When the dust settled on today’s first-line lung cancer showdown at the AACR meeting one clear winner was left standing: Keytruda. With the confirmatory Keynote-189 study showing staggeringly good hazard ratios for survival – beating even those seen in Keynote-021G – the Merck & Co drug looks certain to secure full approval.
This is not to say that Bristol-Myers Squibb’s Opdivo failed; indeed, further data presented at AACR today from the Checkmate-227 Yervoy combination trial were about as good as could have been hoped for. But Keytruda simply has such a strong position that it is difficult to see where Opdivo might now fit in.
Bristol, having earlier claimed a progression-free survival advantage for Opdivo plus Yervoy in patients with high tumour mutation burden (TMB) in the overhauled Checkmate-227 trial, had a dream scenario: that TMB benefit would hold up in PD-L1-negative subjects, and that in Keynote-189 Keytruda plus chemo would either underwhelm numerically or show an effect driven by PD-L1 positives.
Unfortunately for Bristol only one half of that dream came true today. At AACR the group said Opdivo plus Yervoy reduced risk of progression among TMB-high subjects by 42%, while an exploratory analysis of OS showed a 21% reduction in risk of death.
But the day belonged to Merck. In Keynote-189 Keytruda plus chemo reduced risk of death by a staggering 51% versus chemo alone, while risk of progression was cut by 48%. This numerically beat the OS and PFS hazard ratios – 0.59 and 0.53 respectively – seen in Keynote-021G, the uncontrolled trial that served as the basis for Keytruda plus chemo’s accelerated US all-comers approval.
And for Merck it got better: hopes that Opdivo plus Yervoy might have a significant place in PD-L1-negative first-line NSCLC subjects were dealt a blow by Keytruda/chemo’s hazard ratios for OS and PFS in PD-L1-negative patients: 0.59 and 0.75 respectively (AACR preview – an existential crisis for Opdivo, April 13, 2018).
Bristol did tout a 52% reduction in risk of progression among TMB-high, PD-L1-negative subjects, but this analysis comprised just 38 subjects in the active arm; indeed, just 139 subjects given Opdivo plus Yervoy were confirmed as TMB-high, out of a total 1,739 enrolled into Checkmate-227 under its original design.
Bristol’s problem is that Checkmate-227 data effectively comprise a subgroup. The trial had stratified subjects by PD-L1 status, but that design was overhauled with the help of Foundation Medicine, which has developed a TMB assay.
David Fabrizio, leader of cancer immunotherapy at Foundation, said this involved having to go back and reanalyse patients’ tumour biopsies, which were available for only 58% of Checkmate-227’s 1,739 enrolees; 44% of this subset was TMB-high, meaning that the law of small numbers is a major drawback of the redesigned Checkmate-227.
Still, the limitations of relying on PD-L1 status have driven work on novel biomarkers, and the fact that PD-L1 and TMB tend to be mutually exclusive – something supported by Checkmate-227 – makes TMB status important.
Foundation is also trying to head off the problem of non-standardisation. While four different PD-L1 immunohistochemistry tests are used for PD-L1, Foundation has formed a TMB working group, and “we ultimately hope we can publish a paper that delivers best practices for TMB definition and validation standards”, Mr Fabrizio told EP Vantage.
Still, AACR experts were split on TMB's importance. Memorial Sloan Kettering’s Dr Matthew Hellmann, presenting Checkmate-227, said Bristol’s data “validated TMB as an independent and reliable biomarker”, but the trial’s discussant, Dr Naiyer Rizvi, of Columbia University Medical Center, said: “We’re not entirely ready to use TMB on its own; we still need to measure PD-L1.”
Roy Baynes, Merck’s chief medical officer, called TMB an interesting biomarker, but insisted that PD-L1 had performed incredibly well, with a test that was “widely deployed, has a rapid turnaround time, and is inexpensive. TMB is ... an unvalidated biomarker in lung cancer, has not shown survival benefit broadly, and it is an expensive test conducted by very few laboratories,” he told EP Vantage.
Where does this leave Opdivo plus Yervoy in first-line NSCLC? Arguably, if a tumour does not express PD-L1, and the patient cannot tolerate chemotherapy, they might benefit from Opdivo plus Yervoy – at the cost of Yervoy toxicity, of course – if they are TMB-high.
Merck will use Keynote-189 to convert Keytruda/chemo’s accelerated US approval into formal approval, and for an EU filing. Merck was up 3% today, while Bristol stock fell 9%, meaning that some $5bn was wiped from the latter group’s market cap.
A separate consideration is where Keynote-189 leaves Roche. The Swiss group’s Impower-150 study had already reported promising OS and PFS results for a combo of Tecentriq, Avastin and chemo, and today’s AACR presentation focused on new analyses by PD-L1 status.
These showed a benefit across PD-L1 expression, but this was nevertheless clearly driven by PD-L1-positives. It cannot be denied that Tecentriq too might now be sidelined by Keytruda, though Roche might argue that its comparison against chemo plus Avastin set Impower-150 a higher bar than Merck did for Keynote-189.
Mr Baynes disagreed, saying that while Avastin was approved first-line it was not broadly used, and that in non-squamous NSCLC the dominant first-line regimen was Alimta, which is what Keytruda was paired with and compared against in Keynote-189.
|Cross-study comparisons in 1st-line NSCLC|
|Overall survival||Progression-free survival||Enrolled in active arm|
|Keynote-189 (Keytruda+chemo vs chemo; n=616)|
|NR vs 11.3mth||8.8mth vs 4.9mth|
|15.2mth vs 12.0mth||6.1mth vs 5.1mth|
|Checkmate-227 (Opdivo+Yervoy vs chemo; n=1,739)|
|23.0mth vs 16.4mth||7.2mth vs 5.4mth|
|TMB high, PD-L1 negatives||HR=0.48||38|
|7.7mth vs 5.3mth|
|Impower-150 (Tecentriq+Avastin+chemo vs Avastin+chemo; n=1,202)|
|19.2mth vs 14.4mth||8.3mth vs 6.8mth|
|PD-L1 negatives (SP142 assay)||HR=0.77||50|
|8.2mth vs 7.0mth|
|HR=hazard ratio; NR=not reached.|
At least Impower-150 was followed through and analysed as initially designed. While Dr Rizvi remained positive on what Checkmate-227 had shown with regard to TMB he called on Bristol to conduct a confirmatory trial.
What the markets really want to know now is whether, in NSCLC at least, Keytruda can now be said to be a better drug than Opdivo. Mr Baynes said the molecules were clearly different, with different affinities and binding epitopes, but would not be drawn on this central question for lack of head-to-head data.
But he did have one withering putdown for Bristol: “Our trials have been conducted in a very simple way. They asked a very simple question. I would encourage one just to look at [Checkmate]-227 and, if you can figure out what’s going on there, please let me know.”