>and I've yet to go to the Emergency Room and NOT get a CT
Gaa! A whole body CT scan? 10-20 mSv? 5 years of background radiation? Yes, it is still a factor of 50 below what would cause radiation illness, but it still doesn't seem like something to do lightly...
Having recently worked with Kaiser on a medical issue, I would say I tend toward a similar line of thought to the banned user. Insurance companies do have control of pieces of the process, and hospital (company) lawyers also have control over pieces of the process.
I have heard from doctors inside Kaiser and elsewhere when trying to get a diagnosis for a condition that "that's not really how modern medicine works" and was told dozens of times "but how about we treat the symptoms." I turned to private scanning (which the hospital system won't let their radiologists read) and while it did not ultimately lead to a clear signal, it did drop some items in the differential, which made me *less* anxious, not more.
From a cost perspective, every hospital is incentivized *not* to provide patients with imaging that may reveal false positives. This is I believe why I was sent to X-ray although it is not actually the recommended scan for the condition I was going for (it's cheaper, but I don't think that's it--once you have the machine, marginal costs are minimal).
But I think there is truth to the fact that the medical system is interested in treating the symptoms and is not interested in chasing down diagnoses. Without a robust private scanning option, I believe my only option to reduce the medical anxiety I was having would be some pills they were pushing.
Further, I believe that there is a culture of non-diagnosis that results in things like "heart disease" being listed as the major cause of death when in actuality there are probably at least three or four (or five or ten) other major causes that lead to heart failure, get poorly reported as heart disease, and that then in a culture of non-diagnosis lead to a series of conditions across the culture that are never understood or diagnosed. So I would say this analysis that sits squarely in the single-patient/doctor/hospital cost-benefit eludes the question of what scientific power could be brought to bear on our health that will not be--all positioning as if we're doing the individual patients a favor.
"But I think there is truth to the fact that the medical system is interested in treating the symptoms and is not interested in chasing down diagnoses. Without a robust private scanning option, I believe my only option to reduce the medical anxiety I was having would be some pills they were pushing."
The alternative, of course, would be a national health care policy that incentivizes health care providers to invest in preventive care. If they got paid for catching conditions early on, diagnoses would suddenly become a lot more popular (and yes, that has it's own issues).
"If they got paid for catching conditions early on, diagnoses would suddenly become a lot more popular" Yup, which ties our under-diagnosis problem directly to the one place we see the opposite: over-diagnosis in mental health
> But 899 hours of scans would look like going to your doctor’s office for one hour every weekday for three years!
Not to mention that, you can't just schedule just 1 hour in your day to teleport to the doctor and be seen with 0 waiting time, going to your doctor for a 1 hour procedure could cost in addition:
- another hour in the waiting room
- an hour or two in traffic to get there
- some margin before and after the traffic time because you don't want to be late to the doctor, or late to the thing you want to do after the doctor
I think this point pushes the opposite direction: you should only rationally devote 899 hours to this, so if it took 3 hours to get a 1 hour doctor appointment, you should only be willing to go on one-third of days.
I'm just saying the estimated numbers are off. Understandably, the time and place of appointments are optimized for the doctor (whose labor is very expensive) not the patient. So 1 hour of doctor's time typically translates into several hours of patient's time, inflating the lost QALYs from extra procedures.
I think the problem is double-counting your health benefits. In a world where cancer is the only possible way to die, it might make sense to devote 899 hours to screening. But there are many ways to die, and you can't spend 899 hours trying to prevent each of them. And if you succeed to get rid of cancer, but the next day get hit by a bus, the effect wasn't as big as you expected.
I always figure going to the doctor involves some nonnegligible chance that I pick up something else from one of the other patients. I come in with a headache and depart with a headache plus the flu.
AI claims your paper found the opposite effect of what you are claiming:
the paper cited completely contradicts this argument. The research by Lowen et al. shows that influenza virus transmission is highly dependent on relative humidity and temperature, with cold and dry conditions favoring transmission.  Specifically, at 30°C (86°F), no transmission was detected at all, while at 5°C, transmission was significantly more frequent than at 20°C.  The authors explicitly conclude that influenza virus transmission indoors could potentially be curtailed by simply maintaining room air at warm temperatures (above 20°C) and either intermediate or high relative humidity.
You don't sound like an outlier to me. You remind me of a Chinese friend who, studying at an American university, caught a cold and went to health services for a comforting intravenous injection of saline solution. (This is a common practice in China.)
When they explained to her that they didn't do that because it has no medical benefit, she was quite upset, complaining that it felt like they didn't even care that she was sick.
I was suggesting this kind of practice isn’t limited to Chinese culture with that. The US did a similar thing that was (“obviously”) worse. The US also decided to change the practice.
That is, I was expecting people (here, especially) to get I saying this sort of thing isn’t isolated to a particular culture, could have been worse, might not be without downsides, and can be changed.
I’m not sure if saline injections are harmless. (I’m not even sure promoting the idea it’s doing something when it isn’t is harmless.)
That argument can't work. They're not a sudden or unforeseeable stress to the system. If you want to make this kind of argument, you'd have to say that the extra demand for medical services increases the number of medical providers, damaging society by drawing people to become doctors and nurses who would otherwise have done something more productive.
In the event that there 𝘸𝘢𝘴 a shock to the system that called for more serious medical care than usual, the extra slack provided by these extra practitioners would be a benefit - there's no obstacle to them declaring that they're busy with the emergency and saline injections are being temporarily paused.
Well, this is just logistics though. If the approach really worked and really was cheap enough, you'd try to set up cheap scan-booths you can simply book with an app and then pop into at the given time.
But I also don't think it makes sense to equate "time spent going to the doctor" as a loss of QALY to the same extent that being dead or having cancer is.
I spend above-average (for my age) time dealing with medical stuff, and it's not my favorite way to spend time, but it's a lot of waiting around and I bring a book or a game or do stuff on my phone. It's going to vary from person-to-person how 'effective' that time is, but I have to imagine the number of people who basically just zone out and stare at a wall while in the waiting room is fairly low.
'People should not buy the product currently on offer!'
'That's absurd! This product I am imagining which is similar to the product currently on offer but better in many specific ways and with none of the downsides would obviously be good!'
I think he's saying "only do this if there's a reasonable expectation that it will find something". That's a bit of a different statement. Many tests have a false positive that is high enough that unless there's a reason you shouldn't take the test.
Part of the discussion is that a broad set of ppl in the position to improve the product in the ways that could make it good argue that the entire concept is bad. Or at least appear to do so.
"Could this product be improved and is that a better focus of our efforts, or are the detractors correct that this is intrinsically unworkable" is a question worth asking
I work in genetics. This sort of comes up with whole genome sequencing.
We're actually quite conservative about what gets routinely reported back to the patients. In general incidental findings are not reported back unless they're on a shortlist that meets certain conditions re: certainty, whether there's anything that can be done etc.
You mention "anxiety" but that really doesn't carry enough gravitas.
After Chernobyl they caught enough cancers that the background cancer death rate dipped, but a lot of people were so convinced they were gonna die anyway that the alcohol related death rate shot up.
you need to be really careful when it comes to things that scare the shit out of people. Despair can have worse health effects than the things people are despairing about.
I sort of agree with this, but it's also infuriating when I can't get health data I know perfectly well how to use because someone thinks I might freak out to hear I have slow 2d6 metabolism or something.
pro tip: a lot of services have some kind of general workflow to comply with the european GDPR rules.
If you get your genome done it's your data. Typically they'll have some kind of process so you can do a GDPR request for your raw data. (very few people actually do)
You can always hop on a VPN and make it looke like you've moved to europe.
it's easy enough to annotate once you have it.
Though one of the early lessons was how many annotations look impactful but aren't really and how many variants get annotated as "pathogenic" or "HIGH" impact but are actually nothing-burgers. .
By "annotate" I assume you mean some kind of SNP/variant association mapping, and not gene prediction, right ? AFAIK gene prediction on humans is a solved problem...
Annotation as in taking a list of variants and for each one pulling from various databases of known info and predictions to try to figure out its effect, if any of note.
Re: gene prediction, you might be surprised. The current canonical databases of genes are great for highly expressed genes but definitely incomplete when it comes to one's expressed at a low level in small subsets of tissues.
Inside known genes the number of alt isoforms has been going up sharply since the advent of long read rna sequencing.
And genes being recategoriesed from non-coding to coding is pretty steady.
> Annotation as in taking a list of variants and for each one pulling from various databases...
Right, that's what I thought you meant, thanks.
> Inside known genes the number of alt isoforms has been going up sharply since the advent of long read rna sequencing.
I know this is the case in plants, but I'm surprised to hear it is still happening in humans -- I thought human genes were pretty well studied by now. Live and learn !
I'm sitting there with data from long read RNA sequencing, predicted ORF's and confirmation that we found the proteins predicted by those ORF's in mass spec data....
I'm sitting there thinking "this must be a big deal" but then I get chatting to my supervisor and she's like "oh we see that all the time, but it's a nightmare to try to convince people it's a real gene"
Some of it is because most historical gene finding didn't use live brain tissue, (people have this weird aversion to researchers taking samples from their living brain)
You also get into this whole almost philosophical thing of "what is a gene, *really* " because some genes physically next to each other that have been classified as different genes you'll find will actually splice together sometimes, some of the tools for looking at splicing auto-discard those but biologically some are consistent across samples.
There's also absolute shitloads of RNA-genes that don't seem to code for a protein but do seem to show up consistently across multiple individuals.
Noble intention to shield from harm, genuinely unclear whether there’s a net cost or benefit here. More fundamental point is a right to know. This is why there are laws in many jurisdictions (US, globally) that give a patient the right to receive (upon request) all health data that’s been recorded and stored about them.
Sure, but having a right doesn't mean that exercising that right is going to be ultimately beneficial; you also have the right to control what you consume, whether it's vegetables, aspirin, or cigarettes. It seems complicated as doctors have to balance the needs of their patients without perfect insight into their mentality. It's exacerbated since patients who are going to exercise this right to know are going to be a higher risk demographic for the harms described above (Although who's to say what the root cause is? It'd be hard to filter out pre-existing or complex conditions.).
Yes I think this is really the annoying part. Ultimately the policy is calibrated on a median user which means that people who COULD get value out of it STILL DON'T because no one bothers to account in any way for their individuality. Which is why it's generally irritating when others make decisions for your own good by assuming things about you. If it's about years of my life, I'd like to be able to do the best I can!
FWIW I work in a field that is tangential to plant genomics, and it seems like the trend is leaning away from whole genome sequencing / GWAS in there as well. That is to say, people are still actively sequencing whole genomes of all kinds of plans, but the next step in the process is no longer "mwa ha ha we now have all the data we need to make drought-resistant corn once and for all, behold as our stonk price goes to the moon", but rather, "we now have enough data to begin constructing a pangenome model that might one day eventually allow us to learn enough about plant metabolism to maybe increase drought resistance if we're lucky, hopefully".
The Soviet Union (and then the early Russian Federation, though perhaps also to this day) seemed to have a much bigger problem with alcohol related deaths than the US.
I agree that it is very hard to distinguish causal effects. But Chornobyl was really one of the factors that pushed people into despair. At that time I met some young people from Belarus, completely healthy and who lived quite far from Chornobyl when it happened and yet they were feeling fatalistic – that now my health and future is completely ruined; they always reminded this to others requiring a special caring attitude from them.
In Latvia there was a special disability status – a participant of Chornobyl cleanup operation – that entitled to certain benefits especially in healthcare. I personally think that this status alone was more damaging to person's health than actual effects from radiation from that work.
Any discussion with the general public that health effects from Chornobyl disaster were not as bad as people often think, is immediately ridiculed. You would look like an antivaxer trying to suggest to evaluate scientific data. The irony is that in Ukraine vaccine hesitancy is also very high, mostly due to strong distrust of public heath authorities.
Yep, met a lady in Kharkiv - around 2010 - who had been "participant of Chornobyl cleanup operation " at 18, and had been sure; her cancerous end would be near. 25 years later: a healthy mother of 2 healthy grown-ups. - And sure: many others saw no reason not to become an "alkashki" instead of trying for a career.
If I hire you to sequence my genome I want the data. All of it. I do not want you deciding what I get to see. I'd rather be worse off than have you deciding what is best for me.
Biostatistician Frank Harrell says that most effects in medicine are additive, which would imply that, conditional on the value of a whole-body MRI being ~= 0, the benefits of blood test + cardiac test + whole-body MRI are ~= the benefits of just the blood test + cardiac test. But that's a generalization; of course you'd need real data to look at specific conditions.
Incidentally there *are* datasets which have this level of richness, UK BioBank is one: genomes, whole body MRI, blood testing, a bunch of other stuff.
The degree of anxiety that people here in Taiwan get from a screening scan that shows “could be a tumor site, too small to be definitive” is noticeably smaller than friends or family back home in the US when they receive the same. Because a significant fraction of the adult population is getting scanned every two years, and has been wide scale for over a decade, I suspect that people have just compared notes enough to realize “a whole lot of people get this result, and hardly anybody dies from this ‘incedentaloma’ result.”
There’s probably also some psychological safety benefit to knowing (here, not there) that if you want to get a follow up, you can just make an appointment directly with an oncologist, within a week or so. No PCP referral, no clearing it first with anyone, and it won’t cost a ton (making an appointment with basically any specialist can be done directly, here).
Singapore is actually pretty similar to the US as far as the extent of regulation, it’s significantly more liberal/free regs-wise than any of the other OECD countries over here.
It’s not super surprising for me any more though it was 15 years ago (I also expect Singapore regulation-heavy;, I have a team in Singapore and I’ve been doing business (b2b SaaS, 12k business customers) there and 8 other countries between Australia and Japan for 16 years. SG MoM (Ministry of Manpower) not require employers to provide health exams, there are some lesser requirements on covering doctors fees or consultations.
@ruffienne, yep, a little more than half of the adult population here in Taiwan gets either MRI or CT scans or both every two years, and this has been going for about 20 years.
Very interesting. Your explanation makes intuitive sense. But what do you think of the theory that culturally people in Taiwan might be more stoic than people in the US?
How we are expected to behave often drives how we actually feel. Or do you have a good enough sense of the friend's and family's reaction to over events to discount this element?
It has absolutely not been my experience that Taiwanese people are less stoic than those in the US.
As someone who's experienced 'this is probably nothing, let's keep checking every year' I think a large part of why Taiwanese are calmer is that doctors and the medical system in general feel more accessible and trustworthy. The most it tends to cost is around US$16 per appointment, usually including tests, so people feel confident that they will be able to find out what's going on. I had a concerning symptom, went to see a specialist, got an ultrasound (immediately) which came back probably-fine, got blood tests which came back with unrelated precancer markers so got scheduled for an CT scan and then got told that it was probably nothing but come back once a year. The entire experience cost around US$20.
The only thing is that while most screenings are cheap or free cancer *treatment* can get quite expensive, so there are people who can't or don't (out of concern at imposing on the family) pursue it. That no doubt plays a significant role in the acceptance.
Yup, I really think this is just a transient effect. It needs to become mundane enough for people to learn to start writing it off, and then they'll start writing it off. People coexist with weirder things.
I think this is the most important comment. And it's one reason I don't give a lot of credence to the position that knowing might upset someone who's bad at evaluating risk. We genuinely do not know how much of a problem it will be if we never run the experiment. Which it sounds like Taiwan has done for us. Thanks Taiwan!
I don't think you give the age caveat enough weight. To me, it obviates the entire analysis because I (asymptomatic, no bad family history, late 20s, physically active, healthy bmi, nonsmoker, nondrinker, exercise, outdoors etc.) know that I am substantially healther than the "Average Adult", but can't really quantify it; superficially I would assume 10x? More? I've done road trips through America and "Average Adult" isn't doing very well. Starting with your $150,000 I'll happily 10x it and say $1,500,000 cannot possibly be worth it.
Also, "only ~half of people bother to go to regular checkups, which takes one hour per year and probably have effects nearly this big."
Do you have any citation on the actual benefit of regular checkups? If so, I'd love to read up on it. My impression is that as long as I remain asymptomatic the benefit of getting a yearly checkup would be effectively zero. I haven't been to a doctor for a general checkup in at least a decade and assume the next time I see one will be from a traumatic sports injury, I assume when I'm there I can get my bloodwork and whatnot done on the side. Should I actually get regular checkups (and for that matter have a primary care doctor or know the name of a doctor in my state?)
Tbh you could probably ask Claude or ChatGPT and it'd give you a good response; I would guess they would corroborate that you need minimal regular, if any care.
Your next age-gated care check is likely a colonoscopy at 45, maybe some bloodwork at 35 or 40.
I have done that, that's why I was asking for a citation. I am skeptical of the " bother to go to regular checkups, which takes one hour per year and probably have effects nearly this big"
Fascinating! It seems like the literature agrees with me that, given I am already living in the healthy way I would be counseled to live, a regular checkup would do essentially nothing and could in fact be harmful. Scott, I think a correction or rebuttal would be warranted.
It's not a great service and its not a direct answer to your question but...
In the UK on the NHS, annual check-ups are not a thing. You're invited for more specific screenings when you get older - like the above poster mentioned but a 20s male would not be expected to visit his doctor.
The system is overloaded, as you'd expected when free at the point of supply but they do still do their sums on the value of preventative care so this suggests those sums say annual check-ups for the young are at least not money saving vs waiting for issues to arise.
Same in the Netherlands. Annual check-ups aren't a thing; older women get checked for breast cancer until 75, and older people in general get asked to send in a stool sample like every 2 years or so to check for intestinal cancer.
I know someone who moved from Singapore to the Netherlands.
She tried to get a regular annual check-up from her GP. The GP told her: after doing all the tests, I'd just tell you to sleep more and exercise more. So how about we save on the testing?
To my mind, no matter how many health effects are transferred upon me by healthy behaviors (like those you mentioned), I cannot take for granted that I am necessarily and will remain healthy for those reasons specifically. If I understand correctly, cancer rates, to name one example, vary regionally for a whole host of reasons unrelated or indirectly related to individual behavioral choices (like air pollution, genetics, etc.). To my mind, having a relationship with a PCP and engaging in check ups and screenings once or twice a year is (1) trivially costly in terms of time and money, assuming that I’m insured, (2) at least somewhat protective in the short term, and (3) probably more protective in long term through the generation of base-level data against which my PCP can compare and interpret changes. I’ll also add that having a relationship with a PCP has two additional benefits: (a) a relatively frictionless way of accessing prescription drugs in the case of low-probability, high-risk events like infections; I once caught a pretty nasty bacterial infection while spending some time in a low-population, out-of-state, low-health-infrastructure area (as a result of me exercising out there, no less) and was able to have the issue diagnosed and an antibiotic immediately prescribed such that I started it the same day that I noticed the symptoms, and (b) a relatively frictionless way of discussing potentially concerning changes in my health with someone who has expertise in medicine and, ideally, knowledge of my medical history; it kind of kicks ass that, when I have established a relationship with a PCP, I can just give them a call if I feel like I have too. Frictionless-ness isn’t everything, but I assume that even healthy people would, overall, find some protection from low-probability negative health events if they had a frictionless relationship with a PCP, especially one who had some knowledge of and long-term data about their medical history. Idk. I’m just spitballing here. But my point, I guess, is that it sounds like you’re putting just a touch too much faith in your healthy behavioral choices to protect you against negative health outcomes. Your behavioral choices kick ass, but my prior is that no set of behavioral choices can be 100% protective and, therefore, that a relationship with a PCP and a few hours and few dollars a year still confer considerable benefits to an otherwise healthy person. I’m not telling you what to do, of course—and you’re already doing much more for your health than the average person, which is great, I’m just explaining how I think about that particular choice and the potential benefits of it.
Interesting perspective! I see the benefits on antibiotics, but my impression (admittedly secondhand) was that as an asymptomatic patient my generic wellness checkup would consist of me going in, stripping, spending a few minutes telling the doctor about my lifestyle habits and lack of concerns, getting poked a bit, the doctor taking my blood and saying I have slightly low blood pressure, agreeing but saying I've always had slightly low blood pressure, and leaving. Where in that process would the doctor actually catch that I have an early stage of cancer? I don't think I would be getting a full body scan, right?
In the US my regular checkups were pretty trivial and similar to what you describe. Height/weight/abdominal girth, blood pressure, stethoscope exam.
In Japan, the yearly checkups additionally include: some standard blood panel (including e.g. lipids), urine and fecal analysis (collected beforehand and brought into the appointment), ECG, abdominal/heart/carotid ultrasound, chest X-ray, lung capacity testing, vision testing, hearing testing. Plus the infamous barium-swallow stomach cancer screening, which I skip since there are less-annoying tests and the prevalence is much less high in white people. And then I ticked a few boxes to get some extra blood tests for things specific to my known issues.
They then surface all your results with a grading system: A = no abnormalities, B = minor findings, no need for concern; C3/6/12 = follow-up recommended in X months, D = consult physician, F = consult physician urgently.
I've found this to be helpful in the way the grandparent poster describes. E.g. it detected a few early-stage things that have me modifying my lifestyle habits and monitoring for followup (gallbladder polyps slightly below the danger line, mild carotid arteriosclerosis, trends in cholesterol numbers and lung capacity as I modify my diet and exercise programs). And I'm glad they're monitoring things like hearing and vision, as I want to know ASAP if those slip.
With regards to the question about whole-body screening, I suspect these sorts of East Asian yearly checkups have found a good medium between whole body screening and US-style test very little: they pick the most worthwhile tests, and subsidize or mandate them on a population-wide basis.
My experience in the US has been that it varies by doctor, but almost everyone I've gone to does at least a basic metabolic panel (lipids, cholesterol, glucose, etc) with the annual physical and most also do a comprehensive metabolic panel (also tests liver and kidney function, etc), a urine analysis, or both.
I have maintained relationships with primaries since 2016? 2017? In both the Phoenix area and the Wichita, KS, area (I now live in the Pittsburgh, PA, area, and don’t have insurance at the moment, but I plan on establishing with a primary the moment I have insurance, and expect the quality of the care I receive to be similar). My yearly/ twice-yearly checkups have involved blood tests, given at almost every stage of care. Idk. Maybe this is a relatively recent development in standards of care, but my perception is that relatively extensive and relatively useful amount of base-level data has been collected about me, and I assume that that data will be useful at some point down the line. Maybe I’m wrong, but it still puts my mind at ease knowing that someone with some amount of expertise is capable (at some level) of monitoring long-term changes in a certain, somewhat meaningful amount of my base-level data about me. Likewise, there is some short-term amount of value to the data they collect for me personally. For example, I was having some medication-related issues related to dehydration which I recently resolved, which I wouldn’t have been able to identify were it not for the regular blood tests to which I was (consensually) being subjected. I can’t say that I was given the best advice about how to resolve it, but the blood tests helped me to identify the problem, for which I’m very grateful. Again, just a perspective. But my experience with regular check ups in the U.S. has been, uniformly, very good.
Thanks! Only 3.5 months in, but I’m liking Pittsburgh so far! I know there are ACX meet ups here, and I’ve been meaning to make it over to one, but they’re always held on Sundays at 2, and I start work every Sunday at 3pm :[
There are typically also annual labs, probably bloodwork, maybe urinalysis. My primary care doctor also does certain routine bloodwork. I get a complete blood count annually, and when I turned 30 she added the standard metabolic panel. It turned out that despite my general good practices re: diet and exercise, I had a marker that indicated a slight increase in risk for heart disease and metabolic disease. So I can actively work to improve that marker before it becomes a problem. And I learned that because of a routine checkup
Many diseases do not present with symptoms until it's too late. Many of them can be cured if caught by early screening. It's like saying we shouldn't screen people for blood pressure because it can be asymptomatic
I’m guessing you’re male and/or have never had a Pap smear. That’s an example of routine annual screening that catches a ton of cancer cases early, which used to be fatal.
Similarly, if you don’t go to checkups at least some of the time, you’re probably not going to be up-to-date on current best practices re: routine screening. Right now that might not be a big deal, but in 10 years, you might need to keep track of when you need colonoscopies or prostate exams.
> because I (asymptomatic, no bad family history, late 20s, physically active, healthy bmi, nonsmoker, nondrinker, exercise, outdoors etc.) know that I am substantially healther than the "Average Adult", but can't really quantify it;
Assuming you maintain these habits, all cause mortality effect size for you vs the average american is ~4-5x. As in, past 40 they'll have something like a ~1% chance of mortality every year (and going up as they age), while you'll have a ~0.2 - 0.25 % chance per year (and going up as you age).
In expectation, this will net you 5-15 extra years of life, with 7 the median expectation, and 10-20 years of lower morbidity (ie you won't have diabetes, CVD, or other medical conditions lowering quality of life and needing ongoing medication for 10-20 years that the average american will have such).
The majority of the effect is driven by exercise, the next largest is not being obese / overweight, and then things like smoking and drinking. You didn't mention your diet, if you eat actual food that you cook from ingredients (instead of opening packages and microwaving), you're closer to the 5x end of things, and that's next in terms of effect size.
I could see this going the other way. If you are old and falling apart and will not realistically make the life style changes to adjust there is less value in knowing the precise ways you are falling apart.
If you are young and healthy you may be in better position to shrug off minor findings in your scan and not suffer the phycological costs of false positives but stand to benefit from the early detection of a treatable cancer. The distribution of QALYs may be different than the distribution of findings over a large sample
As someone in largely the same position who does get annual checkups one thing I would point out is that my blood tests occasionally catch childhood vaccines that have worn off and I have to get redone.
In general it’s kind of surprising that there’s not actually anything you can spend your money on to directly improve health outcomes beyond what a good doctor will already do for you. (Iirc Robin Hanson said the value of the marginal healthcare dollar is 0). Makes me wonder where the dramatic benefit to life expectancy from wealth actually comes from
There could be survival bias kicking in too. You may found rich people with various strange diseases you won't find in poor people. Because those poor people would already be dead.
Having a bit of savings, or higher disposable income, makes it easier to do some things that pay off long term - turning down a risky gig that pays a bonus, paying for the slightly more expensive apartment that isn’t next to the particulate factory, paying a personal trainer once or twice to give you advice at the gym, eating meals that are tasty in more subtle ways than just large amounts of fat and salt.
You can absolutely spend money to improve health outcomes, if not directly then:
Having a gym membership improves health outcomes. Having a swimming pool, basketball court, home gym set also improves health outcomes. Better quality food (fewer preservatives or just tastier healthy options), cleaner living environment, etc.
While any one of these may not do much, having better doctors, better screening, and more specific treatments should absolutely improve health outcomes in aggregate.
Having better education substantially improves health outcomes. This is probably the most important thing.
