Possibly two of these problems solve each other? If access to the US market becomes a no-go for new drugs because of the FDA, then drug companies need to find a way to be profitable without such access, which would create pressure to reduce the cost of drug development including clinical trials? Americans would be left behind, while development might continue but more streamlined?
I wonder how long our political/regulatory idiocy would continue, if it became the case that companies just... stopped bothering to make treatments available here, while the rest of the world got healthier for less money?
On observational studies' estimates of dosage effects from exercise:
There are obvious selection effects. Your capacity and inclination to move is caused by health. There's also harm; exercise depletes your resources, which eventually compromises health when it overcomes your recovery capacity. Since the forwards causality is complex, and the selection effects are straightforward, the early minimum in these graphs is mostly information about the point at which direct harm starts to become more important than selection effects on the margin; seems like at the right end there's further selection for elite athlete levels of health in many cases.
An exercise type selection effect: Swimming is frequently chosen as a low-impact exercise tolerated well by sick and elderly people.
The main mechanistic argument I've seen for exercise type and quantity is Chris Masterjohn's argument that maintaining basic competence at many movements is helpful both for local soft tissue health and for injury prevention:
https://chrismasterjohnphd.substack.com/p/what-everyone-should-be-doing-for
See also Masterjohn's appearance on Joe Rogan episode 2420: https://open.spotify.com/episode/0ME2AXkR3e6nQQYMjbewjt
He cites as empirical support the longevity advantage of elite gymnasts vs other athletes: https://pmc.ncbi.nlm.nih.gov/articles/PMC11979035/
I made a similar argument here on Twitter: https://x.com/ben_r_hoffman/status/2015616670450987316
ETA: My opinion is possibly a steelman of Masterjohn's "local soft tissue" mechanism argument, influenced by conversations with Valentin Rozlomii and personal experience of his work. The gist is that using or otherwise perturbing a muscle improves circulation and therefore oxygenation in not only that muscle but organs adjacent to it. In addition, if a muscle is chronically tight because weak and underused, it can chronically impinge on a nerve, causing referred pain and chronic tension elsewhere, which then impairs circulation and therefore oxygenation around the referred site.
The main mechanistic argument I've seen for exercise type and quantity is Chris Masterjohn's argument that maintaining basic competence at many movements is helpful both for local soft tissue health and for injury prevention:
If your goal is competence for many movements signing up for a Feldenkrais class is probably better than most of what's traditionally called exercise. The Feldenkrais class is much better optimized for that goal than traditional exercise.
Feldenkrais doesn't give you cardio benefit or muscle hypertrophy but if movement competence is your goal it's superior.
I agree, and for those with enough literacy and capacity to generalize independently, the books, especially Awareness Through Movement, can be as good or better, with the caveats I mention in the Republic of the Self section of Levels of Republicanism. I also discuss some alternatives.
As a doctor, I strongly endorse the usage of LLMs to help collate your medical issues before seeing a doctor. Now, I would be more careful about endorsing this to a general audience, but I have sufficient faith in the LW audience to not feel too conflicted about it. Knowing the average IQ in these parts probably hovers in the 120s-130s is excellent for my peace of mind.
To be specific:
I am a psychiatry resident, and very recently, LLMs correctly helped me question a diagnosis made by other human doctors a year or two back. I had my own reservations, but it was in a speciality that I am not an expert in. It turned out that the previous diagnosis was incorrect, and missed by two human clinicians (leaving aside myself). I may or may not have sought a third opinion myself, but the consensus was sufficient to push me into being less lazy, and it was worth it. Of course, my own medical knowledge made it easy to present the information well to both the LLMs and the human doctor I consulted, but I don't think that was decisive.
(I am pleased to say that the new diagnosis was much less severe. I went from being genuinely concerned about the possibility of losing my vision to learning I have an annoying but not incapacitating condition)
Also note that many doctors have a negative attitude towards patients telling them about ChatGPT-suggested diagnoses. They may become annoyed, dismissive, or condescending.
I am not one of them, I appreciate patients doing their own research as long as they're willing to keep an open mind when it comes to my advice, but caveat emptor.
(If demand exists, I can post a relatively short/concise post I wrote about how to use LLMs for medical advice without shooting your foot off, directed at laymen.)
