Adjuvant nivolumab vs placebo in resected oesophageal or gastroesophageal junction cancer following chemo: Biomarker analysis

Share :
Published: 7 Jul 2023
Views: 70
Dr Ronan Kelly - Baylor University Medical Center, Texas, USA

Dr Ronan Kelly speaks to ecancer about a new biomarker analysis from the CheckMate 577 study that he presented at WCGIC 2023.

The study evaluated adjuvant nivolumab vs placebo in resected oesophageal or gastroesophageal junction cancer following chemotherapy.

He reports that there was an improved disease free survival in inflammatory high and proliferation high patients that received the adjuvant nivolumab.

Dr Kelly hopes that these gene expression signatures can be adopted and brought to the mainstream and assessed in future trials.

The original CheckMate 577 data was presented in ESMO a few years ago, it was then published in The New England Journal of Medicine. The study was basically looking at the role of adjuvant nivolumab post-trimodality therapy in stage 2/3 oesophageal and GE junction cancers that had an R0 resection. What we did in that study was we randomised patients up to about 16 weeks later to get adjuvant nivolumab and there were certain time points when patients started the drug. What we saw was a doubling in the median disease free survival from just over 11 months to 22 months. The hazard ratio there was 0.69. Then as we looked at longer follow-up we saw a directionally improved hazard ratio, decreasing there to 0.67. 

So what we did in this particular data and abstract was we looked at the first comprehensive biomarker analysis of this particular study. Now, remember, this is the only immune checkpoint inhibitor approved in GI cancers in early-stage disease. So it’s a very precious database, it’s the largest biomarker database of its size that exists at the moment. So we did look at the groups to see if there were any specific biomarkers that were indicative of improved efficacy or not.

What did you find with this new analysis?

We looked at a variety of biomarkers; the first ones we looked at were gene expression signatures and they were assessed by RNA sequencing of the baseline post-chemoradiotherapy surgical tissue. We looked at five different categories: an inflammatory gene expression signature; proliferative; M2 macrophages; endothelial and fibroblast. The genes are available in the appendix but each of these gene expression signature scores were generated from the literature and from internal Bristol Myers Squibb data. They had previously been shown in an advanced setting, in the CheckMate 649 trial, to be indicative of an improved survival with the use of nivolumab or ipilimumab. 

So here now, in the early-stage setting, we looked at similar signatures to see if there was a benefit. We divided those into patient groups, into high, medium or low subgroups defined by signature score tertiles. Then we also looked at a select number of relevant genetic mutations or gene mutations, should I say, to see if there was a difference in response based on gene mutation status. Those particular ones we looked at were NOTCH1, CDKN2A, PIK3CA and ARID1A. 

Then finally we looked at tumour mutation burden and microsatellite instability. We did, however, have a unique opportunity, because this is the only phase III study that has an adjuvant IO approved post-chemoradiation, to see if radiation could alter the immune microenvironment to make it more hospitable to an adjuvant immuno-oncology approach. So that was the final dataset that we looked at.

So, if you look at the results, we showed that there was an improved disease-free survival benefit for patients that were inflammatory high and proliferation high, which makes sense because that’s the hypothesis that if tumours are hot they may respond to immune checkpoint inhibitor. Then, similarly, we showed that the lower the M2 macrophage, the fibroblasts and the endothelial signatures, also patients did better. So that was an interesting result that gene expression signatures may be able to predict for disease free survival benefit. 

We did not see that any of those four specific mutations made a significant difference, although the CDKN2A and the PIK3CA mutations, if you were mutated in those, seemed to do better which is something that could, or should, certainly be looked at in future studies to see if that’s an existing phenomenon that’s occurring over and over. 

With regard to MSI high and tumour mutation burden, unfortunately the numbers of patients were very small with those. We only had 13 patients, or 3%, that were TMB high so there really wasn’t enough of a group to be able to make any definitive conclusions. Similarly, only five patients, or 1%, had MSI high tumours so, again, it wasn’t enough to be able to make any definitive decisions. 

Finally, with regards to PD-L1 changes, we looked at the hypothesis, as I said, that chemoradiation could alter the immune microenvironment. Here we had paired the diagnostic tissue from pre-chemoradiation so we had the true baseline PD-L1 score before we intervened at all. Then we looked at the post-chemoradiation PD-L1 score, we looked at that in CPS and TPS. What we showed was that there was a pretty significant dynamic change in PD-L1 expression at the CPS level with 51% of patients having a higher PD-L1 CPS score post-chemoradiation. 

Now, does that make a difference? Well, what we showed was, again it was a small number of patients, it was approximately 80 patients. So in those patients that had CPS upregulation post-chemoradiotherapy, compared to what their PD-L1 score was pre-radiation, we actually saw a very interesting disease benefit from the use of subsequent nivolumab with a hazard ratio of 0.30, which was probably the best disease free survival benefit we’ve seen in this subgroup. So, again, hypothesis generating but this is one of the first studies, in fact, in my mind it’s the first in a phase III setting, to be able to show that chemoradiation may alter that immune microenvironment. So, again, we need to look at that in future prospective studies.

What are the implications of this biomarker analysis?

What we know for certain is that adjuvant nivolumab is a new standard of care in resected patients post-trimodality therapy for oesophageal and GE junction cancers. So that exists now across numerous mutations. TMB high, MSI high, were too small but, again, there’s nothing that would indicate that that would not be an optimal strategy for those patients if they didn’t get an immune checkpoint inhibitor in the neoadjuvant setting, for example, which is being investigated at the minute. 

We showed that the higher inflammatory and proliferation gene expression signature score was beneficial and a lower M2 macrophage, endothelial and fibroblast gene expression score. So hopefully these gene expression signatures can now be adopted and brought into the mainstream and assessed prospectively in future trials as we try to move away from just simply relying on PD-L1 to determine who is likely to benefit. 

We need to keep an eye on whether the CDKN2A or the PIK3CA does result in improved survival, that was a hint here we saw. So we need to look at that moving forward. Then the role of radiation in altering the immune microenvironment – a fascinating hypothesis, we’ve never been able to show it in a phase III setting. Here we showed hazard ratio of 0.30 if you had an upregulation of your PD-L1 CPS score. It wasn’t so much for TPS but there’s a difference between the tumour proportion score and the combined positive score in that tumour proportion score is more driven by oncogenic changes within the tumour whereas the CPS score is driven by the cytokines and inflammatory milieu that can result post-radiation. So it just needs to be looked at in future studies. 

Anything else to add?

Again, it’s small numbers of patients so we need to take the interpretation of these results carefully and the clinical utility of these biomarkers and the optimal cutoffs will need to be validated in future studies. But, as per the norm for these biomarker presentations, they are hypothesis generating but this, as I said, is probably the most robust and largest biomarker dataset we have in this particular disease setting. So the results are quite interesting indeed.