We have a new paper in the NEJM from the Eli Lilly effort on monoclonal antibodies against the coronavirus. And there’s no reason not to be up front about it: it’s disappointing.
This is the BLAZE-1 trial (mentioned in this recent post), which is studying non-hospitalized patients. It was another trial (ACTIV-3) where this antibody was found to be ineffective in hospitalized patients, but the numbers from this trial are not all that impressive, either. I’ll be adding my voice to the chorus – here’s Jason Mast at Endpoints with a roundup of the reactions so far. The trial (which is still continuing) is evaluating single infusions of 700mg, 2.8g, or 7g of the monoclonal in people with mild-to-moderate infection. The primary endpoint is change in viral load at Day 11 after dosing, and what we’re seeing here is the results from an interim analysis in early September.
Here’s the problem: only the middle dosed reached statistical significance on that endpoint. That’s not great from a dose-response standpoint – as usual, the easiest result to understand is seeing the effects increasing with increasing dose. There are most certainly “U-shaped” dose responses, with the U facing either up or down, but with three doses it’s hard to know what that U is like or why it’s shaped like it is. You can bet that Lilly would rather not have seen the results come out the way that they did. One problem was that by Day 11 even the placebo standard-of-care group had a trend towards viral clearance, but the bigger problem is that the antibody wasn’t able to distinguish itself very well from that.
And that’s especially true since the company has asked the FDA for an Emergency Use Authorization for the 700mg dose, arguing that the viral load data – which was the primary endpoint, remember – isn’t the most important thing to consider. But that immediately makes you want to ask them why they made it the focus of the study, and makes this argument sound (unavoidably) like ex post facto special pleading. You also have to wonder what dose the FDA might approve for an EUA, if the FDA is minded to give them one at all. The effective dose in the trial was obviously four times as large as what Lilly is asking for.
When you look at other outcomes, there are trends for symptoms clearing a bit faster, a trend for slightly fewer people to be hospitalized at Day 29, and so on. But the antibody therapy (at any dose) does not distinguish itself much. I really hoped for better, but (all together now) this is why we run the trials. Sometimes we find things that work better than we expected, and sometimes we find a perfectly reasonable idea doesn’t seem to do as much as anyone would have predicted. I think the best hope for a really compelling use for this antibody is in the ongoing prophylaxis trial, but after seeing these data I wonder how that one’s going to read out, too.
Regeneron has also released more data on their antibody cocktail (although in another press release and not in a journal manuscript). This is another non-hospitalized-patient study, in about 800 people. They’re not seeing much difference either in their two doses (2.4g and 8g), so they’re asking for an EUA for the lower one, naturally. The numbers overall look more interesting than the Lilly ones (although it has to be said that these are different trials with different enrollments, etc., so you can’t necessarily just line things up head-to-head). That said, the Regeneron therapy does show significant results for reducing viral load and for follow-ons like reducing medical visits. The best results were in the most at-risk patients, whether that’s defined as higher viral load, pre-existing risk factors, or weak antibody response. That last one was the group highlighted in Regeneron’s earlier look at their data, you may recall. But the Lilly data also look best in these cohorts – that, at least, is a result that’s what you’d have expected.
Does this mean that an EUA would feature a recommendation to use either of these therapies preferentially in high-risk patients? Or a call to get an antibody readout before starting therapy, prioritizing those people who aren’t mounting as good a response? Or is there going to be one at all?
My guess is that yes, we’ll see such a move. But how much difference it makes, that’s another question. As Matthew Herper details in this column at Stat, we have not prioritized production of these antibodies the way that we have vaccines, so there are (at the moment) only about 50,000 doses of the Regeneron cocktail available. A look at the map will suggest that the entire stockpile, even if given immediate authorization and Star Trek transporter-style distribution, would vanish into the national epidemic like a rock thrown into a pond. You’d have to look very hard to see its effects at all. We had around 80,000 new cases yesterday, and that number is going up. 50,000 doses of an antibody that seems to help cut down on hospital visits is not going to make much of a dent in that situation.
Update: and now Regeneron has halted dosing of their antibody cocktail in hospitalized patients with high oxygen requirements, which fits in with the general “dose early” ideas. This one seems to have been both a safety signal and lack of benefit. . .