You sound like a good doctor. Unfortunately I have seen very bad patterns repeatedly:
Organizing your statement helps yourself and the clinician.
3 page summary produced at great cost by a distressed person outlining 1. medication history, 2. course of treatment and effects, 3. background medical problems. Ignored, even with repeated requests on the basis of having trouble keeping it straight.
Think thrice about withholding information from the doctor
Doctors doing wink wink nudge nudge to get the person to not say things that might imply liability in order to avoid just stonewalling or giving a halfhearted referral elsewhere.
I am genuinely sorry to hear about your negative experiences with other doctors, either personally or second-hand. I would not be so bold as to call myself the best clinician around, but I try and make up for that with good bedside manner, patience and an open mind.
Doctors are not a homogeneous population, unfortunately there are those who react poorly to perceived challenge. I can't bring myself to hate them, I have felt my patience running short when someone with uncontrolled diabetes shows up with their toes on the verge of falling off and reveals that they refused to take medication as prescribed and decided to use a combination of old Google, influencers and other questionable sources to opt for homeopathy. This is has happened more than once, though I didn't keep the toes. Fortunately, even free ChatGPT is a clear improvement in terms of quality of information and presentation to laymen. I would be genuinely surprised if it defended homeopathy without very significant nudging.
I will note that I have long become accustomed to other doctors taking me extra seriously, both from presumed priors about my clinical knowledge and professional courtesy. It's one of the few perks of the profession, I am unfortunately not an American doctor and paid far less than I'd like.
>Doctors doing wink wink nudge nudge to get the person to not say things that might imply liability in order to avoid just stonewalling or giving a halfhearted referral elsewhere.
True, but I am writing for a well-informed lay audience here, and I did say to reflect on it very hard instead of attempting to forbid it entirely. I have not yet had any LLM tell me to not disclose something to another doctor, so I can't judge if they are potentially justified if and when doing so. I would strongly recommend the average LW user not do it, if they're entirely using their own judgment instead of being advised by a human doctor or an AI to do so.
Things are relatively quiet on the AI front, so I figured it’s time to check in on some other things that have been going on, including various developments at the FDA.
Table of Contents
FDA Reformandum Est
In lieu of plan A, how about plan B?
Senator Bill Cassidy released a new report on modernizing the FDA. Alex Tabarrok approves, which means it’s probably good.
The FDA chief has an even better idea.
You’d also want a carveout for antibiotics, but yes. Your offer is acceptable, LFG.
That was the good news. There’s also the bad news, which was the main event.
FDA Delenda Est
Settle in for the tale of Moderna, and its attempt to get approval for a flu vaccine.
They were told their study design was acceptable. Then they got rug pulled. Then a lot of people were very loud and the FDA backtracked a bit while pretending not to.
It could have been one way. It’s the other way.
We’ve also lost Jim O’Neill, who was basically our last best advocate for reasonable policy, but on the bright side he’s going to the National Science Foundation. That’s also a good spot for him, but I worry his value over replacement will be a lower there.
This is what happens when you burn the term supervillain on people who want lower prices on drugs. What do you then call people who outright destroy one of the most promising medical advances of our time for no benefit whatsoever, via torching a $750 million study for decisions the FDA had explicitly already approved.
In this case, via withholding review for a drug that they had no business requiring the study for in the first place, in order to retroactively demand an absurd standard?
The FDA claims no, they didn’t say this study design was valid, but they’re lying, of course they did, there’s no way you spend $750 million dollars here without checking.
Per STAT, Prasad overruled FDA scientists to reject the application outright, and we have confirmation that staff was ready to review.
As an obvious technical note, we already have extensive data on how different existing vaccines relate to each other, so even if you thought that we shouldn’t allow a new offering until it was proven strictly superior and you changed your mind about the current standard, it is not statistically difficult to back out the necessary information.
The full slides are available here. It is not obvious this particular vaccine is a good deal for most people, but it clearly works, and they did what the FDA said to do. The FDA pulled the plug anyway, not even examining the findings.
It is hard to overstate how destructive this is. This meaningfully hurts your quality of life, the number of days per year you can expect to be sick, and your long term life expectancy.
That’s because if Moderna has to worry about the FDA retroactively withdrawing its word on what studies will be accepted, after the work is done, and they are facing an mRNA-hostile FDA in general, they have no choice. They were already abandoning much of the work.
