A couple of weeks ago, Genentech announced that interim data were forthcoming (ESMO? SABCS seems a long way off) from IMpassion130, their P3, 900-subject, 1:1 randomized atezo + Abraxane v. placebo + Abraxane study in first-line, triple-negative metastatic breast cancer. To tide us over, they offered up a teaser: the study met its co-primary PFS endpoint in the intent-to-treat and PD-L1+ populations, and the company reported a positive trend in OS in the PD-L1+ population. The language was not subtle. In addition to touting 130 as the “first positive Phase III immunotherapy study in triple negative breast cancer,” Genentech said they were going to “submit to authorities globally with the aim of bringing this combination to people with triple negative breast cancer as soon as possible.”
What kind of data might be backing this up? Merck were slow to combine pembro and chemotherapy, so we don’t have a lot of context for what successful immunotherapy looks like in metastatic TNBC. But let’s mine abstract libraries past for details anyway.
Without chemo, which was initially thought to suppress an immune response if given concurrently with immunotherapy, the results are pretty grim. The most mature, public data we have is from KEYNOTE-086, which was a pembro monotherapy study that stratified subjects by prior therapies and PD-1 status. That was the one where, in previously-treated non-selected patients, we got the response rate of 5% and a PFS of 2.0 months. Ouch. The rest of what’s available from that study is summarized below:
Even when the patients were treatment-naive, they bested the previously-treated patients on PFS by, what, three days? Single-agent pembro has not been able to shake that PFS; the earlier KEYNOTE-012, which enrolled previously-treated, PD-L1+ TNBC patients, had an ORR of 18.5%, but again, a 1.9-month PFS.
Things started looking up in combination with chemo (don’t they always?). Eisai combined eribulin with pembro in ENHANCE-1, where the combo boosted the response rate to 26% and PFS to 4.2 months in a cohort of mixed treatment-naive and previously-treated patients. The atezo experience is limited; a TNBC cohort from a P1 study found a 26% response rate to the single agent in first-line, unselected metastatic TNBC patients, which dropped to 7% when the drug was given 2L or later. Interestingly, there are also results from an all-solid-tumors study that seemed to provide the foundation for IMpassion130. This phase 1b study was an alphabet soup of arms combining atezo with every possible agent in a variety of tumor types, and among nine first-line (of 24 total TNBC) patients in an atezo + Abraxane cohort, there was a 67% response rate. The overall response rate was 42% for the TNBC cohort of that study, as summarized here, along with the other results discussed in this paragraph. The ORR on atezo + Abraxane is an outlier, but otherwise, they’re not so far apart heading into P3:
Table 2. ORR by line of therapy and mono/combo therapy.
Now. I don’t think we’re going to see a 67% response rate in IMpassion130. But when the RRs for a more typical, pre-treated patient population are hovering around 25%, there’s a lot of room for a game-changer. I think we can all agree TNBC is due, and we’re actually expecting a real wave of P3 results:
Table 3. Phase 3 PD-1/PD-L1 trials in metastatic TNBC.
KEYNOTE-119 is the outlier, design-wise (monotherapy v. monotherapy), and the main benefit I can see to conducting it is P3-level validation that it’s okay to have a two-month break from chemo, assuming the forthcoming combo labels are so restrictive and insurers so rigid that patients in later treatment lines can’t get a combo. Because I think we can reasonably expect these combo trials to knock 119 off its feet.
You’ll also notice that IMpower130 and KEYNOTE-355 are virtually identical, not counting the safety lead-in on 355 – though some Googling has me intrigued by the mix of chemo options in 355. The flexibility may have been to secure labeling options, or to appeal more to centers and patients compared to IMpower (on drug costs alone, Taxol is cheaper than Abraxane; you generally don’t lose your hair on carbo/gem, and many patients wouldn’t have had it in the adjuvant setting) but I’m curious about whether that has opened up some vulnerability in the results. Celgene put out a press release a week after Genentech’s announcement where they thanked themselves for the Abraxane and being a “preferred chemotherapy partner” for immunotherapy, which I found kind of endearing. When I checked to see if there was any non-business basis for this comment, I stumbled on a retrospective subgroup analysis from a decade-old P3 study comparing the performance of various first-line chemotherapies in metastatic breast cancer. In TNBC, Abraxane had an edge … over Taxol. It could be a bizarre statistical blip, but TNBC patients did better on Abraxane compared to Taxol (PFS 7.4 months versus 6.4 months; OS 21 months versus 15.3 months), and hormone-positive patients did better on Taxol than Abraxane (PFS 12.2 months versus 9.6). Ixempra screwed everyone, which we already knew from a toxicity/QOL perspective. That trivia aside, could Genentech have opened the door to the slightest possible efficacy edge by virtue of choosing Abraxane over another agent? (Caveat: this analysis looks about sixteen subgroups deep, and thus pretty flimsy.) And unless pembro is secretly planning for a duel at ESMO (?), it looks like Genentech will have the benefit of being first over the line, despite the various KEYNOTEs cooking in TNBC for a long, long time.
We have a lot to watch. When was the last time we had a real race in TNBC?