philh

Wiki Contributions

Comments

Sorted by
philh20

Complex Systems (31 Oct 2024): From molecule to medicine, with Ross Rheingans-Yoo

When you first do human studies with a new drug, there's something like a 2/3 chance it'll make it to the second round of studies. Then something like half of those make it to the next round; and there's a point where you talk to the FDA and say "we're planning to do this study" and they say "cool, if you do that and get these results you'll probably be approved" and then in that case there's like an 85% chance you'll be approved; and I guess at least one other filter I'm forgetting. Overall something like 10-15% of drugs that start on this pipeline get approved, typically taking at least 7 years.

A drug that gets approved needs to make about $2 billion, to make up for the costs of all those trials plus the trials for the drugs that didn't get approved. And it has about 10 years to do that before patent protections expire, because you filed the patent before doing the first human studies and you only get 20 years from that point.

Typically what happens is someone forms a company for a specific drug, and while it's in fairly early trials the company gets bought by a big pharma company. The trials themselves are done by companies that specialize in running clinical trials.

Ross says Thalidomide was sort of the middle of a story. The story started with Upton Sinclair's The Jungle, which he wrote as a "look at the horrible conditions meat packers have to endure" but what the public took from it is "excuse me, there are human fingers in my sausages?" So after that was the pure food and drug act which said that anything had to be just the thing it said it was.

But then a drug came which was exactly what it said it was, and that thing was bad for people. So after that you needed to do studies to show safety, but they were less rigorous than they are now?

And then thalidomide happened, which was fine for most people but caused birth defects when taken by a pregnant person. When it came up for approval, the beurocrat looking at it happened to have previously looked at rabbits and seen that drug uptake and metabolization could be different in pregnant rabbits, making something otherwise non-toxic become toxic. And so she said the company needed data about safety in pregnant people, even though this was a non-standard requirement at the time. The company tried to avoid that, she insisted, and it never got approved in the US. But standards still got stricter.

(It did get approved in Europe. It's relevant that Germany didn't like tracking birth defects due to previous history, so the problems weren't noticed as early as they might have been.)

One of the times when regulations got stricter, part of the story is that at the same time as public outrage, there also happened to be a bill in progress for reasons of punishing pharma for something something, so that's the bill that got through.

At some point you started to get patient advocacy groups, saying "we are dying while you hold this drug up", and the FDA would pay attention to that. And then the pharma companies would get involved in those groups, and now it's at the point where you kinda need one of those or the FDA will be like "why would we prioritize you?"

There are drugs that the FDA wants to encourage but which aren't profitable, e.g. helping with diseases common in the third world and rare in the US. One motivation is if you make one they'll give you a priority review voucher, good to help another drug of yours get approved faster. These vouchers are transferrable, and the market price is... I think $100k or $200k?

With covid vaccines, the government said "if you produce a thing that satisfies these criteria, we will buy X amount of it for sure". That took some uncertainty out of the process and helped things get made.

Sometimes a drug will succeed in trials but not get pushed forward for various reasons, sometimes just falling through cracks. One drug this happened with was a covid treatment, which seemed to reduce hospitalizations by 70% in vaccinated people. When it was in development the FDA said it was unlikely to get emergency use authorization, and the company dropped it.

(Related: VaccinateCA got a lot of funding for a while, and then after the funders themselves got vaccinated, it got less funding.)

Later the company was going bankrupt, and they sold off their assets, which included "drugs that seemed promising but we never went anywhere with", including this one. Ross was involved in some other company buying up that drug.

You can do trials for covid much cheaper than for cancer drugs. For cancer you'll often have a list of a smallish number of people and try to find the specific individuals who give you the best chance of a statistically significant result based on comorbidities and such, and have someone specifically approach the people you want. For covid the cheap thing to do is: everyone who comes to your clinic with a cough gets the drug and gets a covid test. Later you find out if they had covid and (thanks to a phone call) what happened to their symptoms. And you can do this sort of thing somewhere like Brazil, instead of doing it in the most prestigious hospital (where there are a bunch of other studies going on distracting people). But it's kind of a weird thing to do, and if trials fail your investors might be like "why didn't you do the normal thing?"

philh20

Though this particular story for weight exfiltration also seems pretty easy to prevent with standard computer security: there’s no reason for the inference servers to have the permission to create outgoing network connections.

