Posts in: science

Here are a few unrelated articles that crossed my inbox this morning:

ChatGPT, the blog expert

The latest episode of The Talk Show was with Taegan Goddard, who all the way back in 1999 founded the blog Political Wire which is apparently a continuous intravenous drip for people interested in US politics. Now, I’ve had other preocupations back then and not being an American citizen still have little to no interest, so this blog wasn’t even on my radar until listening to the episode. But now I wonder: are there any more relevant blogs I’ve missed out on, about medicine and biotechnology in particular?

ChatGPT’s first pass was mediocre. I’ll save you the verbalist padding, but here are its suggestions in response to my prompt: “Is there a website/blog like politicalwire.com or daringfireball.net but for biotechnology?”

It’s a 20% hit rate: only Derek Lowe’s In the Pipeline comes close to what I asked for. The others are all medium to big news outlets that yes, focus on biotech, but that’s not what I asked for. The second try, after I asked for more like Lowe’s, was a tad better:

That’s more like it! 80% now, and if I were feeling generous I’d give it a full 100% since In the Pipeline is, in fact, a Sci Trans Med blog. But then I asked for too much, and it hallucinated 3 more, two of which were hallucinations (BioPunk and BiotechBits, which were at least plausible names) and one was a sub-blog of Endpoints that also didn’t exist.

So, now I have two new blogs to follow (Timmerman Report and The Niche; Biotech Strategy is behind a paywall and I’ve already been following the others), and an ever-increasing urge to update the Blogroll, which has been under construction for the past five months with no end in sight.

I had Linus Lee’s blog The Sephist filed under “Paused and Defunct” for a while now, but he is back at it. Although most of the subject matter is out of my wheelhouse this mental model of Motivation as a function of Exploration (or was it the other way around) rang true — certainly truer to the scientific method than what my 6th-grader has been hearing at school.

An interesting entry to the big and ever-growing book of unintended consequences:

Chernobyl caused many more deaths by reducing nuclear power plant construction and increasing air pollution than by its direct effects which were small albeit not negligible.

(ᔥAlex Tabarrok)

Andrew Gelman has a new — and free — textbook out, Regression and Other Stories. From the cover:

Many textbooks on regression focus on theory and the simplest of examples. Real statistical problems, however, are complex and subtle. This is not a book about the theory of regression. It is a book about how to use regression to solve real problems of comparison, estimation, prediction, and causal inference. It focuses on practical issues such as sample size and missing data and a wide range of goals and techniques. It jumps right in to methods and computer code you can use fresh out of the box.

Between that, his Bayesian Data Analysis and many other freely available lectures and books, has there ever been a better time for high school students bored out of their minds by the pedestrian curriculum? But I am now just projecting to myself from 20-some years ago — I am sure high school students of today would rather spend time on their PS5, and my past self would probably have joined them. (↬Andrew Gelman)

Good quote today from Adam Mastroianni’s latest newsletter:

When I see someone salivating over the idea of a Science Gestapo, I have to marvel at their faith that authorities only ever prosecute guilty people.

Applies more broadly than science, of course.

Why are clinical trials expensive?

Why haven’t biologists cured cancer? asks Ruxandra Teslo in my new-favorite Substack newsletter, and answers with a lengthy analysis of biology, medicine and mathematics. Clinical trial costs inevitably come up, and I know it is a minor point in an otherwise well-reasoned argument but this paragraph stood out as wrong:

Clinical trials, the main avenue through which we can get results on whether drugs work in humans, are getting more expensive. The culprits are so numerous and so scattered across the medical world, that it’s hard to nominate just one: everything from HIPAA rules to Institutional Review Boards (IRBs) contribute to making the clinical trial machine a long and arduous slog.

What happened here is the classical question substitution, switching out a hard question (Why are clinical trials getting more and more expensive?) with an easy one (What is the most annoying issue with clinical trials?). Yes, trials involve red tape, but IRB costs pale in comparison to other payments. Ditto for costs of privacy protection.

If we are picking out likely reasons, I would single out domain-specific inflation fueled by easy zero-interest money flowing from whichever financial direction into the biotech and pharmaceutical industries, leading to many well-coined sponsors competing for a limited — and shrinking! — pool of qualified sites and investigators. It is a pure supply-and-demand mechanic at heart which is, yes, made worse by a high regulatory burden, but that burden does not directly lead to more expensive trials.

