In August, when BMS announced that nivolumab failed to meet its primary endpoint in Checkmate 026, I brushed it off. The flaw was obvious: Checkmate 026 randomized treatment-naive advanced non-small cell lung cancer (NSCLC) subjects with PD-L1 of 1+% to nivo monotherapy or chemo. The competing Merck pembrolizumab study targeted the same previously untreated advanced NSCLC patient population, with the significant difference that Merck’s patients were enrolled based on 50+% PD-L1 expression (note: the studies used different diagnostics/PD-L1 thresholds).
So BMS over-reached. They went head-to-head with Merck for the broader label (not requiring positive PD-L1 expression for treatment, which dramatically increases the patient population), and hubris makes fools of us all. Like a lot of people, I figured the results would improve on subgroup analyses.
As you’ve seen by now, the results did not improve, and Checkmate 026 appears unsalvageable.
- PFS was 4.2 months in nivo-treated subjects with 5+% PD-L1 expression and 5.9 months with platinum-based doublet chemotherapy (stratified hazard ratio [HR]=1.15 [95% CI: 0.91, 1.45, p=0.25]).
- Overall survival was 14.4 months in the nivo group versus 13.2 months for chemotherapy (HR=1.02 [95% CI: 0.80, 1.30])
Merck handily won the first-line indication, and BMS will likely proceed with a nivo/ipilimuab combo in first-line NSCLC. That data will need to be pretty impressive, because the addition of ipi is going to raise toxicity as well as cost.
Let this be a reminder that there is no shame in a narrow label. You can always add indications. This is something patients don’t understand: a restrictive label is perceived as an example of corporate greed (something about cherry-picking patients; because the evil drug companies make more money by selling … less drug?), and there is zero acknowledgement of what happens when, like BMS, you have to start over after your phase III blows up. It means patients wait even longer for a novel therapy, as opposed to what happens when it just gets to market, where it can be used immediately off-label (sometimes with the blessing of NCCN guidelines [see Perjeta and its first-line label], which virtually ensures that payers will cover the drug). Pembro is on-label for advanced lung and melanoma patients with tumors expressing PD-L1, but I’m aware of it being used in all kinds of indications, with or without confirmation of PD-L1 expression – and getting paid for by insurers. The enthusiasm for PD-1/PD-L1 inhibitors isn’t going anywhere. BMS’s public flogging is punishment for their hubris, nothing more.
The flogging is richly deserved (a phase III is no place for an experiment), but the more generous read is that it was not outside the realm of possibility that Checkmate 026 might have performed okay. It’s not like BMS just rolled the dice on a terrible study design. The company has long insisted that PD-L1 expression does not predict response to treatment with PD-1/PD-L1 inhibitors (see Checkmate 017 for an example, which found minimal difference in efficacy between PD-L1-positive and -negative tumors, a result that was underscored with an 057/017 press release yesterday). There is something to this, and it would be an unfortunate consequence of this fiasco to see it brushed aside: we’re just starting to explore the other markers and factors contributing to immunotherapy response. Tumor mutational burden is suspected to play a role in response to checkpoint inhibitors, and regular old radiation therapy can promote immune activity, potentially sensitizing tumors to immunotherapy. We can revel in our schadenfreude, for now, but there’s work to be done here.