AACR 2014

Biomarker identifies melanoma patients who may respond to immunotherapy MK-3475

Dr Adil I. Daud - University of California San Francisco School of Medicine, San Francisco, USA

So our study focusses on the correlative data arising from this phase I trial looking at this anti-PD1 monoclonal antibody MK-3475 in melanoma. Here are my disclosures, I have financial support, research support, from Merck, Bristol-Myers, Genentech, GSK, Pfizer and I’m an advisory board member for the drug companies reported.  I will discuss off-label use of MK-3475 and this was supported by Merck.

So it’s increasingly recognised now that the immune system plays a major role in controlling cancer and one of the ways cancer cells evade the immune system is by using the PD1/PDL1 pathway. Basically what they do is they co-opt this pathway and they have to, in some models, express PD ligand and thereby blocking T lymphocytes that come to kill them. So it’s a kind of signal that blocks T lymphocyte action. One way to defeat this is by using a PD1 antibody and MK-3475 is an anti-PD1 antibody and one of the things about it that’s notable is that it’s extraordinarily potent. It was constructed by grafting variable sequences from a parental mouse antibody which was an IgG1 antibody onto a human IgG4 framework preserving the affinity. So you’ll notice that the KD or the affinity of the antibody for PD1 is preserved and stays at this picomolar level. MK-3475 blocks both PD ligand 1 and PD ligand 2 binding to PD1 so this is complex, evolving field and both ligands are thought to play some role in triggering the PD1 pathway. So MK-3475 blocks both of these.

Objectives of the current analysis were to evaluate tumour PD ligand expression and its relationship with outcome in patients with advanced melanoma treated with MK-3475. Another objective was to evaluate the effects of MK-3475 on peripheral T-cell distribution as a pharmacodynamic marker. So this is a complex trial and has many different sub-parts to it so I just want to take a moment to go through this. Basically, our data derives from the non-randomised cohorts, of which there are five of them, looking at several different drug doses and patients were either ipilimumab treated or ipilimumab naïve, all with advanced melanoma. Some patients from the randomised cohorts at the bottom accrued in late 2012. Cut-off of this current analysis was October 18th 2013.

So, here’s the study design and I think the notable point about the study design is that all patients were required to have a mandatory biopsy within two months of study entry. This biopsy had to be either a core biopsy, a core needle biopsy, or an incisional or excisional biopsy, so an FNA was not adequate, so a fair amount of sample was required. That sample was then formalin fixed paraffin embedded and used for all the immunohistochemistry analysis that I’ll share with you in a minute. The first response assessment was done at week 12 and this was a very rigorous assessment using RECIST 1.1 with independent central review. Patients were treated with one of these three doses: 10mg/kg every two weeks, 10 Q3 or 2 Q3. Patients who responded were allowed to stay on, or stable disease, and patients with progressive disease or unacceptable toxicity were discontinued.

So here are the key eligibility criteria. Basically patients had to have advanced melanoma, measurable disease. If you had previous ipi there was no restriction on therapies; if you had no previous ipi then you could have had a maximum of two prior therapies. Patients had to have an ECOG performance status of 0 or 1 and of course they had to consent to undergo the pre-treatment tumour biopsy. The major exclusions: you couldn’t be on systemic steroids or have active autoimmune disease or untreated CNS mets.

These are just some representatives of the samples using our assay. On the very left hand side is a PDL1 negative tumour and you can see various degrees of positivity in the samples.

Here’s how we arrived at our population, starting off with 195 patients in the training set. Some patients were rejected, some tumour samples were rejected because there was insufficient tumour, excessive melanin pigment which obscured this assay, or no sample was provided or other reasons. 125 patients had evaluable PD ligand expression. Of those, 89 had PD ligand positive with the assay that I’ll just show you, with a cut point that I’ll show you in a minute. 36 had negative tumours. Of the 89, 83 had measurable disease by central review and of the negatives 30 had measurable disease. So that’s a total of 71% PD ligand positivity.

