mutation actually fared better on chemotherapy.
Dr. Sholl calls for a much more robust PD-L1 biomarker in the first-line setting. “Or maybe we just need to set the bar a little higher?” she asks. “I think as the data evolves we’ll become better at anticipating who will and who will not respond.” The EGFR story, she recalls, took at least half a dozen years to unfold to the point where “everyone fell in line and agreed we need testing before we put somebody on a TKI.”
“Even though we’ve been working with immunotherapies for several years now,” she continues, “we need some perspective on how long it takes to nail down a biomarker, especially one that’s as complex as PD-L1.” What is the tumor cell doing to increase PD-L1? What is the microenvironment doing to induce PD-L1? “There are so many variables that could potentially affect this. It’s not a one-to-one biomarker: Are you mutation positive or mutation negative?”
Lenin, as it turns out, was (among other things) a prolific writer. He also authored a book titled What Is To Be Done?
Laboratories will be kept busy answering that question for the foreseeable future. The most recent push to use pembrolizumab as a first-line therapy will likely renew demand for PD-L1 testing.
“Labs need to be prepared to have this conversation,” Dr. Sholl says, even if they lack all the answers. “Honestly, we haven’t quite figured out how we’re going to handle this.” Labs might want to follow her lead: “We’re talking right now with our oncologists. And each individual lab is going to need to think about the financial considerations, because the companion diagnostic is significantly more expensive than any LDT.” Indeed, the growth in antibodies and companion diagnostics may be making pathologists feel somewhat testy as they survey the plentiful, and pricey, options that are springing up around them.
Fortunately, she says, pathologists are adept at handling such matters. “It’s important that we just keep our eyes open in terms of improving the biomarkers. That’s where we can offer a lot of value—figuring out ways to offer practical solutions to good biomarker detection in this complex scenario.”
Matters will only get more complicated, she cautions. “We keep moving the goalposts.” New clinical trials are looking at combined immunotherapy in the front-line. Multiple drugs and multiple diagnostics have become the norm in the second line; now they’re heading to the first line.
But if it’s any consolation, Dr. Sholl sees practical reality winning out. “I don’t think we’re expecting anybody to be running all the different companion and complementary diagnostics to decide on the right drug. Ultimately, the decision is going to be driven by side-effect profile, by the patient’s preexisting health conditions, and, potentially, by cost and what their insurance is willing to pay for.”
Nothing revolutionary about that.
[hr]
Karen Titus is CAP TODAY contributing editor and co-managing editor.