But it might be convenient to have that setting configured through some file stored in Github, which the execution server has access to.

philh20

Yeah, if that was the only consideration I think I would have created the market myself.

philh198

Launching nukes is one thing, but downvoting posts that don't deserve it? I'm not sure I want to retaliate that strongly.

philh2712

I looked for a manifold market on whether anyone gets nuked, and considered making one when I didn't find it. But:

  • If the implied probability is high, generals might be more likely to push the button. So someone who wants someone to get nuked can buy YES.
  • If the implied probability is low, generals can get mana by buying YES and pushing the button. I... don't think any of the generals will be very motivated by that? But not great.

So I decided not to.

philh20

No they’re not interchangeable. They are all designed with each other in mind, along the spectrum, to maximize profits under constraints, and the reality of rivalrousness is one reason to not simply try to run at 100% capacity every instant.

I can't tell what this paragraph is responding to. What are "they"?

You explained they popped up from the ground. Those are just about the most excludable toilets in existence!

Okay I do feel a bit silly for missing this... but I also still maintain that "allows everyone or no one to use" is a stretch when it comes to excludability. (Like, if the reason we're talking about it is "can the free market provide this service at a profit", then we care about "can the provider limit access to people who are paying for it". If they can't do that, do we care that they can turn the service off during the day and on at night?)

Overall it still seems like you want to use words in a way that I think is unhelpful.

philh20

Idk, I think my reaction here is that you're defining terms far more broadly than is actually going to be helpful in practice. Like, excludability and rivalry are spectrums in multiple dimensions, and if we're going to treat them as binaries then sure, we could say anything with a hint of them counts in the "yes" bin, but... I think for most purposes,

  • "occasionally, someone else arrives at the parking lot at the same time as me, and then I have to spend a minute or so waiting for the pay-and-display meter"

is closer to

  • "other people using the parking lot doesn't affect me"

than it is to

  • "when I get to the parking lot there are often no spaces at all"

I wouldn't even say that: bathrooms are highly rivalrous and this is why they need to be so overbuilt in terms of capacity. While working at a cinema, did you never notice the lines for the womens' bathroom vs the mens' bathroom once a big movie let out? And that like 99% of the time the bathrooms were completely empty?

My memory is we didn't often have that problem, but it was over ten years ago so dunno.

I'd say part of why they're (generally in my experience) low-rivalrous is because they're overbuilt. They (generally in my experience) have enough capacity that people typically don't have to wait, and when they do have to wait they don't have to wait long. There are exceptions (during the interval at a theatre), but it still seems to me that most bathrooms (as they actually exist, and not hypothetical other bathrooms that had been built with less capacity) are low-rivalrous.

None of your examples are a counterexample. All of them are excludable, and you explain how and that the operators choose not to.

I'm willing to concede on the ones that could be pay gated but aren't, though I still think "how easy is it to install a pay gate" matters.

But did you miss my example of the pop-up urinals? I did not explain how those are excludable, and I maintain that they're not.

philh42

Thing I've been wrong about for a long time: I remembered that the rocket equation "is exponential", but I thought it was exponential in dry mass. It's not, it's linear in dry mass and exponential in Δv.

This explains a lot of times where I've been reading SF and was mildly surprised at how cavalier people seemed to be about payload, like allowing astronauts to have personal items.

philh42

Sorry, I didn't see this notification until after - did you find us?

philh20

I agree that econ 101 models are sometimes incorrect or inapplicable. But

I don’t know how much that additional cost is, but seemingly less than the benefit, because three months later, the whole of Germany wants to introduce this card. The introduction has to be delayed by some legal issues, and then a few counties want to introduce it independently. So popular is this special card!

The argument here seems to be that the card must satisfy a cost-benefit analysis or it wouldn't be so popular, and I don't buy that either.

Load More