> Makes me wonder where the dramatic benefit to life expectancy from wealth actually comes from
1. Rich people, generally speaking, do not work in mines or as lumberjacks or any of the other dangerous jobs out there.
2. Drug addiction has a negative correlation to both wealth and health. The same is true of a lot of maladaptive behaviors.
3. Poor people are more likely to have short time horizons, with causality going both ways: they're poor, so they need to focus on making this month's rent, not what's going to happen in six years' time, and they're bad at investing money so they remain poor. Short time horizons mean you do shit that's bad for your health but feels good in the moment (comfort eating into obesity, smoking, casual sex with strangers, etc).
4. Stress is unhealthy. If the chemicals responsible for stress were good to have constantly flooding your system, they already would be. It's supposed to help you in crises where you might get eaten by a lion if you fuck up, so it's worth the health debuff, but in modernity there are no lions but plenty of stress - especially if you're poor.
5. Rich people can and will pay a premium so they don't have to live in areas where there's lead in the pipes, pollution in the air, etc.
There was an interesting paper published in JAMA Internal Medicine last year which highlighted the finding that eight cancers have doubled in incidence since 1992 in patients younger than 50 years, while aggregate mortality has remained flat. Here's the citation: JAMA Intern Med. 2025;185(11):1370-1374. doi:10.1001/jamainternmed.2025.4917 Published online September 29, 2025. The authors suggest that "Much of the increase appears to reflect increased diagnostic scrutiny and overdiagnosis. Interpreting rising incidence as an epidemic of disease may lead to unnecessary screening and treatment while also diverting attention from other more pressing health threats in young adults." It wouldn't surprise me if a rigorous study of whole body scanning, using whatever modality, revealed parallel findings related to diagnosis and mortality.
Meanwhile, the notion that there is an epidemic of rising cancer in young people has trickled down into the popular culture. I hear it regularly from friends and acquaintances, with attributions to whatever the cause du jour is (e.g., microplastics, pesticides, etc.).
It actually seems that they are right - though the cancers aren’t just going up in young people, but in people of all ages, and young people are reaching levels where it’s relevant to think about it, while older people already had to: https://dynomight.substack.com/p/crc-rates
> Any of these would be a more grounded hope than “of course more data will always be good and worth it”.
Would this line of reasoning not also apply to other claims about the efficacy of AIs in other fields -- and thus (in aggregate) to the probability of reaching AGI/ASI via increased LLM training ?
I don't see why. The claim about AI is that we're extending scaling laws and extrapolating known trends, not just a fuzzy prior that data has to be generically good.
I guess I'm confused about your reasoning. AFAICT, in your post (above the line I quoted) you list all the reasons why Midjourney's ultrasound scanner is unlikely to rival MRI on quality, and in fact unlikely to be extremely useful other than as a stimulus for other research. Your arguments sound reasonable to me; it would appear that ultrasonic scanning technology has some built-in limitations which increased amounts of data just cannot overcome. But would you say that scaling laws *could* in fact overcome these limitations, or would you say that ultrasonic scanning had reached its theoretical peak (or is very close to it) -- or am I misunderstanding you completely ?
If it's the latter (ultrasonic scanning is reaching a peak that cannot be overcome by increased intelligence), then could this not also apply to other technologies ?
Whether gathering more data is worth it is an empirical question that depends on whether gathering more data has thus far led to marginal benefits. We see that for AI, it has, but not so much for medical scanning.
Right, but it seems like ultrasonic scanning has built-in physical limitations, and thus getting a full-body ultrasonic scan is of marginal utility at best. If so, then a thousandfold improvement in AI would still yield little to no improvements in ultrasonic scanning. You could ask Hyper-Claude 9000, "help me build a better ultrasound machine", and it would come back with, "sorry, you've gone about as far as you can go, maybe consider switching to MRI instead". If that's true, then perhaps there are other technologies that follow a similar pattern (and I suspect that there are).
You don’t seem to understand that "the scaling laws" being referred to here is a specific empirical finding regarding the loss function in the pretraining stage of LLMs. It’s not a generic statement that "scale is good", and it certainly does not apply to this Midjourney ultrasound thing.
Right, but that is my point: it seems that we have found at least one domain where even applying a 1000x improved AI is unlikely to yield more than marginal benefits. I would argue that there exist many such domains, i.e. scaling up LLMs is not an automatic recipe for success in general.
I think you’re conflating things, the scaling laws tell you exactly how much improvement in the loss function you can expect for any additional compute and data. In fact the scaling laws tell you that you need exponentially more compute and data for any additional improvement in loss. If anything they tell you there’s diminishing returns.
But the scaling laws are only about the loss function in the pretraining step, and aren’t why people are bullish on LLM as promising technology. People are bullish because LLMs keep having exponential increases in capability, as measured in real world usefulness.
Right, I understand everything you'd said so far. So does this mean that LLMs are likely to produce substantial increases of the capability of ultrasonic whole-body scanners, specifically ? If not, why not ? If yes, then what might this increase look like in practice ?
This is an interesting line of reasoning, though I disagree with the conclusion. You’re saying, “Here’s a problem AI won’t solve”, i.e. there are some problems where a better-trained LLM won’t give a better answer, therefore “more data won’t always be worth it”. I agree with the narrow conclusion, if you only care about things like getting data out of ultrasounds. But the reason LLMs can’t help with ultrasounds is that actual intelligence won’t help with ultrasounds, either. Your argument is basically “LLMs won’t help if you only care about problems that a team of smart humans couldn’t solve, either”. I’d agree with that, but I also care about lots of things which a team of smart people *could* solve.
> I’d agree with that, but I also care about lots of things which a team of smart people *could* solve.
Yes, I agree, of course. There are still lots of problems that smart people and/or LLMs could solve, and once those problems are solved, they will open the way to other problems to be solved. Nonetheless, if there are many domains similar to ultrasonic scanning, where more "intelligence" and more training data cannot yield substantial benefits, then the concept of a "superintelligent AGI" is put into jeopardy. The chief feature of such ASI is that it's capable of significantly outperforming humans across many (if not all !) domains, but what if most domains are just like ultrasonic scanning ?
Great post, the fifth point in the problems list is probably the core of a lot of this discussion.
Neurotic, worried-well people with no symptoms who are <50 years old are the exact type of person where the cost-benefit analysis makes asymptomatic screening a poor idea and it's basically just a luxury good for this segment. And indeed this is also often the loudest group clamoring for it.
Are there awful stories of physicians messing up and undertesting? Absolutely - doesn't mean an abundance of asymptomatic imaging is the answer.
Can AI help make more screening programs effective? Absolutely - in part because AI can do a better job understanding your medical history and symptoms and can corroborate the screening findings / identify reasons to get a screening or imaging order.
> Neurotic, worried-well people with no symptoms who are <50 years old are the exact type of person where the cost-benefit analysis makes asymptomatic screening a poor idea and it's basically just a luxury good for this segment. And indeed this is also often the loudest group clamoring for it.
This sounds like the exact demographic that would benefit from a clean MRI result. About 2/3rds of the study participants had a clean result. That... sounds like a good thing? I don't think we should ignore the benefits of giving peace of mind to the majority of people getting screened.
The thing about highly anxious people is that reassurance-seeking is actually counterproductive in the long run. Sure, you might in the short-term feel less anxious because you got a clean result, but at a deeper level you've validated your hypochondria instead of facing your anxiety and learning to tolerate it.
not to mention if they do get a positive result (which, just on base rates, is extremely likely to be a false positive), the downside impact of additional testing and anxiety is enormous
Would it actually give peace of mind? The point of many mental disorders is that the brain *isn't* working properly or behaving rationally. I would not be surprised if the response to a clean scan is something like "the scan might have missed something, what else can I do?"
Also, 2/3 success means 1/3 failure rate--is the impact on the latter group the same magnitude? I have no data here, but again it would not surprise me if they were worse off by a lot more than the first group were better off.
> Neurotic, worried-well people with no symptoms who are <50 years old are the exact type of person where the cost-benefit analysis makes asymptomatic screening a poor idea
It actually genuinely surprised me that Scott estimated the anxiety at .01 QALY. So being anxious for a month or two while you get followups is equivalent to losing 3 days of your life, *permanently??*
That just sounds crazy to me! Like you would rather die 3 days earlier to not now this information? I'm not sure I believe that.
But then, I'm extremely non-neurotic (I think I was <20th percentile on OCEAN) and non-anxious, so I'm definitely talking from the other side of that spectrum.
So anxious people here, you would literally die 3 days sooner to not worry about something over a few months where otherwise your base health and the way you feel physically is the same it's been all year?
I imagine it depends on the results you're validating - if it's something like cancer, it's probably pretty large, esp. for a high neuroticism segment.
Most of the people who are indifferent to medical costs are not rich. They are people like me who consume a lot of medical services and are on medicare & supplemental insurance, or they are on medicaid, or get VA benefits.
I think it's the difference between public health thinking from the part of the doctors (looking holistic at impact on the whole healthcare system including costs) vs private individual expected value calculation (moar information always moar better). A lot of public health interventions recommend people do things which individually on the margins look bad for them but look good on aggregate and this feels weird to individualistic thinking
Doesn't this article just reaffirm the 'adjust the thresholds' point it's responding to? The cost/benefit 'back of the envelope' depends critically on both the proportion of people being sent for follow-up, and the proportion of that group who _do_ find something.
I was given a whole body MRI after a particularly nasty mountain bike crash a few years ago. It found all the normal broken bones and punctured livers that go along with some things, but it also detected a significant growth on my kidney.
The default advice was "wait 6+ months for someone in the public system to see you about it because we're kinda busy right now" (I live in Australia) but as a minorly rich person I paid out of pocket for a specialist to review. He said it looked really bad and ordered a biopsy, which was inconclusive.
I had lost my first wife to cancer at 29 years old, and my current wife was pregnant, so we decided to just get it removed at significant cost privately.
When they finally completed the expensive surgery they decided it wasn't cancer after all, but maybe it might have possibly taken over my whole liver, even if it couldn't spread elsewhere?
I continue to be unsure if I made the right decision or not to keep investigating. I probably wouldn't have, except I really didn't want to be told it was bad at the public system appointment right as my wife was due to give birth.
Just adding some colour to "what can happen when you get a whole body scan and things look weird".
In contrast, scans I had for real acute issues that also flagged a nearby asymptomatic incidentaloma. Only it turned out not to be incidental and did turn out to be cancer when subsequent scans for the ongoing acute issues served as follow-ups, observed concerning changes, and a surgical biopsy confirmed this. Better yet, it was a (so far!) very treatable cancer using existing medications with low side effect profiles.
I feel very lucky that it was noted by the radiologists doing the other scans and prompted, with some prodding from me, excellent response from the medical establishment leading to correctly identifying and treating the condition. I will note that it remains asymptomatic but was growing rapidly when untreated and could have lead to very serious complications although likely not death (organ damage, nerve damage, etc) that would severely impact my quality of life and would likely have done so by the time any symptoms would have arisen.
So here was a case where a large benefit was obtained through diligent surveillance. Furthermore, as a very rational actor and non-anxious person in tune with false positives and statistics, I am sure the only anxiety-related side effects were to less rational family members I told, not to me personally, and now I have peace of mind in knowing my prognosis and monitoring my improving condition through occasional follow ups.
My experience makes me a big believer in whole body scans. At the very least, maybe when you’re 40 you get a scan, and if you want to worry about unnecessary costs or anxiety, do nothing with it unless something really bad is obvious to the doctor. But then you have a baseline scan, so when you do the next one a few years later, your doc can more clearly find any changes from that one, and those are much more likely to be meaningful and merit investigation (or if nothing changes, give you more confident peace of mind).
Your story is interesting, but it's not obvious from the telling that you benefitted from early detection. For some cancers (I'm thinking especially of myeloma but there are others), early detection doesn't affect the cure rate much. If not detected early the cancer will eventually manifest through organ damage. However, the harm of this is often overstated. Most of the time, the organ damage is reversible and quite unlikely to lead to death in the acute phase. So while it may sound like early detection is a benefit, I'm not sure that ostensible benefit would show up in a population study.
And of course, what you say doesn't change the point of the post, which is that most people who get a full body MRI do in fact have an incidentaloma. You are the exception.
In my case it sounded like the most likely outcomes from untreated progression, whether detected or not, would have been partial leg paralysis, a permanent nephrostomy tube, and some kidney damage. I am very glad I don’t have to deal with the daily life challenges of those conditions even if they are unlikely to impact mortality assuming the rest of the treatment could be managed at that point.
I understand the incidentaloma issues, and have had other scans for acute conditions that were inconclusive, gotten 1-2 follow-ups from the scan results after the acute problem was resolved, and the issue declared to be nothing material just looking slightly unusual in a place no one would have looked had something else not prompted the scan.
I saw an interview, perhaps with Attia, saying basically if you do a whole body scan, be prepared to find stuff that you will either need to ignore or do a few rounds of follow ups on that are likely nothing especially the first time you do such a scan without a baseline. So I was prepared for these scans to see things and not especially worried until there were changes in an already suspicious area.
The biggest problem I have with the criticisms is that everyone assumes that the results aren't going to get better and the prices aren't going to get cheaper. They throw up their hands and say "This is wasting time and money for only 4 people per 1000!"
The whole point is that the tests are supposed to get better and more accurate. So the number "300 have mildly concerning findings" will go down especially with the proliferation of AI. That's why I have been doing full body MRI scans for the last several years. I pay $2000 which I can claim from FSA and any changes over the course of the year they can see.
As money gets put into this, there's no question that the technology will get better and more accurate. Yes, 1000 scans may only save 4 lives, but if the prices drop down to $200 and the no or mild findings increase dramatically and are accurately quantified and the 20 people with extremely concerning findings are accurately identified and only the people who need it get biopsies, then we are living in a better world.
Whether or not this is possible is simply a matter of opinion, but I believe in technology and I believe in the brilliant people who want to make this technology better.
That could all be true, but it would still be consistent with the notion that a full body screening MRI is not going to benefit the marginal person considering doing it.
Consider too that even if devoting a lot of resources to improving this technology will eventually make it useful, that doesn't mean that we couldn't have done even better by developing some other more effective technology with those same resources.
There's a lot of "could"s and "would"s in your response. If there's a better way to look inside the body that is cheaper and faster and safer, I would be all for it. Right now, MRIs are the best we have so why shouldn't we double down on it to make it better and cheaper and learn how to get much more reliable results?
You could say that about anything. There's an infinite number of things that have a small benefit that is canceled out by an equal or greater cost. It seems like you're saying we should just spend the money on all those things even though we know it has negative expected value because this will make the technology advance and eventually make the thing have positive expected value.
Unless we run out of things that have positive expected value, I think we should do those rather than do things that have negative expected value in hopes that this will eventually change.
There is no other technology that can scan our insides and let us know if we have a mass growing inside us. If 1000 people take it and there is no mass for 990 of them, then that's great. Then it's like a blood test that I take several times a year, all of which never show any results. No one seems to suggest that blood tests are useless or cause unnecessary anxiety.
Of those 10 people that do find something, it means it was detected far sooner than it normally would have by "traditional" means.
The only thing we need to do is make the tests more accurate and informative so that only the 10 people who need it get follow ups. That means better testing and better AI to discern.
The idea that we should completely ignore this gift from science to look inside our bodies because *right now* it's mathematically inconvenient is insane. We have a way to detect tumors years earlier, especially if everyone gets a scan once a year, like the dentist. What we need is more research to make it more effective, not ignore it and pooh pooh the idea just because it's not perfect right now.
There is nothing special about scanning and early detection. A QALY is a QALY. If we can get one for $108k/year (Scott's estimate, not saying it's right, just using as an example) by screening MRI, or we could get one for $50k/year by giving effective treatment to the people who actually have cancer, then it would be better to do the latter. I'm not saying that I have absolute confidence about the cost per QALY of all these different things. I'm just saying that if you accept the premise that screening MRI doesn't have a very favorable cost per QALY, it doesn't make sense to say that we should do it anyway so that it gets better. Sure, spend money to make the process cheaper. But don't encourage individual patients to spend their money on it when it's not worth it. There are plenty of other bigger fish to fry.
It's also silly to suggest that the medical field has been ignoring or pooh-poohing cancer screening. Colonoscopy, chest CT, PSA, and mammography have all been the subject of enormous study and billions of dollars of funding for several decades. Recently, there has been more pushback as the idea that the harms may outweigh the benefits has gained some currency. For example, it is not obvious that screening mammography achieves anything of value, even though we know with certainty that it detects many breast cancers. Unfortunately, population statistics suggest that the diagnosis of early stage breast cancers has risen enormously, while the diagnosis of late-stage breast cancers has not changed much. This strongly suggests that many of the newly detected early stage breast cancers were not going to manifest themselves or affect the lifespan of the patient, and there is no benefit to detecting them, whereas there has been harm because they end up getting a series of expensive and unpleasant treatments. It's true that more rational decision rules would shield us from a lot of his harm, because we would rationally choose to ignore the screen detected cancers a lot of the time. But the fact is that hasn't been the case, and patients and doctors have been led to overtreat many cancers through irrationality, enabled by the screening imaging.
The problem is, until the technology has positive utility for an individual patient you’re effectively asking people to donate their QALYs to Midjourney’s shareholders.
If Midjourney('s VCs) are functionally subsidizing the scans (à la Uber rides in its early days) then it might nevertheless be a positive-sum transaction.
I think my prior is so high that gathering more data is good that I'm still mostly unconvinced that gathering more data could be bad. I think the intuition pump of "your skin is an organ, it has weird growths and zits and things all the time, most of which don't matter, and it would be bad and annoying to get surgery every time you got a zit" did the most to shift my priors somewhat. But even here, I feel like this is downstream of a medical establishment that a priori assumes that tests will be expensive and so simply refuses to order more tests. Even though the best way to make something cheaper per unit is to massively increase demand and then let the market do its thing.
More bluntly: right now we assume all of the above is expensive and hard. But if you could wave a magic wand and make it so that these scans are actually super easy to do and you can just do them at the gym, then that would be awesome and a lot of the negative risks go away. The way you get to "machine you can run in the gym" is by first creating demand for that kind of thing, which justifies the innovation cycles necessary to make them cheaper
That intuition pump is what got me on team "more data is good". Or am I reading you backwards? We have lots of data on what our skin does, that’s *why* we don’t get surgeries for zits all the time.
this is a really boring simple point, and I'll only say it because I don't see it in the article: the empirical benefits/downsides of MJ medical don't really matter. They're building a tech-themed wellness spa; this has PMF regardless of the science.
Yeah, that's my opinion too. They'll either set up their own clinics or go into partnership with the kind of chain of clinics that already do these scans, and you go private and pay for the entire package, only NOW WITH AI!
Maybe make it an entire package where it's the new version of taking the waters:
(1) Well that doesn't seem creepy and cult-like, at all!
(2) If the guy could string two words together without needing filler like "y'know" and "like" and "um", I'd feel more confident in the wunnerful wunnerful tek. What, he couldn't write down (or get a chatbot to write it for him) a speech/blessing beforehand?
Always cellos for the musical invocations. I like cellos but c'mon David, this is Revolutionising Healthcare With AI, mix it up some - missed opportunity for a theremin! or the lirone/lira da gamba! or having your pet AI create a musical theme of its own utilising bricks and broken MRI scanners!
Lirone - an instrument I just discovered the other day:
I read a book by an oncologist a while back called "How We Do Harm," which I think is relevant. One aspect that is highlighted in that book, and which you mention briefly in your piece, is the profit motive and the set of bad actors waiting in the wings to make money from unnecessary additional screenings, unnecessary / non evidence based treatments, etc. Some of these treatments even cause harm themselves (e.g. chemo for an early stage cancer increasing the risk of leukemia). Opportunities for these types of bad doctors abound in a world where everyone is being screened for everything all the time and people are not perfectly rational actors.
Another piece to consider is the strain that additional screenings + procedures place on the medical system as a whole. There are pretty hard limits on the rate at which doctors are being produced in any given specialty: with a few notable exceptions, medical school costs an arm and a leg (you get to reattach them once you become a doctor); and whether we agree with their stated reasoning or not, the AMA limits the number of residencies per year. So we have to ask ourselves how resources get reallocated when we open up another tap of demand for limited doctors' time (radiologists, oncologists, etc) to start pouring in from elective screenings. I suspect the answer is similar to what in practice ends up happening with charter schools & vouchers: a (further) reallocation of resources away from the working class and to the wealthy. That means a step back from a society where medical care is accessible based on need, or more simply put, a less healthy population.
> But if the calculations above are right, and only ~8 people out of 1,000 benefit, then it would be hard for high-quality studies to measure the effect. You’d probably need hundreds of thousands of participants to find a signal.
Wouldn't you be able to estimate just the negative effects with a much smaller study?
I feel like, if someone told me "maybe you have cancer, come back in six months and we'll do another test and tell you for certain," that would cause me *much* more than -0.01 QALYs of anxiety.
I had an optometrist tell me that I might have glaucoma and needed an urgent appointment with an opthalmologist once. I spent the month between those appointments worrying about going blind.
Turns out my optic nerves just look kinda weird and I followed up every 6 months for three years. My mom told me she has the same thing, funny looking optic nerves but normal visual field and pressure, and she sees an eye doctor every 6 months and nothing has come of it.
They've detected polyps in my gallbladder for a while now, just below the threshold of "let's operate and get that thing removed". After the initial screening, I spent a week being very anxious about the possibility of surgery, the recovery, reading up on all the side effects of losing your gallbladder, thinking about how disruptive it was going to be on my work and life, etc. At the specialist appointment, they did more imaging and said "let's just monitor, come back in 6 months". It's now been ~20 months, and every 6 months we check back in and it's fine, same size. But I vividly remember that week of anxiety.
Another time I went to physical therapy for some hip pain. The PT said "oh this is concerning, it might be bone stress". Again, several days of severe anxiety about how this would impact my life and exercise program, how the treatment was 6 weeks on crutches loading it as little as possible, negotiating with Claude as to whether it'd be OK to walk just 2000 steps/day or whether that's too much, etc. And then the imaging doctor finds out it's a hip impingement, much less serious, and I go back to the PT to get some strengthening exercises.
I'm hopeful that with these experiences under my belt, I'll be more sanguine the next time something potentially concerning shows up. But you never know. Health stuff can just be really scary!
I remember this was what the influencer Attia was pushing for on the 60 minutes interview. What he said showed a lack of understanding of human psychology. Yes, those who want to should get it, but why should everyone get it? False positives can be very costly, especially emotionally.
A lot of the concerns about how people will react to likely benign internal zits can probably be solved by so scans so cheap they become mundane. Sure some people freak out about moles and rashes, but most are satisfied with "Don't worry about it, just let me know if it changes".
Exactly. If the scans became super cheap, that would solve a part of the problem. Plus once you have a suspicion about a specific body part, maybe next time it is enough to scan only that body part, which could be even cheaper.
None of these, for me, "epistemically defeat" the straightforward first principle argument that in single player games (cancer can't change its behaviour to thomas schelling gotcha you, and I'm assuming you hide this from your insurance company), it's straightfowardly good to get more information.
Most discusison about this is just emotionally biased cope, which is almost unavoidable for a subject that involves money, health policy, and has the touch of "thing health bilionaires do"
I wonder if the problem could be solved by doing some kind of rationality checks.
I imagine a general system where you could sign up for a lesson, do a test afterwards, and it unlocks access to some knowledge. We already have something like this for driving licenses, now imagine a more general thing that also unlocks e.g. access to drugs, or medical information, or investing money, etc. Simply, some actions require a knowledge check first.
> Simply, some actions require a knowledge check first.
I think in real life this basically happens in real time, with the doctor judging from your character and rationality from your past history of interactions, the types of questions you ask, your affect, and what they know about your background. I have a doctor who basically treats me like another doctor, because they know how much pubmed I read and the questions I ask over the ~10 years we've been interacting, and will readily prescribe pretty much anything I want to try, and will share any information or prognosis or treatment plan with no padding or worries about anything.
Also, this avoids the very salient failure mode that Jeffrey Soreff points out - in America, it would definitely turn into some sort of ridiculous tribal thing, just like "well *I* believe in *science™,* sirrah!"
I know you’re focusing on whole body here, but the existing organ-specific screening data also seems relevant to this analysis, since that would seem to represent a best case scenario in terms of suitability of the screening modality and risk in the targeted population. Despite this optimization, it’s surprisingly difficult to establish an effective screening program. If the WB program could achieve comparable performance to each specific modality simultaneously (unlikely), what does that upper bound calculation look like?
This doesn't help answer the question, but it seems like the scanner is a an ineffective technology. In most fields a test where 98.7% of flagged results didn't need attention would be seen as a failure and wouldn't be used.
These scanners have much better than 98.7% specificity when they're being used the way they're intended. Whole-body screening isn't one of those ways which is why doctors recommend against using it.
Scott, your ledger's two most generous assumptions are that being diagnosed with an indolent cancer costs ~nothing, and that treating it costs ~nothing. Both depend on a clinician being able to tell, at the moment of diagnosis, which of your screen-detected cases are the 4 who benefit and which are the 6 who don't. That's precisely the thing we can't do. (cf. the birds/rabbits/turtles framing of Crile and Welch, which you allude to in your post - which is itself a statement that the heterogeneity is invisible in advance.)
You don't actually get to tell the turtle "just be chill"; you treat it like a rabbit, because you can't distinguish them, and the patient absorbs the full treatment harm. As a clinical researcher who has worked in exactly the field of cancer treatment-related harm, this is non-negligible. Screening-related false-positives still get the extra treatment-related harms, just not the benefits, but this can't be known in advance. More people die of heart disease than cancer, but cancer if often the scarier diagnosis, because treatment sucks so much.
PSA is a clean example, and not an abstract one for me: my grandfather believed his whole life that a PSA test made him a cancer survivor, when what it demonstrably gave him was a prostatectomy and lifelong incontinence, for a cancer there's no good evidence ever needed treating. The ERSPC trial showed how common that is: at 16 years, ~18 men had to be diagnosed (and nearly all treated) to prevent one prostate-cancer death, for a ~20% drop in prostate-specific mortality - and no detectable change in all-cause mortality!
Bretthauer et al.'s 2023 meta-analysis (JAMA Internal Medicine) found that the established, trial-backed single-organ screens barely extend overall lifespan at all. Every prior time a plausible cost-benefit case got deployed ahead of the trial - the radical mastectomy stood for most of a century before NSABP B-04 showed it saved no lives over simple resection - the unmeasured harms turned out bigger than expected. We struggle to make a solid argument for single-disease scans, let alone general scans that might pick up ... something we're not sure about yet.
This is my concern. Not that we're spending a bunch of money that some people are totally willing to burn on health screening. I'm not opposed to rich people wasting their money on things that might well make them less healthy - let freedom ring! But that we're repeating the exact same mistakes we've made time and again with screening, just with a brand-new modality. Can we please stop selling these untested screenings as "potentially beneficial with almost no chance of harm" when the history of screening in medicine should point us to exactly the opposite conclusion?
I appreciate it. I think you're referring to this new paragraph, which strengthens your case: "Fifth, it assumes that when all the most accurate tests are done and all the best doctors have been consulted, it’s at least possible to distinguish the cancers that should be treated from the slow growing ones that won’t hurt them. This isn’t always true, and sometimes people will be completely rational and still make the wrong choice. We can get this number arbitrarily low by setting stricter and stricter thresholds for which cancers we treat, but the stricter we go, the lower we drive number of true positives caught, until eventually our 32 QALYs gained number becomes a large overestimate. This decreases the benefits some amount, but I don’t have a good sense of how big an effect it is."
My prior is that new tests should be treated as a potential net detriment (i.e. the potential benefit is often swamped by the potential harm from treating a healthy person - and more so when looking at multiplex tests with lots of potential sources of (false) positives) until we can demonstrate net benefit.
Vinay Prasad talks about a case in his own clinic, where an otherwise healthy heavy smoker in his mid-90's came into the clinic for a checkup. He had no specific complaints, but they ran a few routine tests on him anyway, because of his long smoking history. Those tests brought up concerns, that were followed by more tests, treatments, scans, add-on treatments, etc. Eventually, Prasad realized this poor man had almost certainly received worse quality of life from his interaction with his health provider than if he'd stayed home and let the asymptomatic disease run its course. He had a strong probability of dying of something else. Risk factors, after all, should never be treated as certainties at the level of an individual.
Look, I like prevention. But that old saying that "an ounce of prevention is worth a pound of cure" often describes the opposite of what people are being sold. People don't come into the clinic to get their HbA1C down, or because they're really suffering from an eGFR that's too low. We track those things to prevent worse harms, sure. But it's always necessary in medicine that we be able to point to the hard endpoints - like overall survival or all-cause mortality - that prove we're doing more good than harm. And yes, interaction with the health care system is real, tangible harm - more so the more serious the disease, because we're more willing to take drastic steps. "Show me a healthy person, and I'll show you someone who hasn't had enough tests," is more than just an adage. It's a cautionary tale, when the test could throw an otherwise healthy person into a life-or-death struggle.
Finally, while I'm often critical of people applying Bayesian statistics in places where they don't belong, this subject is one of the best/easiest places for a Bayesian approach. If your cancer risk is high enough (you're older, maybe have family history of cancer death) there are good arguments that the net benefit of screening is positive. But this isn't being rolled out with those qualifiers or that intended target audience. It's being rolled out as potentially good data that anyone could benefit from and maybe there are harms, but the treatment couldn't possibly be worse than the disease. (It can, when there's no disease and all treatment.) A rational 35yo otherwise healthy male with a base rate of 0.07%/yr. chance of new malignancy (many of which would be turtles or birds, not rabbits), who gets a positive screening screening with 99% specificity like this is raising their chances of iatrogenic harm, without sufficiently tipping the scales in favor of intervention. Our prior should be that they are more likely to be better off NOT knowing the results of the scan than knowing them.
Thank you for this; I think many people here are really underestimating the anxiety impact and real harms cancer treatments cause, and you've put it much better than I could.
I appreciate your comment. Yeah, there's something about a cancer diagnosis that's really different.
And the treatment is horrible. A rare side effect of cancer treatment is ... cancer. Even if you cure it, you'll still come back every 3 months for scans, then every six months, then every year... Each scan a new expense and a reminder of the cancer.
Even a diagnosis of a relatively benign cancer gives you years of anxiety. It's not just the resection and adjuvant chemo (as bad as they are). It's the year 5 scan that reminds you to not get too comfortable...
I'm not sure how we'd handle this with QALYs, but in finance when making calculations like this you discount the future. You can usually use the risk free rate or expected market returns or whatever, depending on the exact situation. But the point is you don't value $1 10 years from now the same as $1 today.
Does the same apply here? I certainly wouldn't want to waste 899 hours of my prime life to save 900 hours decades from now.
It complicates the calculations, but I think at a high level the costs are realized sooner and the benefits later which nudges the result further in the "not worth it" direction.
Separately, I've sometimes wondered if it would be useful to get a full body scan and not look at it, but then I'd have it as a baseline if standard medicine asked me to get one in the future. I feel like that might have some of the benefits but without the anxiety and false positive costs. But I'm not sure if that would actually help in any medical protocols.
"The other six either have such slow-growing cancers that they would never have noticed before dying of something else"
^^^ This is a major source of overtreatment that could substantially change the calculations depending on which cancer is being detected. For annual PSA screening for prostate cancer, for instance, the QALY benefit ranges from -21 to 97 per 1000 (https://pmc.ncbi.nlm.nih.gov/articles/PMC4982868/) meaning there's a possibility screening is net harmful. A big challenge for screening programs is being most sensitive to the lowest risk cases. This issue comes up for genetic screening / polygenic scores in surprising ways: the people most likely to participate in cancer GWAS are the mildest cases and so the score is being trained on the least aggressive subtypes.
The possible harms don't stop at anxiety and biopsies, though. You also need to consider the risk of significant harms from over-treating indolent cancers that we can't reliably differentiate from aggressive cancers, like what can happen with impotence and urinary incontinence from over-treatment of prostate cancer in response to prostate specific antigen screening.
Ah yeah both my father and I had life threatening conditions that would have been easily caught with whole body scans. He has permanent damage to his bladder and has to catheter 3 times a day now because his doc did not send him for a scan. In my case I had to go to the er and demand a scan after my insurance rejected it. As it turned out I was going into organ failure and would have been dead within 12 hours without that scan. So yeah I am a big fan of whole body scans.
CT scan with contrast. I had a burst appendix, peritonitis, and was going septic. The ER surgeon said he could not explain how I was a) still alive, b) had walked through the front door of the hospital on my own power.
Although it has been suggested, I do not have an autism diagnosis but probably have ADHD or some sort of neurodivergence. I definitely have a very high pain tolerance. Initially, when it burst, I had a lot of pain and nausea but thought I had food poisoning from some fish tacos at a dive bar. The next day I just felt sick, no fever, no pain. By the next night I felt better and had some soup, then had bad shakes and very high fever (yes I should have called 911 at that point) but took some advil, had a cold shower, fever came down I sweated it out overnight and thought "Yes I have an awesome immune system." So on the 4th day I thought it odd that I still had a low level fever, went to urgent care, they took some blood tests they thought were odd and wanted me to get a scan but insurance denied it. I did some grocery shopping and walked 30 blocks home. Took a nap. Woke up to a call from the urgent care doctor who said "I've been thinking about this and I think you should go to the ER and ask for a scan." I said okay, got up went to the bathroom and pissed blood because my kidneys were shutting down. Took an Uber to the ER. After the scan the doctor asked me if I had a medical condition that prevented me from feeling pain (by that point I was actually in a lot of pain, and they gave me 2 shots of morphine). I asked why and that's when he said he had never seen someone as sick as me still conscious. Anyway, 10 days in the ICU, and 3 surgeries later, doing okay.
This analysis reminded me of the work of psychologist Gerd Gigerenzer. He has written extensively on the harms, overdiagnosis, and poor statistical communication surrounding prostate-specific antigen (PSA) screening for prostate cancer. He wrote "Cancer screening does not save lives." This editorial published in Healthcare in Europe, details how patients are frequently misled about the true benefits and harms of cancer screening.
Other publications along these lines include "Public knowledge of benefits of breast and prostate cancer screening: A cross-validation in nine European countries" – Published in the Journal of the National Cancer Institute. This article analyzes how vast numbers of men grossly overestimate the benefits of PSA tests while being uninformed of the risks of overdiagnosis.
"Effect of Tabular and Icon Fact Box Formats on Comprehension of Benefits and Harms of Prostate Cancer Screening: A Randomized Trial" – Published in Medical Decision Making, this study explores how presenting clear, transparent statistical facts impacts patient decisions.
Wait back up, this implies the average seemingly-healthy person has a 1% chance of currently having cancer? That seems distressingly high
(It passes the smell test - if about half the people ever get cancer and life expectancy is 80 years that's about 1/160 of getting a new cancer every year. But this includes old and unhealthy people and assumes you could have cancer for a year without knowing about it until you took the full body scan. I'm guessing most cancers affect the elderly more? Man, fun new source of constant medical anxiety).
Take the prostates of men (who died of something other than prostate cancer) at autopsy and check them for undetected prostate cancer. Results: latent prostate cancer present in 5% of men aged <30 years, increasing to 59% at age >79 years.
As a male in the USA your lifetime risk of symptomatic prostate cancer is ~10%, and you have a 3% chance of dying from the disease. Risk is greater for African Americans. Anyone you treat for prostate cancer risks incontinence and impotence for the rest of their lives.
20+-year oncologist here: the data clearly and repeatedly shows that we treat too many breast cancers that would never be dangerous, and 5x that many never-life-threatening prostate cancers (among many others, but those are the top two in women and men respectively, so a good, non-cherry-picked sampling).
And the QOL-cost of this treatment is truly massive, beyond what you report here Scott. Specifically, for both of those cancers we give treatments that take away people’s estrogen/testosterone, and in so doing take away a big part of who those people are, how they function, and even how they see themselves as humans. We change their lives, permanently, and not for the better.
The reason experts advise against asymptomatic whole body scans (and avoid them ourselves) is because we believe the data, not to mention our own eyes.
> "[We] advise against asymptomatic whole body scans ... because we believe the data, not to mention our own eyes"
I understand and am sympathetic to the pragmatic logic, but remain stuck on the question:
Why can't the people who "treat" the innocuous cancers believe the data, and make good decisions?
Roughly:
- inertia + cultural presumption that aggressive early intervention would be positive
- general bias toward action / involvement (surgical industrial complex has incentive survival mechanisms?)
- pressure from relatively low-info high-anxiety patients
- ignored positives present higher legal liabilities than treated negatives
The first two should be easy enough to overcome if we believe medicine has anything like a scientific or technocratic arc.
The third - evidenced in this comment section by a Korean anecdote - I believe is a cultural artifact that would settle on a new equilibrium pretty quickly if mass-market scans revealed that everybody's uncle, cousin, and neighbour was carrying weird internal anomalies.
The fourth is furthest from a natural / automatic solution, but it does seem tractable in terms of documented evidentiary decision making? Individual patients will be outliers in so far as their conditions are actually much worse / more urgent than our understanding lets us guess.
My case: I was constipated for years after a motorcycle accident. Got worse and worse until I was going about once every two weeks and was basically living on soup. I saw multiple doctors who just told me to eat more fiber (I was in my 20s). Out of desperation: I got a prenuvo scan - I had an internal hernia. Got surgery and felt loads better. Night and day. No regrets!
This makes me wonder what the QALY yield would be from a protocol of "use whole body MRI as a backup diagnostic tool for people who have weird persistent debilitating/painful/annoying symptoms not resolved by the easy/simple stuff." Where you set the threshold for debilitating/painful/annoying probably matters, but seems plausible that it might target a higher yield population than the worried well.
So you sought and received a scan to identify the cause of a symptom, right? This is an argument against receiving scans when there are no symptoms or problems to focus on.
Is there no value for the ~2/3rds of patients who got a full body MRI, and it came back clean? Why would giving "peace of mind" to someone who might be nervous count for zero QALYs, while giving the remaining ~1/3rd "anxiety" count for negative QALYs?
If the roi of promethease were calculated this way, what would it show?
No, because 1) although it gives people many more negative hits of info 2) people have adjusted
I don't think it's right to port over the costs unchanged from a world where patients believe "tests are all meaningful and expensive and only docs can request them, so a potential problem is serious" and then plug them into a world where scans are way cheaper and faster and don't need a doctor.
By not accounting for adaptation, this kind of analysis seems like it would show that most tech or system changes should not be or have been made.
First of all, this analysis tries to minimize the effect of actual irrationalities and cultural problems, and show that even if all of those get assumed away, this is still low yield.
Second, I don't think we should assume those away. Segways might be a revolutionary invention if everyone switched them all at once and we replaced all of our car infrastructure with Segway infrastructure, but the fact that they require this unlikely thing means they probably aren't worth you personally buying unless you're feeling incredibly optimistic. Likewise, every dating site would be great and life-changing if the culture changed so that all of the highest-value singles in the world used that dating site, but that doesn't mean that dating site is good today, or that you should use it, or that it's a genius revolutionary invention. Coordination problems are often the hard part, and if you've solved everything except the coordination problem then you often haven't done very much.
What do you think uptake would look like? I modeled a gradual rollout with those especially needing it (people with DNA or family history reasons to rationally want to be scanned more often) self selecting in.
And in the background as a pressure on the mainstream, all good things to me. A broad day one rush of likely false positives would be bad - but I'm curious to know if that's something we couldn't adapt to?
The value "try to know more things about health, medicine etc" is so close to what I care about, I didn't expect the reactions against it so strongly. (I'm only referring to the case it works and eg raises scan rates 100x due to cost but with 2x worse false positives. if it doesn't work at all then it'll fail w/out opposition.) What could happen which would really lead to big problems?
I'd expected docs to be more scared to appear openly against generic "progress" measured by naive abilities like speed, etc.
I think you alluded to this in your first post, but the base rate for "newly expanded screening procedure for healthy adults improves health outcomes" is very poor. Are there *any* screening procedures (of any kind) that are currently recommended for healthy adults aged 18-40? HIV test for sexually active adults maybe? Even a lot of the traditional ones (mammograms for example) are not slam dunk in terms of cost benefit in the general population.
Also I think you downplay the risk of biopsies a little bit: we are not always talking about minor side effects; you can get antibiotic-resistant infections, sustain serious brain or organ damage, or die from a biopsy. Rare, yes, but it does happen -- sometimes I think people just brush it aside into the category of "the band-aid might cause skin irritation."
I think the "people won't do the basics" is outdated. Almost 50% of adults get 150 minutes of mild exercise a week, and 25% do "muscle strengthening". 20% of people go to a gym/fitness class in a given week. These numbers include all adults, including decrepit elderly people. Among 18-30 2/3rds of people meet exercise requirements. Basically every food product line is updating as people learn that protein and fiber are good for you. The same doctors who complain that people won't lose weight or change anything are annoyed that people keep coming to them proactively for things they deem not "real" health issues like trouble sleeping or minor pain.
There are a ton of people willing to make pretty large sacrifices for health and wellness.
How does the doctor's liability concerns fit into this calculation?
In spherical cow land, the doctor is thinking Pr[this finding is treatable cancer] vs Pr[this finding is no biggie/not treatable/doesn't need to be treated]. But it seems like in reality, at least in the US, the doctor is also thinking Pr[this finding looks innocent but isn't and I get sued by this guy's heirs for a gazillion dollars and my malpractice insurance goes up by 200%].
The massive QALY burn from chasing underspecified MRI anomalies is an exact manifestation of a broader, systemic failure in how we apply Bayes. The medical system consistently forces continuous, structurally underspecified reality into discrete, fully-specified models.
The mathematical boundary of this error, what I refer to as the Observational Sufficiency Principle (OSP), is by reverse-engineering the most notorious example of Bayesian overconfidence: the textbook 2/3 consensus on the Monty Hall Problem.
The standard Monty Hall solution commits is a Fully Specified Stochastic Process (FSSP) error. To force the math to work, the analyst hallucinates a generative protocol for the host (q=0.5) that was physically unobservable. They import a spurious rule of selection to complete the model.
This is exactly what is happening with whole-body screening. When an MRI flags a 'mildly concerning' shadow, the generative process of that shadow is entirely underspecified. But instead of treating it as an equivalence class, the diagnostic model forces an FSSP onto it, hallucinating a precise risk likelihood that generates the 300 false-positive follow-ups Scott calculated.
When you apply a rigorous quotienting step to an actual underspecified observation space - restricting the σ-algebra to only what is admissible (the equivalence class under permutation π) - the posterior is mathematically forced to respect the indistinguishability, dropping the overconfident probabilities back to 1/2.
The insistence on completing underspecified models with post-hoc likelihoods is why predictive models hemorrhage value, whether on a game show stage or in an MRI machine.
The full π-invariance quotienting proof and taxonomy of the FSSP category error are mapped out here:
Ah, apparently your "Monty Hall Problem" is the default, non-verifiable general consensus existence of a vanilla Fully Specified Stochastic Process (FSSP), calling itself "The Monty Hall Problem". I agree; under that model, the correct answer is 2/3.
But the historic, textbook-verified (via Priority Rule) Monty Hall Problem, one which bears everyone's intention to clarify (not from those providing the consensus answer), actually refers to themselves in a one-shot scenario with the uncertainty to be quantified for the expected payout. Because they imagine themselves standing now in front of two doors instead of three, wanting to win the car and asking themselves which door has the advantage. That Monty Hall Problem's correct answer is 1/2.
I have now written over 60 long-form posts regarding the Monty Hall Problem. Below is the 61st, specifically for you. Understand that this was done not only out of pure love for the problem, but also because I simply enjoy every opportunity to create a bespoke explanation.
>"The game starts by you picking any door you wish."
At this point you have a 1/3 chance of picking correctly, and a 2/3 chance it's behind one of the other doors. Nothing that happens afterwards can change that fact.
>"Monty's testimony since 1975 that he doesn't do so."
Irrelevant. There's a reason it's called "the Monty Hall problem" and not "Let's Make a Deal": it's inspired by but not isomorphic to the game show.
"A few years ago, my aunt had some concerning symptoms and got a scan. The scan showed she probably had cancer. The doctor inexplicably ignored it for several months and she died. RIP."
Not to comment on Scott's aunt, but I think doctors prefer not to jump straight off to "it could be cancer" because of the fear of terrifying the patient. I had that with an appointment for a particular minor surgical procedure, where I could tell the consultant suspected it might be cancer but refused to tell me, so of course when I got home I immediately looked it up online and yes, the result came back "with this set of symptoms in this type of patient, first suspicion is cancer".
I didn't freak out because I knew it wasn't cancer (for various reasons too detailed to go into here) so I was more annoyed the doctor wouldn't be honest with me, but to be fair, had he said "it could be cancer" and had I been less certain of my grounds, I probably would have worried myself into oblivion over what turned out to be, when the results of the procedure came back, "no it's not cancer".
This is an impressive ICER exercise. But as you’ve alluded to, any ICER is predicated on the intervention actually being effective…and I’m not aware of any outcome evidence of whole body MRI to that effect. You’ve made some best (and no doubt learned) guesses here but of course that doesn’t replace outcome evidence.
In the end, I suspect this will appeal to the subset where money is no object….the folks who do stuff because they can, even absent evidence that they should (probably considerable overlap with the longevity types also). In that way, this is part and parcel of the whole boutique/concierge/executive medicine trend that exists to cater to the worried well.
You say "In the past when doctors have come up with clever cost-benefit analyses like the one in this post, these sorts of known unknowns and unknown unknowns have predominated, and the real effects have been much different from the expected effects - usually much worse."
As opposed to whole body MRI for which there is no outcome data, and whole body US which has even less, what are your feelings about PSA and/or colonoscopy, given the body of evidence available regarding those tests/procedures?
Thanks for that. That’s a good summary and discussion of NordICC. Intention to treat analysis is the gold standard in RCT as it preserves the effect of randomization (as opposed to on-treatment or per-protocol)…and the screening strategy IS the “treatment” here…so I don’t understand the fuss. If someone did an RCT about a surgical procedure and only included those in the “treatment” arm who actually underwent the surgery, they would be roundly mocked, and deservedly so.
I'm not super optimistic about full body scans as the best investment in medical technology, but Midjourney's plan before this was to marginally improve image generation models. Image generation is mostly zero-sum imo (a bit of a hot take, but can we at least agree a lot of the rewards to Midjourney come from competing with other mildly inferior image sources instead of providing actually new and good services?) So it's better for them to be working on full body scans even if they're very overhyped, as long as they somehow won't do a bad enough job to set the field back worse than had they never entered. The only exception is if the availability of full body scans is somehow negative, and even if they're bad for most people there has to be some way to restrict it to people who can benefit, like those who can't have an MRI for whatever reason.
I think it's important to consider that we have arguably already caused great harm and wasted a great many resources through overzealous screening. H. Gilbert Welch has some nice work suggesting that screening mammography and other popular forms of cancer screening have not actually saved lives but have led to many, many more people receiving toxic therapy that couldn't have helped them.
So while I think it's peobably true that screening which costs nothing and is undertaken by perfectly rational actors cannot harmful, we already have several decades of history of screening being harmful.
Is there a good writeup somewhere on screening tests, false positives, mean time to survival stats, and the like? Many years ago, the blog Respectful Insolence had a really excellent sequence of posts explaining this, but I couldn't find them the last time I looked. (Notably, you get the fun paradox where mean survival time goes up because you are doing more early detection of some untreatable cancer--nobody actually lives longer, they just find out about their cancer a year or so earlier.)
I'm not aware of a really good written summary of the entire field, but a lot of H.G. Welch papers are quite readable and describe the phenomenon convincingly. I'd recommend:
"Effect of Three Decades of Screening Mammography on Breast-Cancer Incidence"
Authors: Archie Bleyer, M.D., and H. Gilbert Welch, M.D., M.P.H. Publication: New England Journal of Medicine, 2012
"The Rapid Rise in Cutaneous Melanoma Diagnoses." Authors: H. Gilbert Welch, M.D., M.P.H., Benjamin L. Mazer, M.D., M.B.A., and Adewole S. Adamson, M.D., M.P.P. Publication: New England Journal of Medicine, 2021
"Are Increasing 5-Year Survival Rates Evidence of Success Against Cancer?" Authors: H. Gilbert Welch, MD, MPH; Lisa M. Schwartz, MD, MS; Steven Woloshin, MD, MS Publication: JAMA, 2000
This suffers from spherical cow in another direction. Even if everything above is correct and even optimistic: we're moving FAST in a direction where the capacity to analyze data increases by orders of magnitude (plural). What Claude Code can do with your genome, clinical data and MRI history is already more than any single doctor will ever have the patience to do. It currently needs guidance and common sense provided by a human - but It's a matter of months until you get harnesses which prevent its worse mistakes. And then you get Mythos gen models and better harnesses.
So not else is being equal. 3 years ago this all was, at the very least, debatable. But now the "more data" camp wins hands down.
I always thought one of the main objections by doctors to scanning healthy patients is because MRI machines, operators of the machines, and interpreters of the scans, are all constrained resources, that are better distributed to patients who are sicker.
I also think that if such scanning and the knowledge of "minor cancers" became much more commonplace, the general sentiment and level of anxiety regarding them would calibrate itself. re the comment on Taiwan
If the patient is paying, that's not a big problem, since the money from the patients can be spent buying more scanners and hiring / training more technicians and radiologists and such.
1. If the cost/benefit is so close, why do doctors insist that multiple large trials of preventative screening have been done and failed, across multiple types of cancer?
2. Why are we even talking about cancer in response to the Midjourney announcement? AFAICT they said nothing about cancer or tumors in any of their announcements. All this made sense several months ago when we were talking about Prenuvo, but all Midjouney has said they will do is provide body composition info and then later apply to the FDA for more stuff.
I think the elephant in the room is the value of the data, independently of how they are immediately used to benefit individuals patients. Imagine you collect a few millions of full body scans and you also follow up the patients, learning long term outcomes. You will be able to train a predictive model that may turn out to be invaluable for future patients. Whether you see this as a data grab or as a great opportunity for future patients, this is likely to matter more than the impact on current patients.
>Imagine you collect a few millions of full body scans and you also follow up the patients, learning long term outcomes.
Is any organization in a legal position to do this? There is the NIH All of Us research program, but that has a pool of volunteers. Given health information privacy laws (which exist for good reasons), I'm a bit skeptical that millions of follow ups are feasible.
I find those tweets curiously enraging, and I think it's because these are the types who would otherwise be smugly "trust the science, bro; look, just hand over control of society to the technocrats and it'll be perfect" but when it comes to this, then they're all "waaah, why is the mean ol' Harvard physician not letting me, a tech head, play with the shiny new toy I want to play with? what does he know, anyway!". Also perhaps a hint of "I am so well-established in life that I can afford the kind of health insurance that will pay for fripperies, because I am Smart and Educated and High Value Human Capital" at play as well.
You know what, Midjourney? I was highly dubious about this entire project but now I say "Go right ahead. Slap AI on everything! Soak these pigeons for every dollar you can squeeze out of their high value backsides! Make it an entire spa package and charge 'em $$$$$$$ for what in effect is a Seltzer bath with knobs on!"
I figure these are the types who probably were all over NFTs as well, so if they're going to throw money away, why not scoop it up in a pivot from "we help you draw your own porn tailored to your specific kinks" to "ah yes, the field of medicine, a vast plain of opportunity for quackery dressed up in the latest buzzwords".
"If you don't otherwise have symptoms, ignore the scan", forsooth! Hey, Amanda, how's about if you don't otherwise have symptoms, you don't get the scan in the first place, and save all the time and bother?
If these people want 24/7 monitoring of their health so they can gather data, then I say let 'em have it, hot and heavy. Blood draws every hour on the hour. All the poking and prodding and hanging around and "take off all your clothes and put this skimpy paper gown on and then sit around waiting until a doctor deigns to poke his head around the door" galore. Tests, tests, and more tests, till they don't know whether they're coming or going.
They should get their blood the old-fashioned way; go to a kosher butcher and drink it fountaining straight from the slashed throat of the animal.
(Years and years back I read something that this was a cure recommended to consumptives, to drink animal blood, and that there was a difference between blood from animals slaughtered the usual way and kosher slaughter).
Well, if it actually works, the next step is to figure out what _specifically_ is the useful component in the 18-yr-old blood plasma, and gene-engineer yeast to synthesize it, at which point it will be generally useful to all of us. _IF_ it actually works. Look at it this way: Those guys have volunteered to be lab rats for the rest of us. Remember that virtually _all_ of medicine looked "creepy" when first introduced - look up the original comments on the first smallpox vaccine, for instance.
And the creepiest thing of all are the politics of envy.
You don't think it's at all slightly creepy to be treating your own son as a pomme de sang (to borrow a term from the Anita Blake novels)?
There's guinea pigs, and there's 'holy hell, what???' Organ transplant because family member is dying otherwise, and there's no better match - okay. "Dad is scared spitless he'll get old and die like every other human so far, so he's scrabbling about grasping at straws" - nope, sorry.
If the guy is doing it to 'look 18' as per the article, it ain't working. He looks very sallow and anxious, and if he's this scared of getting old (and eventual mortality) no wonder he looks like he's living on his nerve ends.
Swear to me aunt, "Good Lady Ducayne" to the life:
"For Mr Bryan Johnson, the plasma exchange is not an unusual happening.
He has been to the Dallas-area clinic for several consecutive months and received plasma–not from a family member but from a young, anonymous donor.
He carefully screened the donor to make sure the person had an ideal body mass index, lived a healthy lifestyle and was free of diseases."
Lady Ducayne:
"Bella went in fresh, blooming, a living image of youth and hope, and before she looked at the Person her gaze was riveted by the owner of the chariot.
Never had she seen anyone as old as the old lady sitting by the Person's fire: a little old figure, wrapped from chin to feet in an ermine mantle; a withered, old face under a plumed bonnet--a face so wasted by age that it seemed only a pair of eyes and a peaked chin. The nose was peaked, too, but between the sharply pointed chin and the great, shining eyes, the small, aquiline nose was hardly visible..'This is Miss Rolleston, Lady Ducayne.'
Claw-like fingers, flashing with jewels, lifted a double eyeglass to Lady Ducayne's shining black eyes, and through the glasses Bella saw those unnaturally bright eyes magnified to a gigantic size, and glaring at her awfully.
'Miss Torpinter has told me all about you,' said the old voice that belonged to the eyes. 'Have you good health? Are you strong and active, able to eat well, sleep well, walk well, able to enjoy all that there is good in life?'
'I have never known what it is to be ill, or idle,' answered Bella.
'Then I think you will do for me.'
'Of course, in the event of references being perfectly satisfactory,' put in the Person.
'I don't want references. The young woman looks frank and innocent. I'll take her on trust.'
'So like you, dear Lady Ducayne,' murmured Miss Torpinter.
'I want a strong young woman whose health will give me no trouble.'"
>You don't think it's at all slightly creepy to be treating your own son as a pomme de sang (to borrow a term from the Anita Blake novels)?
>There's guinea pigs, and there's 'holy hell, what???' Organ transplant because family member is dying otherwise, and there's no better match - okay.
Aging _is_ a lethal disease. As things stand, it eventually kills all of us. As I said, and I continue to stand by it:
>Remember that virtually _all_ of medicine looked "creepy" when first introduced - look up the original comments on the first smallpox vaccine, for instance.
To add another example, since it is now part of this thread: There has been (and, to a smaller extent, _continues_ to be) a _lot_ of wringing of hands about transplants. Now, there are reasonable questions about a lot of medicine: efficacy? side effects, short term and long term? limited supplies of various sorts? etc. But 'creepiness' is now, and has always been, a lousy criterion for evaluating medical experiments.
As with any slowly moving malady, attempted treatments for aging _do_ have a very high bar to clear. Ensuring that some novel treatment of e.g. a 50 year old to prevent them dying in 30 years _doesn't_ have a side effect that kills them in 20 years is a very tough problem.
Yet again, over-anxious people with more money than sense turning "Good Lady Ducayne" (published 1896) into a 'how-to' manual rather than a work of supernatural fiction:
"She pushed up her loose lawn sleeve, and exhibited a scar, which he scrutinized intently, with a surprised and puzzled look.
'This is no mosquito bite,' he said.
'Oh, yes it is--unless there are snakes or adders at Cap Ferrino.'
'It is not a bite at all. You are trifling with me. Miss Rolleston--you have allowed that wretched Italian quack to bleed you. They killed the greatest man in modern Europe that way, remember. How very foolish of you.'
'I was never bled in my life, Mr Stafford.'
'Nonsense! Let me look at your other arm. Are there any more mosquito bites?'
...Stafford examined both her arms in the broad sunlight, scars new and old.
'You have been very badly bitten, Miss Rolleston,' he said, 'and if ever I find the mosquito I shall make him smart. But, now tell me, my dear girl, on your word of honour, tell me as you would tell a friend who is sincerely anxious for your health and happiness--as you would tell your mother if she were here to question you--have you no knowledge of any cause for these scars except mosquito bites--no suspicion even?'
...'How long do you mean to stop with Lady Ducayne, Miss Rolleston?' Herbert Stafford asked, after a thoughtful silence, breaking suddenly upon the trivial talk of the two girls.
...'Even if you feel your health breaking down in her service?'
...'Then you mean to go on till you break down; to die at your post?'
'Like the other two companions? No! If ever I feel seriously ill--really ill--I shall put myself in a train and go back to Walworth without stopping.'
'What about the other two companions?'
'They both died. It was very unlucky for Lady Ducayne. That's why she engaged me; she chose me because I was ruddy and robust. She must feel rather disgusted at my having grown white and weak. By-the-bye, when I told her about the good your tonic had done me, she said she would like to see you and have a little talk with you about her own case.'
...'You have begun upon my companion, she tells me.'
'I have prescribed for her, certainly, and I am happy to find my prescription has done her good; but I look upon that improvement as temporary. Her case will require more drastic treatment.
'Never mind her case. There is nothing the matter with the girl--absolutely nothing--except girlish nonsense; too much liberty and not enough work.'
'I understand that two of your ladyship's previous companions died of the same disease,' said Stafford, looking first at Lady Ducayne, who gave her tremulous old head an impatient jerk, and then at Parravicini, whose yellow complexion had paled a little under Stafford's scrutiny.
'Don't bother me about my companions, sir,' said Lady Ducayne. 'I sent for you to consult you about myself--not about a parcel of anæmic girls. You are young, and medicine is a progressive science, the newspapers tell me. Where have you studied?'
'In Edinburgh--and in Paris.'
'Two good schools. And you know all the new-fangled theories, the modern discoveries--that remind one of the mediæval witchcraft, of Albertus Magnus, and George Ripley; you have studied hypnotism--electricity?'
'And the transfusion of blood,' said Stafford, very slowly, looking at Parravicini.
'Have you made any discovery that teaches you to prolong human life--any elixir--any mode of treatment? I want my life prolonged, young man. That man there has been my physician for thirty years. He does all he can to keep me alive--after his lights. He studies all the new theories of all the scientists--but he is old; he gets older every day--his brain-power is going--he is bigoted--prejudiced--can't receive new ideas--can't grapple with new systems. He will let me die if I am not on my guard against him.'
"If any real person went to their doctor for an hour daily for years to decrease their risk of cancer by 5%, we would call them an insane hypochondriac!"
The good old-fashioned term for doctor-botherers is valetudinarian, and the likely market for this procedure. Perhaps we should call them Sholtos, in honour of the example of the species as per Conan Doyle's "The Sign of the Four":
“This is Mr. Sherlock Holmes, and this is Dr. Watson.”
“A doctor, eh?” cried he, much excited. “Have you your stethoscope? Might I ask you—would you have the kindness? I have grave doubts as to my mitral valve, if you would be so very good. The aortic I may rely upon, but I should value your opinion upon the mitral.”
I listened to his heart, as requested, but was unable to find anything amiss, save indeed that he was in an ecstasy of fear, for he shivered from head to foot. “It appears to be normal,” I said. “You have no cause for uneasiness.”
“You will excuse my anxiety, Miss Morstan,” he remarked, airily. “I am a great sufferer, and I have long had suspicions as to that valve. I am delighted to hear that they are unwarranted. Had your father, Miss Morstan, refrained from throwing a strain upon his heart, he might have been alive now.”
...“That is well! That is well!” said he. “May I offer you a glass of Chianti, Miss Morstan? Or of Tokay? I keep no other wines. Shall I open a flask? No? Well, then, I trust that you have no objection to tobacco-smoke, to the mild balsamic odour of the Eastern tobacco. I am a little nervous, and I find my hookah an invaluable sedative.” He applied a taper to the great bowl, and the smoke bubbled merrily through the rose-water.
...I stammered out some few halting words of congratulation, and then sat downcast, with my head drooped, deaf to the babble of our new acquaintance. He was clearly a confirmed hypochondriac, and I was dreamily conscious that he was pouring forth interminable trains of symptoms, and imploring information as to the composition and action of innumerable quack nostrums, some of which he bore about in a leather case in his pocket. I trust that he may not remember any of the answers which I gave him that night. Holmes declares that he overheard me caution him against the great danger of taking more than two drops of castor oil, while I recommended strychnine in large doses as a sedative."
> So in total, 8 of the 1,000 people have benefited. The average benefit among those 8 people is 4 quality-adjusted life-years, so the total benefit from the scans is ~32 QALYs.
> What’s the total cost?
I don't think this is a good framework in which to view the policy issue. One benefit of doing regular scans of normal people is that you develop a sense of what normal scans look like. (And also a sense of what abnormal scans look like!)
This means that the scans become more valuable over time, because we can interpret them more reliably.
Arguing that they're not very valuable 𝗻𝗼𝘄 misses this point.
> A onetime snapshot of a "could be benign or slow-growing" cancer is unlikely to be better in the future.
This is the opposite of what I'm saying. You're wrong in two ways:
1. A onetime snapshot of a mysterious lump is in fact likely to be better in the future;
2. Cancers are not the only thing we can see on internal scans. Cancers are what motivates the scans now. With better understanding of what we can see, we'll look at more things.
Is QALYs the right measurement for understanding this? Are the doctors who recommend patients not get screening procedures optimizing for QALYs? Like if suggesting screenings gets a doctor 8 5-star reviews and 992 1-star reviews,and suggesting no screenings doesn't, then the ""best"" doctors will suggest no screenings and have plausible reasons why.
How I know I'm not The Exception: routine checkups eventually surfaced that "you probably have prolactinoma, but it's a benign brain tumor, don't worry"...and this freaked me the hell out for years. Every headache that seemed to originate from one particular location, every time vision on one side would go weird, every time "brain fog" would inexplicably descend - it's the brain cancer, I'm dying! And the main reason I never followed up with neurology, despite a recommendation, was indeed cost...attending free city clinics on spotty/no insurance is not a choice a rich person makes. This would have turned out to be a mistake anyway, since things cleared up on their own over time with medication adjustments.
That's the direction I thought the MidScanners section would go - holding the same can opener assumptions steady, let's minimize both the cost and time variables, and also add a further assumption that AI ultrasound magically reaches MRI levels of efficacy. If such a technology had been available back during my scare, would that have actually been a good thing? I don't know! Even if they're all quibblesome numbers, the odds as given don't look especially promising in the median case...so I'm mostly still interested from an R&D Yay Progress standpoint. AI has already done hat tricks with far less, or so they say.
Surprised you didn't address the "if we have squillions more scans, then AI can process them and get vastly more capable at discriminating stuff that matters from the rest" claim. More broadly, it's a mistake to think that the medicine of 5 years ago looks anything like the medicine of 10 years from now.
There’s a Nate Soares quote that breaks my brain ‘there may well come a day when humanity would tear apart a thousand suns to prevent a single death.’ It’s part of his discussion of how the cost of a life isn’t the value of a life.
But while money is a mostly-adequate measure of costs and values, it’s still an artefact to try and measure something else that we actually care about. I think this is the same point as not actually wanting to spend 800 hours in scans for 4 more years of life, or the old joke ‘Exercise doesn’t make you live longer, it just feels like it.’ If building a scanner involves torturing small children to work in heavy metal mines, and going for a scan involves hours of waiting in hospitals and months of anxiety, whatever cost the economics has put on that might not be a good measure of the cost. On the other hand, if getting a scanner involves some bored teenagers and an AI being super happy doing a summer project, and getting scanned involves walking into Starbucks through a gate that snoops your health and never tells you if you don’t need to worry, that’s different.
I think medics have a very good pragmatic feeling for ‘what is worth it’ and a really good sense that we’re all going to go of something anyway (doctors’ approach to end of life care / chemo etc is really interesting) and that’s why they come down more anti whole body scanning than the numbers suggest.
With the infinitely rich man, the cost of the scan is trivial, but the value / opportunity cost of his time is more complicated - at the moment we end up with the cost of a QALY being about ‘if a normal person worked normally for a year, they could buy a year of life’, which feels a normal thing for the market to do. But if you are infinitely rich, you’re saying ‘I’d rather be sitting in this scanner and having these tests than be going into low earth orbit or eating grapes peeled by the attractive people of my choice’ - they might value their time now more than their expected time aged 80 (which is where the 4 extra years are likely to go)
I somewhat disagree that "more scans" reduces the marginal benefit. In the extreme (which to be clear I don't believe is still even remotely close to possible) where I can every day just get in the bath and have a whole-body scan done, same as I could step on a scale and weigh myself, then the benefits clearly go back to being positive. There is no "is this zit dangerous?" question any more. I can just wait a few days and monitor its size precisely; its trajectory will say whether I need to worry or not. The anxiety is greatly reduced. For a comparable example of something I *can* already do - all the little defects would become equivalent to the defects I can already see: the moles on my skin. I look at my moles and check them. I know one of them COULD be cancer. Occasionally I get a scare by one that really looks a bit weird and keep a close eye on it for some time, but then I see it doesn't really behave any more weirdly than the others and relax. Very occasionally I go to a doctor to have it checked. And very very occasionally (this has happened twice to me) the mole is annoying, painful, or worrying enough to get it surgically removed. But ultimately, the mundanity of it all means my bar adjusts; and I'm a comparatively rather anxious person.
There probably is a minimum somewhere in there - an amount of scans that is the worst possible amount, enough to cause worry and not enough to deliver the ideal benefits. But at very very high frequency, daily scans are obviously better than no scans.
I’m a rich person who doesn’t care about time or money, and is immune to anxiety, and won’t make irrational decisions.
The reason I would get a Full Body Scan (FBS) is so that I have a documented reference point for any future health issues that may/or may not be easily identified. The hope is that having a comparative scan at a previous point in time could potentially help identify an issue sooner and track the growth rate. For example detection of an Aortic Root Dilation change could be the difference between living 1 year or 20 years if not diagnosed separately.
I would rather have the opportunity to spend my savings on holidays and memories with my partner than die and leave it to the government to not spend wisely.
Is there no way to have better thresholds for when to escalate care and when not to? Maybe the scan shows a small, likely insignificant finding. Then you don't do followup. Or it shows a really big, bad finding. Then you do followup.
1) This seems like a pretty long-winded way of saying that it's all about the cost and accuracy of the technology. More information is always good, but current tech is too costly. Well... isn't that exactly midjourney's pitch?
2) regarding the negative QALYs from anxiety. I don't buy the model. People are afraid of death, bodily malfunction / decay, etc. Their first encounter with it will always be scary. But this is a *fixed* cost. There's only so much shit to give about death.
Anecdotally, people get more relaxed about it once they've faced it in some form. It follows that the total QALY loss is simply time shifted. It doesn't matter whether you have to work through those feelings based on a medical scan, your neighbor suddenly dying, or what have you. It's unfair to put this cost entirely on the balance sheet of the scan itself.
Furthermore, the culture will eventually adjust to the new reality. An incidentaloma wouldn't be scary if you had literally dozens of people around you treating it as a normal thing that happens. There are many well-understood dangers in the world that don't induce constant anxiety.
A comment on the 'anxiety' aspect, which I think most healthy people will discount as overblown, or paternalistic. You probably feel that in no universe would you ever want to have less information - certainly that was the prevailing opinion on Twitter in most threads I encountered on this debate.
I have multiple sclerosis, and I know something about medical anxiety. I was diagnosed in my late 20s, just after completing the long process of schooling and exams to practice law. I went from being on top of the world, at the height of my optimism and ambition for life, to sitting in a room being told I had lesions on my brain, feeling the walls closing in on me.
MS is indeed a motherfucker of a disease, but three characteristics make it particularly insidious: it is neurodegenerative, progressive, and incurable. The day you are diagnosed, you are told that you will live with it forever, you will almost assuredly get worse over time, and the rate of your worsening is unpredictable.
Most MS patients become obsessed with the future, because we don't know what kind of future we will have. Will we lose our sight? Our ability to walk? We will be exhausted all the time? Or perhaps a mix: we’ll lose 30% of power in one leg, sight in on eye, and be tired most of the time. But even if we accept that these things will happen, it’s difficult to go through life not knowing when. That anxiety over an uncertain future begins to dominate, and become a source of pain all on its own separate from your actual disease course. But I suppose in some deep sense, health anxiety due to the diagnosis is part of the disease too.
An important caveat is that some small percentage of MS sufferers (3-6%) get through life with a benign course that does not affect their lifespan/healthspan. So when you first get scanned, you hope you are one of those lucky few whose initial neurological episode is deemed Clinically Isolated Syndrome - which never develops into full blown MS - or you will continue to get episodes but never progress into real, life-changing disability.
About 6 years ago, I had a stem cell transplant (HSCT) which is the most aggressive treatment for MS, involving high dose chemotherapy and a host of other drugs. It’s risky, it’s very hard on the body, but it resulted in remission for me. For the first 3 years post-transplant, I would regularly have mental breakdowns and periods of depression because I would experience an uptick in some symptoms, and I feared that it must be a sign that the disease had re-activated, and I was back on the descent into hell that was my pre-transplant life.
Eventually, my transplant doctor emailed me back after one of these episodes and just plain told me that I needed to drop my obsession with my condition, and learn to face those fears, because they were doing more damage than anything else. He was right.
It has taken a lot of time to gain some footing against health anxiety. It’s still there, though not as powerful as it once was. HSCT is the best and most effective treatment for my type of MS, but for most people the remission doesn’t last forever. So statistically, I am still expecting a return to active disease. And however much I can tell myself that I have grown, that my psyche is more resilient - I know that deep down, that question mark still hangs heavy over me.
Whatever happens, it is undeniable that I have lost a lot of enjoyment of life that I "should" have had in the last several years because I was worrying about something that had not yet come to pass. And if it does come to pass, then I’ll have just wasted the good years worrying about the thing that came anyway later on anyway.
To Scott's point, if you are a completely rational agent with newly diagnosed MS, you would simply run through the following process: you collect the medical information (how long they think you've had it, how much damage is done), you research the best treatment options available, then pick one, adjust lifestyle factors to minimize severity of disease course, and that's it. That's all you can do. After that, worrying only adds to the level of inflammation your body is already dealing with. But in my experience with fellow MS patients, it's rare to find such a person. There is too much at stake to really just 'let go', even though holding on is not accomplishing anything.
I think 99% of people would find it very difficult to be handed a report saying “You might have a cancerous growth but it’s probably nothing”, and then just trust that report and sleep soundly at night. You can read all these stats that Scott posted, and if you were told that the finding on your full body scan had a 0.1% chance of being something dangerous, you would insist on investigating it fully and completely. I imagine the same for 0.01%, and perhaps even lower.
The first neurological episode I had began with numbness in my feet and palms, and quickly spread outwards, down my limbs and towards my torso. I went to the ER, waited 7 hours to see a doctor, and was sent home with an appointment to see a neurologist 4 months later.
I had to go to the ER three nights in a row to finally get admitted. Nobody took my symptoms seriously, and even after admission, they wouldn't send me for an MRI because it was Thanksgiving weekend, and i wasn't a pressing enough concern.
I sat in the hospital for 5 days with zero treatment and zero answers, with every doctor telling me it was nothing, until finally the MRI showed those lesions. Then they told me it's likely MS or a handful of other things, none of them any better.
Soon after, when I got assigned an MS specialist doctor, I felt at ease... he suggested to me a treatment course, I followed his advice, and it later turned out he prescribed me a medicine which does almost nothing (copaxone). I spent 2 years getting worse while the medicine silently failed me, and then when I finally just completely broke down with a big relapse, he said we need to go more aggressive and suggested a real drug.
At that point I took it into my own hands, stopped trusting what the doctors told me, and started reading. I found out about the stem cell transplant and all of the other drug options. That same doctor refused to even refer me for the transplant, said it wasn't the right treatment for me... despite the fact that every single journal article/study I read said that this was actually the very best option for young patients who were not already badly disabled....and here I am 8 years after that conversation, still in remission, feeling better than I was in my late 20s.
I've seen both sides of the coin, both trusting and distrusting doctors.
Thanks for doing the sums for me and justifying my, almost certainly irrational, decsion not to have one. I think the self selection problem is huge here - Old men are notoriously reluctant to go to the doctor (in the UK) unless their leg falls off or they grow a third head.
One thing you are missing is the fact that everyone is going to die anyways.
If humans were generally immortal *unless* they get cancer, the benefit to screening for cancer goes way up.
This is a slightly different point than saying that some slow cancers won’t affect you because you’ll be dead before then.
It’s something I’ve thought a lot about because my grandma had ALS, which is mostly not a heritability risk (family members are more likely to have other brain conditions like depression, autism, and schizophrenia, but not ALS), but for like 10% of cases there is a heritable mutation. The heritable form usually starts earlier and takes longer to kill you. At this point I think it’s unlikely given the age of my dad and my aunts and uncles that our family has the heritable form, but when my dad was younger it certainly weighed on my mind a lot. Would I want the genetic testing to know if I had it?
The answer I came to was “no.” I decided that people aren’t meant to know what kills them that early, that psychologically, it is not a developmentally appropriate thing for someone in their 20s to know. If you knew what was going to kill you, you start thinking about what is going to kill your future kids or if you should even have kids. But nobody is supposed to know what will kill their kids.
Developmentally, your early 30s are the time to have an existential crisis, when you start to realize you aren’t going to live forever and need to do something meaningful. But you can’t create meaning about the future thing that will kill you, because by doing so, you are not living in the moment and you are trying to postpone the inevitable, which kind of ruins the whole existentialism vibe.
Since I will die anyways, I do not want to know how until it starts affecting me.
2K per MRI has got to be a huge underestimation. You're going to get bills from both the ordering physician and the radiologist.
Side story: My kid has a bicuspid aortic valve which is an extremely minor heart defect that about 1% or the population has. You can live into old age with no problem, but they say about 50% of people with it will need a valve replacement (which has become a pretty simple non-invasive surgery) by age 50. But in rare cases you can have an aortic aneurysm related to this that causes instant death.
So the protocol is for my son to get an echocardiogram every 18 months. Each time we get bills for around 4K (usually one from the "outpatient hospital" one from the cardiologist and one from the radiologist).
On the one hand--worth it to prevent a catastrophic outcome for my son. On the other hand, over his childhood that's about 24K that could have gone to his college savings account and we have zero proof that any of this screening benefited him at all, as this defect rarely causes problems in childhood and if anything, may cause him problems in old age. 🤷♀️
Among the dozen caveats and assumptions, I would add one more: that measuring this makes sense at all, and that the units used for the measurement are the correct ones. Even within broadly utilitarian framework, I would argue this sort of calculation strongly relies on hedonistic utilitarianism with no discountings. Given preference utilitarianism for example, it doesn't strike me at all as obvious that if individuals assigned 100 utils to a lifetime of 100 years, that they would have to assign 105 utils to a lifetime of 105 years (with healthspan correspondingly elongated). Or, even from hedonistic utilitarianism point of view, I think there are reasons for discounting that I find perfectly sound, but that aren't included in the usual QALY calculations, such as (at least in expectation, a couch potato centenarian *could* in principle pick up extreme sports) reduced amount of novel experiences likely contributing to faster passage of subjective perception of time, which would lead to temporal discounting.
Or, more broadly, while it seems to me that a wise and prudent person would usually follow sound advice from evidence-based medicine, "a life well-lived" (which, as above, even a preference-utilitarian might value) might look less like QALY-maxxing and more like living in the moment and taking a cancer diagnosed too late in your stride. I can see an argument for the opposite ("healthy mind in a healthy body"), but even that would seem to suggest flourishing involving exercise, a healthy diet, etc, not so much delaying the death from preventable disease.
TL;DR - tragedy of the commons. What is arguably rational individual behaviour is a collectively bad outcome.
It's a decent QALY analysis at micro-level; it's also blind to macro-level constraints and equilibrium effects. I think it relies entirely on three unexamined assumptions:
1) Scan volume is elastic and can scale with no effects/costs/requirements on units, staffing, or quality.
2) The capacity of the broader medical system for the follow-up cascade is elastic.
3) Neither (1) nor (2) above will displace other patients with the medical system, or that there will be zero QALY costs of such displacements.
This argument has been used here in the UK to justify no mass screening for prostate cancer, with the predictable backlash from those who haven't even googled the side-effects of false positives.
Then again, this is early days and when we have enough data AI will be able to catch 90% of the significant cases with no further tests.
The real question we should be asking is: given $x to spend what gives the best health outcomes?
As a pathologist who deals with this kind of discussion regularly, I have mixed feelings about it. I got the Galleri test myself. But one thing is certain: there is a downside to a false positives beyond simple cost and anxiety. People seem to be taking the approach of, maybe this finds something useful (positive), maybe it finds nothing (neutral), maybe it finds something to watch (slight downside). But that paints an overly rosy picture. Whether by bad luck or by incompetence, bad things happen, and I've seen things like a 40 year old man who died on the table getting a second biopsy that probably wasn't necessary. A spleen coming out (you need those!) because someone overthought a report. Once in a while I'll make the call myself after numerous failed biopsies and tell the treating doc to stop chasing something, that they should watch and wait because it's almost assuredly nothing. Sometimes it's like they need permission to stop. These are the kinds of things that are difficult to document in a formal way as part of a study. If people want to do this, I don't really have a problem with it, but "Rare and minor side effects of the testing might contribute -0.005 QALYS per person" is cold comfort to someone in the hospital with a collapsed lung after biopsy misadventure chasing nothing.
I think it's underrated how fucking annoying it is dealing with medical things. As a middle-aged woman with some health issues, my minimum number of appointments per year is 10 (2 dentist, 2 visits to my normal doctor, 2 blood draws, 2 visits with my cardiologist, 1 echocardiogram, 1 mammogram), assuming no follow-ups. I've had a slightly elevated white blood cell count for years and for the past couple of visits, my doctor has been trying to refer me to a hematologist about it in case I had leukemia. So I finally did that. One visit, one blood draw, no leukemia and it ended there.
Was it likely I had leukemia? No. And I got off easy on the testing this time. But there's always a chance that a hematologist will check you for 20 things, one will be out of range (this is similar to the M&Ms xkcd thing, if you think about it), that will lead to further testing, and then in the end it will either be nothing, or you'll turn out to have some syndrome that requires regular checking for the rest of your life but probably wouldn't have done anything to you if you'd never known about it. Of course there's a small chance your life will be saved, which is why I didn't opt out of the whole thing.
Because I got scanned for atherosclerosis, I know I have a small aneurysm, so I have to monitor that annually in case it grows and kills me. I guess I could choose to ignore it, but that seems hazardous. I guess it's good to know about it. But it also sucks to know about it and it sucks to have a whole echocardiogram about it every year and wonder when it's going to kill me. Probably it would never have killed me.
And I think this is all for screenings that are proven beneficial. Do I really want to know my left ovary is 2% larger than normal? Or there's some funny area on my spleen? I kind of don't.
One of my friends is thinking about getting a whole body scan because he wants to have a baseline. I think he might be able to handle hearing about small anomalies.
One item to consider in the US healthcare system is that, for many people on third-party insurance plans, there comes a point at least once per year where their healthcare is "free" (to them) and they are effectively cost-insensitive billionaires. This occurs any time they have exceeded their deductible (for HSA / HRA plans) or their out-of-pocket maximum has eliminated their copays (for more traditional plans). This isn't everyone, but it's tens of millions of people every year that, come June or August or November, don't care at all about healthcare cost.
As a tangential matter, I was extremely interested to hear about actual literature quantifying (or attempting to quantify) QALYs for anxiety and psychological stress, as well as QALY numbers for monetary cost. Of particular note, has anyone used this literature to reassess the Covid response? Because one of the narratives about, e.g., cloth masking in the US was that it was "cost-free" and thus unassailable even though its detractors' argument was that it helped spread panic, anxiety, and psychological trauma (especially for younger kids) and also that it primed adults to perceive extra risk everywhere, leading to added anxiety and also increased openness to harsh response tactics. There was also additional massive stress and anxiety from lockdowns, as people lost their livelihoods and their social connections. If this type of added anxiety were near 0.01 QALYs per person, across the US population that would be over 3 million QALYs. Based on the age profile of decedents, that would suggest NPI anxiety alone would have nearly matched all the QALYs lost to ALL deaths in the pandemic, and many times that of the alleged QALYs "saved" by such NPIs. The lost money would loom even larger.
Also, once you start giving bad advice to calm people down, you lose credibility. Next time, you'll be giving actually useful advice and a bunch of people will ignore you, figuring this is like that nonsense with the cloth masks.
I think there is one little error in attributing doctors some wisdom of erring on the side of caution by not recommending screens.
There are all kinds of situations where doctors over screen, and those are also defended as erring on the side of caution. And if the guidance were flipped, I bet the rationale would remain.
It can’t be both!
Also, I need to read it again, and I’m on mobile so I can’t scroll as I write this, but was the very slightly below the bar QALY per dollar figure a combined number for the four out of ten really severe and also the four out of 300 from follow-ups? If it was, isn’t the natural interpretation that the four out of 300 are well below the bar and shouldn’t actually be followed up upon?
Another point about how many people will get these scans and how they will be distributed, that I think buttresses the anti-scan argument, is that the US system tends to massively balloon recipients of any initially niche intervention. If the scans are OK for some subgroup (e.g. "rational billionaires"), the political impetus is that "this is a life-saving technology, it should be available to everyone." Then there is political pressure to add it to Medicare coverage and mandate coverage by private insurance. Enterprising politicians, eager medical groups, and aggrieved patient organizations will highlight the 4 saved out of 1000, then contrast with a child or young mother whose cancer "could have been caught," to push for mandatory widespread adoption. So even if in isolation the "rational billionaire" could use the tech, at the societal level all of the political incentives will push and push until it suddenly becomes "life saving care" that "shouldn't be restricted" and usage balloons to tens of millions because of coverage mandates. So normalizing it even for niche cases has predictable consequences for the public fisc and public health. I say this as someone who favors letting boutique health and patients do what they want, but I understand why the opponents are saying "niche = public" because that is the pattern of pressure for US health care interventions.
Cost aside, part of the problem is current medical ethics. If I search for "headache" as a symptom online, I'm told that it *could* be cancer. But if I got to a doctor they don't give me a scan or a biopsy, they don't tell me it *could* be cancer. They tell me I'm fine and prescribe an OTC pain killer.
It seems like this set of ethics isn't impossible for scanning. Tell the 300 people they are perfectly fine. (I think the risk with current ethics is that they can sue, or they can claim moral outrage that doctors don't care, etc.)
The dystopian version of this is forced yearly scans where a large corporation collects a ton of data but doesn't show patients the results, reassures them they are fine, and carefully/surreptitiously gathers more data in select individuals. Direct communication to the patient only happens in extreme cases. This is probably not outside the realm of possibility for the brand on a large healthcare company in the right regulatory/cultural environment.
I don't think "We’ll assume that the extra cost of treating the treatable cancers is zero. Why assume this? Because if we didn’t detect the cancer early, we would have detected it later, and still had to treat it, and the cost of treating small early cancers is probably no worse - and likely better" holds at all.
This may make sense for the costs of immediate treatment, both financial (e.g. surgery costs) or non-financial (e.g. post-surgical pain). But there are also long-term costs from treatment that add up over time if the patient remains alive. If someone gets erectile dysfunction from prostate surgery at age 50, this is not the same as getting it at 70s, and you should definitely subtract some QALYs for earlier tretament in that case. Similarly, if someone starts on hormonal blockers for prostate cancer a decade earlier, this will increase the cost of treatment rather than decrease it.
The other case in which the assumption doesn't hold is in the case of cancers that will be treated because of screening but would otherwise never manifest because the patient ended up dying from something else before it caused any trouble (a situation usually known as overdiagnosis). Again, prostate cancer is the classic example where a significant fraction of cancers detected by screening will end up being overdiagnosed, but this is probably non-negligible for other cancers such as thyroid, breast or kidney as well.
You'd have to look up the numbers in overdiagnosis of individual cancers to make a good estimate of how much treatment is directly attributable to screening and would not have happened in its absence. But it is certainly not zero, so "zero is an upper bound for the extra costs of treatment" is probably an incorrect assumption.
I think the analysis under-emphasises how much the benefit would vary depending on the individual (apart from financial situation).
"The average benefit among those 8 people is 4 quality-adjusted life-years"
An otherwise fatal cancer which is detected early in a 50 year old should result in an additional 30 quality-adjusted years. It would be a very different calculation for an 80 year old.
For a 50 year old, It is not unreasonable to pay a fairly high cost for even a small possibility of such a large benefit. Even if the expected value may be zero, or slightly negative, to save someone from "ruin" it may be a highly rational choice.
I don't think you can substract from both sides AND then look at their ratio to figure out the cost.
Suppose you can get $101 for if you give me $100 and one apple. This indeed looks as if I am offering $ per apple. Having ten apples you might expect to gain $10. But what if you are short on money, and don't have $1000 dolars to pair with 10 apples? In real world you might borrow them for a moment.
This is a lot for a 'patient' to digest and I confess I have not read the entire article word-by-word. Is it reasonable to assume that in the absence of a FBS, the responsibility for discovering negative health findings is left to individual physicians, typically during routine wellness checks? (That seems to be the state-of-the-art presently.) As an apparently healthy 72-year-old patient, I do not have confidence that today's crop of physicians is particularly adept at diagnosing or effectively treating medical problems. Some are not even capable of making appropriate referrals. I speak from the personal experiences of my late wife and experiences of friend and associates. In today's medical marketplace, it next to impossible to find physicians who are interested, competent, and inquisitive to the point of effectively delving in to a patient's concerns. With the rise of PAs and NPs, it is difficult to even obtain an appointment with an MD. Most are now employed by hospital conglomerates, who, in turn, rely on insurance company payments for their financial viability. In today's medical world, it's easy to see why FBS would be attractive, and prospective customers would not be inclined to view the potential personal benefits through the lens of an analysis such as presented in this article.
I think a potential problem here might be that the incentives of doctors, due to malpractice, the insurance system and everything, is going to be to drive towards investigating absolutely everything to death. Defensive medicine. The doctors know this. They don't want to waste their time being compelled to investigate a bunch of stuff that is low-likelihood of being an issue. But they don't want to say "if you do this, I'm going to be forced to investigate every little thing". They don't want to acknowledge how much medicine is driven by risk aversion due to malpractice risk. And they also don't want scarce resources to be consumed and diverted away from people who are higher risk. So the safer route is to recommend against doing it. Because in a lot of ways, it's just throwing up more noise.
Doctors often prescribe a benign medicine just to make the patient feel better. We need something similar for slow growing ignorable cancers. What comes to mind for me is one of the prostate cancer treatments, which is to insert a small radioactive ball into the interior of the cancer to shrink it. Something similar for other cancers might be a small ball that emits something/anything for 10 years that might have a mild physical benefit but at least does no harm. The real benefit would be emotional tho, that at least the doctor did something, so the patient can relax and be happy.
The number of women who think they are above average drivers: 55%. The number of men who think they are above average drivers: 80%. No source, stat told to me by a (female) friend because they thought it was funny. Almost certainly falls into the 88% of statistics made up on the spot. But somehow holds a bigger truth. I haven’t looked it up - finding it false would destroy the pleasure I take in it.
Slight mindbender: people have different conceptions of driver quality, such that all 55%/80% could conceivably be correct (relative to their idiosyncratic metric).
It is also important to consider the potentially substantial medicolegal costs associated with widespread whole-body screening. Diagnostic errors by radiologists, complications arising from biopsies, delays in obtaining timely follow-up appointments, and inadequate communication of a finding’s clinical significance can all expose physicians and healthcare systems to litigation. These legal risks contribute to rising malpractice insurance premiums, which are ultimately passed on to patients through higher healthcare costs.
The real trick pulled by Midjourney is to get people discussing all these random numbers and questions like “are doctors evil actually?” instead of just asking “is this Midjourney thingy a Theranos like pivot?”
Like I get it, there’s lots of low hanging fruit out there, but why tf is a random AI image generation company doing a supposedly amazing, novel, medical breakthrough? Could it just be that… they’re running out of cash due to OpenAI/Gemini eating their lunch and now they’re trying to pivot into a hot area to get a bunch of VC money? Why isn’t **that** the main question vs all these (useful!) Bayesian calculations?
Something that few people are mentioning - lawsuits. It's all well and good to say we'll just raise our threshold for intervention to avoid harm. But even if a physician correctly calculated that something is 99% chance of being inconsequential, then for every 100 calls they'll miss one finding that leads to harm which they can get sued for.
This is mostly an American problem. Other countries either make it very difficult to sue doctors, or the payout is small, or the payout comes from a government fund.
Left out of this analysis is the value of time. It's the one thing you can never get back. All kinds of time: time spent making ( & unmaking) schedules, time lost in transportation, time sitting in offices, time in follow-up appointments, time worrying & etc. I' m sure everone can add to the list.
>But sometimes harmless things have the same size, shape, density, and location as dangerous things, and even after you’ve wrung as much information as possible out of those parameters, you’re still not sure. Mr. X is 5’8, overweight, and lives in Chicago. Is he a terrorist? Would it help if I told you his height to five decimal places?
>(“But if you get multiple scans at different times, you can see how fast it’s growing!” Okay, Mr. X gained three pounds since last year, ready to answer now?)
Agreed. Which is why I find Yudkowsky's "That Alien Message" essay https://www.lesswrong.com/posts/5wMcKNAwB6X4mp9og/that-alien-message so irritating. We build specialized instruments _for good reasons_ , and additional precision on the _wrong_ measurement plus even unbounded thinking time on the wrong measurement, is not going to make up the deficit.
( One minor quibble with:
>So, converting everything to the same units, whole-body screening costs $108,000 to save one quality-adjusted life year. Usually in health economics, $100,000 per year of healthy life saved is considered the bar for a good cost-effective intervention (though other sources say $150,000). So this is somewhere around the bar, not unambiguously good or bad.
Since the dominant cost is financial, large scale screening MRIs might come out more favorably. The learning curve for going from millions of MRI screenings to hundereds of millions. Quoth Google/Gemini:
>When cumulative MRI manufacturing volume scales from10^7 to 10^9 units, learning curve theory predicts a cost reduction of approximately 14% to 27%
which isn't huge, but enough to considerably shift the decision. )
There is a second half of this equation that is also difficult for me to approximate: every hour spent chasing incedentaloma is an hour that is not spent on another patient. I don't know how I would begin to quantify this, but it has a nonzero cost in time and attention.
Interestingly, the main arguments given against whole body scanning (anxiety and unnecessary time spent) don't move the needle sufficiently to argue against it. With 32 QALY benefit, and only 7 QALY time cost, those arguments are only a relatively small factor and the financial cost is the main cost.
So decreasing sensitivity would plausibly be a bad thing. You could lose some of the 32 QALY benefit, without reducing the main cost.
I think your analogy with the person getting an hour of scans every day for 3 years misleads our intuitions to judge it less rational than it is.
First, for scans in the actual world, there is very little benefit from each scan after the first, when you're getting them daily. So we rightly judge someone doing this to be crazy.
Even if we imagine that each of the 899 hours were necessary, there is also the further problem that when you calculated the cost effectiveness, you priced in stuff like driving to and from appointments. But if you get 1 hour of scans a day, you will have to add that back in again, double counting.
Also, if you're getting 1 hour a day, that will likely interfere with your life a lot more than the raw time, since it'll probably consistently make you unable from participating in activities certain times a day.
A more accurate test would probably have to imagine that the hours were more concentrated in bigger lumps. Or even better that the cost were, say, staring at a wall 1 hour a day, at *some* point throughout the day, on average over 3 years (don't ask me how that would work). If we do that, then I don't think it's nearly as intuitively crazy.
Add to that that we're bad at probabilities so that 5% of 4 years doesn't intuitively register as 900 waking hours.
All that is to say that I don't think that particular analogy was a particularly good gauge of how rational the scan is for a rich person.
But what you say after that point might still be right
Aren't there a lot of non-catastrophic benefits to the scan? Seems like you could get meaningful data on things like muscle/fat composition, tendon/joint health, and many others. This would be particularly helpful for people who are trying to optimize their health through different supplements and practices.
Perhaps this is too far afield for your analysis, but I wonder whether the multi-cancer screening blood tests (e.g., Galleri (GRAIL)) have better QULYs and make more sense than whole-body imaging?
Yes, it's about threshold values, but what values? Threshold values are conceptual, not only clinical, if ever. Shouldn't we be more cautious about how our very own language is used? I would like to see more intense scrutiny in science about what gradients constitute something being allowed to be something, e.g., a word, a meaning and so on. In other words, an immensely thorough operationalization when it comes to our own narrative. Otherwise, we just might end up scanning threshold values for our own narratives rather than anomalies/diseases, etc.
Gosh. I’ve been a doctor for 40 years, trying to help my patients live longer and better, and you, with no experience, are so smart. It just seems obvious, just plain obvious
The shape of our problem is that there is a set of data that could improve decision making, but knowing the data by the party most affected has costs. We currently lack a technology to a) generate the data b) make decisions on it c) dont share it with the party that suffers.
In the good ol days I think doctors would hide data like this from patients. That was determined to be bad for a number of reasons.
An AI third party could be a technology for solving this problem? Midjourney would not be a third party but it could make the introduction of one easier. The third party needs to be able to consent on your behalf to follow ups without disclosing specifics to you. You need to be conditioned to not worry about what your AIs actions suggest about the data.
Actually, the only problem I care about is what extent the rest of us are paying additional costs due to unnecessary medical procedures. This seems to me that it could be a much more serious issue, with society wide implications for gov't healthcare policy and the average cost of insurance premiums. What fraction of the rise in medical costs are due to such procedures? In what ways are the rest of us paying for it? That's the conversation I would like to see.
Great piece, thank you for digging into this. I am a cancer survivor who benefited from super aggressive intervention following a surprise scan finding. Amazing, wish more people benefited in this way. On the other hand, I am now closely monitored by constant scans of different types and despite frankly being one tough as nails mfer, it is often psychologically torturous. I consider myself pretty immune to anxiety. Work and interpersonal matters do not faze me. But having an ambiguous finding (as you often do! the human body is confusing!) and waiting weeks or months to discern whether it’s anything serious is NOT for the faint of heart. It is IN your body, invisible to you and entirely beyond your control. People vastly underestimate how hard this is to face, psychologically.
All that is to say, given the collective mental health levels of our population I’m not particularly optimistic people will handle frequent scanning well.
Scanning good. Waiting for scan or biopsy results… bad.
The most important reason for having a whole body screen is to get a baseline for evaluating further changes. After I was hospitalized for a serious, life-threatening condition, I was subjected to every test imaginable over 3 days.
Most came out normal, but in some cases, the correct interpretation depended on whether I was just an outlier, or whether something had gotten worse.
If they had an earlier baseline, doctor could have seen if something had actually CHANGED or if I was just an outlier on some reading.
I notice you've been using AI estimates more frequently for your research. Have you tried using them to estimate values that you're already knowledgeable about, to see how reliable they are?
I notice you've been using AI estimates more frequently for your research. Have you tried using them to estimate values that you're already knowledgeable about, to see how reliable they are?
I'm grateful for this analysis, which pretty much confirms my preexisting views about overzealous screening.
What a lot of people don't seem to recognize is that interfacing with the medical system in any way is dangerous. Many of the other comments here illustrate this fact very well. When it's worth the risk, it's worth the risk, but when it's not, it's not, so the decision to subject yourself to it should not be made lightly. You definitely shouldn't just assume that more is better.
My recommended heuristic for most people is: don't overthink it. Go to the doctor if you have troubling symptoms, but as far as screenings go, just get the normal ones recommended for someone with your risk profile. I'm sure the official recommendations aren't calibrated perfectly, but I think it's very unlikely that the average layperson could beat them with confidence.
In addition, consider spontaneous remission rates: https://pubmed.ncbi.nlm.nih.gov/19029493/ . This study estimates a 22% higher incidence in those under active screening.
In addition, consider spontaneous remission rates: https://pubmed.ncbi.nlm.nih.gov/19029493/ . This study estimates a 22% higher incidence in those under active screening.
The most interesting part of this is doctors asking people not to bother, waste unnecessary money or time and the people behaving they know better!!
As an Indian I can say with absolute certainty NO DOCTOR in India will tell people not to overtest and spend money on shit they dont need. This technology is god send for Indian docs.
Rethinking about this this morning, in a perfect market, which is what your QALY sums are sort of assuming, people will choose to do a thing when it gives them more extra life than it takes them to do (whether that is physically doing the thing, or working to get the money to pay for the thing.)
But the reductio ad absurdum of that is that people ought to be happy spending an entire year working to get an extra year and a second of life, but not to get an entire year less a second of life.
Imagine a hypothetical where the cost of buying extra life was sitting in a doctors waiting room, (and all the actual health saving happened by the pure magic of being in the room, not any more inconvenient tests or scans or treatments.) This is a just-better-than-marginal intervention, sitting in the room gains you a as-much-life-plus-a-minute as you spent sitting in the room.
Which would mean over the years of life you end up with, half of them (or at least half of the awake hours, or the productive hours) would be spent sitting in a doctors waiting room.
If you frame it as ‘do you want to live for a year doing whatever you want, or live for two years but for 8 hours a day you have to sit in this one spot thinking about your health’ it feels a much starker thought experiment to me.
We already have various DNA tests with exactly the same market niche and “more data will make it so much better” mantra. I still don’t know a single person who actually get anything useful from unprompted genetic test. You are more likely to find criminal among your relatives than unnoticed disease. All medically significant results is some form of “3 times higher risk of X, follow healthy diet and exercise” and only useful for such a thin slice of people I doubt there is more than 100 of them in the world
I just hope by the time there's an actual functioning procedure they still only need water. Imagine the cleanup required after hanging out naked in a vat of ultrasound gel for a while!
>and I've yet to go to the Emergency Room and NOT get a CT
Gaa! A whole body CT scan? 10-20 mSv? 5 years of background radiation? Yes, it is still a factor of 50 below what would cause radiation illness, but it still doesn't seem like something to do lightly...
Many Thanks! That sounds a lot less concerning.
Many Thanks! Yes, 10 mSv is enough that one doesn't want to do it lightly.
Banned for this comment - highly emotive controversial content without any factual backup or useful contribution.
Having recently worked with Kaiser on a medical issue, I would say I tend toward a similar line of thought to the banned user. Insurance companies do have control of pieces of the process, and hospital (company) lawyers also have control over pieces of the process.
I have heard from doctors inside Kaiser and elsewhere when trying to get a diagnosis for a condition that "that's not really how modern medicine works" and was told dozens of times "but how about we treat the symptoms." I turned to private scanning (which the hospital system won't let their radiologists read) and while it did not ultimately lead to a clear signal, it did drop some items in the differential, which made me *less* anxious, not more.
From a cost perspective, every hospital is incentivized *not* to provide patients with imaging that may reveal false positives. This is I believe why I was sent to X-ray although it is not actually the recommended scan for the condition I was going for (it's cheaper, but I don't think that's it--once you have the machine, marginal costs are minimal).
But I think there is truth to the fact that the medical system is interested in treating the symptoms and is not interested in chasing down diagnoses. Without a robust private scanning option, I believe my only option to reduce the medical anxiety I was having would be some pills they were pushing.
Further, I believe that there is a culture of non-diagnosis that results in things like "heart disease" being listed as the major cause of death when in actuality there are probably at least three or four (or five or ten) other major causes that lead to heart failure, get poorly reported as heart disease, and that then in a culture of non-diagnosis lead to a series of conditions across the culture that are never understood or diagnosed. So I would say this analysis that sits squarely in the single-patient/doctor/hospital cost-benefit eludes the question of what scientific power could be brought to bear on our health that will not be--all positioning as if we're doing the individual patients a favor.
"But I think there is truth to the fact that the medical system is interested in treating the symptoms and is not interested in chasing down diagnoses. Without a robust private scanning option, I believe my only option to reduce the medical anxiety I was having would be some pills they were pushing."
The alternative, of course, would be a national health care policy that incentivizes health care providers to invest in preventive care. If they got paid for catching conditions early on, diagnoses would suddenly become a lot more popular (and yes, that has it's own issues).
"If they got paid for catching conditions early on, diagnoses would suddenly become a lot more popular" Yup, which ties our under-diagnosis problem directly to the one place we see the opposite: over-diagnosis in mental health
This seems needlessly harsh.
Agreed. It's a perspective, at least.
And it's how talking to at least some doctors genuinely feels.
> But 899 hours of scans would look like going to your doctor’s office for one hour every weekday for three years!
Not to mention that, you can't just schedule just 1 hour in your day to teleport to the doctor and be seen with 0 waiting time, going to your doctor for a 1 hour procedure could cost in addition:
- another hour in the waiting room
- an hour or two in traffic to get there
- some margin before and after the traffic time because you don't want to be late to the doctor, or late to the thing you want to do after the doctor
That ends up being a big hole in every day!
I think this point pushes the opposite direction: you should only rationally devote 899 hours to this, so if it took 3 hours to get a 1 hour doctor appointment, you should only be willing to go on one-third of days.
I'm just saying the estimated numbers are off. Understandably, the time and place of appointments are optimized for the doctor (whose labor is very expensive) not the patient. So 1 hour of doctor's time typically translates into several hours of patient's time, inflating the lost QALYs from extra procedures.
I think the problem is double-counting your health benefits. In a world where cancer is the only possible way to die, it might make sense to devote 899 hours to screening. But there are many ways to die, and you can't spend 899 hours trying to prevent each of them. And if you succeed to get rid of cancer, but the next day get hit by a bus, the effect wasn't as big as you expected.
In a world where you drive to the doctor every day, the additional traffic hazard might outstrip any QALYs gained.
The cancer can't get you if the Hummer gets you first.
Do not underestimate the destructive power of a Kia.
I’m an outlier but I actually enjoy going to the doctor. In the same way I enjoy getting my hair cut it’s like an asmr/ being taken care of vibe
You must have a very, very, very nice doctor :-(
I always figure going to the doctor involves some nonnegligible chance that I pick up something else from one of the other patients. I come in with a headache and depart with a headache plus the flu.
AI claims your paper found the opposite effect of what you are claiming:
the paper cited completely contradicts this argument. The research by Lowen et al. shows that influenza virus transmission is highly dependent on relative humidity and temperature, with cold and dry conditions favoring transmission.  Specifically, at 30°C (86°F), no transmission was detected at all, while at 5°C, transmission was significantly more frequent than at 20°C.  The authors explicitly conclude that influenza virus transmission indoors could potentially be curtailed by simply maintaining room air at warm temperatures (above 20°C) and either intermediate or high relative humidity.
You don't sound like an outlier to me. You remind me of a Chinese friend who, studying at an American university, caught a cold and went to health services for a comforting intravenous injection of saline solution. (This is a common practice in China.)
When they explained to her that they didn't do that because it has no medical benefit, she was quite upset, complaining that it felt like they didn't even care that she was sick.
>> “… went to health services for a comforting intravenous injection of saline solution.”
Hell no!
We do not differ in our opinion of this practice.
No, I would prefer a shot of placebo. That always makes me feel better!
I mean, that's the whole idea.
Though it is not widely known among the Chinese that the saline solution has no effect.
And these are the people we are losing to.
I don’t think this is healthy (mentally) or good for the culture to reinforce/pander to it.
(There was an old practice of passing out antibiotics for colds which made them less useful.)
> The[re] was an old practice of passing out antibiotics for colds which made them less useful.
I'm not really following you here. By this metric, the saline injections are harmless.
I was suggesting this kind of practice isn’t limited to Chinese culture with that. The US did a similar thing that was (“obviously”) worse. The US also decided to change the practice.
That is, I was expecting people (here, especially) to get I saying this sort of thing isn’t isolated to a particular culture, could have been worse, might not be without downsides, and can be changed.
I’m not sure if saline injections are harmless. (I’m not even sure promoting the idea it’s doing something when it isn’t is harmless.)
Heroin junkies can shoot up so much that their arm veins are damaged to the point that it's basically impossible to get a needle in.
I think it's safe to say the Chinese aren't getting saline injections quite that often.
They increase utilization of medical care which is bad for the people who actually need it.
That argument can't work. They're not a sudden or unforeseeable stress to the system. If you want to make this kind of argument, you'd have to say that the extra demand for medical services increases the number of medical providers, damaging society by drawing people to become doctors and nurses who would otherwise have done something more productive.
In the event that there 𝘸𝘢𝘴 a shock to the system that called for more serious medical care than usual, the extra slack provided by these extra practitioners would be a benefit - there's no obstacle to them declaring that they're busy with the emergency and saline injections are being temporarily paused.
Well, this is just logistics though. If the approach really worked and really was cheap enough, you'd try to set up cheap scan-booths you can simply book with an app and then pop into at the given time.
But I also don't think it makes sense to equate "time spent going to the doctor" as a loss of QALY to the same extent that being dead or having cancer is.
I spend above-average (for my age) time dealing with medical stuff, and it's not my favorite way to spend time, but it's a lot of waiting around and I bring a book or a game or do stuff on my phone. It's going to vary from person-to-person how 'effective' that time is, but I have to imagine the number of people who basically just zone out and stare at a wall while in the waiting room is fairly low.
Why would you go to a doctor's office for a commonplace scan? Just pop into the scanning parlour in your local mall for it.
'People should not buy the product currently on offer!'
'That's absurd! This product I am imagining which is similar to the product currently on offer but better in many specific ways and with none of the downsides would obviously be good!'
'...yes?'
My favorite so far has been "if we altered your brain so that you enjoyed this, would you enjoy this?"
I think he's saying "only do this if there's a reasonable expectation that it will find something". That's a bit of a different statement. Many tests have a false positive that is high enough that unless there's a reason you shouldn't take the test.
Part of the discussion is that a broad set of ppl in the position to improve the product in the ways that could make it good argue that the entire concept is bad. Or at least appear to do so.
"Could this product be improved and is that a better focus of our efforts, or are the detractors correct that this is intrinsically unworkable" is a question worth asking
Actually, the question is "Where should limited investment dollars be spent?" And "In whose interest?"
I work in genetics. This sort of comes up with whole genome sequencing.
We're actually quite conservative about what gets routinely reported back to the patients. In general incidental findings are not reported back unless they're on a shortlist that meets certain conditions re: certainty, whether there's anything that can be done etc.
You mention "anxiety" but that really doesn't carry enough gravitas.
After Chernobyl they caught enough cancers that the background cancer death rate dipped, but a lot of people were so convinced they were gonna die anyway that the alcohol related death rate shot up.
you need to be really careful when it comes to things that scare the shit out of people. Despair can have worse health effects than the things people are despairing about.
I sort of agree with this, but it's also infuriating when I can't get health data I know perfectly well how to use because someone thinks I might freak out to hear I have slow 2d6 metabolism or something.
pro tip: a lot of services have some kind of general workflow to comply with the european GDPR rules.
If you get your genome done it's your data. Typically they'll have some kind of process so you can do a GDPR request for your raw data. (very few people actually do)
You can always hop on a VPN and make it looke like you've moved to europe.
it's easy enough to annotate once you have it.
Though one of the early lessons was how many annotations look impactful but aren't really and how many variants get annotated as "pathogenic" or "HIGH" impact but are actually nothing-burgers. .
By "annotate" I assume you mean some kind of SNP/variant association mapping, and not gene prediction, right ? AFAIK gene prediction on humans is a solved problem...
Annotation as in taking a list of variants and for each one pulling from various databases of known info and predictions to try to figure out its effect, if any of note.
Re: gene prediction, you might be surprised. The current canonical databases of genes are great for highly expressed genes but definitely incomplete when it comes to one's expressed at a low level in small subsets of tissues.
Inside known genes the number of alt isoforms has been going up sharply since the advent of long read rna sequencing.
And genes being recategoriesed from non-coding to coding is pretty steady.
> Annotation as in taking a list of variants and for each one pulling from various databases...
Right, that's what I thought you meant, thanks.
> Inside known genes the number of alt isoforms has been going up sharply since the advent of long read rna sequencing.
I know this is the case in plants, but I'm surprised to hear it is still happening in humans -- I thought human genes were pretty well studied by now. Live and learn !
it surprised me too when I moved labs.
I'm sitting there with data from long read RNA sequencing, predicted ORF's and confirmation that we found the proteins predicted by those ORF's in mass spec data....
I'm sitting there thinking "this must be a big deal" but then I get chatting to my supervisor and she's like "oh we see that all the time, but it's a nightmare to try to convince people it's a real gene"
Some of it is because most historical gene finding didn't use live brain tissue, (people have this weird aversion to researchers taking samples from their living brain)
You also get into this whole almost philosophical thing of "what is a gene, *really* " because some genes physically next to each other that have been classified as different genes you'll find will actually splice together sometimes, some of the tools for looking at splicing auto-discard those but biologically some are consistent across samples.
There's also absolute shitloads of RNA-genes that don't seem to code for a protein but do seem to show up consistently across multiple individuals.
Noble intention to shield from harm, genuinely unclear whether there’s a net cost or benefit here. More fundamental point is a right to know. This is why there are laws in many jurisdictions (US, globally) that give a patient the right to receive (upon request) all health data that’s been recorded and stored about them.
Sure, but having a right doesn't mean that exercising that right is going to be ultimately beneficial; you also have the right to control what you consume, whether it's vegetables, aspirin, or cigarettes. It seems complicated as doctors have to balance the needs of their patients without perfect insight into their mentality. It's exacerbated since patients who are going to exercise this right to know are going to be a higher risk demographic for the harms described above (Although who's to say what the root cause is? It'd be hard to filter out pre-existing or complex conditions.).
"A little knowledge can be a dangerous thing."
Yes I think this is really the annoying part. Ultimately the policy is calibrated on a median user which means that people who COULD get value out of it STILL DON'T because no one bothers to account in any way for their individuality. Which is why it's generally irritating when others make decisions for your own good by assuming things about you. If it's about years of my life, I'd like to be able to do the best I can!
I'm a strong believer in people's right to access their data *if they seek out that knowledge* then on their own head be it.
But thrusting knowledge upon people against their will has some problems.
Are you sure it’s being calibrated on the median user rather than the 80th percentile user?
FWIW I work in a field that is tangential to plant genomics, and it seems like the trend is leaning away from whole genome sequencing / GWAS in there as well. That is to say, people are still actively sequencing whole genomes of all kinds of plans, but the next step in the process is no longer "mwa ha ha we now have all the data we need to make drought-resistant corn once and for all, behold as our stonk price goes to the moon", but rather, "we now have enough data to begin constructing a pangenome model that might one day eventually allow us to learn enough about plant metabolism to maybe increase drought resistance if we're lucky, hopefully".
The Soviet Union (and then the early Russian Federation, though perhaps also to this day) seemed to have a much bigger problem with alcohol related deaths than the US.
I agree that it is very hard to distinguish causal effects. But Chornobyl was really one of the factors that pushed people into despair. At that time I met some young people from Belarus, completely healthy and who lived quite far from Chornobyl when it happened and yet they were feeling fatalistic – that now my health and future is completely ruined; they always reminded this to others requiring a special caring attitude from them.
In Latvia there was a special disability status – a participant of Chornobyl cleanup operation – that entitled to certain benefits especially in healthcare. I personally think that this status alone was more damaging to person's health than actual effects from radiation from that work.
Any discussion with the general public that health effects from Chornobyl disaster were not as bad as people often think, is immediately ridiculed. You would look like an antivaxer trying to suggest to evaluate scientific data. The irony is that in Ukraine vaccine hesitancy is also very high, mostly due to strong distrust of public heath authorities.
Yep, met a lady in Kharkiv - around 2010 - who had been "participant of Chornobyl cleanup operation " at 18, and had been sure; her cancerous end would be near. 25 years later: a healthy mother of 2 healthy grown-ups. - And sure: many others saw no reason not to become an "alkashki" instead of trying for a career.
If I hire you to sequence my genome I want the data. All of it. I do not want you deciding what I get to see. I'd rather be worse off than have you deciding what is best for me.
It's your data, if you want a full copy of the raw data then all you have to do is ask. We have to provide it under GDPR.
At that point you're choosing to delve into it.
You're free to do that, but it won't be thrust upon you if you don't ask.
>Despair can have worse health effects than the things people are despairing about.
There's nothing to fear but fear itself!
And lions. Those guys are scary.
What does adding additional tests / screenings (blood tests, cardiac stress test, etc.) alongside the MRI do?
Biostatistician Frank Harrell says that most effects in medicine are additive, which would imply that, conditional on the value of a whole-body MRI being ~= 0, the benefits of blood test + cardiac test + whole-body MRI are ~= the benefits of just the blood test + cardiac test. But that's a generalization; of course you'd need real data to look at specific conditions.
Incidentally there *are* datasets which have this level of richness, UK BioBank is one: genomes, whole body MRI, blood testing, a bunch of other stuff.
The degree of anxiety that people here in Taiwan get from a screening scan that shows “could be a tumor site, too small to be definitive” is noticeably smaller than friends or family back home in the US when they receive the same. Because a significant fraction of the adult population is getting scanned every two years, and has been wide scale for over a decade, I suspect that people have just compared notes enough to realize “a whole lot of people get this result, and hardly anybody dies from this ‘incedentaloma’ result.”
There’s probably also some psychological safety benefit to knowing (here, not there) that if you want to get a follow up, you can just make an appointment directly with an oncologist, within a week or so. No PCP referral, no clearing it first with anyone, and it won’t cost a ton (making an appointment with basically any specialist can be done directly, here).
Thank you for sharing this - it provides a really interesting cultural/contextual difference.
I was not aware that the procedure was so common in Taiwan.
Ditto. I wouldn't have been surprised to learn this about Singapore, but I was for Taiwan.
Singapore is actually pretty similar to the US as far as the extent of regulation, it’s significantly more liberal/free regs-wise than any of the other OECD countries over here.
It’s not super surprising for me any more though it was 15 years ago (I also expect Singapore regulation-heavy;, I have a team in Singapore and I’ve been doing business (b2b SaaS, 12k business customers) there and 8 other countries between Australia and Japan for 16 years. SG MoM (Ministry of Manpower) not require employers to provide health exams, there are some lesser requirements on covering doctors fees or consultations.
@ruffienne, yep, a little more than half of the adult population here in Taiwan gets either MRI or CT scans or both every two years, and this has been going for about 20 years.
Comment in the OG article with links to health center screening pricing sheets, and the law that requires health exams (but does not require MRIs or CTs): https://www.astralcodexten.com/p/preliminary-thoughts-on-the-midjourney/comment/278955137
Very interesting. Your explanation makes intuitive sense. But what do you think of the theory that culturally people in Taiwan might be more stoic than people in the US?
How we are expected to behave often drives how we actually feel. Or do you have a good enough sense of the friend's and family's reaction to over events to discount this element?
It has absolutely not been my experience that Taiwanese people are less stoic than those in the US.
As someone who's experienced 'this is probably nothing, let's keep checking every year' I think a large part of why Taiwanese are calmer is that doctors and the medical system in general feel more accessible and trustworthy. The most it tends to cost is around US$16 per appointment, usually including tests, so people feel confident that they will be able to find out what's going on. I had a concerning symptom, went to see a specialist, got an ultrasound (immediately) which came back probably-fine, got blood tests which came back with unrelated precancer markers so got scheduled for an CT scan and then got told that it was probably nothing but come back once a year. The entire experience cost around US$20.
The only thing is that while most screenings are cheap or free cancer *treatment* can get quite expensive, so there are people who can't or don't (out of concern at imposing on the family) pursue it. That no doubt plays a significant role in the acceptance.
Yup, I really think this is just a transient effect. It needs to become mundane enough for people to learn to start writing it off, and then they'll start writing it off. People coexist with weirder things.
I think this is the most important comment. And it's one reason I don't give a lot of credence to the position that knowing might upset someone who's bad at evaluating risk. We genuinely do not know how much of a problem it will be if we never run the experiment. Which it sounds like Taiwan has done for us. Thanks Taiwan!
I don't think you give the age caveat enough weight. To me, it obviates the entire analysis because I (asymptomatic, no bad family history, late 20s, physically active, healthy bmi, nonsmoker, nondrinker, exercise, outdoors etc.) know that I am substantially healther than the "Average Adult", but can't really quantify it; superficially I would assume 10x? More? I've done road trips through America and "Average Adult" isn't doing very well. Starting with your $150,000 I'll happily 10x it and say $1,500,000 cannot possibly be worth it.
Also, "only ~half of people bother to go to regular checkups, which takes one hour per year and probably have effects nearly this big."
Do you have any citation on the actual benefit of regular checkups? If so, I'd love to read up on it. My impression is that as long as I remain asymptomatic the benefit of getting a yearly checkup would be effectively zero. I haven't been to a doctor for a general checkup in at least a decade and assume the next time I see one will be from a traumatic sports injury, I assume when I'm there I can get my bloodwork and whatnot done on the side. Should I actually get regular checkups (and for that matter have a primary care doctor or know the name of a doctor in my state?)
Tbh you could probably ask Claude or ChatGPT and it'd give you a good response; I would guess they would corroborate that you need minimal regular, if any care.
Your next age-gated care check is likely a colonoscopy at 45, maybe some bloodwork at 35 or 40.
Obvious caveat, I am not a doctor.
I have done that, that's why I was asking for a citation. I am skeptical of the " bother to go to regular checkups, which takes one hour per year and probably have effects nearly this big"
Fascinating! It seems like the literature agrees with me that, given I am already living in the healthy way I would be counseled to live, a regular checkup would do essentially nothing and could in fact be harmful. Scott, I think a correction or rebuttal would be warranted.
It's not a great service and its not a direct answer to your question but...
In the UK on the NHS, annual check-ups are not a thing. You're invited for more specific screenings when you get older - like the above poster mentioned but a 20s male would not be expected to visit his doctor.
The system is overloaded, as you'd expected when free at the point of supply but they do still do their sums on the value of preventative care so this suggests those sums say annual check-ups for the young are at least not money saving vs waiting for issues to arise.
Same in the Netherlands. Annual check-ups aren't a thing; older women get checked for breast cancer until 75, and older people in general get asked to send in a stool sample like every 2 years or so to check for intestinal cancer.
I know someone who moved from Singapore to the Netherlands.
She tried to get a regular annual check-up from her GP. The GP told her: after doing all the tests, I'd just tell you to sleep more and exercise more. So how about we save on the testing?
To my mind, no matter how many health effects are transferred upon me by healthy behaviors (like those you mentioned), I cannot take for granted that I am necessarily and will remain healthy for those reasons specifically. If I understand correctly, cancer rates, to name one example, vary regionally for a whole host of reasons unrelated or indirectly related to individual behavioral choices (like air pollution, genetics, etc.). To my mind, having a relationship with a PCP and engaging in check ups and screenings once or twice a year is (1) trivially costly in terms of time and money, assuming that I’m insured, (2) at least somewhat protective in the short term, and (3) probably more protective in long term through the generation of base-level data against which my PCP can compare and interpret changes. I’ll also add that having a relationship with a PCP has two additional benefits: (a) a relatively frictionless way of accessing prescription drugs in the case of low-probability, high-risk events like infections; I once caught a pretty nasty bacterial infection while spending some time in a low-population, out-of-state, low-health-infrastructure area (as a result of me exercising out there, no less) and was able to have the issue diagnosed and an antibiotic immediately prescribed such that I started it the same day that I noticed the symptoms, and (b) a relatively frictionless way of discussing potentially concerning changes in my health with someone who has expertise in medicine and, ideally, knowledge of my medical history; it kind of kicks ass that, when I have established a relationship with a PCP, I can just give them a call if I feel like I have too. Frictionless-ness isn’t everything, but I assume that even healthy people would, overall, find some protection from low-probability negative health events if they had a frictionless relationship with a PCP, especially one who had some knowledge of and long-term data about their medical history. Idk. I’m just spitballing here. But my point, I guess, is that it sounds like you’re putting just a touch too much faith in your healthy behavioral choices to protect you against negative health outcomes. Your behavioral choices kick ass, but my prior is that no set of behavioral choices can be 100% protective and, therefore, that a relationship with a PCP and a few hours and few dollars a year still confer considerable benefits to an otherwise healthy person. I’m not telling you what to do, of course—and you’re already doing much more for your health than the average person, which is great, I’m just explaining how I think about that particular choice and the potential benefits of it.
Interesting perspective! I see the benefits on antibiotics, but my impression (admittedly secondhand) was that as an asymptomatic patient my generic wellness checkup would consist of me going in, stripping, spending a few minutes telling the doctor about my lifestyle habits and lack of concerns, getting poked a bit, the doctor taking my blood and saying I have slightly low blood pressure, agreeing but saying I've always had slightly low blood pressure, and leaving. Where in that process would the doctor actually catch that I have an early stage of cancer? I don't think I would be getting a full body scan, right?
I think this might be cultural.
In the US my regular checkups were pretty trivial and similar to what you describe. Height/weight/abdominal girth, blood pressure, stethoscope exam.
In Japan, the yearly checkups additionally include: some standard blood panel (including e.g. lipids), urine and fecal analysis (collected beforehand and brought into the appointment), ECG, abdominal/heart/carotid ultrasound, chest X-ray, lung capacity testing, vision testing, hearing testing. Plus the infamous barium-swallow stomach cancer screening, which I skip since there are less-annoying tests and the prevalence is much less high in white people. And then I ticked a few boxes to get some extra blood tests for things specific to my known issues.
They then surface all your results with a grading system: A = no abnormalities, B = minor findings, no need for concern; C3/6/12 = follow-up recommended in X months, D = consult physician, F = consult physician urgently.
I've found this to be helpful in the way the grandparent poster describes. E.g. it detected a few early-stage things that have me modifying my lifestyle habits and monitoring for followup (gallbladder polyps slightly below the danger line, mild carotid arteriosclerosis, trends in cholesterol numbers and lung capacity as I modify my diet and exercise programs). And I'm glad they're monitoring things like hearing and vision, as I want to know ASAP if those slip.
With regards to the question about whole-body screening, I suspect these sorts of East Asian yearly checkups have found a good medium between whole body screening and US-style test very little: they pick the most worthwhile tests, and subsidize or mandate them on a population-wide basis.
My experience in the US has been that it varies by doctor, but almost everyone I've gone to does at least a basic metabolic panel (lipids, cholesterol, glucose, etc) with the annual physical and most also do a comprehensive metabolic panel (also tests liver and kidney function, etc), a urine analysis, or both.
I have maintained relationships with primaries since 2016? 2017? In both the Phoenix area and the Wichita, KS, area (I now live in the Pittsburgh, PA, area, and don’t have insurance at the moment, but I plan on establishing with a primary the moment I have insurance, and expect the quality of the care I receive to be similar). My yearly/ twice-yearly checkups have involved blood tests, given at almost every stage of care. Idk. Maybe this is a relatively recent development in standards of care, but my perception is that relatively extensive and relatively useful amount of base-level data has been collected about me, and I assume that that data will be useful at some point down the line. Maybe I’m wrong, but it still puts my mind at ease knowing that someone with some amount of expertise is capable (at some level) of monitoring long-term changes in a certain, somewhat meaningful amount of my base-level data about me. Likewise, there is some short-term amount of value to the data they collect for me personally. For example, I was having some medication-related issues related to dehydration which I recently resolved, which I wouldn’t have been able to identify were it not for the regular blood tests to which I was (consensually) being subjected. I can’t say that I was given the best advice about how to resolve it, but the blood tests helped me to identify the problem, for which I’m very grateful. Again, just a perspective. But my experience with regular check ups in the U.S. has been, uniformly, very good.
Thanks! Only 3.5 months in, but I’m liking Pittsburgh so far! I know there are ACX meet ups here, and I’ve been meaning to make it over to one, but they’re always held on Sundays at 2, and I start work every Sunday at 3pm :[
There are typically also annual labs, probably bloodwork, maybe urinalysis. My primary care doctor also does certain routine bloodwork. I get a complete blood count annually, and when I turned 30 she added the standard metabolic panel. It turned out that despite my general good practices re: diet and exercise, I had a marker that indicated a slight increase in risk for heart disease and metabolic disease. So I can actively work to improve that marker before it becomes a problem. And I learned that because of a routine checkup
> I assume when I'm there I can get my bloodwork and whatnot done on the side.
I wouldn't necessarily make this assumption. Doctors like to do different things at different appointments.
They also might not do bloodwork in house and send you somewhere else to get your blood drawn.
Many diseases do not present with symptoms until it's too late. Many of them can be cured if caught by early screening. It's like saying we shouldn't screen people for blood pressure because it can be asymptomatic
I’m guessing you’re male and/or have never had a Pap smear. That’s an example of routine annual screening that catches a ton of cancer cases early, which used to be fatal.
Similarly, if you don’t go to checkups at least some of the time, you’re probably not going to be up-to-date on current best practices re: routine screening. Right now that might not be a big deal, but in 10 years, you might need to keep track of when you need colonoscopies or prostate exams.
My bad, thanks for the fact check
> because I (asymptomatic, no bad family history, late 20s, physically active, healthy bmi, nonsmoker, nondrinker, exercise, outdoors etc.) know that I am substantially healther than the "Average Adult", but can't really quantify it;
Assuming you maintain these habits, all cause mortality effect size for you vs the average american is ~4-5x. As in, past 40 they'll have something like a ~1% chance of mortality every year (and going up as they age), while you'll have a ~0.2 - 0.25 % chance per year (and going up as you age).
In expectation, this will net you 5-15 extra years of life, with 7 the median expectation, and 10-20 years of lower morbidity (ie you won't have diabetes, CVD, or other medical conditions lowering quality of life and needing ongoing medication for 10-20 years that the average american will have such).
The majority of the effect is driven by exercise, the next largest is not being obese / overweight, and then things like smoking and drinking. You didn't mention your diet, if you eat actual food that you cook from ingredients (instead of opening packages and microwaving), you're closer to the 5x end of things, and that's next in terms of effect size.
I could see this going the other way. If you are old and falling apart and will not realistically make the life style changes to adjust there is less value in knowing the precise ways you are falling apart.
If you are young and healthy you may be in better position to shrug off minor findings in your scan and not suffer the phycological costs of false positives but stand to benefit from the early detection of a treatable cancer. The distribution of QALYs may be different than the distribution of findings over a large sample
I'm not sure about the benefit of regular general health checkups, but regular dental checkups might be worth it?
(For me they usually pair them up with a cleaning, and getting rid of the build-up coffee stains is nice by itself.)
As someone in largely the same position who does get annual checkups one thing I would point out is that my blood tests occasionally catch childhood vaccines that have worn off and I have to get redone.
In general it’s kind of surprising that there’s not actually anything you can spend your money on to directly improve health outcomes beyond what a good doctor will already do for you. (Iirc Robin Hanson said the value of the marginal healthcare dollar is 0). Makes me wonder where the dramatic benefit to life expectancy from wealth actually comes from
It's easier to become rich and stay rich if you're healthy. It's easy to become poor if you suffer a serious health problem.
So if you look at a random bunch of rich and poor people you'll find wealthier people are healthier on average.
Add in some genetic effects, the same applies to heritable health problems and caring for sick family can put people into poverty.
There could be survival bias kicking in too. You may found rich people with various strange diseases you won't find in poor people. Because those poor people would already be dead.
"good doctors" are pretty rare, and often expensive
People who make good choices financially are also those who make good choices health-wise.
Having a bit of savings, or higher disposable income, makes it easier to do some things that pay off long term - turning down a risky gig that pays a bonus, paying for the slightly more expensive apartment that isn’t next to the particulate factory, paying a personal trainer once or twice to give you advice at the gym, eating meals that are tasty in more subtle ways than just large amounts of fat and salt.
You can absolutely spend money to improve health outcomes, if not directly then:
Having a gym membership improves health outcomes. Having a swimming pool, basketball court, home gym set also improves health outcomes. Better quality food (fewer preservatives or just tastier healthy options), cleaner living environment, etc.
While any one of these may not do much, having better doctors, better screening, and more specific treatments should absolutely improve health outcomes in aggregate.
Having better education substantially improves health outcomes. This is probably the most important thing.
> Makes me wonder where the dramatic benefit to life expectancy from wealth actually comes from
1. Rich people, generally speaking, do not work in mines or as lumberjacks or any of the other dangerous jobs out there.
2. Drug addiction has a negative correlation to both wealth and health. The same is true of a lot of maladaptive behaviors.
3. Poor people are more likely to have short time horizons, with causality going both ways: they're poor, so they need to focus on making this month's rent, not what's going to happen in six years' time, and they're bad at investing money so they remain poor. Short time horizons mean you do shit that's bad for your health but feels good in the moment (comfort eating into obesity, smoking, casual sex with strangers, etc).
4. Stress is unhealthy. If the chemicals responsible for stress were good to have constantly flooding your system, they already would be. It's supposed to help you in crises where you might get eaten by a lion if you fuck up, so it's worth the health debuff, but in modernity there are no lions but plenty of stress - especially if you're poor.
5. Rich people can and will pay a premium so they don't have to live in areas where there's lead in the pipes, pollution in the air, etc.
There was an interesting paper published in JAMA Internal Medicine last year which highlighted the finding that eight cancers have doubled in incidence since 1992 in patients younger than 50 years, while aggregate mortality has remained flat. Here's the citation: JAMA Intern Med. 2025;185(11):1370-1374. doi:10.1001/jamainternmed.2025.4917 Published online September 29, 2025. The authors suggest that "Much of the increase appears to reflect increased diagnostic scrutiny and overdiagnosis. Interpreting rising incidence as an epidemic of disease may lead to unnecessary screening and treatment while also diverting attention from other more pressing health threats in young adults." It wouldn't surprise me if a rigorous study of whole body scanning, using whatever modality, revealed parallel findings related to diagnosis and mortality.
Meanwhile, the notion that there is an epidemic of rising cancer in young people has trickled down into the popular culture. I hear it regularly from friends and acquaintances, with attributions to whatever the cause du jour is (e.g., microplastics, pesticides, etc.).
It actually seems that they are right - though the cancers aren’t just going up in young people, but in people of all ages, and young people are reaching levels where it’s relevant to think about it, while older people already had to: https://dynomight.substack.com/p/crc-rates
This seems to be true for some cancers, but for others, the rates are rising: https://dynomight.substack.com/p/crc-rates
> Any of these would be a more grounded hope than “of course more data will always be good and worth it”.
Would this line of reasoning not also apply to other claims about the efficacy of AIs in other fields -- and thus (in aggregate) to the probability of reaching AGI/ASI via increased LLM training ?
I don't see why. The claim about AI is that we're extending scaling laws and extrapolating known trends, not just a fuzzy prior that data has to be generically good.
I guess I'm confused about your reasoning. AFAICT, in your post (above the line I quoted) you list all the reasons why Midjourney's ultrasound scanner is unlikely to rival MRI on quality, and in fact unlikely to be extremely useful other than as a stimulus for other research. Your arguments sound reasonable to me; it would appear that ultrasonic scanning technology has some built-in limitations which increased amounts of data just cannot overcome. But would you say that scaling laws *could* in fact overcome these limitations, or would you say that ultrasonic scanning had reached its theoretical peak (or is very close to it) -- or am I misunderstanding you completely ?
If it's the latter (ultrasonic scanning is reaching a peak that cannot be overcome by increased intelligence), then could this not also apply to other technologies ?
Whether gathering more data is worth it is an empirical question that depends on whether gathering more data has thus far led to marginal benefits. We see that for AI, it has, but not so much for medical scanning.
Right, but it seems like ultrasonic scanning has built-in physical limitations, and thus getting a full-body ultrasonic scan is of marginal utility at best. If so, then a thousandfold improvement in AI would still yield little to no improvements in ultrasonic scanning. You could ask Hyper-Claude 9000, "help me build a better ultrasound machine", and it would come back with, "sorry, you've gone about as far as you can go, maybe consider switching to MRI instead". If that's true, then perhaps there are other technologies that follow a similar pattern (and I suspect that there are).
But... the medical scanning uses AI.
You don’t seem to understand that "the scaling laws" being referred to here is a specific empirical finding regarding the loss function in the pretraining stage of LLMs. It’s not a generic statement that "scale is good", and it certainly does not apply to this Midjourney ultrasound thing.
Right, but that is my point: it seems that we have found at least one domain where even applying a 1000x improved AI is unlikely to yield more than marginal benefits. I would argue that there exist many such domains, i.e. scaling up LLMs is not an automatic recipe for success in general.
