Astrazeneca is today pointing investors focused on its Mystic study towards Imfinzi monotherapy and possible subgroups analyses, but in reality this first-line NSCLC trial is a bust.
Having failed its primary PFS analysis last year, Mystic today showed no benefit for Imfinzi plus tremelimumab versus platinum doublet chemo on the OS endpoint either. Thus, like the flop of Bristol-Myers Squibb’s Opdivo plus Yervoy in the Checkmate-227 trial, the Mystic failure will only increase the grip of Merck & Co’s Keytruda on first-line NSCLC.
Mystic had been overhauled and expanded some time ago to promote various analyses to primary endpoints and give Astra what it kept referring to as “optionality”. But its first analysis failed to show a PFS benefit for the Imfinzi plus tremelimumab combination (Mystic falls at the first hurdle, July 27, 2017).
Both the PFS and OS primary analyses had looked at patients whose tumours expressed PD-L1 at 25% or greater as the first hierarchical test of the combo. The fact both have now failed makes redundant three more stringent PD-L1 expression cuts: PD-L1 expressers at over 5% and 1%, and all-comers.
A third co-primary endpoint, OS for Imfinzi monotherapy, has shown a numerically positive 24% reduction in risk of death in PD-L1 >25% expressers, but caution is required here: firstly the upper bound of this confidence interval exceeds 1.00, and secondly the better result with monotherapy is hard to square unless one argues that the addition of tremelimumab causes patients to die more quickly.
|Readouts of Astra's Mystic study (NCT02453282)|
|Study arm||Primary endpoint||Result|
|Imfinzi + tremelimumab, >25% PD-L1||PFS*||Failed|
|Imfinzi + tremelimumab, <25% PD-L1||PFS||Not relevant; only capable of showing significance if >25% level had been positive|
|Imfinzi + tremelimumab, >25% PD-L1||OS*||Failed (HR=0.85; nominal p=0.202)|
|Imfinzi + tremelimumab, <25% PD-L1||OS||Not relevant; only capable of showing significance if >25% level had been positive|
|Imfinzi monotherapy, >25%**||OS||Numerical benefit (HR=0.76; nominal p=0.036)|
|Note: *first hierarchical analysis; **PFS, a secondary endpoint, was deemed not have shown a benefit.|
It could be argued that, even if the Astra combo had shown a benefit at >25% PD-L1 expression this would have been commercially underwhelming: Keytruda monotherapy is already available in >50% expressers, and will likely be approved soon in the >1% population, based on the positive readout of the Keynote-042 study.
When Astra chose 25% as its first cutoff Keynote-042 had not yet read out, so the group was likely doing anything possible to find a niche for Imfinzi plus tremelimumab. In any case, the Keytruda chemo combo is approved in all first-line NSCLC patients irrespective of PD-L1 expression.
Investors who still think Imfinzi has a chance in this NSCLC setting will look to subgroup analyses Astra hinted at today. These could include the controversial tumour mutational burden (TMB) biomarker on which Bristol is pinning hopes in the Opdivo plus Yervoy filing that will see a US FDA decision by May 20, 2019.
However, Bristol was recently dealt a blow when it revealed that patients’ TMB status was in no way predictive of an OS benefit, so this approval looks like a long shot. Astra had previously said that its upcoming Neptune trial could be redesigned to look at various TMB cuts, and the company might use a subgroup analysis of Mystic to inform such a move.
Neptune tests Imfinzi plus tremelimumab in combination versus chemo, and should read out for OS in March. Even if this study somehow yields a positive result this looks like being too little, too late.