Baseline demographics are kind of difficult in a lot of melanoma clinical trials. I think what’s notable is that 76% of patients were BRAF wildtype and a fair number of patients had advanced disease, what’s called M1c disease with advanced visceral mets. Patients could have multiple lines of therapy, pretreatment ipilimumab. There are a couple of imbalances here which you might have noticed: in the patients who had no prior treatment 47% of them were PD ligand negative while those who had previous ipilimumab you’ll notice that 40% of them are PD ligand positive while only 19% were PD ligand negative. So there are a couple of imbalances in the two arms.

Here’s the waterfall plot and tumour shrinkage is shown on the left hand side and tumour growth is shown on the right hand side. So you’ll notice that a lot of the blue lines, which are the PD ligand positive tumours, are experiencing shrinkage while the yellow lines, which are PD ligand negative, are showing tumour growth.

Looked at another way, if you look at unselected patient samples, 40% of those will have a response by RECIST version 1.1 by central review so this is a very rigorous way of measuring response. If you look at PD ligand positive tumours 49% will have a response versus 13% who are negative. If you look at disease control rate 58% unselected patients, 65% PD ligand positive and 37% negative. Both of these are highly statistically significant in terms of comparing the positive PD ligand patients to the negative PD ligand patients.

Looking at the Kaplan-Meier progression free survival curves, again there’s a major difference between the PD ligand positive patients which are in the yellow curve on top and the PD ligand negative ones on the bottom. Median progression free survival is 10.6 months versus 2.9 months. If you look at progression free survival at six months it’s 57 and 35 or at a year 45 versus 18. This is statistically highly significant, the p-value is 0.0051.

Overall survival not so much. Median hasn’t been reached, partly because of the activity of this drug and somewhat also to the lack of maturity of this data, there haven’t been enough survival events to see but it’s interesting that there isn’t a separation of survival curves.

We also looked at the impact of MK-3475 on the peripheral T-cell population and this is interesting because one of the mechanisms by which it is postulated that the anti-PD1 antibodies work is by causing activation of T-cells, including helper T-cells, or CD4 T-cells, and CD8 T-cells, or killer T-cells. What you can see is that at this six week point, which we picked because the six week point was the first common point for both the two week and three week dosing so we had to pick a point that was common across, we saw 14.6% increase in T-cell activation or CD8 T-cell activation, 15.7% for CD4 activation but no changes in overall numbers. We also didn’t see a correlation between T-cell activation and response. So while this is a pharmacodynamic marker it does not appear to be correlating with response like PDL1 does.

So, to conclude, this is an extraordinarily active antibody in melanoma. A durable objective response of 40% in unselected patients is high for immunotherapy. 71% of evaluable patients had PD ligand positive tumours, as evidenced by staining in greater than 1% of cells. Although the objective response rate was higher in patients with PD ligand positive tumours, responses were observed in patients with PD ligand negative tumours also and actually the quality of the response was just as good in PD ligand negative tumours, it was just that fewer PD ligand negative patients had responses.

Given the high prevalence of PD ligand positive tumours, the clinical utility of PD ligand expression in melanoma is not clear at this point just because unselected patients have such a high level of response. Changes in T-cell subtype distribution as a pharmacodynamic marker support the proposed MK-3475 mechanism of action.

We think that this is a good foundation to build on and we think that ongoing studies looking at larger clinical trials will help clarify the role of PD ligand expression in terms of correlating with response of patients treated with MK-3475. Clinical development of MK-3475 is ongoing, both in multiple solid tumours and hematologic malignancies and these are just a couple of the trials that are currently completed enrolment and we are waiting for results from these. One of these is what’s known as KEYNOTE-002, which is a comparison of MK-3475 at two different doses versus chemo in ipi refractory patients. And also KEYNOTE-006 which is a front line setting comparing MK 10 Q2, 10 Q3 versus ipi at 3 Q3.

I want to take this opportunity to thank my co-investigators and especially patients and their families who have contributed not just in participating in this study but also undergoing the biopsy which, as you know, is a very difficult thing to do in clinical trials. Thank you.