There are some indirect effects of too much regulation, and at the very least it may have contributed to more investigators quitting their jobs and decreasing supply. They also contributed to regulatory capture: part of the reason why industry has been overtaking academia for the better part of this century is that it’s better at dealing with dealing with bureaucracy. But again, these costs pale in comparison to direct clinical trial costs.

Another nit I could pick is the author’s very limited view of epigenetics: if more people read C.H. Waddington maybe we could find a better mathematical model to interrogate gene regulatory networks, which are a much more important part of the epigenetic landscape than the reductionists' methylation and the like. But I’d better stop before I get too esoteric.

If there had been a webpage monitoring the progress of the actual moonshot in the 1960s, it would have said stuff like “we built a rocket” and “we figured out how to get the landing module back to the ship.” In 1969, it would have just said, “hello, we landed on the moon.” It would not have said, “we are working to establish the evidence base on multilevel interventions to increase the rates of moon landings.”

This is Adam Mastroianni skewering the “science moonshot” initiatives, and rightfully so. If all we have to show for them are 2,000 papers full of mealy-mouthed prose, it was a ground shot at best.

Failure of imagination, drug cost edition

Here are a few facts we should all be able to agree on:

  1. The US has the highest drug prices in the world
  2. The US has the highest clinical trial costs in the world
  3. American drug R&D is so developed that it subsidizes the rest of the world

The disagreement lies in how these three are connected. If I interpret the Marginal Revolution school of thought correctly, drug prices are high (1) because America subsidizes all R&D (3) while having the highest clinical trial costs (2). So, 3 + 2 → 1 and limiting 1 to make drugs more affordable will lead to a negative feedback loop which would limit 3. This is why people who want to regulate down drug prices are “Supervillains” This implies that most drugs approved in the US are (1) life-saving and (2) have no other alternative. My gut reaction, backed by no direct research but some insight in cancer drug effects and mechanism of action, is that this is not the case and that if someone were to perform a rigorous review of approved drugs they would see that most are me-too drugs with marginal benefit. That is not, however, the argument I’m trying to make here which is why it is relegated to the margin. responsible for future deaths of millions of people who won’t be able to benefit from the never-developed drugs.

But of course, if 3 + 2 → 1 were true, there are two more ways to lower 1: limiting the scope of R&D (3) — which wouldn’t be the first time — or, preferably, lowering clinical trial costs (2), which have ballooned out of all proportion thanks to a potent mix of To expand on this, on the margin for now and in a separate post later: pre-clinical and phase 1/2 startups get billion-dollar valuations based on but a dream of success, which gets them hundreds of millions of dollars in their accounts, which in turn gets them to spend like drunken sailors on what should be low-single-digit million-dollar trials, which gets you to $>10M phase 1 trial, which borders on clinical trial malpractice. regulatory burden and oodles of money floating around the pharmaceutical/biotech space. The reason for all that money flying around? The promise of high payout guaranteed by unregulated drug prices! So: 1 + 3 → 2.

And if both of those relationships are true, well then there is a positive feedback loop in play, also known as a vicious cycle, and if there is one word that encapsulates the American drug cost landscape “vicious” is better than most.

The problem with high clinical trial costs isn’t only that they serve as an excuse for/lead to high drug prices. They also pose an impossibly high barrier for disconfirmatory trials that could get hastily approved but ultimately ineffective drugs out of the market (see Ending Medical Reversal). Because even in the world of Marginal Revolution’s hyper-accelerated approvals and early access to all, patients and physicians alike would need formal measures of safety and efficacy as a guide, and clinical trials are the ultimate way to do it. So we’d better have a quick-and-dirty way to do those too.

Unless the failure of imagination is mine, and in the Marginal Revolution world it would be an artificial intelligence sifting through “real-world” data for safety and efficacy of the thousands of new medical compounds and procedures blooming in this unregulated Utopia, perhaps even recommended and/or administered by LLMs who would finally bypass those pesky rent-seeking doctors. You have to see item number 6 to believe it. I did a double-take. Then again, maybe I shouldn’t take economists so seriously.

The Junk Charts blog sometimes links to charts that are not junk at all, today being one of those times. The link is to some beautiful storytelling on the many neighborhoods of New York City and now I wish the Washington Post had something similar for DC.