It has become the norm for late-breaking studies highlighted at major medical meetings’ press conferences to be described as “practice-changing”, and at Esmo today this accolade was bestowed on Merck & Co’s Keynote-048 trial of Keytruda in first-line head and neck cancer.
However, the dataset is highly equivocal, and comes on the heels of Keytruda’s failure in Keynote-040, a second-line head and neck study that was meant to have confirmed the drug’s accelerated approval. Today Merck said it was withdrawing its US filing for full second-line registration, and was moving to file for front-line use based on Keynote-048.
Merck has not said in which subgroups of patients such approval would be sought, and anyone looking at the Keynote-048 data to provide the answer might be left unsatisfied. Meanwhile, Keytruda retains its second-line accelerated approval status (Merck’s Keynote falls flat, and that’s just fine, July 25, 2017).
Hard to fathom
The problems with Keynote-048 start with figuring out why some patient cohorts did better than others. The study had three arms – Keytruda monotherapy, Keytruda plus chemo, and chemo alone – and four primary endpoints: progression-free and overall survival, in PD-L1-positive subjects and in all-comers.
In terms of overall survival Keytruda monotherapy beat chemo alone in patients expressing PD-L1 at above 20% and 1%, but not in all-comers. Keytruda plus chemo, meanwhile, worked in all-comers versus chemo alone; but paradoxically the combo showed no benefit in PD-L1 expressers.
None of the active arms or their subgroups showed a PFS benefit versus chemo control. There was also no benefit in terms of overall response, Professor Barbara Burtness, from Yale School of Medicine, told Esmo.
To be fair, where an overall survival benefit has been seen it looks impressive and, theoretically at least, in subgroups where it has not been observed it might yet appear at Keynote-048’s final analysis. Merck revealed back in July that the trial had hit its overall survival endpoint.
|Summary of Keynote-048 (NCT02358031) results|
|Keytruda + chemo||Keytruda monotherapy|
|mOS in PD-L1 >20%||No benefit vs chemo*||14.9mth vs 10.7mth (p=0.0007)|
|mOS in PD-L1 >1%||No benefit vs chemo*||12.3mth vs 10.3mth (p=0.0086)|
|mOS in all-comers||13.0mth vs 10.7mth (p=0.0034)||No benefit vs chemo*|
|mPFS in any group or all-comers||No benefit vs chemo||No benefit vs chemo|
|ORR in any group or all-comers||No benefit vs chemo||No benefit vs chemo|
|Source: Esmo 2018. *Described as non-inferior to chemo, with final superiority analysis yet to be carried out.|
What to make of all this? “Our hypothesis ... was that [Keytruda] has a higher response rate in patients who are PD-L1-rich,” said Professor Burtness. “Chemotherapy might be more necessary for the patients who didn’t express as much PD-L1.”
She suggested that the lack of a PFS benefit might be explained by many progressing patients crossing from Keytruda to chemo and deriving a survival benefit – a known phenomenon. “Early exposure to an immune checkpoint inhibitor improves overall survival. This interacts with chemotherapy either concurrently or subsequently,” she explained.
Another confounding factor might be pseudo-progression, which is known to occur in immunotherapy of this cancer type. Professor Burtness also refused to accept that Keynote-040 had failed, pointing to its numerical survival advantage favouring second-line Keytruda, and putting the lack of significance down to a “quirk of statistical design”.
It cannot be denied that head and neck cancer is a difficult area for immuno-oncology; Astrazeneca’s Imfinzi, for instance, had spent some time on clinical hold in this indication owing to the emergence of bleeding events. Professor Burtness said there had been bleeds in Keynote-048, but that they were equivalent among the study arms.
Commenting on the data, Université Catholique de Louvain’s Professor Jean-Pascal Machiels pointed to long-duration responses seen in Keytruda-treated subjects.
Still, this suggests the influence of as-yet undetermined biomarkers. Professor Burtness said tumour mutation burden would be investigated, and is known to be high in head and neck cancer, though its pattern differs depending on a patient’s human papillomavirus (HPV) status. HPV is one cause of head and neck cancer.
Head and neck cancer is expected to account for $698m of Keytruda’s 2024 sales of $14.4bn, according to EvaluatePharma sellside consensus. Around 85% of all patients with this cancer express PD-L1 above 1%, said Professor Burtness.
The task of unravelling the Keynote-048 dataset now falls to the US FDA.