Rarely have scientific data presentations been so keenly anticipated than the disclosures on two Phase III trials of checkpoint inhibitors in first-line non-small cell lung cancer (NSCLC) that were delivered at Esmo yesterday. Those presentations were of course for the Keynote-024 study of Merck’s (NYSE:MRK) Keytruda and Checkmate 026 trial of Bristol-Myers Squibb’s (NYSE:BMY) Opdivo.
The first was known to be positive, the second negative, as topline results were reported in June and August this year (EP Vantage – Bristol swings for the fences and strikes out, August 5, 2016). At the time, the unexpected divergence in outcome for what were then considered two similar agents led to stock market movements for those companies that were measured in tens of billions of dollars. Today's movement was of a similar magnitude for Bristol, which fell 10% following its disappointing Esmo presentation.
The two studies were largely similar in design except to the extent to which patients were selected by their degree of expression of PD-L1. Merck, wisely as it turned out, focused only on high expressers, or PD-L1 >50%, representing around 30% of all patients, and has achieved a stunning result.
Median progression free survival (PFS), the primary endpoint, was extended by 4.3 months, with a hazard ratio of 0.5 – almost unprecedented for NSCLC – and equivalent to a 50% improvement in the relative risk of progression or death. Data for Keynote-024 were published simultaneously in the NEJM, and show an extraordinary separation of curves for Keytruda, known generically as pembrolizumab, on the Kaplan-Meier plot after about five months (see picture).
By contrast, Bristol-Myers, which recruited patients into its study based on a lower >1% PD-L1 cut-off, but conducted its primary analysis based on a >5% figure, obtained results that were clearly worse than feared. Its PFS at the 5% level trended in the wrong direction and the hazard ratio suggests patients were probably better off on chemotherapy. The Kaplan-Meier plot also confirms that a significant proportion of patients in the study perform better on chemo than on Opdivo, known generically as nivolumab.
Worse still, Bristol-Myers disclosed >50% PD-L1 cutoff figures in an investor presentation last night and these were as bad as those in the broader population. The like for like hazard ratio for PFS for Opdivo was 1.07. Indeed the analysis suggests only one subgroup of 11 where there was a hazard ratio below 1.0, with one, never smokers, achieving an unusual ratio of 2.5 favoring chemotherapy.
Analysts who had speculated since August that exploratory data from CheckMate 026 at higher PD-L1 cut-offs would show effective equivalence for Opdivo and Keytruda will now have to revisit those assumptions. That hope is now looking very doubtful, and it seems unlikely that Opdivo, at least as a monotherapy, will be used outside its indicated second-line setting in NSCLC.
Combination with chemotherapy
Bristol-Myers will have to place its hopes on combinations, in particular with Yervoy, to have any chance of traction in first-line NSCLC. However, combinations of chemotherapy with checkpoint inhibitors are also looking promising, an approach being notably pioneered by Roche with Tecentriq.
This approach also received a significant boost at Esmo yesterday with data from a combination phase II trial of Keytruda in first-line NSCLC. Researchers reported data that the addition of Keytruda to a doublet of carboplatin and pemetrexed boosted overall response rate from 29% to 55% (HR=0.53, p=0.0016). While patients were not selected by PD-L1 expression levels, a higher response rate (~80%) was observed in the combination arm in patients with the highest PD-L1 expression (>50%).
Some 11 phase III studies are underway of checkpoint inhibitors in first line NSCLC (see below), the first of which could start to read out next year. Unfortunately for Bristol-Myers, these are all for its competitors.