KEYNOTE-427 Brings Into Question Potential Biomarkers in Clear Cell Renal Cell Carcinoma

In an interview with Targeted Oncology, Scott Tykodi, MD, PhD, discussed findings from the analysis from the KEYNOTE-427 trial in patients with clear cell renal cell carcinoma.

Updated follow-up of cohort A in the KEYNOTE-427 study indicated that pembrolizumab (Keytruda) monotherapy was tolerable with promising antitumor activity in patients with clear cell renal cell carcinoma (ccRCC). There was a trend towards improved overall survival (OS) among patients who had greater reductions in target lesions, as well.1

The overall objective response rate (ORR) was 36.4% (95% CI, 27.4%-46.1%), which included 3 complete responses (CRs) and 37 partial responses. The median progression-free survival (PFS) was 7.1 months (95% CI, 5.6-11.0), and the median overall survival (OS) was not yet reached. The 18-month OS rate was 80.0% and the 18-month PFS rate was 26.6%.

Patients enrolled to this cohort had to have clear cell histology. These data were presented at the 2020 American Society of Clinical Oncology (ASCO) Virtual Scientific Program, and a second study, a biomarker analysis on these data2, was also presented at this year’s virtual meeting.

The objective of this research was to determine if there was an association between the depth of response and OS in cohort A. At the time of data cut-off, the median duration of response had not yet been reached.

In an interview with Targeted Oncology, Scott Tykodi, MD, PhD, a physician with the Seattle Cancer Care Alliance, and associate professor in the Division of Medical Oncology at University of Washington Medicine, and an associate professor in the Clinical Research Division at the Fred Hutchinson Cancer Research Center, discussed findings from the analysis from the KEYNOTE-427 trial in patients with ccRCC.

TARGETED ONCOLOGY: What was the background for KEYNOTE-427, and what were the findings presented at this year’s meeting?

Tykodi: KEYNOTE-427 has been presented previously. It’s a single-arm frontline study of pembrolizumab monotherapy for advanced ccRCC. It has 2 cohorts, so cohort A was clear cell histology only and cohort B was non–clear cell. Cohort A had 110 patients, and the clinical outcomes have been presented previously. The response rate was 36%, including a 3% CR rate. PFS was about 7 months in the total cohort. These are definitely interesting data, and this was the first study looking at monotherapy with anti–PD-1 in kidney cancer, so it established this is a highly effective drug but realizing that the current frontline regimens that are FDA-approved are all doublets, such as PD-1 plus CTLA-4 blockade or PD-1/PD-L1 with a VEGF tyrosine kinase inhibitor (TKI). The focus of the abstract 5024 is now taking this data set with well-established clinical end points and digging deeper and trying to look for markers that might guide you to who are the patients that are most likely to respond. This was a popular topic at this year’s ASCO, with other similar attempts to look at the large clinical databases for biomarker discovery. There are some parallels to point out, but 1 of the techniques here was taking the tissue and doing RNA sequencing, then clustering the gene expression into coherent pathways and looking to see if there is a signature of a biological pathway that associates [with] the outcomes data.

There was 1 signature, out of 11 different signatures that were looked at, that [included] T inflammation, T effector function, [and an] 18-gene panel that did associate with better ORR. There was not a strong signal associated with PFS or OS. The other 10 gene signatures did not have a positive association. The other issue that was looked at was using DNA to do gene panel analysis, then looking at some of the common gene abnormalities to see if there was an association with outcome. Out of a 110-patient cohort, recognizing that only a minority of the patients, if any, had abnormalities, it winds up being a small group. Even though there were some trends for differences in outcomes, none of them had a statistical significance. Perhaps this is material for further study with a larger cohort of patients.

We like to think that [it makes] biological sense that a tumor that is already inflamed and has T cells is energized in the tumor microenvironment could be mobilized to give you an antitumor effect and kill tumor progression by applying PD-1 blockade. That theme has cut across some of the other larger data sets as well.

TARGETED ONCOLOGY: What does the prognosis typically look like in this patient population?

Tykodi: Typically, with the new generation of doublets, most of the large studies are still fairly immature so we do not have a lot of conventional end points. However, the longest running doublet study is the CheckMate 014 study of nivolumab (Opdivo) and ipilimumab (Yervoy), where the median OS in the intermediate- to poor-risk cohort was 47 months. This essentially shows 4-years median OS, which, looking backwards overtime, the median OS with the TKIs is about 2 years, the survival with cytokine therapy is 14 to 16 months at best, so I think everyone is convinced there is progress. It doesn’t feel as fast as we want but these newer generation drugs are certainly better than what we had in the past.

TARGETED ONCOLOGY: What were the findings and implications of these data?

