Interpreting the 'Messy Early Days' of Immunotherapy in Breast Cancer

Kathy D. Miller, MD; Peter Schmid, FRCP, PhD, MD


December 13, 2018

Kathy D. Miller, MD: Hi. I am Kathy Miller, professor and the associate director of clinical research at the Indiana University Simon Cancer Center. Welcome to Medscape Oncology Insights. Joining me today is Dr Peter Schmid, professor of cancer medicine and lead of the Centre for Experimental Cancer Medicine at the Barts Cancer Institute in London. Welcome, Peter.

Peter Schmid, FRCP, PhD, MD: Thank you very much.

Miller: You might be the most popular man in San Antonio this year as the lead of the IMpassion study, which is what we have been waiting for. I think we felt a little left out when it comes to immunotherapy in breast cancer when we talked to our colleagues who deal with non–small cell lung cancer, bladder cancer, and kidney cancer. Tell us about the IMpassion130 trial.[1]

IMpassion130 Trial

Schmid: The IMpassion trial is an important study because it puts immunotherapy on the map for breast cancer, and it provides (in my opinion) clear proof of principle that immunotherapy works. The IMpassion study is a large randomized phase 3 study [using immunotherapy as] first-line treatment for metastatic triple-negative breast cancer. The 902 patients were randomized between atezolizumab (an immune checkpoint inhibitor) and placebo, both in combination with nab-paclitaxel chemotherapy. We have some final data for progression-free survival (PFS), which [was one of two] primary endpoints, and interim data for overall survival (OS), with only about 40% of the events. What we have seen in the analysis published recently and updated here in San Antonio[2] was that there is a significant benefit in PFS in both the intention-to-treat (ITT) population and the anti–programmed cell death ligand 1–positive (PD-L1+) population. However, if you look at the actual dimension of the benefit, the hazard ratio (HR) in the ITT population is 0.8, whereas the HR in the PD-L1+ population is 0.62. When we do the further subgroup analysis, it becomes very clear that patients with PD-L1-negative (PD-L1–) tumors did not derive a benefit in this trial at this time for PFS or OS. The test of interaction was positive. PD-L1 staining in triple-negative breast cancer is very different to other cancer types. In lung cancer, for example, the focus is on tumor cells, whereas in this trial, staining was focused on PD-L1 expression on the immune cells, not on tumor cells.

PD-L1 Staining

Miller: I want to ask you about that difference. We now have at least three immune checkpoint inhibitors used in different diseases that are studied a bit differently. They each have different selection criteria and a different antibody. Some use staining on tumor cells and some on immune cells. Why focus on staining on the immune cells instead of the tumor cells?

Schmid: Breast cancers have a different biology. If you look at the data of IMpassion130, 41% of patients were classified as positive based on immune cells staining at a cutoff of 1%. And only 9% of patients were classified as positive on tumors cells when we looked at this as well. If you look at those patients, most of them are also immune-cell positive. Only 2% of patients were tumor-cell positive but not immune-cell positive.

We also do not have patients, in contrast to lung cancer, with very high tumor cell PD-L1 expression. We won't have those groups of 25+ or 55+, as has been defined in lung cancer. So it is a very different underlying biology. We had to take that into consideration when we developed the assay system for this trial.

Miller: I'm imagining that for a practicing oncologist working with a community pathology laboratory, this now becomes a confusing mess with different antibodies and different tests for different disease settings. It feels like it's going to take a bit for us to get the systems in place to make certain that we are getting the right test in the right setting.

