A return to form from Vinay Prasad with two short and sweet articles on AI in oncology:

• Terrence Tao does not understand cancer medicine
• Can AI ‘hack’ a cancer clinical trial

If mathematicians want to complain that computers are replacing them, they should seek the comfort of chess Masters complaining that they have been bested. But they should not look for comfort from doctors. We would do anything to help our patients. We would do anything to extend survival. We already do things on a wish and a prayer and a mechanism of action and a hint of bioplausibility.

True.

As heard at the NIH IRB retreat, which I sadly had to miss:

“For the incredible gift of looking at the most chaotic study designs and effortlessly deciphering their hidden meaning, this goes to Milos Miljkovic!”

Not into sharing accolades here, but I’m super-proud of this one.

Frequent Infinite Regress linkee Ruxandra Tesslo wrote an op-ed for The New York Times about why we can’t get new cancer cures. Unsurprising for those who have read her past work, she places the bulk of the blame on the glacial speed by which clinical trials are designed, set up and run, at least in the US. Biotechs are looking at other countries — yes, including China — for their next first-in-human trial. All the cures we need are right there in the lab, you see, it is these darn trials that are too slow, keeping patients away from life-saving medicines. There is a nice n-of-one story about a patient with KRAS-mutated lung cancer who received a RAS inhibitor in a phase 1 trial to set the tone.

Vinay Prasad calls out a few big omissions in Tesslo’s article, and the number one point he brings out is the number one point I have issue with as well: the science is, in fact, lacking. There are no miracle cures locked in a drawer somewhere, not reaching patients because of government bureaucracy or a big pharma conspiracy.

But how can that be, with all the money thrown into cancer research ever since Dick Nixon’s War on Cancer? Well. Though I don’t know this for a fact, I have a strong suspicion that the answer to the curious lack of cancer cures is the same as the answer to that other question no one dares ask: if trials being so slow and expensive is a huge bottleneck, why are they so slow and expensive in the first place? It’s also the reason why a discerning reader of this blog may have noticed that my writing has taken a slight but unexpected anti-market turn.

Financialization, once the great enabler of garage-based startups, has turned every field of American endeavor into a casino and given one too many a bullshitter enough money to choke the competition. Biomedicine is particularly susceptible to bullshit large and small. Even if decently financed, the honest players — of which there are quite a few — have to fight for other limited resources in the ecosystem, from PhDs to do the actual research and clinical trial sites to attention. The result is hyperinflation of trial costs and a veritable bullshit bonanza in the biomedical literature, and this is not even taking into account the more personal motives of academics to fill the field with dreck.

My big fear is that no one is talking about the negative impact of financialization because it is water. Kudos to Tesslo for not talking about billion dollar drugs, but that seems to be the default, even among science journalists. Mr. Market will give you cheaper 4K TVs — surveillance comes for free — what it won’t give you is cancer cures.

A note on the title: I have come to detest The NYT’s two-sentence titles that go out of their way to dumb down the subject matter while piquing people’s interest, but thought it would be fun to write one for a quick post that’s a comment on a NYT article. Promise it won’t happen again.

• Saloni Dattani: Every disease is a policy failure. A good overview of how far we’ve come in treating many diseases, but also a thoroughly depressing perspective and one that I can’t endorse. It’s like saying “poverty is a policy failure.” No. Poverty and disease are the default state of the human condition. We should celebrate and cherish each step away from them; beating ourselves up over our imperfections in eliminating them is the road to madness.
• Hugh Rienhoff for the FT: Biotech VCs have become like bankers and we might all pay the price. With the above in mind, I think we can agree that financialization of drug development has brought asymmetric returns to the parties involved, and the closer you are to the actual science the further you are from the rewards. The bankers always win.
• Rahul Parikh and Jonathan Mates for Nautilus: The Trouble with Cancer Screening. There is a school of thought which views information as the ultimate good: the more you have the better. Here is a nice example of when that may not be the case. A pox on every doctor who sends out a screening PSA without first discussing it with their patient.

