I wanted NSI-189 to be real so badly.
Pharma companies used to love antidepressants. Millions of people are depressed. Millions of people who aren’t depressed think they are. Sell them all a pill per day for their entire lifetime, and you’re looking at a lot of money. So they poured money into antidepressant research, culminating in 80s and 90s with the discovery of selective serotonin reuptake inhibitors (SSRIs) like Prozac. Since then, research has moved into exciting new areas, like “more SSRIs”, “even more SSRIs”, “drugs that claim to be SNRIs but on closer inspection are mostly just SSRIs”, and “drugs that claim to be complicated serotonin modulators but realistically just work as SSRIs”. Some companies still go through the pantomime of inventing new supposedly-not-SSRI drugs, and some psychiatrists still go through the pantomime of pretending to be excited about them, but nobody’s heart is really in it anymore.
How did it come to this? Apparently discovering new antidepressants is really hard. Part of it is that depression has such a high placebo response rate (realistically probably mostly regression to the mean) that it’s hard for even a good medication to separate much from placebo. Another part is that psychopharmacology is just a really difficult field even at the best of times. Pharma companies tried, tried some more, and gave up. All the new no-really-not-SSRIs are the fig leaf to cover their failure. Now people are gradually giving up on even pretending. There are still lots of exciting possibilities coming from the worlds of academia and irresponsible self-experimentation, but the Very Serious People have left the field. This is a disaster, insofar as they’re the only people who can get things through the FDA and into the mass market where anyone besides fringe enthusiasts will use them.
Enter NSI-189. A tiny pharma company called Neuralstem announced that they had a new antidepressant that worked on directly on neurogenesis – a totally new mechanism! nothing at all like SSRIs! – and seemed to be getting miraculous results. Lots of people (including me) suspect neurogenesis is pretty fundamental to depression in a way serotonin isn’t, so the narrative really worked – we’ve finally figured out a way to hit the root cause of depression instead of fiddling around with knobs ten steps away from the actual problem. Irresponsible self-experimenters managed to synthesize and try some of it, and reported miraculous stories of treatment-resistant depressions vanishing overnight. Someone had finally done the thing!
There are many theories about what place our world holds in God’s creation. Here’s one with as much evidence as any other: Earth was created as a Hell for bad psychiatrists. For one thing, it would explain why there are so many of them here. For another, it would explain why – after getting all of our hopes so high – NSI-189 totally flopped in FDA trials.
I don’t think the data have been published anywhere (more evidence for the theory!), but we can read off the important parts of the story from Neuralstem’s press release. In Stage 1, they put 44 patients on 40 mg NSI-189 daily, another 44 patients on 80 mg daily, and 132 patients on placebo for six weeks. In Stage 2, they took the people from the placebo group who hadn’t gotten better in Stage 1 and put half of them on NSI-189, leaving the other half on placebo – I think this was a clever trick to get a group of people pre-selected for not responding to placebo and so avoid the problem where everyone does well on placebo and so it’s a washout. But all of this was for nothing. On the primary endpoint – a depression rating instrument called MADRS – the NSI-189 group failed to significantly outperform placebo during either stage.
Neuralstem’s stock fell 61% on news of the study. Financial blog Seeking Alpha advised readers that Neuralstem Is Doomed. Investors tripped over themselves to withdraw support from a corporation that apparently was unable to handle the absolute bread-and-butter most basic job of a pharma company – fudging clinical trial results so that nobody figures out they were negative until half the US population is on their drug.
From last month’s New York Times:
The first thing you feel when a [drug] trial fails is a sense of shame. You’ve let your patients down. You know, of course, that experimental drugs have a poor track record – but even so, this drug had seemed so promising (you cannot erase the image of the cancer cells dying under the microscope). You feel as if you’ve shortchanged the Hippocratic Oath […]
There’s also a more existential shame. In an era when Big Pharma might have macerated the last drips of wonder out of us, it’s worth reiterating the fact: Medicines are notoriously hard to discover. The cosmos yields human drugs rarely and begrudgingly – and when a promising candidate fails to work, it is as if yet another chemical morsel of the universe has been thrown into the dumpster. The meniscus of disappointment rises inside you: That domain of human biology that the medicine hoped to target may never be breached therapeutically.
And so the rest of us gave a heavy sigh, shed a single tear, and went back to telling ourselves that maybe vortioxetine wasn’t exactly an SSRI, in ways.
But the reason I’m writing about all of this now is that Neuralstem has just put out a new press release saying that actually, good news! NSI-189 works after all! Their stock rose 67%! Investment blogs are writing that Neuralstem Is A Big Winner and boasting about how much Neuralstem stock they were savvy enough to hold on to!
