Much ink has been spilled trying to puzzle out why Bristol-Myers Squibb’s star immunotherapy Opdivo (nivolumab) failed a key study in non-small cell lung cancer last summer. The stunning setback surprised many across the industry and upended the pecking order in immuno-oncology.
So far, prevailing wisdom points to Bristol-Myers’ decision to test Opdivo in a broader patient population as a fatal misstep that washed out a potentially positive result.
In many ways, though, the failure of Checkmate-026 underscores a crucial question that remains unanswered even as big pharmas pile into the fast-growing field.
Checkpoint inhibitors, such as Opdivo, have delivered exciting and, in some cases, unprecedented responses in patients with skin, lung and bladder cancers. But many patients see little or no benefit over standard chemotherapies. Identifying those patients most likely to respond, then, has become the next hurdle for immuno-oncology.
Clues from Checkmate
Although Checkmate-026 missed its primary endpoint, Bristol-Myers continues to trawl through data from the trial in hopes of gleaning clues to why Opdivo failed to show benefit.
Currently, patients in clinical trials testing PD-1 or PD-L1 inhibitors are stratified by the level of PD-L1 expression - a biomarker thought to predict response. In Checkmate-026, for example, Bristol-Myers sought to show Opdivo’s benefit in patients with greater than 5% PD-L1 expression.
But emerging research appears to show that response to PD-1/PD-L1 inhibition may also be predicated by what’s known as tumor mutational burden (TMB), a quantitative score measuring (roughly speaking) the number of mutations present in a patient’s tumor.
As measured by TMB, the Opdivo arm in Checkmate-026 may not have been as similar to the chemo arm as Bristol-Myers would have liked, according to a retrospective analysis presented at the annual meeting of the American Association of Cancer Research earlier this month.
In Checkmate-026, a little less than 30% of patients in the Opdivo arm had a high TMB, compared to 39% of patients receiving placebo, the abstract shows. Among patients with high TMB scores, both median progression-free survival and objective response rates were higher with Opdivo treatment than with chemo.
The results, taken with other research, hint that TMB may be a more fruitful way of stratifying patients than the standard immunohistochemistry (IHC) approach to measuring PD-L1 levels.
Even still, the analysis should be taken with a grain of salt. Higher mutational burden did not correlate with better overall survival, wrote Umer Raffat, a biotech analyst at Evercore ISI, in a research note in early April.
Collaboration with Foundation Medicine
Bristol-Myers, though, was intrigued enough by TMB to strike a collaboration with Foundation Medicine, Inc., a molecular information company. The partnership is aimed at identifying predictive biomarkers such as TMB in patients enrolled in Bristol-Myers’ immunotherapy studies.
Foundation Medicine calculates TMB by sequencing about 2 million base pairs of a tumor specimen’s protein coding region, searching for mutations in about 300 genes. That count can then be extrapolated across the entire coding region, giving a score that represents the total number of somatic mutations in a tumor genome.
"Something like TMB is a good indicator of how likely your immune system has responded to the tumor itself," explained David Fabrizio, immuno-oncology leader at Foundation Medicine. "What we have found is it tends correlate really well with response to the PD-1 or PD-L1 checkpoint inhibitors."
That is because TMB itself is actually a proxy for another biomarker, which Foundation Medicine terms neoantigenic burden, he explained. Neoantigens, while perhaps not responsible for oncogenesis, act as red flags drawing the immune system in to attack tumor cells.
Higher levels of neoantigens make those tumor cells more visible to immune cells, potentially predicting a greater response when checkpoint inhibitors knock out the tumor’s defenses.
Foundation Medicine’s Fabrizio believes TMB will eventually prove itself to be a more valuable biomarker than PD-L1 expression alone.
"[There is] little doubt in my mind where this field is heading; it is to either complement or try to replace [immunohistochemistry] as the companion diagnostic for these checkpoint inhibitor therapies," Fabrizio said.
"While IHC was certainly a very good start, there are complications with the method itself, with the different types of assays that have been adopted that make it difficult to standardize."
Variation in assay type has been one challenge in comparing the approved checkpoint inhibitors against each other. To date, the Food and Drug Administration has approved PD-L1 assays alongside some immunotherapies but so far there is no standard test to measure PD-L1 expression. And each assay may have differences that can affect reported expression levels.
"We found that TMB can identify response independent of your PD-L1 expression and we do get cases where patients are TMB high and PD-L1 low or negative and they still respond," explained Fabrizio. "The reason for that, likely, is in those specific cases probably due to the inability of IHC to determine true expression level."
Currently, Foundation Medicine includes TMB scores in readouts for its Foundation One and Foundation Heme genomic profiles, in addition to other testing methods for PD-L1 expression.
As with other promising biomarkers, much work still needs to be done to validate tumor mutational burden as predictive of response to PD-1/PD-L1 inhibition. Given the potential rewards of solving the puzzle of why only certain patients respond to immunotherapy, though, TMB and other methods are likely to continue to draw pharma attention.
"There is nothing more important for immuno-oncology, in my opinion, than identifying the likely responders," Fabrizio said.