Alpelisib Summits Mt. PI3K

ESMO was a snooze, right? No new drugs out of nowhere, no hotly-anticipated studies knocking it out of the park (cough, IMpassion, cough). Which I guess frees us up to talk about alpelisib.

Novartis announced results from their SOLAR-1 study, which succeeded where buparlisib (also NVS) and taselisib (Roche) failed. In patients with PI3K-mutated, hormone-positive and HER2-negative advanced breast cancer, alpelisib plus fulvestrant achieved a PFS of 11 months, compared to 5.7 months for fulvestrant alone, and an ORR of 36% for alpelisib + fulvestrant, which compares to 16% in fulvestrant alone. These are hormone-positive patients, so we won’t have OS data for another few decades, but these are good numbers. It wasn’t Ibrance/letrozole, which blew out a wild PFS in the first-line setting (notably, the SOLAR-1 patients will have progressed on an AI as well as possibly a CDK 4/6 inhibitor), but it’s good, and the first real validation we’ve seen of PI3K targeting in breast.

This would have been a perfect announcement if not for the AEs. Hyperglycemia was reported for 64% of alpelisib patients (compared to 10% among controls); in 37%, hyperglycemia was grade 3/4. Grade 3 hyperglycemia is defined as blood sugar of 250-500 mg/dL; grade 4 is >500 mg/dL. Which sounds challenging, even in patients who theoretically have bigger problems. These are patients that are poised to live a really long time with a good quality of life. Is any onc out in the community, dealing with real patients, psyched about managing blood sugar of 400? The press release didn’t break out the rate of dose reductions, but what does that inevitability do to efficacy? 

To their credit, Novartis has been fairly relaxed about enrolling diabetics in its PI3K studies. SOLAR-1 excluded type 1, but allowed controlled type 2; BELLE-2 and BELLE-3, the buparlisib studies in HR+ MBC, didn’t appear to restrict based on diabetes or blood glucose, though other, investigator-led studies did. SANDPIPER, the P3 taselisib, excluded Type 1 or Type 2 diabetes requiring anti-hyperglycemic medications. Still, not everyone is going to be represented in the SOLAR-1 population, and there’s a reasonable likelihood that physicians will be prescribing for patients with high baseline blood glucose or less well-controlled diabetes. (And while there’s an argument that metformin and PI3K/AKT/mTOR inhibitors are synergistic, and that by adding metformin to one of these drugs you’d get both protection against a blood sugar crisis and enhanced anti-cancer efficacy, there’s at least some evidence with everolimus that it’s not so straightforward. Though these diabetic neuroendocrine patients who were on metformin when starting evero did better than their non-diabetic counterparts.)

Everolimus is an older, HR+ breast-approved mTOR inhibitor that popped up in a couple of ESMO abstracts (here and here) that aren’t really interesting enough to discuss on their own but that roughly echoed what we know from BOLERO-2, which is that evero doesn’t really care about PI3K positivity. This is notable since this drug for patients with targetable, frequently-mutated PI3K has been one of the prime examples of precision medicine in practice. It’s been offered to me a couple of times based on genomic sequencing (never in a large center, obvs), which I always declined because I was waiting for one of the next-gen PI3Ks. Taselisib turned out to be a disaster, despite the all the strategizing that preceded signing that consent. I think this is a good example of the kind of hold that sequencing has over patients; those mutations feel a lot like capital, and I put a huge amount of effort into deciding when and how to cash it in. Diabetes aside, if alpelisib works, and exclusively in PI3K-mutated patients (there is PI3K-wild-type cohort in SOLAR-1 being tracked for a secondary endpoint), it could help reverse what has felt like a big disappointment in personalized medicine.