Schmid: I think you are absolutely right. What we have is the result of one trial, and that gives us a relatively clear signal. Obviously, there are other trials ongoing with different checkpoint inhibitors, different biomarkers, and different chemotherapy backbones, and even some with the same checkpoint inhibitor and the same biomarker. We will have to see how those results pan out and then wait for the dust to settle. In my opinion, if you want to use atezolizumab, you do need to use the test that was applied in this clinical trial. Of course, not every pathology lab is using this test at the moment. If we want to adopt this, before we want to consider getting approval for drugs in this indication, one key step is that we need to find out how we can establish testing. That is technically not difficult. Of course, there are operational issues. Initially, I could imagine that different regions in different countries might come up with one central lab, and over time, if this biomarker remains the key biomarker for selecting patients, local pathology labs would gradually adopt the test. If you go back 20 years in time, it reminds me a bit of HER2 testing.

Miller: It sounds like the messy early days of the Wild West of HER2 testing.

Schmid: We need to learn from that experience and not get into the Wild West, as you say. We need to quickly establish some central labs. There are already some central labs, and in a way there were central labs used for the clinical trial. If we want to adopt the therapy, we obviously need to share the experience and training relatively quickly to establish more central labs. I agree 100% with you that the test needs to be in good hands because it's important which antibody we use, what cells we look at, what cut-off is used; and pathologists need to be trained on this particular platform.

Outcomes Data

Miller: I know our listeners are interested in a couple of important questions. The OS data look really interesting but are also really early. When are we expecting to next look at the OS data with more numbers and greater confidence?

Schmid: I hope that we see an update of the OS data possibly in the next 6 months. It's obviously an event-driven analysis and therefore we can only give out predictions of when these events may occur. We presented data with 40% of events, and that is clearly not a mature trial. It leaves us room for speculation and different interpretations at this point in time. But, of course, we need to wait for the second interim and then for the final OS analysis as well.

Miller: It also puts us in a bit of a quandary because we want the data quickly. Then you start feeling evil because that really means you are hoping that patients succumb to the disease.

Schmid: Absolutely. And it's obviously something very positive to say that only 40% of the events have occurred, because ultimately behind each event is a patient. In other words, 60% of patients are, fortunately, still alive—which is important to consider, knowing what the normal course of disease in triple-negative breast cancer often can be.

Elucidating Differences Between PD-L1+ and PD-L1

Miller: I want to ask you about those roughly 60% of patients who were PD-L1–, by this assay, in the immune cells, who really did not seem to derive substantial benefit. Do we have any understanding of the difference in their biology and how we might shift their tumors to be more immunogenic and to take advantage of this?

Schmid: That is a fantastic question, and to be honest, we have just obtained this result. Now we are going through all of the correlative studies; some of the data were presented here in San Antonio by Dr Leisha Emens. We are learning how different the PD-L1+ group is from the PD-L1– group and what other possible drivers there are. We are already working on triplet studies, adding in a third drug that will hopefully help to change the biology, especially in those PD-L1– tumors. A lot of this will be down to having adequate trial design and being able to rapidly evaluate which extra drug can make those tumors slightly more immunogenic, which is the underlying problem.

Miller: Do those PDL-L1+ patients fare differently, even in the absence of an immune checkpoint inhibitor?

Schmid: There is possibly a small difference in outcomes; there may be a small prognostic value. If you look at all four Kaplan-Meier curves in one figure, you essentially see at the top of the curve the patients with PD-L1+ tumors who received atezolizumab. You see patients with PD-1+ tumors who received placebo at the bottom of the curve. And there are two curves completely superimposed, which are kind of in the middle. These are the patients who had PD-L1– tumors either receiving placebo or atezolizumab. If you compare the control/immunotherapy-free arms—which is always tricky to do because it's obviously not preplanned—for PD-L1+ and PD-L1– patients, there is a small numerical difference in there. But I would not overinterpret these results at this point in time.

Next Steps

Miller: What is next for immunotherapy in breast cancer? There are certainly lots of trials in the can that are awaiting results. Where do you see our next step?