What are these new results? Can we believe them?
I’m still trying to figure out exactly what’s going on; the results themselves were presented at a conference and aren’t directly available. But from what I can gather from the press release, this isn’t a new trial. It’s new secondary endpoints from the first trial, that Neuralstem thinks cast a new light on the results.
What are secondary endpoints? Often during a drug trial, people want to measure whether the drug works in multiple different ways. For depression, these are usually rating scales that ask about depressive symptoms – things like “On a scale of 1 to 5, how sad are you?” or “How many times in the past month have you considered suicide?”. You could give the MADRS, a scale that focuses on emotional symptoms. Or you could give the HAM-D, a scale that focuses more on psychosomatic symptoms. Or since depression makes people think less clearly, you could give them a cognitive battery. Depending on what you want to do, all of these are potentially good choices.
But once you let people start giving a lot of tests, there’s a risk that they’ll just keep giving more and more tests until they find one that gives results they like. Remember, one out of every twenty statistical analyses you do will be positive at the 0.05 level by pure coincidence. So if you give people ten tests, you’ve got a pretty good chance of getting one positive result – at which point, you trumpet that one to the world.
Statisticians try to solve this loophole by demanding researchers pre-identify a primary endpoint. That is, you have to say beforehand which test you want to count. You can do however many tests you want, but the other ones (“secondary endpoints”) are for your own amusement and edification. The primary endpoint is the one that the magical “p = 0.05 means it works” criteria gets applied to.
Neuralstem chose the MADRS scale as their primary endpoint and got a null result. This is what they released in July that had everybody so disappointed. The recently-released data are a bunch of secondary endpoints, some of which are positive. This is the new result that has everybody so excited.
You might be asking “Wait, I thought the whole point of having primary versus secondary endpoints was so people wouldn’t do that?” Well…yes. I’m trying to figure out if there’s any angle here besides “Company does thing that you’re not supposed to do because it can always give you positive results, gets positive results, publishes a press release”. I am not an expert here. But I can’t find one.
The pattern of positive results shows pretty much the random pattern you would expect from spurious findings. They’re divided evenly among a bunch of scales, with occasional positive results on one scale followed by negative results on a very similar scale measuring the same thing. Most of them are only the tiniest iota below p = 0.05. Many of them only work at 40 mg, and disappear in the 80 mg condition; there are occasional complicated reasons why drugs can work better at lower doses, but Occam’s razor says that’s not what’s happening here. One of the results only appeared in Stage 2 of the trial, and disappeared in Stage 1 and the pooled analysis. This doesn’t look exactly like they just multiplied six instruments by two doses by three ways of grouping the stages, got 36 different cells, and rolled a die in each. But it’s not too much better than that. Who knows, maybe the drug does something? But it sure doesn’t seem to be a particularly effective antidepressant, even by our very low standards for such. Right now I am very unimpressed.
Except…why did their stock jump 67%? We just got done talking about the efficient market hypothesis and the theory that the stock market is never wrong in a way detectable by ordinary humans.
First of all, maybe that’s wrong. My dad is a doctor, and he swears that he keeps making a lot of money from medical investments. He just sees some new medical product, says “Yeah, that sounds like the sort of thing that will work and become pretty popular”, and buys it. I keep telling him this cannot possibly work, and he keeps coming to me a year later telling me he made a killing and now has a new car. Maybe all financial theory is a total lie, and if you get a lucky feeling when looking at a company’s logo you should invest in them right away and you will always make a fortune.
Or maybe the it’s that it’s not investors’ job to answer “Does this drug work?” but rather “Will investing in this stock make me money?”. Neuralstem has mentioned that they’ll be bringing these new results in front of the FDA, presumably in the hopes of getting a Phase III trial. FDA standards seem to have gotten looser lately, and maybe a fig leaf of positive results is all they need to give the go ahead for a bigger trial anyway – after all, they wouldn’t be approving the drug, just saying more research is appropriate. Then maybe that trial would come out better. Or it would be big enough that they would discover some alternate use (remember, Viagra was originally developed to lower blood pressure, and only got switched to erectile dysfunction after Phase 1 trials). Or maybe Neuralstem will join the 21st century and hire a competent Obfuscation Department.
I don’t know. I’m beyond caring. The sign of a really deep depression is abandoning hope, and I’ve abandoned hope in NSI-189…
…which just leaves me even more time to be excited about SAGE-217, the novel GABA-A positive allosteric modulator that just passed Phase 2 trials! This one is going to be great!
[EDIT: Wait, is SAGE-217 just a weird attempt to rebrand benzodiazepines? Surely it’s got to be more than that, right?]