Tykodi: For this particular study, the T-cell inflammation signature is interesting. The question is because people have already looked at PD-L1 on tumors consistently throughout these studies, [in[ a lot of tumors probably, RCC including, PD-L1 is upregulated because of T-cell inflammation and interferon-gamma release in the tumor microenvironment. It is just a different iteration of looking at T-cell inflammation cytokine release in the tumor. These are more complex gene signatures kind of touching on the same biology, and although it is interesting, the question is if it is enough to pick out for checkpoint blockade, then the people who do not have the gene signature or PD-L1 do so poorly that you would convert them to some other therapy. That hasn’t been the case; people who do have the marker still do pretty well, and the comparison data sets do well compared with the targeted drug. We don’t have an emerging biomarker that is splitting our population in a way that is going to change their clinical therapy. It is certainly interesting and trying to understand the biology better is relevant, but there is nothing emerging that will clearly impact our clinical management in the short term.

I think the other interesting feature is the data from lung cancer and melanoma, which are some of the mutated tumors that respond to checkpoint therapy, there has been association between the density of antigens and tumor mutation with outcomes. That has been looked at in our data set as well as the upcoming larger frontline data sets. There does not seem to be an association between neoadjuvant antigens and mutational burden. That is interesting, and there is a lot of excitement about neoantigens, there are cancer vaccines based on identifying cancer neoantigens, there are emerging T-cell therapy platforms talking about harvesting T-cell receptors and genes from the tumor microenvironment and making a cellular therapy product, all built on the back of the concept that the neoantigen recognition is essential for tumor response. RCC does not appear to follow the same rules as lung and melanoma, so I think we need to take a step back and make sure we are thinking more broadly about what is the antigen repertoire in kidney cancer. Neoantigens can certainly still be a part of that response, but they do not seem to be a dominant feature when we analyze the datasets.

TARGETED ONCOLOGY: What are the next steps for this research?

Tykodi: For KEYNOTE-427, some of our patients are frail and we are concerned about giving them doublet therapy, so it comes up from time to time that we want to give a patient just PD-1 monotherapy to start. This provides a nice framework because we know we are giving that patient anti-cancer therapy, we have a dataset to refer to so you know what to expect, and then the pembrolizumab drug has gone forward into adjuvant therapy as a monotherapy approach.

KEYNOTE-427 is a very relevant study. We do not expect pembrolizumab to move forward and get its own indication as monotherapy, but it is proven in combination. We can tailor [regimens] for our patients. It is a very relevant dataset that is definitely instructing what we are doing with our patients in the clinic.

TARGETED ONCOLOGY: What are your key takeaways from these 2 studies?

Tykodi: For the KEYNOTE-427 datasets, we touched on the biomarker analysis, where we only observed the single T-cell association but with the commonality of some of the larger datasets I think I see a theme emerging that preceding T-cell inflammation seems beneficial to mounting anti-cancer response with checkpoint blockade. It is interesting amongst solid tumors that preceding T-cell inflammation is usually a good prognostic factor across most solid tumors, and in kidney cancer, it is a strong negative feature. Why that is the case isn’t completely understood. PD-L1 is also a negative prognostic marker for RCC, but you can convert that phenotype to your advantage with PD-1 blockade, so you can kind of reverse that pathway being turned on, which is interesting.

In the other abstract, we broke up the cohort by depth of response: progressors, patients who had 0% to 30% tumor regression fell into the stable disease category, and regression from 30% to 60%, 60% to 80%, or 80% to 100%. When you start splitting 37 patients that responded into smaller and smaller subgroups, the numbers get tiny. The trend did suggest that patients with a deeper response did better for OS, which I think is a very intuitive outcome that the very good responders had smaller disease burden and having an active therapy in place will likely have better OS. That theme of tumor burden being prognostic or potentially predictive of other data sets have been observed. We need a larger number of patients to tease out more carefully because there is a transition point where if you exceed a certain amount of tumor regression, your OS perhaps looks as good as the complete responders. Everyone loves the CR and we would love to get all our patients to that end point, but unfortunately only a tiny fraction achieved a CR. However, maybe patients who get 60% to 80% regression of disease can have very long control of disease and may begin to behave like they had a CR. It’s more of a hypothesis-generating set as it is relatively small, but it is showing a trend that 1 might expect, and I expect that some of the larger datasets will be looked at in the same fashion.

Reference

1. McDermott DF, Lee JL, Bjarnason GA, et al. First-line pembrolizumab (pembro) monotherapy in advanced clear cell renal cell carcinoma (ccRCC): Updated follow-up for KEYNOTE-427 cohort A. J Clin Oncol. 2020;38(suppl):5069. doi:10.1200/JCO.2020.38.15_suppl.5069

2. McDermott DF, Lee JL, Donskov F, et al. Association of gene expression with clinical outcomes in patients with renal cell carcinoma treated with pembrolizumab in KEYNOTE-427. J Clin Oncol 38: 2020 (suppl; abstr 5024). doi:10.1200/JCO.2020.38.15_suppl.5024