And everyone else has to have the same worry everywhere else. Oh no. At minimum, this forces ballooning costs as studies have to be designed defensively in case the wrong person at the FDA suddenly changes their mind.
Good luck implementing risk-based monitoring (RBM) or otherwise cutting costs, and many potential drugs will be abandoned.
The pivot by Moderna to not invest in new Phase 3 trials was announced at Davos in January in the face of other symptoms of FDA hostility. Every person who gets Epstein-Barr, herpes or shingles more than a few years from now should know exactly who to blame for this outcome. That is only the start. Other companies, even those not focused on mRNA, have to now factor in this risk.
The good news is that we managed to yell loudly enough about this one that they may refuse course.
Especially when you’re simultaneously arguing the exact opposite:
Mr. Kennedy long raised concerns as a vaccine activist that the inactive ingredients in vaccines could cause harm and should not be used in human trials.
They say Kennedy was not involved here, but he’s head of the agency and has been furiously acting to undercut Moderna and its mRNA research at every turn.
One can try to frame this all as business as usual, and there are advantages to everyone coming away thinking about it that way whether or not it’s true:
Venk Murthy tries to frame this as ‘rebuilding trust’ by the public, but I expect this only further inflames popular paranoia about vaccines, because of the reversal, and the key ‘rebuilding trust’ that is suddenly necessary is the trust in the FDA by those researching vaccines. Without that, no new vaccines.
Alex Tabarrok shares this, which summarizes the ‘medical freedom’ movement driving much of this. Do you really think you can ‘rebuild trust’ with such folks?
It is important, when people’s objections are Obvious Nonsense, to point this out.
IN MICE
Progress is being made.
We now have a single vaccine could protect against all coughs, cold and flus, researchers say. Well, okay, IN MICE. But it’s a start. The problem is, do they have the regulatory confidence to fully invest in that?
Stanford researchers identified 16-PGDH as a root cause of cartilage loss, claiming the mechanism works IN MICE, and also in human tissue samples. Epirium for now is doing Phase 1 trials for sarcopenia. FDA Delenda Est so we’re looking even in an optimistic case at 5-7 year timelines, but if it all works it’s possible this solves not only arthritis but a wide range of quality of life issues. Alas, this won’t do that much for life expectancy on its own since these mostly aren’t the things that kill you, but it could be a big win in QALY terms if it works out.
Doctor, Doctor
How should you describe your situation to a doctor? You want to be respectful and honest with them, but if you dump a bunch of information they can get easily distracted in predictable ways. I buy Carl Feynman’s advice here to start with your diagnosis and key facts that seem most relevant and unlikely to mislead, then only later give the rest. Think of it as (polite) prompt engineering. LLMs can help you.
Trust The Process
This is one way to do plastic surgery, but you’d better be damn sure you’ve solved the associated principal-agent problems and the surgeon has good taste:
It seems right that most people are not good judges of what would or would not make them look hotter, and that experts will know better, but also those experts have incentives. Perhaps you should focus more on establishing trust than trying to solve the problem yourself.
Also the person who most has to look at you will be you, so your preferences matter. Don’t fully ignore them. And yes, there is cumulative risk of backfire if you overdo it.
Cancer Screening
AI but not LLMs: New large study (n = 105,934) of AI in cancer screening that started in April 2021 finds strong results from Transpara Version 1.7.0. Gains plausibly were largely about workflow integration and use of risk scores, rather than AI as a standalone second reader. Descriptively, there were overall 16% fewer invasive, 21% fewer large (T2+), and 27% fewer non-luminal A interval cancers in the intervention group compared with the control group, and the detection rate of cancer rose 29%, while enabling a 44% reduction in radiologist ‘screen readings.’ They got increased sensitivity without loss of specificity.
I am deeply confused by the whole ‘we still don’t know if this would be worth the cost’ response. What cost?
That’s for old school AI that was available back in 2021. Diffusion not going great.
Autism Everywhere All At Once
Autism as it is now super broadly defined and aggressively diagnosed may hit boys and girls in roughly equal measure. The marginal diagnosis clearly does net harm, and as explained previously there has been no actual increase in autism rates to explain.
