The markets refuse to forgive Infinity
The company’s second foray with a PI3K inhibitor shows signs of additive activity, but investors punish it all the same.
Infinity has had quite the journey in PI3K inhibition, having given up on its most advanced project, duvelisib, and sold it to Verastem for zero up front in 2016. The focus turned to the gamma-selective asset eganelisib, which today had its judgement day; the markets did not like what they saw.
The stock opened off 30% after the group revealed survival data from two PD-(L)1 combo studies, Mario-275 in urothelial cancer and Mario-3 in triple-negative breast cancer, that underwhelmed statistically and numerically. However, at least in urothelial cancer this is surely a function of small patient numbers, and there appear to be clear signs that eganelisib is biologically active.
That activity, Infinity claims, is down to eganelisib’s effect on macrophages, which it can apparently switch from an immune system suppressing (M2) type to an antitumour (M1) one.
The more important of the studies unveiled today is Mario-275, which compared eganelisib plus Opdivo against Opdivo alone in urothelial cancer patients who had failed at least one prior treatment. This trial had previously shown eganelisib’s promise especially in PD-L1-negative patients, albeit only in terms of remission rates.
Today Infinity backed this up with median overall survival, which in all-comers came in at 15.4 months for the combo and 7.9 months for Opdivo, resulting in a 38% reduction in risk of death. The comparison was also supported by Bristol Myers Squibb’s Checkmate-275 trial in the same setting, in which Opdivo monotherapy yielded median OS of 8.6 months.
But Infinity offered no stats, and the extremely wide confidence intervals cited for the OS hazard ratio, 0.28-1.36, suggest that there was no statistical effect for this measure. However, it would have been foolish to expect one: OS was one of many secondary endpoints, and Mario-275 enrolled just 49 subjects.
The company seems justified in believing that it is seeing biological activity, and is now planning a registrational study for eganelisib in urothelial cancer. Infinity also highlighted the Mario-275 result in PD-L1-negative patients, in whom mOS was also 15.4 months for the combo versus 7.9 months for Opdivo.
Apparently finding a role for eganelisib, which is strongly selective for PI3K gamma, could be important. PI3K alpha inhibitors like Novartis’s Piqray have clear activity in breast cancer, while pan-PI3K agents like the earlier duvelisib and Gilead’s Zydelig have effects on B-cell signalling but ultimately proved too toxic.
One reason why Infinity ran Mario-275 was that post hoc analyses of Checkmate-275 had revealed a particularly poor showing in patients with high levels of myeloid-derived suppressor cells, including macrophages.
...breast cancer miss
But perhaps the group’s case was not helped today by it also revealing a negative dataset from Mario-3, a separate uncontrolled trial in first-line TNBC. Here eganelisib was combined with Roche’s Tecentriq, and across 38 evaluable patients yielded progression-free survival of 7.4 months and ORR of 54%.
This clearly missed expectations of an ORR at least 15 points higher than the 53% seen in Roche’s Impassion-130 study of Tecentriq plus Abraxane. Worst of all is that Impassion-130 showed median PFS of 7.2 months, so eganelisib appears to be adding little in this tumour type.
Why eganelisib should have additive benefit in urothelial cancer but not in TNBC will vex some investors. The finding, if true, is unlikely to be explained away by the fact that beta, delta and gamma PI3K subtypes are not normally associated with breast cancer, since eganelisib’s mechanism is on macrophages, which do play a role.
Nevertheless, the market looks to have overreacted. Perhaps after the duvelisib debacle investors have run out of patience, but Infinity’s sub-$200m valuation prices in virtually nothing. The company is right to press on.
This story has been updated to correct the PFS in Impassion-130 all-comers.