I think you’re conflating things, the scaling laws tell you exactly how much improvement in the loss function you can expect for any additional compute and data. In fact the scaling laws tell you that you need exponentially more compute and data for any additional improvement in loss. If anything they tell you there’s diminishing returns.
But the scaling laws are only about the loss function in the pretraining step, and aren’t why people are bullish on LLM as promising technology. People are bullish because LLMs keep having exponential increases in capability, as measured in real world usefulness.
Right, I understand everything you'd said so far. So does this mean that LLMs are likely to produce substantial increases of the capability of ultrasonic whole-body scanners, specifically ? If not, why not ? If yes, then what might this increase look like in practice ?
This is an interesting line of reasoning, though I disagree with the conclusion. You’re saying, “Here’s a problem AI won’t solve”, i.e. there are some problems where a better-trained LLM won’t give a better answer, therefore “more data won’t always be worth it”. I agree with the narrow conclusion, if you only care about things like getting data out of ultrasounds. But the reason LLMs can’t help with ultrasounds is that actual intelligence won’t help with ultrasounds, either. Your argument is basically “LLMs won’t help if you only care about problems that a team of smart humans couldn’t solve, either”. I’d agree with that, but I also care about lots of things which a team of smart people *could* solve.
> I’d agree with that, but I also care about lots of things which a team of smart people *could* solve.
Yes, I agree, of course. There are still lots of problems that smart people and/or LLMs could solve, and once those problems are solved, they will open the way to other problems to be solved. Nonetheless, if there are many domains similar to ultrasonic scanning, where more "intelligence" and more training data cannot yield substantial benefits, then the concept of a "superintelligent AGI" is put into jeopardy. The chief feature of such ASI is that it's capable of significantly outperforming humans across many (if not all !) domains, but what if most domains are just like ultrasonic scanning ?
Great post, the fifth point in the problems list is probably the core of a lot of this discussion.
Neurotic, worried-well people with no symptoms who are <50 years old are the exact type of person where the cost-benefit analysis makes asymptomatic screening a poor idea and it's basically just a luxury good for this segment. And indeed this is also often the loudest group clamoring for it.
Are there awful stories of physicians messing up and undertesting? Absolutely - doesn't mean an abundance of asymptomatic imaging is the answer.
Can AI help make more screening programs effective? Absolutely - in part because AI can do a better job understanding your medical history and symptoms and can corroborate the screening findings / identify reasons to get a screening or imaging order.
> Neurotic, worried-well people with no symptoms who are <50 years old are the exact type of person where the cost-benefit analysis makes asymptomatic screening a poor idea and it's basically just a luxury good for this segment. And indeed this is also often the loudest group clamoring for it.
This sounds like the exact demographic that would benefit from a clean MRI result. About 2/3rds of the study participants had a clean result. That... sounds like a good thing? I don't think we should ignore the benefits of giving peace of mind to the majority of people getting screened.
The thing about highly anxious people is that reassurance-seeking is actually counterproductive in the long run. Sure, you might in the short-term feel less anxious because you got a clean result, but at a deeper level you've validated your hypochondria instead of facing your anxiety and learning to tolerate it.
not to mention if they do get a positive result (which, just on base rates, is extremely likely to be a false positive), the downside impact of additional testing and anxiety is enormous
Would it actually give peace of mind? The point of many mental disorders is that the brain *isn't* working properly or behaving rationally. I would not be surprised if the response to a clean scan is something like "the scan might have missed something, what else can I do?"
Also, 2/3 success means 1/3 failure rate--is the impact on the latter group the same magnitude? I have no data here, but again it would not surprise me if they were worse off by a lot more than the first group were better off.
A lot of the worst I hear is from people who have symptoms for years or decades, but they can't get tested.
> Neurotic, worried-well people with no symptoms who are <50 years old are the exact type of person where the cost-benefit analysis makes asymptomatic screening a poor idea
It actually genuinely surprised me that Scott estimated the anxiety at .01 QALY. So being anxious for a month or two while you get followups is equivalent to losing 3 days of your life, *permanently??*
That just sounds crazy to me! Like you would rather die 3 days earlier to not now this information? I'm not sure I believe that.
But then, I'm extremely non-neurotic (I think I was <20th percentile on OCEAN) and non-anxious, so I'm definitely talking from the other side of that spectrum.
So anxious people here, you would literally die 3 days sooner to not worry about something over a few months where otherwise your base health and the way you feel physically is the same it's been all year?
I imagine it depends on the results you're validating - if it's something like cancer, it's probably pretty large, esp. for a high neuroticism segment.
Most of the people who are indifferent to medical costs are not rich. They are people like me who consume a lot of medical services and are on medicare & supplemental insurance, or they are on medicaid, or get VA benefits.
I think it's the difference between public health thinking from the part of the doctors (looking holistic at impact on the whole healthcare system including costs) vs private individual expected value calculation (moar information always moar better). A lot of public health interventions recommend people do things which individually on the margins look bad for them but look good on aggregate and this feels weird to individualistic thinking
Doesn't this article just reaffirm the 'adjust the thresholds' point it's responding to? The cost/benefit 'back of the envelope' depends critically on both the proportion of people being sent for follow-up, and the proportion of that group who _do_ find something.
I was given a whole body MRI after a particularly nasty mountain bike crash a few years ago. It found all the normal broken bones and punctured livers that go along with some things, but it also detected a significant growth on my kidney.
The default advice was "wait 6+ months for someone in the public system to see you about it because we're kinda busy right now" (I live in Australia) but as a minorly rich person I paid out of pocket for a specialist to review. He said it looked really bad and ordered a biopsy, which was inconclusive.
I had lost my first wife to cancer at 29 years old, and my current wife was pregnant, so we decided to just get it removed at significant cost privately.
When they finally completed the expensive surgery they decided it wasn't cancer after all, but maybe it might have possibly taken over my whole liver, even if it couldn't spread elsewhere?
I continue to be unsure if I made the right decision or not to keep investigating. I probably wouldn't have, except I really didn't want to be told it was bad at the public system appointment right as my wife was due to give birth.
Just adding some colour to "what can happen when you get a whole body scan and things look weird".
In contrast, scans I had for real acute issues that also flagged a nearby asymptomatic incidentaloma. Only it turned out not to be incidental and did turn out to be cancer when subsequent scans for the ongoing acute issues served as follow-ups, observed concerning changes, and a surgical biopsy confirmed this. Better yet, it was a (so far!) very treatable cancer using existing medications with low side effect profiles.
I feel very lucky that it was noted by the radiologists doing the other scans and prompted, with some prodding from me, excellent response from the medical establishment leading to correctly identifying and treating the condition. I will note that it remains asymptomatic but was growing rapidly when untreated and could have lead to very serious complications although likely not death (organ damage, nerve damage, etc) that would severely impact my quality of life and would likely have done so by the time any symptoms would have arisen.
So here was a case where a large benefit was obtained through diligent surveillance. Furthermore, as a very rational actor and non-anxious person in tune with false positives and statistics, I am sure the only anxiety-related side effects were to less rational family members I told, not to me personally, and now I have peace of mind in knowing my prognosis and monitoring my improving condition through occasional follow ups.
My experience makes me a big believer in whole body scans. At the very least, maybe when you’re 40 you get a scan, and if you want to worry about unnecessary costs or anxiety, do nothing with it unless something really bad is obvious to the doctor. But then you have a baseline scan, so when you do the next one a few years later, your doc can more clearly find any changes from that one, and those are much more likely to be meaningful and merit investigation (or if nothing changes, give you more confident peace of mind).
Your story is interesting, but it's not obvious from the telling that you benefitted from early detection. For some cancers (I'm thinking especially of myeloma but there are others), early detection doesn't affect the cure rate much. If not detected early the cancer will eventually manifest through organ damage. However, the harm of this is often overstated. Most of the time, the organ damage is reversible and quite unlikely to lead to death in the acute phase. So while it may sound like early detection is a benefit, I'm not sure that ostensible benefit would show up in a population study.
And of course, what you say doesn't change the point of the post, which is that most people who get a full body MRI do in fact have an incidentaloma. You are the exception.
In my case it sounded like the most likely outcomes from untreated progression, whether detected or not, would have been partial leg paralysis, a permanent nephrostomy tube, and some kidney damage. I am very glad I don’t have to deal with the daily life challenges of those conditions even if they are unlikely to impact mortality assuming the rest of the treatment could be managed at that point.
I understand the incidentaloma issues, and have had other scans for acute conditions that were inconclusive, gotten 1-2 follow-ups from the scan results after the acute problem was resolved, and the issue declared to be nothing material just looking slightly unusual in a place no one would have looked had something else not prompted the scan.
I saw an interview, perhaps with Attia, saying basically if you do a whole body scan, be prepared to find stuff that you will either need to ignore or do a few rounds of follow ups on that are likely nothing especially the first time you do such a scan without a baseline. So I was prepared for these scans to see things and not especially worried until there were changes in an already suspicious area.
This is a better outcome than death, though
The biggest problem I have with the criticisms is that everyone assumes that the results aren't going to get better and the prices aren't going to get cheaper. They throw up their hands and say "This is wasting time and money for only 4 people per 1000!"
The whole point is that the tests are supposed to get better and more accurate. So the number "300 have mildly concerning findings" will go down especially with the proliferation of AI. That's why I have been doing full body MRI scans for the last several years. I pay $2000 which I can claim from FSA and any changes over the course of the year they can see.
As money gets put into this, there's no question that the technology will get better and more accurate. Yes, 1000 scans may only save 4 lives, but if the prices drop down to $200 and the no or mild findings increase dramatically and are accurately quantified and the 20 people with extremely concerning findings are accurately identified and only the people who need it get biopsies, then we are living in a better world.
Whether or not this is possible is simply a matter of opinion, but I believe in technology and I believe in the brilliant people who want to make this technology better.
That could all be true, but it would still be consistent with the notion that a full body screening MRI is not going to benefit the marginal person considering doing it.
Consider too that even if devoting a lot of resources to improving this technology will eventually make it useful, that doesn't mean that we couldn't have done even better by developing some other more effective technology with those same resources.
There's a lot of "could"s and "would"s in your response. If there's a better way to look inside the body that is cheaper and faster and safer, I would be all for it. Right now, MRIs are the best we have so why shouldn't we double down on it to make it better and cheaper and learn how to get much more reliable results?
You could say that about anything. There's an infinite number of things that have a small benefit that is canceled out by an equal or greater cost. It seems like you're saying we should just spend the money on all those things even though we know it has negative expected value because this will make the technology advance and eventually make the thing have positive expected value.
Unless we run out of things that have positive expected value, I think we should do those rather than do things that have negative expected value in hopes that this will eventually change.
There is no other technology that can scan our insides and let us know if we have a mass growing inside us. If 1000 people take it and there is no mass for 990 of them, then that's great. Then it's like a blood test that I take several times a year, all of which never show any results. No one seems to suggest that blood tests are useless or cause unnecessary anxiety.
Of those 10 people that do find something, it means it was detected far sooner than it normally would have by "traditional" means.
The only thing we need to do is make the tests more accurate and informative so that only the 10 people who need it get follow ups. That means better testing and better AI to discern.
The idea that we should completely ignore this gift from science to look inside our bodies because *right now* it's mathematically inconvenient is insane. We have a way to detect tumors years earlier, especially if everyone gets a scan once a year, like the dentist. What we need is more research to make it more effective, not ignore it and pooh pooh the idea just because it's not perfect right now.
There is nothing special about scanning and early detection. A QALY is a QALY. If we can get one for $108k/year (Scott's estimate, not saying it's right, just using as an example) by screening MRI, or we could get one for $50k/year by giving effective treatment to the people who actually have cancer, then it would be better to do the latter. I'm not saying that I have absolute confidence about the cost per QALY of all these different things. I'm just saying that if you accept the premise that screening MRI doesn't have a very favorable cost per QALY, it doesn't make sense to say that we should do it anyway so that it gets better. Sure, spend money to make the process cheaper. But don't encourage individual patients to spend their money on it when it's not worth it. There are plenty of other bigger fish to fry.
It's also silly to suggest that the medical field has been ignoring or pooh-poohing cancer screening. Colonoscopy, chest CT, PSA, and mammography have all been the subject of enormous study and billions of dollars of funding for several decades. Recently, there has been more pushback as the idea that the harms may outweigh the benefits has gained some currency. For example, it is not obvious that screening mammography achieves anything of value, even though we know with certainty that it detects many breast cancers. Unfortunately, population statistics suggest that the diagnosis of early stage breast cancers has risen enormously, while the diagnosis of late-stage breast cancers has not changed much. This strongly suggests that many of the newly detected early stage breast cancers were not going to manifest themselves or affect the lifespan of the patient, and there is no benefit to detecting them, whereas there has been harm because they end up getting a series of expensive and unpleasant treatments. It's true that more rational decision rules would shield us from a lot of his harm, because we would rationally choose to ignore the screen detected cancers a lot of the time. But the fact is that hasn't been the case, and patients and doctors have been led to overtreat many cancers through irrationality, enabled by the screening imaging.
The problem is, until the technology has positive utility for an individual patient you’re effectively asking people to donate their QALYs to Midjourney’s shareholders.
If Midjourney('s VCs) are functionally subsidizing the scans (à la Uber rides in its early days) then it might nevertheless be a positive-sum transaction.
Funny enough I posted about something similar not that long ago in an open thread, which Scott commented on.
In case you want to read about my experience hiccuping non stop for 72 hours and losing my faith in the medical establishment, here you go:
https://theahura.substack.com/p/why-are-doctors-so-unwilling-to-run?utm_source=share&utm_medium=android&r=5sutf
I think my prior is so high that gathering more data is good that I'm still mostly unconvinced that gathering more data could be bad. I think the intuition pump of "your skin is an organ, it has weird growths and zits and things all the time, most of which don't matter, and it would be bad and annoying to get surgery every time you got a zit" did the most to shift my priors somewhat. But even here, I feel like this is downstream of a medical establishment that a priori assumes that tests will be expensive and so simply refuses to order more tests. Even though the best way to make something cheaper per unit is to massively increase demand and then let the market do its thing.
More bluntly: right now we assume all of the above is expensive and hard. But if you could wave a magic wand and make it so that these scans are actually super easy to do and you can just do them at the gym, then that would be awesome and a lot of the negative risks go away. The way you get to "machine you can run in the gym" is by first creating demand for that kind of thing, which justifies the innovation cycles necessary to make them cheaper
Midjourney Bath Tubs are for everyone and every home! In 20 years, the patent expires, and everyone gets one.
That intuition pump is what got me on team "more data is good". Or am I reading you backwards? We have lots of data on what our skin does, that’s *why* we don’t get surgeries for zits all the time.
what is the opportunity cost of this much time, attention, and money being allocated from known and potential other issues?
this is a really boring simple point, and I'll only say it because I don't see it in the article: the empirical benefits/downsides of MJ medical don't really matter. They're building a tech-themed wellness spa; this has PMF regardless of the science.
"They're building a tech-themed wellness spa;"
Yeah, that's my opinion too. They'll either set up their own clinics or go into partnership with the kind of chain of clinics that already do these scans, and you go private and pay for the entire package, only NOW WITH AI!
Maybe make it an entire package where it's the new version of taking the waters:
https://www.baden-baden.com/en/wellness/thermal-baths
yea they are building their own. see https://x.com/DavidSHolz/status/2068431559439642912
(1) Well that doesn't seem creepy and cult-like, at all!
(2) If the guy could string two words together without needing filler like "y'know" and "like" and "um", I'd feel more confident in the wunnerful wunnerful tek. What, he couldn't write down (or get a chatbot to write it for him) a speech/blessing beforehand?
Always cellos for the musical invocations. I like cellos but c'mon David, this is Revolutionising Healthcare With AI, mix it up some - missed opportunity for a theremin! or the lirone/lira da gamba! or having your pet AI create a musical theme of its own utilising bricks and broken MRI scanners!
Lirone - an instrument I just discovered the other day:
https://www.youtube.com/watch?v=ybig-0BBihA
I read a book by an oncologist a while back called "How We Do Harm," which I think is relevant. One aspect that is highlighted in that book, and which you mention briefly in your piece, is the profit motive and the set of bad actors waiting in the wings to make money from unnecessary additional screenings, unnecessary / non evidence based treatments, etc. Some of these treatments even cause harm themselves (e.g. chemo for an early stage cancer increasing the risk of leukemia). Opportunities for these types of bad doctors abound in a world where everyone is being screened for everything all the time and people are not perfectly rational actors.
Another piece to consider is the strain that additional screenings + procedures place on the medical system as a whole. There are pretty hard limits on the rate at which doctors are being produced in any given specialty: with a few notable exceptions, medical school costs an arm and a leg (you get to reattach them once you become a doctor); and whether we agree with their stated reasoning or not, the AMA limits the number of residencies per year. So we have to ask ourselves how resources get reallocated when we open up another tap of demand for limited doctors' time (radiologists, oncologists, etc) to start pouring in from elective screenings. I suspect the answer is similar to what in practice ends up happening with charter schools & vouchers: a (further) reallocation of resources away from the working class and to the wealthy. That means a step back from a society where medical care is accessible based on need, or more simply put, a less healthy population.
> But if the calculations above are right, and only ~8 people out of 1,000 benefit, then it would be hard for high-quality studies to measure the effect. You’d probably need hundreds of thousands of participants to find a signal.
Wouldn't you be able to estimate just the negative effects with a much smaller study?
I feel like, if someone told me "maybe you have cancer, come back in six months and we'll do another test and tell you for certain," that would cause me *much* more than -0.01 QALYs of anxiety.
I had an optometrist tell me that I might have glaucoma and needed an urgent appointment with an opthalmologist once. I spent the month between those appointments worrying about going blind.
Turns out my optic nerves just look kinda weird and I followed up every 6 months for three years. My mom told me she has the same thing, funny looking optic nerves but normal visual field and pressure, and she sees an eye doctor every 6 months and nothing has come of it.
I've had similar experiences.
They've detected polyps in my gallbladder for a while now, just below the threshold of "let's operate and get that thing removed". After the initial screening, I spent a week being very anxious about the possibility of surgery, the recovery, reading up on all the side effects of losing your gallbladder, thinking about how disruptive it was going to be on my work and life, etc. At the specialist appointment, they did more imaging and said "let's just monitor, come back in 6 months". It's now been ~20 months, and every 6 months we check back in and it's fine, same size. But I vividly remember that week of anxiety.
Another time I went to physical therapy for some hip pain. The PT said "oh this is concerning, it might be bone stress". Again, several days of severe anxiety about how this would impact my life and exercise program, how the treatment was 6 weeks on crutches loading it as little as possible, negotiating with Claude as to whether it'd be OK to walk just 2000 steps/day or whether that's too much, etc. And then the imaging doctor finds out it's a hip impingement, much less serious, and I go back to the PT to get some strengthening exercises.
I'm hopeful that with these experiences under my belt, I'll be more sanguine the next time something potentially concerning shows up. But you never know. Health stuff can just be really scary!
I remember this was what the influencer Attia was pushing for on the 60 minutes interview. What he said showed a lack of understanding of human psychology. Yes, those who want to should get it, but why should everyone get it? False positives can be very costly, especially emotionally.
A lot of the concerns about how people will react to likely benign internal zits can probably be solved by so scans so cheap they become mundane. Sure some people freak out about moles and rashes, but most are satisfied with "Don't worry about it, just let me know if it changes".
Exactly. If the scans became super cheap, that would solve a part of the problem. Plus once you have a suspicion about a specific body part, maybe next time it is enough to scan only that body part, which could be even cheaper.
None of these, for me, "epistemically defeat" the straightforward first principle argument that in single player games (cancer can't change its behaviour to thomas schelling gotcha you, and I'm assuming you hide this from your insurance company), it's straightfowardly good to get more information.
Most discusison about this is just emotionally biased cope, which is almost unavoidable for a subject that involves money, health policy, and has the touch of "thing health bilionaires do"
Even this extremely theoretical argument is only true if the information has zero cost.
If it has zero cost, then this is true in principle as long as you're perfectly rational, but most people aren't.
Yes both true, and I should have included them in my original message
Also, I should have included "once you learn how to use the information" part of it
I wonder if the problem could be solved by doing some kind of rationality checks.
I imagine a general system where you could sign up for a lesson, do a test afterwards, and it unlocks access to some knowledge. We already have something like this for driving licenses, now imagine a more general thing that also unlocks e.g. access to drugs, or medical information, or investing money, etc. Simply, some actions require a knowledge check first.
I hear you, but either a rationality check or a knowledge check could _very_ easily turn into a political correctness check / ideology check.
> Simply, some actions require a knowledge check first.
I think in real life this basically happens in real time, with the doctor judging from your character and rationality from your past history of interactions, the types of questions you ask, your affect, and what they know about your background. I have a doctor who basically treats me like another doctor, because they know how much pubmed I read and the questions I ask over the ~10 years we've been interacting, and will readily prescribe pretty much anything I want to try, and will share any information or prognosis or treatment plan with no padding or worries about anything.
Also, this avoids the very salient failure mode that Jeffrey Soreff points out - in America, it would definitely turn into some sort of ridiculous tribal thing, just like "well *I* believe in *science™,* sirrah!"
I know you’re focusing on whole body here, but the existing organ-specific screening data also seems relevant to this analysis, since that would seem to represent a best case scenario in terms of suitability of the screening modality and risk in the targeted population. Despite this optimization, it’s surprisingly difficult to establish an effective screening program. If the WB program could achieve comparable performance to each specific modality simultaneously (unlikely), what does that upper bound calculation look like?
This doesn't help answer the question, but it seems like the scanner is a an ineffective technology. In most fields a test where 98.7% of flagged results didn't need attention would be seen as a failure and wouldn't be used.
These scanners have much better than 98.7% specificity when they're being used the way they're intended. Whole-body screening isn't one of those ways which is why doctors recommend against using it.
Scott, your ledger's two most generous assumptions are that being diagnosed with an indolent cancer costs ~nothing, and that treating it costs ~nothing. Both depend on a clinician being able to tell, at the moment of diagnosis, which of your screen-detected cases are the 4 who benefit and which are the 6 who don't. That's precisely the thing we can't do. (cf. the birds/rabbits/turtles framing of Crile and Welch, which you allude to in your post - which is itself a statement that the heterogeneity is invisible in advance.)
You don't actually get to tell the turtle "just be chill"; you treat it like a rabbit, because you can't distinguish them, and the patient absorbs the full treatment harm. As a clinical researcher who has worked in exactly the field of cancer treatment-related harm, this is non-negligible. Screening-related false-positives still get the extra treatment-related harms, just not the benefits, but this can't be known in advance. More people die of heart disease than cancer, but cancer if often the scarier diagnosis, because treatment sucks so much.
PSA is a clean example, and not an abstract one for me: my grandfather believed his whole life that a PSA test made him a cancer survivor, when what it demonstrably gave him was a prostatectomy and lifelong incontinence, for a cancer there's no good evidence ever needed treating. The ERSPC trial showed how common that is: at 16 years, ~18 men had to be diagnosed (and nearly all treated) to prevent one prostate-cancer death, for a ~20% drop in prostate-specific mortality - and no detectable change in all-cause mortality!
Bretthauer et al.'s 2023 meta-analysis (JAMA Internal Medicine) found that the established, trial-backed single-organ screens barely extend overall lifespan at all. Every prior time a plausible cost-benefit case got deployed ahead of the trial - the radical mastectomy stood for most of a century before NSABP B-04 showed it saved no lives over simple resection - the unmeasured harms turned out bigger than expected. We struggle to make a solid argument for single-disease scans, let alone general scans that might pick up ... something we're not sure about yet.
This is my concern. Not that we're spending a bunch of money that some people are totally willing to burn on health screening. I'm not opposed to rich people wasting their money on things that might well make them less healthy - let freedom ring! But that we're repeating the exact same mistakes we've made time and again with screening, just with a brand-new modality. Can we please stop selling these untested screenings as "potentially beneficial with almost no chance of harm" when the history of screening in medicine should point us to exactly the opposite conclusion?
Thanks, I've added something in about this.
I appreciate it. I think you're referring to this new paragraph, which strengthens your case: "Fifth, it assumes that when all the most accurate tests are done and all the best doctors have been consulted, it’s at least possible to distinguish the cancers that should be treated from the slow growing ones that won’t hurt them. This isn’t always true, and sometimes people will be completely rational and still make the wrong choice. We can get this number arbitrarily low by setting stricter and stricter thresholds for which cancers we treat, but the stricter we go, the lower we drive number of true positives caught, until eventually our 32 QALYs gained number becomes a large overestimate. This decreases the benefits some amount, but I don’t have a good sense of how big an effect it is."
My prior is that new tests should be treated as a potential net detriment (i.e. the potential benefit is often swamped by the potential harm from treating a healthy person - and more so when looking at multiplex tests with lots of potential sources of (false) positives) until we can demonstrate net benefit.
Vinay Prasad talks about a case in his own clinic, where an otherwise healthy heavy smoker in his mid-90's came into the clinic for a checkup. He had no specific complaints, but they ran a few routine tests on him anyway, because of his long smoking history. Those tests brought up concerns, that were followed by more tests, treatments, scans, add-on treatments, etc. Eventually, Prasad realized this poor man had almost certainly received worse quality of life from his interaction with his health provider than if he'd stayed home and let the asymptomatic disease run its course. He had a strong probability of dying of something else. Risk factors, after all, should never be treated as certainties at the level of an individual.
Look, I like prevention. But that old saying that "an ounce of prevention is worth a pound of cure" often describes the opposite of what people are being sold. People don't come into the clinic to get their HbA1C down, or because they're really suffering from an eGFR that's too low. We track those things to prevent worse harms, sure. But it's always necessary in medicine that we be able to point to the hard endpoints - like overall survival or all-cause mortality - that prove we're doing more good than harm. And yes, interaction with the health care system is real, tangible harm - more so the more serious the disease, because we're more willing to take drastic steps. "Show me a healthy person, and I'll show you someone who hasn't had enough tests," is more than just an adage. It's a cautionary tale, when the test could throw an otherwise healthy person into a life-or-death struggle.
Finally, while I'm often critical of people applying Bayesian statistics in places where they don't belong, this subject is one of the best/easiest places for a Bayesian approach. If your cancer risk is high enough (you're older, maybe have family history of cancer death) there are good arguments that the net benefit of screening is positive. But this isn't being rolled out with those qualifiers or that intended target audience. It's being rolled out as potentially good data that anyone could benefit from and maybe there are harms, but the treatment couldn't possibly be worse than the disease. (It can, when there's no disease and all treatment.) A rational 35yo otherwise healthy male with a base rate of 0.07%/yr. chance of new malignancy (many of which would be turtles or birds, not rabbits), who gets a positive screening screening with 99% specificity like this is raising their chances of iatrogenic harm, without sufficiently tipping the scales in favor of intervention. Our prior should be that they are more likely to be better off NOT knowing the results of the scan than knowing them.