Other Mental Problems Everywhere All At Once
Jesse Arm also has a similar view to this, as do I:
Mental health screenings sound like a good idea, but the false positives and overtreatment and medicalization of problems that result in practice mean they end up doing more harm than good. Such screenings look for trouble, and trouble they often find, which can be devastating for kids and families.
Source Data Verification
Adam explains some of the reasons clinical trials are inefficient.
That seems rather obviously crazy. Even the FDA now agrees. Yet Adam claims this is what typically happens.
Easy. Ban the practice, as in have the FDA require you not to do it. Alternatively, Claude thinks the concrete fear is how things will look on audits, so you can create rules to protect you from audits.
Instead we have the opposite, now more than ever. The FDA is announcing it will look for any excuse to deny your application in at least some cases.
The question then is, are there cases where the FDA will look for any excuse to say no, but end up having to say yes anyway? There might not be. A common pattern when seeking approval is that you beat yourself up over details that might matter, and then you get told one of your choices mattered, when really their mind was already made up for other reasons.
Also I saw claims that this particular data point does not hold up. Claude notes SPV is not universal, but thinks the claim of 25%-40% of total costs is plausible.
The other big obvious problem is that our scientists have yet to learn the secret of mass production.
This again seems obviously crazy. There should be a standard product called a clinical trial, you should do it in standard ways over and over again.
Why no fixes?
I don’t buy the ‘no motivation’ argument, as reduced costs drive more profits, and there are at least several big players who do enough iterating to justify this, and if they figured it out they could turn the trials into a new profit center.
This feels like a clear ‘if he wanted to’ situation. You can fix the cultural issue if you care enough. The real problem is indeed cultural, and it’s that they consider everything involved sacred, and everyone demands the everything bagels at all times, and those are sufficient legal and other barriers that no one dares try to fix things.
External Review Board
Why not let researchers choose their own accredited IRB, at the risk of taking some of the internal out of the review board? This is very much a ‘mostly disarm the IRBs’ proposal and I fully support it. That’s because I think IRBs are totally out of control and cause orders of magnitude more harm than they prevent, so letting people shop for the most efficient and accommodating one seems fine. The point would be that someone independent of you has to put themselves and their reputation on the line to say you’re doing something reasonable and not doing any blatant ethical violations, whereas right now no one can do much of anything.
Walk It Off
A new large observational study claims that 45 minutes a day of walking gives you most of the all-cause mortality benefits of exercise, and running doesn’t do better than walking until you do rather a lot of it, whereas too much jogging can actively backfire and swimming and bicycling do little. Doing multiple types of activity was better.
Note that the X-axis on different graphs can be very different here.
It’s an observational and self-reported study, so you can’t assume causation and doubtless ‘willing and able’ correlations should be doing a lot of work. Presumably people who swim or bicycle must have some real problems? But there’s still a lot of explaining to do and in some places that only raises further questions.
The obvious response is to defy the data, or at least some of it, or at least attribute it to selection effects. It doesn’t make any sense for bicycling and swimming to have about zero effect and jogging to start actively backfiring this much, while walking cuts mortality by 20%.
I do buy that this is evidence for ‘if you do a lot of walking that largely counts being active and doing exercise.’
An Unhealthy Weight Can Be Worse Than You Realize
Being in generally poor health causes all sorts of other issues, part quite a lot:
The obvious way to do that is to buy one of the drugs and give it away, or sell it as a generic at marginal cost. It’s fine to pay the NPV of future royalties to get it, it’s a win-win, and you only have to buy one of them so you can play them off against each other.
Our GLP-1 Price Cheap
Or the market can solve this on its own, if this crosses the threshold and becomes worthwhile enough that consumers pay cash. Go go gadget market incentives.
Then again, perhaps the market is solving this via patent infringement? Novo Nordisk says so in their lawsuit against Hims & Hers.
Right To Die Should Include Right To Try
In Canada and want to try an experimental treatment? We can’t allow that, would you instead like to have us help you commit suicide?
I think it’s reasonable to be for or against assisted suicide to varying degrees. But if someone would be allowed to invoke MAID and kill themselves, then you have to also allow them (on their own dime as needed) to then do any other crazy thing that only might kill them. As Alex Tabarrok puts it: If you have the right to die, you should have the right to try, at least if you absolve the system of your future oversized directly related medical expenses.
Meanwhile: