In the controversy over the abbreviated presentation of results from Checkmate-214 at an Esmo press conference there has been little chance to examine the questions posed by the kidney cancer study’s surprising results.
While firm deductions are impossible, it seems that Opdivo plus Yervoy is driving a small subpopulation into complete remission, but that Sutent is better at maintaining a greater proportion of patients in stable disease. Yet there are at least 10 questions that require answers before definitive interpretations can be made – and many could be answered with data that likely already exist outside the public domain (see table below).
Meanwhile, many conclusions remain little more than hypothesis-generating. There are considerable gaps in the data, some as a result of the holding back of items for publication later or a future conference presentation – a frustrating but not abnormal practice.
In Checkmate-214 Bristol-Myers Squibb’s Opdivo plus Yervoy was compared against Pfizer’s standard of care, Sutent, in first-line renal cell carcinoma, ostensibly delivering a highly positive result.
The study achieved statistical significance in two of its three co-primary endpoints, ORR and OS in the intermediate/poor-risk subgroup, missing the third co-primary, PFS, owing to a statistical penalty. Furthermore, significance for OS was maintained in all-comers. And this was largely how the story appeared in the media.
But, as EP Vantage reported last week, after a subgroup analysis was presented at Esmo the story subsequently unravelled. Favourable-risk patients fared much worse on Bristol’s combination in terms of PFS, and there was no benefit, relative to Sutent, in the PD-L1-negative intermediate/poor-risk patients who made up most of the group.
Furthermore, it looks likely that the whole study was rendered positive only by a sizeable effect in a small and well-defined subgroup, with the large majority of patients receiving little or no benefit – or maybe even doing worse – on Opdivo plus Yervoy (Spotlight – Medical disclosure farrago hits Esmo 2017, September 15, 2017).
In addition to the specific questions that this raises, detailed below are several broad issues that at this point have no answers. For instance why is Opdivo plus Yervoy worse than Sutent in favourable-risk patients, when this is precisely the group expected to do best on such a therapy?
And what therapy should be recommended for borderline intermediate/favourable-risk patients? What is the optimum cut-off for PD-L1 status that will best allocate patients to TKI therapy - now most likely Exelixis’s Cabometyx - versus immuno-oncology?
|What is the hazard ratio for OS in favourable-risk pts?||HR>1.0 could confirm Sutent to be the better choice (note HR for PFS=2.18)|
|What is the hazard ratio for OS in PD-L1+ intermediate/poor-risk pts?||Probably better than 0.63, possibly by a large margin, suggesting OS benefit in intermediate/poor-risk pts is driven by PD-L1 positives|
|What is the hazard ratio for OS in PD-L1 negatives (c70% of intermediate/poor-risk pts)?||HR >1.0 would confirm no advantage for Opdivo/Yervoy over Sutent in large majority of intermediate/poor-risk pts|
|Is there any difference between PD-L1- and PD-L1+ pts in terms of PR, CR and SD on Sutent?||If similar to overall intermediate/poor-risk pts this would suggest no relationship between PD-L1 status and Sutent activity|
|What is duration of response for Sutent in PD-L1+ and PD-L1- pts?||If similar to 18.2 mth in overall intermediate/poor-risk pts this would suggest no relationship between PD-L1 status and Sutent activity|
|How do CR, PR and SD for Opdivo/Yervoy compare vs Sutent in favourable-risk pts?||Might help develop hypothesis for unexpected result favouring Sutent in favourable-risk pts|
Finally, there are important considerations as to the continued follow-up from Checkmate-214. Will favourable-risk subjects randomised to Sutent and remaining on study drug be crossed over to Opdivo plus Yervoy as a result of the early efficacy stop?
Subgroup data suggest that this is not something that should be done. It is also still unclear whether Opdivo used second line after Sutent progression might have confounded the OS data. And it is possible for the early efficacy stop to complicate further follow-up or make a conclusive result unlikely.
The lack of hard facts makes it difficult to come to a definitive conclusion about Checkmate-214 at present, although it has not stopped principal investigators from doing so. If anything positive comes out of this it might be that more care is taken in future with data interpretation and use of the soundbite “practice-changing”.