• Gillian Tett for the FT: Welcome to the era of financial candyfloss. ⊕ [Note: Gift link, if you click on time. ] It is the perfect metaphor for American finance: spinning thin and devalued real properties into mounds of tooth-decaying fluff. You know things are bad when it comes not from the Guardian but from a seasoned Financial Times columnist.
• Brett Murphy for USA Today: Rigged. ⊕ [Note: ᔥCorey Doctorow ] An exhaustively sourced, dramatically illustrated, harrowing story about truck drivers working as slave labor, from 2017. If only there were a mechanism by which the richest, most prosperous country in the world could control this and similar excesses of a greedy minority.
• Alice Fleerackers for Nautilus: When Stupid Was a Diagnosis. On the plight of people with intellectual disabilities. In some ways, the 1700s were better to most of them than even now, or rather, they were equally cruel and bad to everyone regardless of their mental prowess. Then came the oh so ironically named Enlightenment.
• David Tuller: Trial By Error: No Benefits for Multidisciplinary Rehabilitation in New ME/CFS Study. “ME/CFS” stands for Myalgic Encephalomyelitis/Chronic Fatigue Syndrome and it seems to be the consequence of viruses playing havoc with people’s immune systems, which leads to their autonomous nervous systems being out of whack. Many of you may not remember, but there was a big viral infection going around 5-6 years ago that left everyone exposed all at once, which led to an explosion of these cases in absolute terms, though in terms of percentages it is thankfully still a vanishingly small fraction. Sadly, you can’t rehabilitate your way out of this particular organic problem, and it puts many young and otherwise healthy people out of commission. If only there was a mechanism to fund more basic research.
• Davide Gioia: Wildlife on the Planet Furaha by Gert van Dijk. A book review. The book in question is an alternative evolutionary history, richly illustrated, from a professor of neurology at Leiden University. An instant pre-order for me, and it is coming out next week in the US!

Derek Lowe is scandalized by the fact that 9 out of 10 drugs that get to a first-in-human trial don’t get approval or, to use some jargon, a clinical failure rate (CFR) of 91% per year. He quotes a recent paper which followed CFR over the last six decades ⊕ [Note: Just to make things confusing, the paper itself calls this percentage “clinical attrition” but Lowe rebrands it to a CFR. Let’s for the sake of consistency stick with the latter, although a pedant may question whether that statistic truly is a rate. ] and noted that it was improved somewhat in the 1970s and 1980s — down to about 80% — but was consistently at or above 88% in every other decade analyzed.

Both the paper and Lowe present this as a Bad Thing. Here is a direct quote from Lowe:

But let’s think about that 91% failure rate for a moment. When I bring this up in presentations, I invite the audience to consider what the auto industry would look like if 91% of new car designs proved unable to roll out of the factory, or if 91% of new airliner models were unable to leave the ground - and if you only found that out after spending all the R&D money to build them at full size and trying to fly them. No cutting-edge restaurant could survive if 91% of its innovative dishes proved inedible or outright poisonous. What other industries operate under these bizarre conditions?

Sounds scary! The paper has a different and a somewhat less alarming spin: see how bad our preclinical testing is if so many of the drug candidates don’t make it to approval for reasons of biology. If there are bad side effects, or the drug doesn’t make it to the targeted tissue, or the liver destroys it too quickly, should we not have picked that up in cell culture or animal testing? Should we not focus our resources on developing more human-relevant preclinical tools, and maybe call them New Approach Methodologies (NAMs)?

At this point in the story I should mention that the paper in question, titled “Need for NAMs: A systematic evidence synthesis revealing over half a century of drug development failure”, was published in NAM Journal. Holy motivated reasoning, Batman.

So now let’s look at the counterfactual: what if the increased CFRs from 1980s to today came by design, from the realization by drug developers that preclinical testing isn’t the greatest at predicting toxicity and is completely useless when it comes to efficacy, so why not just get drugs to clinical as quickly as possible and be as quick in cutting them? My go-to paper outlining this philosophy is one from 2010 which a team from Eli Lilly — as successful at developing drugs as they come — published in Nature Reviews Drug Discovery. To pick out two key points from their above-the-paywall outline:

Reducing late-stage (Phase II and III) attrition rates and cycle times during drug development are among the key requirements for improving R&D productivity.

To achieve the necessary increase in R&D productivity, R&D investments, both financial and intellectual, must be focused on the ‘sweet spot’ of drug discovery and early clinical development, from target selection to clinical proof-of-concept.

Which is to say: since Phase II and especially Phase III trials are an order of magnitude more expensive than Phase Is, let’s quickly do what we must to get drugs into clinic, gather early human data, then make an informed decision. If that makes an arbitrary number go up, so be it.

You can see why the de-emphasis of preclinical efficacy data could have made some lab people unhappy, which I guess is why we now have a journal called NAM. Or was it ‘Nam?

Of course, there are tradeoffs everywhere, and the big tradeoff in the “fail fast” school of drug development is that you are tying your faith, financial and otherwise, to the health care system and contributing to the health care ouroboros. The one where Americans pay so much for health care because drugs are so expensive to develop, and drugs are so expensive to develop because health care in America is generally so expensive. But of course, no one is obligated to run their early-stage trials in the US.

• Tristan Manalac for Biospace: Amgen’s Tavneos troubles continue as NEJM retracts pivotal publication. First came FDA questions about data integrity. Now the academic authors are confirming the data tampering, though per the now retracted study they did “vouch for the completeness and accuracy of the data, the complete reporting of adverse events, and the adherence of the trial to the protocol. All the authors participated with the sponsor in the design of the trial, the analysis of the data, and the writing of the manuscript.” I wonder if people who sold this rotten pile of garbage to Amgen will offer a refund. So it goes…
• Carl Zimmer and Marco Hernandez for The NYT: This Cell Feeds, Grows and Reproduces. And It’s Manmade. “Dr. Adamala named her creation SpudCell, after its potato-like appearance. Rather than patent it, she and Dr. Endy are organizing a community of scientists to focus on making SpudCells more fully alive and adapting them to new kinds of experiments.” Bravo! Maybe science isn’t a complete lost cause.
• M Auerbach, MY Lim, GM Rodgers, and T DeLoughery for Sensible Medicine: When Normal Isn’t Normal: Non-anemic Iron Deficiency. A very academic group of authors makes a very good case for treating low iron levels in the absence of overt anemia. This is in response to an earlier text which poo-pooed the practice as “overmedicalizing normality”. Adam Cifu from Sensible Medicine proposes randomized controlled trials, though if symptoms of low iron are as diverse as restless leg syndrome, “brain fog” and craving for ice it would be a challenging endpoint to establish. Boy, isn’t medicine hard.
• Titotal (pseud): The best cause will disappoint you: An intro to the optimisers curse. ⊕ [Note: ᔥAndrew Gelman ] Mathematical demonstration of why effective altruists are, at the end of the day, a bunch of bozos. And not just them: America is the land of hyperoptimizers and if Taleb’s Incerto wasn’t a convincing enough case for why that doesn’t end up well in the long run, well, here is a much shorter Substack post that shows one facet of the problem.
• Ed Conway: The Apple-Banana Paradox. On the incredible flattening of banana prices in the last 25 years, and the incoming threat of another fungal infection that could wipe out the current Cavendish banana monoculture. On the other, could this be an opening for a Honeycrisp of bananas? One can hope.

• Max Kozlov for Nature: Inside the new political screening that’s stalling NIH grants. The plot to murder American science in the style of Rasputin (poisoned, shot, shot again) is well under way, and this is the slow-release poison. Choke off funding, dissuade people from thinking about federal funding, see the number of scientists drop for years to come.
• Vinay Prasad: Marc Tessier-Lavigne and Jose Baselga. To be clear, there are many things wrong with American science and Prasad in his review of the book How To Rule the World presents some good examples, but you don’t treat a serious diagnosis by killing the patient.
• Andrew Gelman: Bayesian Workflow exists as a physical book! An example of what’s right in science. I will be ordering a copy.
• Ann Lee for the Guardian: ‘David Bowie was a crazy workaholic’: Labyrinth at 40 – an oral history. ⊕ [Note: ᔥBill Harris ] An oral history done the right way, about a genius person in a genius movie.
• Catherine Ruth Pakaluk for The NYT: Life Is Better With Siblings. Sure is — just ask Jennifer Connelly from The Labyrinth! Joking aside, this is a serious case for large families and one with which I largely agree, though the double-digit census of children in some of the families mentioned gave me pause. [Note: The old George Carlin bit about idiots and maniacs applies. ] Regardless, for most people the right number is greater than two, and for almost everyone it is certainly greater than zero.