Schmid: A lot of these steps have already been made but we have to wait for the results. A lot of data are going to come out in the next 12 to 18 months in the first-line metastatic setting from other phase 3 trials using different immune checkpoint inhibitors, different assays, and different chemotherapy backbones, which is important. Does paclitaxel, for example, or platinum-based therapy drive a similar or better benefit, or no benefit, in combination with immune checkpoint inhibitors? The second big piece of data we are expecting is in early breast cancer. As we both know, there are very encouraging signs from small studies of immune therapy added to preoperative chemotherapy in patients with triple-negative breast cancer. The early pathologic complete response (PCR) rates we saw are encouraging and were enough to convince us to do randomized trials. Two randomized, phase 3 trials are expected to read out in the next 6 to 9 months in early triple-negative breast cancer, evaluating the benefit of adding an immune checkpoint to neoadjuvant chemotherapy. We never can predict the results but we are very eager to see them.

Miller: For PCR or powered for long-term outcome? Because the relationship between those two endpoints is imperfect at best.

Schmid: You are absolutely right. There are two different trials. There is another IMpassion trial, which is relatively small, with just over 200 patients. That is powered because of the small numbers for PCR rates only. If that trial is positive, we all know that it would require a second confirmatory trial, which is already in planning. In my opinion, the second trial, KEYNOTE-522, has avoided this dilemma because it has two primary endpoints. PCR will be the first endpoint that is read out. The second primary endpoint is going to be event-free survival because we need to establish the link between PCR and long-term outcome. We know this exists for chemotherapy in triple-negative breast cancer, but we have yet to prove that it exists for biological therapy or immune therapy.

To Steroid or Not to Steroid

Miller: One last practical question. The IMpassion trial used nab-paclitaxel to avoid steroids. In other settings, they’re commonly given with chemotherapies—a short course of steroids as part of the pre-medication, typically for antiemetic prophylaxis. Is the avoidance of steroids for short-term chemotherapy toxicities really critical?

Schmid: We don't have a definitive answer. When we designed the IMpassion130 trial, we were probably more anxious about this than I would be now. There were limited data on chemotherapy/immunotherapy combinations. I think what we wanted to achieve with IMpassion130 was to design a pure trial. We wanted this first trial to be positive and not have other confounding factors. Taking out the confounding factor of steroids is a big step. Had the IMpassion130 trial been negative and had we used steroids, we would not know whether it was the immune therapy not working or whether it was the steroids. We have a positive trial, so our next job is to work out whether steroids are a confounding factor. I'm less convinced that they are. I've seen many patients respond to other combinations in breast cancer but also in other diseases, and I do think that the impact probably was overestimated at the time when we designed the trial. But I still think it was the right thing to do. We will have an answer in the next few months when KEYNOTE-355 will read out, which is a trial that used three different chemotherapy options: nab-paclitaxel, paclitaxel, or platinum-based therapy. Patients did receive steroids, even in the early phase. We will see whether it has an impact. There are also some data using high-dose steroids in other diseases, where we do see that patients may have shorter outcomes. I think the jury is still out, to some degree. The question is: What is high-dose steroids? What we use for antiemetics I wouldn't necessarily consider as high-dose steroids.

Miller: It's a shorter duration rather than a chronic therapy.

Schmid: I can't, at this point in time, say that it's good to do this. All I can say is, my personal expectation is that this does not play a role as strongly as we thought it might at least 4 or 5 years ago. I do hope that we can move on and use steroids in the future.

Miller: As often happens, an early positive study raises a lot more questions than it answers, so we will look forward to the other studies.

Schmid: It also raises interest, and I think that is important. We saw these amazing data on melanoma and lung cancer. Now breast is one of those cancers where we have been able to demonstrate that immunotherapy works. Of course, we have a lot of work to do to optimize this in the PD-L1+ patients and to find useful strategies in PD-L1– patients, and then we need to apply this to other breast cancer subtypes. A lot of work needs to be done, but we have a very important first step, and that is a positive trial of immunotherapy in triple-negative breast cancer.

Miller: Thank you so much for coming in and taking us through some of these issues. We look forward to the other trials which will start to answer those questions one by one. I want to thank you for joining us. This is Dr Kathy Miller from Medscape.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.