• Sasha Gusev: Thoughts on AI in academia. They are good ones. Extra points for leading me to an article from Sam Kriss in Harper’s Magazine about some magnificently agentic stupid people spending away their youth in San Francisco.
• Ruxandra Teslo: Manufacturing requirements are killing cell and gene therapy. The FDA wants companies to make at least two batches of product at the highest standard of manufacturing before approving it for commercial use. You should know this before starting your clinical program, especially if you have manufacturing that’s expensive, so maybe make two small batches instead of a big one? Just a thought. Separately, none of this would be an issue if there was momentum towards considering cell & gene therapy more of a blood bank/cell processing thing than a commercial drug. But then you couldn’t charge as much, could you?
• Regan Penaluna for Nautilus: Lessons in Chemistry, 19th-Century Style. Frustratingly, it takes Penaluna four paragraphs to mention the full name of Jane Marcet, the woman whose book “Conversations on Chemistry” inspired Michael Faraday — first paragraph mention! — to pursue science. The headline is also too broad: this was 1806, pre-Victorian times and barely 19th century. An extraordinary woman. Also: I want that book.
• Kristen French, also for Nautilus: Solving Feynman’s Formula for Eating Well, Parking Your Car, and Finding a Mate. How Feynman’s scribbles in a Thai restaurant lead to a paper in the Proceedings of the National Academy of Sciences, with mathematical proof of a common-sense inkling: more possible choices and more time should lead to more experimentation in order to discover “the best” of anything.
• Adam Mastroianni: Stop eating Lady Gaga’s Oreos. One of Mastroianni’s best, hinged on one key insight: Americans used to see themselves as temporarily embarrassed millionaires. High fortune now being so out of reach for many that it is simply unimaginable, they now see themselves as temporarily anonymous celebrities instead, which is why we have become more tolerant of celebrities hawking Oreos and less tolerant of billionaires. Also, good confirmation that I did not imagine the period when artists were trying very hard not to be labeled as sellouts.

Go to openevidence.com and you will see, right under the elegant logo and a free text box prompting you to ask a medical question, an immodest tag line: “America’s Official Medical Knowledge Platform”. The boast sits above an enviable lineup of official partners: The New England Journal of Medicine, Journal of the American Medical Association, National Comprehensive Cancer Network, Cochrane Systematic Reviews. If you were a clinician in need of information these would be the first places to go, ⊕ [Note: Save, perhaps, for a few journals in the JAMA network, and I write this as someone who has published in and reviewed for JAMA. ] but now there is no need because OpenEvidence will do it for you, for free and — unlike those poor community doctors whose practices can’t afford an NEJM subscription — with full access to all those journals.

Their About page is even more effusive. “Our mission is to help doctors save lives and improve patient care.” Great! It goes on:

This year, more than 100 million Americans will be treated by a clinician using OpenEvidence. As a product, OpenEvidence is an AI copilot for doctors that helps them make high-stakes decisions at the point of care. OpenEvidence is the most widely used medical AI among verified U.S. clinicians. To date, we have supported over 200 million AI-powered clinical consultations from U.S. doctors and other frontline clinicians.

In a remarkably short period of time, OpenEvidence has become the default operating system of medical knowledge in the United States.

Underneath lies the Team, laden with Harvard and MIT affiliations, and long list of medical advisors ranging from Mayo, Hopkins and Mass General staff to prominent YouTubers.

It was a rather obvious idea, to create a specialized LLM chatbot which restricts its data sources to medical literature only, so when I first saw OpenEvidence, the way it presented itself (partnership with NEJM and JAMA, MIT affiliation) and the price (free for everyone with an NPI) I was pleasantly surprised that these institutions came together for the common good, to create our generation’s PubMed.

Hardy har har.

Scroll further down and under another immodest headline — “Supported by the Best” — sit the logos of Sequoia Capital, Kleiner Perkins, Blackstone, Andreessen Horowitz, Nvidia, Google Ventures and the like. Not listed on the website because there is no “Investor relations” page — that may spook the clinicians! — is the financial history. Earlier this year it raised $250 million in a Series D round at $12 billion valuation. Just three months before that it raised $200 million at $6 billion valuation. In total, it has received close to $700 million in funding over its four years of existence.

Yes, OpenEvidence, “the default operating system of medical knowledge in the United States” (their words, emphasis included), is a tech startup zipping through the first phase of enshittification, i.e. attracting users with a high-quality offering. I would argue that even the “high-quality offering” is a bit of a crock, but we’ll come back to that shortly. Let’s, for the purposes of this paragraph, go with the premise that the unique thing that OE provides is the “artificial intelligence” portion. Well, from what I understand the company relies on OpenAI, Anthropic and others for the actual compute and if that is the case they are one-step removed from the absolute carnage whose genesis Ed Zitron and others have been diligently chronicling. The default operating system of American medicine is an earnings miss away from the blue screen of death.

I won’t cry for the billionaires involved. I will, however, mourn the opportunity cost of so many smart physicians and programmers on their medical and technical teams spending their time on point-one-percenter enrichment instead of truly building our generation’s PubMed. It would not even require compute! The true value of OE is the curated collection and unrestricted access to peer-reviewed journals, treatment guidelines, and systematic reviews, supplements and all. Let me google all that — or better yet, look it up on Kagi — and I will not care at all for the LLM-generated veneer glued onto man-made knowledge. But good luck having NEJM, JAMA et al. open their vaults without the VC-backed carrot of (I suspect) God knows how many millions of dollars for access rights combined with the FOMO stick that Anthropic and OpenAI’s PR teams have been so diligently whittling.

Trigger warning for an LLM-sounding phrase: the mounds of AI slop added to OE search results aren’t just wasteful, they are dangerous. Back in the Triassic era when shmucks like yours truly were nursing their middle-finger calluses writing progress notes by hand you knew that every part of that note contained useful knowledge. With the electronic medical record mandate — thanks, Obama — much of it became an unreadable mix of computer-generated charts and copypasta; you had to look at the end of the note to find actual human thought, whether it is in the Assessment and Plan or the Attending Addendum section. Well, I can report from the front lines that much of the time even that one meager paragraph has become a copy/paste job carrying with it that distinct LLM waft.

I am not against using LLMs for progress notes — we have been using human scribes for decades to write up the facts of the doctor-patient encounter. But those are costly and your rural primary care physician certainly won’t have one, so why not delegate that work to AI? The assessment and plan, however, are where you infuse those facts with meaning and then act on them, which is the entire purpose of the physician’s job. Writing is thinking and millions of US medical professionals have decided to delegate the one job they have to AI while keeping all the moral and legal responsibility, reverse-centauring themselves willingly and with eyes wide open.

This may seem like a “the food is horrible and the portions are too small” joke — have I not just wrote that the whole thing will soon be dead? If you are a physician who values their brain and doesn’t copy off a clanker why should you care if either start relying on them and then get a rug-pull? Three reasons:

• Expectation-setting: those who copy will need 15 minutes per encounter, then 10, then 5, continuing to ingest slop and regurgitate it over patient notes even as it gets increasingly bad from more and more expensive compute.
• Asbestos exposure: as in, AI is the asbestos we are shoveling into the walls of our society, only the asbestos here is in the form of regurgitated slop we are putting into patient medical records. That, too, will take our descendants some time to dig out, although human life span being what it is it should be less than a whole generation.
• Thinking of the kids: some of my own highest yield learning moments were reading the attending addendum on my note, or the dictation of a particularly skilled specialist’s consult note; will the incoming generations of medical students and residents have the same opportunity?

So if your mission truly was to help doctors save lives and you weren’t a greedy son of a bitch would you not have made a non-profit to achieve that goal? It may not have been as slick as something coming out of Silicon Valley, but it would also not have the risk of blowing up if the financial winds turn and the funding flywheel stops spinning. After all, there have been many attempts to replace the government-funded Medline/PubMed combo, but none of them were that much (if at all) better to justify the cost.