Thank you for this; I think many people here are really underestimating the anxiety impact and real harms cancer treatments cause, and you've put it much better than I could.
I appreciate your comment. Yeah, there's something about a cancer diagnosis that's really different.
And the treatment is horrible. A rare side effect of cancer treatment is ... cancer. Even if you cure it, you'll still come back every 3 months for scans, then every six months, then every year... Each scan a new expense and a reminder of the cancer.
Even a diagnosis of a relatively benign cancer gives you years of anxiety. It's not just the resection and adjuvant chemo (as bad as they are). It's the year 5 scan that reminds you to not get too comfortable...
I'm not sure how we'd handle this with QALYs, but in finance when making calculations like this you discount the future. You can usually use the risk free rate or expected market returns or whatever, depending on the exact situation. But the point is you don't value $1 10 years from now the same as $1 today.
Does the same apply here? I certainly wouldn't want to waste 899 hours of my prime life to save 900 hours decades from now.
It complicates the calculations, but I think at a high level the costs are realized sooner and the benefits later which nudges the result further in the "not worth it" direction.
Separately, I've sometimes wondered if it would be useful to get a full body scan and not look at it, but then I'd have it as a baseline if standard medicine asked me to get one in the future. I feel like that might have some of the benefits but without the anxiety and false positive costs. But I'm not sure if that would actually help in any medical protocols.
I suspect that the sorts of problems that can be detected by MRI scans don't wait 30 years to kill you, so that this probably doesn't matter.
This post was worth it for "MTHFR mutation" alone.
Not sure what you mean, see https://my.clevelandclinic.org/health/diseases/mthfr-gene-variant
So it was an accident? Hilarious!
Try interpolating some letters: M_TH__F____R
You know who else has a funny abbreviation? 5-HTTLPR
https://slatestarcodex.com/2019/05/07/5-httlpr-a-pointed-review/
"The other six either have such slow-growing cancers that they would never have noticed before dying of something else"
^^^ This is a major source of overtreatment that could substantially change the calculations depending on which cancer is being detected. For annual PSA screening for prostate cancer, for instance, the QALY benefit ranges from -21 to 97 per 1000 (https://pmc.ncbi.nlm.nih.gov/articles/PMC4982868/) meaning there's a possibility screening is net harmful. A big challenge for screening programs is being most sensitive to the lowest risk cases. This issue comes up for genetic screening / polygenic scores in surprising ways: the people most likely to participate in cancer GWAS are the mildest cases and so the score is being trained on the least aggressive subtypes.
Not MRIs, but I found this flyer about the cost/benefit of prostate cancer screening to be very eye opening in terms of the potential harms from unnecessary screening: https://www.annfammed.org/content/annalsfm/21/2/165/F3.large.jpg
The possible harms don't stop at anxiety and biopsies, though. You also need to consider the risk of significant harms from over-treating indolent cancers that we can't reliably differentiate from aggressive cancers, like what can happen with impotence and urinary incontinence from over-treatment of prostate cancer in response to prostate specific antigen screening.
Ah yeah both my father and I had life threatening conditions that would have been easily caught with whole body scans. He has permanent damage to his bladder and has to catheter 3 times a day now because his doc did not send him for a scan. In my case I had to go to the er and demand a scan after my insurance rejected it. As it turned out I was going into organ failure and would have been dead within 12 hours without that scan. So yeah I am a big fan of whole body scans.
CT scan with contrast. I had a burst appendix, peritonitis, and was going septic. The ER surgeon said he could not explain how I was a) still alive, b) had walked through the front door of the hospital on my own power.
Although it has been suggested, I do not have an autism diagnosis but probably have ADHD or some sort of neurodivergence. I definitely have a very high pain tolerance. Initially, when it burst, I had a lot of pain and nausea but thought I had food poisoning from some fish tacos at a dive bar. The next day I just felt sick, no fever, no pain. By the next night I felt better and had some soup, then had bad shakes and very high fever (yes I should have called 911 at that point) but took some advil, had a cold shower, fever came down I sweated it out overnight and thought "Yes I have an awesome immune system." So on the 4th day I thought it odd that I still had a low level fever, went to urgent care, they took some blood tests they thought were odd and wanted me to get a scan but insurance denied it. I did some grocery shopping and walked 30 blocks home. Took a nap. Woke up to a call from the urgent care doctor who said "I've been thinking about this and I think you should go to the ER and ask for a scan." I said okay, got up went to the bathroom and pissed blood because my kidneys were shutting down. Took an Uber to the ER. After the scan the doctor asked me if I had a medical condition that prevented me from feeling pain (by that point I was actually in a lot of pain, and they gave me 2 shots of morphine). I asked why and that's when he said he had never seen someone as sick as me still conscious. Anyway, 10 days in the ICU, and 3 surgeries later, doing okay.
Did you go get a whole body scan on a whim, or was it with a specific complaint?
See my reply to the other comment for details.
This analysis reminded me of the work of psychologist Gerd Gigerenzer. He has written extensively on the harms, overdiagnosis, and poor statistical communication surrounding prostate-specific antigen (PSA) screening for prostate cancer. He wrote "Cancer screening does not save lives." This editorial published in Healthcare in Europe, details how patients are frequently misled about the true benefits and harms of cancer screening.
Other publications along these lines include "Public knowledge of benefits of breast and prostate cancer screening: A cross-validation in nine European countries" – Published in the Journal of the National Cancer Institute. This article analyzes how vast numbers of men grossly overestimate the benefits of PSA tests while being uninformed of the risks of overdiagnosis.
"Effect of Tabular and Icon Fact Box Formats on Comprehension of Benefits and Harms of Prostate Cancer Screening: A Randomized Trial" – Published in Medical Decision Making, this study explores how presenting clear, transparent statistical facts impacts patient decisions.
Wait back up, this implies the average seemingly-healthy person has a 1% chance of currently having cancer? That seems distressingly high
(It passes the smell test - if about half the people ever get cancer and life expectancy is 80 years that's about 1/160 of getting a new cancer every year. But this includes old and unhealthy people and assumes you could have cancer for a year without knowing about it until you took the full body scan. I'm guessing most cancers affect the elderly more? Man, fun new source of constant medical anxiety).
However big of a deal you think cancer is, it’s actually much worse than that.
This is true of fifty-something year olds (the population from whom these stats are drawn) but probably false for people younger than that.
Take the prostates of men (who died of something other than prostate cancer) at autopsy and check them for undetected prostate cancer. Results: latent prostate cancer present in 5% of men aged <30 years, increasing to 59% at age >79 years.
As a male in the USA your lifetime risk of symptomatic prostate cancer is ~10%, and you have a 3% chance of dying from the disease. Risk is greater for African Americans. Anyone you treat for prostate cancer risks incontinence and impotence for the rest of their lives.
All numbers taken from https://doi.org/10.1002/ijc.29538
20+-year oncologist here: the data clearly and repeatedly shows that we treat too many breast cancers that would never be dangerous, and 5x that many never-life-threatening prostate cancers (among many others, but those are the top two in women and men respectively, so a good, non-cherry-picked sampling).
And the QOL-cost of this treatment is truly massive, beyond what you report here Scott. Specifically, for both of those cancers we give treatments that take away people’s estrogen/testosterone, and in so doing take away a big part of who those people are, how they function, and even how they see themselves as humans. We change their lives, permanently, and not for the better.
The reason experts advise against asymptomatic whole body scans (and avoid them ourselves) is because we believe the data, not to mention our own eyes.
> "[We] advise against asymptomatic whole body scans ... because we believe the data, not to mention our own eyes"
I understand and am sympathetic to the pragmatic logic, but remain stuck on the question:
Why can't the people who "treat" the innocuous cancers believe the data, and make good decisions?
Roughly:
- inertia + cultural presumption that aggressive early intervention would be positive
- general bias toward action / involvement (surgical industrial complex has incentive survival mechanisms?)
- pressure from relatively low-info high-anxiety patients
- ignored positives present higher legal liabilities than treated negatives
The first two should be easy enough to overcome if we believe medicine has anything like a scientific or technocratic arc.
The third - evidenced in this comment section by a Korean anecdote - I believe is a cultural artifact that would settle on a new equilibrium pretty quickly if mass-market scans revealed that everybody's uncle, cousin, and neighbour was carrying weird internal anomalies.
The fourth is furthest from a natural / automatic solution, but it does seem tractable in terms of documented evidentiary decision making? Individual patients will be outliers in so far as their conditions are actually much worse / more urgent than our understanding lets us guess.
My case: I was constipated for years after a motorcycle accident. Got worse and worse until I was going about once every two weeks and was basically living on soup. I saw multiple doctors who just told me to eat more fiber (I was in my 20s). Out of desperation: I got a prenuvo scan - I had an internal hernia. Got surgery and felt loads better. Night and day. No regrets!
This makes me wonder what the QALY yield would be from a protocol of "use whole body MRI as a backup diagnostic tool for people who have weird persistent debilitating/painful/annoying symptoms not resolved by the easy/simple stuff." Where you set the threshold for debilitating/painful/annoying probably matters, but seems plausible that it might target a higher yield population than the worried well.
So you sought and received a scan to identify the cause of a symptom, right? This is an argument against receiving scans when there are no symptoms or problems to focus on.
Is there no value for the ~2/3rds of patients who got a full body MRI, and it came back clean? Why would giving "peace of mind" to someone who might be nervous count for zero QALYs, while giving the remaining ~1/3rd "anxiety" count for negative QALYs?
Most people who get these are healthy with no symptoms. I don't think their anxiety decreases very much, but maybe I'm underestimating health anxiety.
If the roi of promethease were calculated this way, what would it show?
No, because 1) although it gives people many more negative hits of info 2) people have adjusted
I don't think it's right to port over the costs unchanged from a world where patients believe "tests are all meaningful and expensive and only docs can request them, so a potential problem is serious" and then plug them into a world where scans are way cheaper and faster and don't need a doctor.
By not accounting for adaptation, this kind of analysis seems like it would show that most tech or system changes should not be or have been made.
First of all, this analysis tries to minimize the effect of actual irrationalities and cultural problems, and show that even if all of those get assumed away, this is still low yield.
Second, I don't think we should assume those away. Segways might be a revolutionary invention if everyone switched them all at once and we replaced all of our car infrastructure with Segway infrastructure, but the fact that they require this unlikely thing means they probably aren't worth you personally buying unless you're feeling incredibly optimistic. Likewise, every dating site would be great and life-changing if the culture changed so that all of the highest-value singles in the world used that dating site, but that doesn't mean that dating site is good today, or that you should use it, or that it's a genius revolutionary invention. Coordination problems are often the hard part, and if you've solved everything except the coordination problem then you often haven't done very much.
What do you think uptake would look like? I modeled a gradual rollout with those especially needing it (people with DNA or family history reasons to rationally want to be scanned more often) self selecting in.
And in the background as a pressure on the mainstream, all good things to me. A broad day one rush of likely false positives would be bad - but I'm curious to know if that's something we couldn't adapt to?
The value "try to know more things about health, medicine etc" is so close to what I care about, I didn't expect the reactions against it so strongly. (I'm only referring to the case it works and eg raises scan rates 100x due to cost but with 2x worse false positives. if it doesn't work at all then it'll fail w/out opposition.) What could happen which would really lead to big problems?
I'd expected docs to be more scared to appear openly against generic "progress" measured by naive abilities like speed, etc.
I think you alluded to this in your first post, but the base rate for "newly expanded screening procedure for healthy adults improves health outcomes" is very poor. Are there *any* screening procedures (of any kind) that are currently recommended for healthy adults aged 18-40? HIV test for sexually active adults maybe? Even a lot of the traditional ones (mammograms for example) are not slam dunk in terms of cost benefit in the general population.
Also I think you downplay the risk of biopsies a little bit: we are not always talking about minor side effects; you can get antibiotic-resistant infections, sustain serious brain or organ damage, or die from a biopsy. Rare, yes, but it does happen -- sometimes I think people just brush it aside into the category of "the band-aid might cause skin irritation."
Presumably getting one's blood pressure checked is useful, or at least really cheap if you happen to be at a doctor's office for any other reason.
In the UK at least, cervical screening is offered every 5 years to relevant adults between 25 and 64.
I think the "people won't do the basics" is outdated. Almost 50% of adults get 150 minutes of mild exercise a week, and 25% do "muscle strengthening". 20% of people go to a gym/fitness class in a given week. These numbers include all adults, including decrepit elderly people. Among 18-30 2/3rds of people meet exercise requirements. Basically every food product line is updating as people learn that protein and fiber are good for you. The same doctors who complain that people won't lose weight or change anything are annoyed that people keep coming to them proactively for things they deem not "real" health issues like trouble sleeping or minor pain.
There are a ton of people willing to make pretty large sacrifices for health and wellness.
How does the doctor's liability concerns fit into this calculation?
In spherical cow land, the doctor is thinking Pr[this finding is treatable cancer] vs Pr[this finding is no biggie/not treatable/doesn't need to be treated]. But it seems like in reality, at least in the US, the doctor is also thinking Pr[this finding looks innocent but isn't and I get sued by this guy's heirs for a gazillion dollars and my malpractice insurance goes up by 200%].
The massive QALY burn from chasing underspecified MRI anomalies is an exact manifestation of a broader, systemic failure in how we apply Bayes. The medical system consistently forces continuous, structurally underspecified reality into discrete, fully-specified models.
The mathematical boundary of this error, what I refer to as the Observational Sufficiency Principle (OSP), is by reverse-engineering the most notorious example of Bayesian overconfidence: the textbook 2/3 consensus on the Monty Hall Problem.
The standard Monty Hall solution commits is a Fully Specified Stochastic Process (FSSP) error. To force the math to work, the analyst hallucinates a generative protocol for the host (q=0.5) that was physically unobservable. They import a spurious rule of selection to complete the model.
This is exactly what is happening with whole-body screening. When an MRI flags a 'mildly concerning' shadow, the generative process of that shadow is entirely underspecified. But instead of treating it as an equivalence class, the diagnostic model forces an FSSP onto it, hallucinating a precise risk likelihood that generates the 300 false-positive follow-ups Scott calculated.
When you apply a rigorous quotienting step to an actual underspecified observation space - restricting the σ-algebra to only what is admissible (the equivalence class under permutation π) - the posterior is mathematically forced to respect the indistinguishability, dropping the overconfident probabilities back to 1/2.
The insistence on completing underspecified models with post-hoc likelihoods is why predictive models hemorrhage value, whether on a game show stage or in an MRI machine.
The full π-invariance quotienting proof and taxonomy of the FSSP category error are mapped out here:
https://trissimondsen.wordpress.com/2026/06/21/the-monty-hall-problem-solved-an-observational-sufficiency-principle-osp-proof-for-1-2/
Your thoughts?
::click link::
As I suspected, it's written in a way that is difficult to understand and triggers my nonsense detector. Would not recommend.
Appreciate your look. Which part can I further clarify and what exactly triggered your detector?
+1
Someone who asserts that the Monty Hall problem probability is 1/2 is not credible.
Ah, apparently your "Monty Hall Problem" is the default, non-verifiable general consensus existence of a vanilla Fully Specified Stochastic Process (FSSP), calling itself "The Monty Hall Problem". I agree; under that model, the correct answer is 2/3.
But the historic, textbook-verified (via Priority Rule) Monty Hall Problem, one which bears everyone's intention to clarify (not from those providing the consensus answer), actually refers to themselves in a one-shot scenario with the uncertainty to be quantified for the expected payout. Because they imagine themselves standing now in front of two doors instead of three, wanting to win the car and asking themselves which door has the advantage. That Monty Hall Problem's correct answer is 1/2.
Stop schizoposting.
Understand the conflation. It's not a fight. It's the solution to an unnecessary paradox that is actually now causing real-life catastrophic failures.
I'm not fighting, I'm mocking.
I have now written over 60 long-form posts regarding the Monty Hall Problem. Below is the 61st, specifically for you. Understand that this was done not only out of pure love for the problem, but also because I simply enjoy every opportunity to create a bespoke explanation.
https://trissimondsen.substack.com/p/the-monty-hall-problem-explained
>"The game starts by you picking any door you wish."
At this point you have a 1/3 chance of picking correctly, and a 2/3 chance it's behind one of the other doors. Nothing that happens afterwards can change that fact.
>"Monty's testimony since 1975 that he doesn't do so."
Irrelevant. There's a reason it's called "the Monty Hall problem" and not "Let's Make a Deal": it's inspired by but not isomorphic to the game show.
"A few years ago, my aunt had some concerning symptoms and got a scan. The scan showed she probably had cancer. The doctor inexplicably ignored it for several months and she died. RIP."
...were they sued into oblivion?
Not to comment on Scott's aunt, but I think doctors prefer not to jump straight off to "it could be cancer" because of the fear of terrifying the patient. I had that with an appointment for a particular minor surgical procedure, where I could tell the consultant suspected it might be cancer but refused to tell me, so of course when I got home I immediately looked it up online and yes, the result came back "with this set of symptoms in this type of patient, first suspicion is cancer".
I didn't freak out because I knew it wasn't cancer (for various reasons too detailed to go into here) so I was more annoyed the doctor wouldn't be honest with me, but to be fair, had he said "it could be cancer" and had I been less certain of my grounds, I probably would have worried myself into oblivion over what turned out to be, when the results of the procedure came back, "no it's not cancer".
> 899 hours at the doctor to save 900
This fails to factor in that I'd rather be unconscious (dead) than be at the doctors.
This is an impressive ICER exercise. But as you’ve alluded to, any ICER is predicated on the intervention actually being effective…and I’m not aware of any outcome evidence of whole body MRI to that effect. You’ve made some best (and no doubt learned) guesses here but of course that doesn’t replace outcome evidence.
In the end, I suspect this will appeal to the subset where money is no object….the folks who do stuff because they can, even absent evidence that they should (probably considerable overlap with the longevity types also). In that way, this is part and parcel of the whole boutique/concierge/executive medicine trend that exists to cater to the worried well.
You say "In the past when doctors have come up with clever cost-benefit analyses like the one in this post, these sorts of known unknowns and unknown unknowns have predominated, and the real effects have been much different from the expected effects - usually much worse."
Do you have any examples?
As opposed to whole body MRI for which there is no outcome data, and whole body US which has even less, what are your feelings about PSA and/or colonoscopy, given the body of evidence available regarding those tests/procedures?
On colonoscopies, I found this article worthwhile reading: https://asteriskmag.com/issues/04/you-re-invited-to-a-colonoscopy .
Thanks for that. That’s a good summary and discussion of NordICC. Intention to treat analysis is the gold standard in RCT as it preserves the effect of randomization (as opposed to on-treatment or per-protocol)…and the screening strategy IS the “treatment” here…so I don’t understand the fuss. If someone did an RCT about a surgical procedure and only included those in the “treatment” arm who actually underwent the surgery, they would be roundly mocked, and deservedly so.
I'm not super optimistic about full body scans as the best investment in medical technology, but Midjourney's plan before this was to marginally improve image generation models. Image generation is mostly zero-sum imo (a bit of a hot take, but can we at least agree a lot of the rewards to Midjourney come from competing with other mildly inferior image sources instead of providing actually new and good services?) So it's better for them to be working on full body scans even if they're very overhyped, as long as they somehow won't do a bad enough job to set the field back worse than had they never entered. The only exception is if the availability of full body scans is somehow negative, and even if they're bad for most people there has to be some way to restrict it to people who can benefit, like those who can't have an MRI for whatever reason.
I think it's important to consider that we have arguably already caused great harm and wasted a great many resources through overzealous screening. H. Gilbert Welch has some nice work suggesting that screening mammography and other popular forms of cancer screening have not actually saved lives but have led to many, many more people receiving toxic therapy that couldn't have helped them.
So while I think it's peobably true that screening which costs nothing and is undertaken by perfectly rational actors cannot harmful, we already have several decades of history of screening being harmful.
Is there a good writeup somewhere on screening tests, false positives, mean time to survival stats, and the like? Many years ago, the blog Respectful Insolence had a really excellent sequence of posts explaining this, but I couldn't find them the last time I looked. (Notably, you get the fun paradox where mean survival time goes up because you are doing more early detection of some untreatable cancer--nobody actually lives longer, they just find out about their cancer a year or so earlier.)
I'm not aware of a really good written summary of the entire field, but a lot of H.G. Welch papers are quite readable and describe the phenomenon convincingly. I'd recommend:
"Effect of Three Decades of Screening Mammography on Breast-Cancer Incidence"
Authors: Archie Bleyer, M.D., and H. Gilbert Welch, M.D., M.P.H. Publication: New England Journal of Medicine, 2012
"The Rapid Rise in Cutaneous Melanoma Diagnoses." Authors: H. Gilbert Welch, M.D., M.P.H., Benjamin L. Mazer, M.D., M.B.A., and Adewole S. Adamson, M.D., M.P.P. Publication: New England Journal of Medicine, 2021
"Are Increasing 5-Year Survival Rates Evidence of Success Against Cancer?" Authors: H. Gilbert Welch, MD, MPH; Lisa M. Schwartz, MD, MS; Steven Woloshin, MD, MS Publication: JAMA, 2000
This suffers from spherical cow in another direction. Even if everything above is correct and even optimistic: we're moving FAST in a direction where the capacity to analyze data increases by orders of magnitude (plural). What Claude Code can do with your genome, clinical data and MRI history is already more than any single doctor will ever have the patience to do. It currently needs guidance and common sense provided by a human - but It's a matter of months until you get harnesses which prevent its worse mistakes. And then you get Mythos gen models and better harnesses.
So not else is being equal. 3 years ago this all was, at the very least, debatable. But now the "more data" camp wins hands down.
I always thought one of the main objections by doctors to scanning healthy patients is because MRI machines, operators of the machines, and interpreters of the scans, are all constrained resources, that are better distributed to patients who are sicker.
I also think that if such scanning and the knowledge of "minor cancers" became much more commonplace, the general sentiment and level of anxiety regarding them would calibrate itself. re the comment on Taiwan
If the patient is paying, that's not a big problem, since the money from the patients can be spent buying more scanners and hiring / training more technicians and radiologists and such.
Good post.
1. If the cost/benefit is so close, why do doctors insist that multiple large trials of preventative screening have been done and failed, across multiple types of cancer?
2. Why are we even talking about cancer in response to the Midjourney announcement? AFAICT they said nothing about cancer or tumors in any of their announcements. All this made sense several months ago when we were talking about Prenuvo, but all Midjouney has said they will do is provide body composition info and then later apply to the FDA for more stuff.
I think the elephant in the room is the value of the data, independently of how they are immediately used to benefit individuals patients. Imagine you collect a few millions of full body scans and you also follow up the patients, learning long term outcomes. You will be able to train a predictive model that may turn out to be invaluable for future patients. Whether you see this as a data grab or as a great opportunity for future patients, this is likely to matter more than the impact on current patients.
Ideally yes, but
>Imagine you collect a few millions of full body scans and you also follow up the patients, learning long term outcomes.
Is any organization in a legal position to do this? There is the NIH All of Us research program, but that has a pool of volunteers. Given health information privacy laws (which exist for good reasons), I'm a bit skeptical that millions of follow ups are feasible.
I find those tweets curiously enraging, and I think it's because these are the types who would otherwise be smugly "trust the science, bro; look, just hand over control of society to the technocrats and it'll be perfect" but when it comes to this, then they're all "waaah, why is the mean ol' Harvard physician not letting me, a tech head, play with the shiny new toy I want to play with? what does he know, anyway!". Also perhaps a hint of "I am so well-established in life that I can afford the kind of health insurance that will pay for fripperies, because I am Smart and Educated and High Value Human Capital" at play as well.
You know what, Midjourney? I was highly dubious about this entire project but now I say "Go right ahead. Slap AI on everything! Soak these pigeons for every dollar you can squeeze out of their high value backsides! Make it an entire spa package and charge 'em $$$$$$$ for what in effect is a Seltzer bath with knobs on!"
I figure these are the types who probably were all over NFTs as well, so if they're going to throw money away, why not scoop it up in a pivot from "we help you draw your own porn tailored to your specific kinks" to "ah yes, the field of medicine, a vast plain of opportunity for quackery dressed up in the latest buzzwords".
"If you don't otherwise have symptoms, ignore the scan", forsooth! Hey, Amanda, how's about if you don't otherwise have symptoms, you don't get the scan in the first place, and save all the time and bother?
If these people want 24/7 monitoring of their health so they can gather data, then I say let 'em have it, hot and heavy. Blood draws every hour on the hour. All the poking and prodding and hanging around and "take off all your clothes and put this skimpy paper gown on and then sit around waiting until a doctor deigns to poke his head around the door" galore. Tests, tests, and more tests, till they don't know whether they're coming or going.
They should get their blood the old-fashioned way; go to a kosher butcher and drink it fountaining straight from the slashed throat of the animal.
(Years and years back I read something that this was a cure recommended to consumptives, to drink animal blood, and that there was a difference between blood from animals slaughtered the usual way and kosher slaughter).
Well, if it actually works, the next step is to figure out what _specifically_ is the useful component in the 18-yr-old blood plasma, and gene-engineer yeast to synthesize it, at which point it will be generally useful to all of us. _IF_ it actually works. Look at it this way: Those guys have volunteered to be lab rats for the rest of us. Remember that virtually _all_ of medicine looked "creepy" when first introduced - look up the original comments on the first smallpox vaccine, for instance.
And the creepiest thing of all are the politics of envy.
You don't think it's at all slightly creepy to be treating your own son as a pomme de sang (to borrow a term from the Anita Blake novels)?
There's guinea pigs, and there's 'holy hell, what???' Organ transplant because family member is dying otherwise, and there's no better match - okay. "Dad is scared spitless he'll get old and die like every other human so far, so he's scrabbling about grasping at straws" - nope, sorry.
If the guy is doing it to 'look 18' as per the article, it ain't working. He looks very sallow and anxious, and if he's this scared of getting old (and eventual mortality) no wonder he looks like he's living on his nerve ends.
Swear to me aunt, "Good Lady Ducayne" to the life:
"For Mr Bryan Johnson, the plasma exchange is not an unusual happening.
He has been to the Dallas-area clinic for several consecutive months and received plasma–not from a family member but from a young, anonymous donor.
He carefully screened the donor to make sure the person had an ideal body mass index, lived a healthy lifestyle and was free of diseases."
Lady Ducayne:
"Bella went in fresh, blooming, a living image of youth and hope, and before she looked at the Person her gaze was riveted by the owner of the chariot.
Never had she seen anyone as old as the old lady sitting by the Person's fire: a little old figure, wrapped from chin to feet in an ermine mantle; a withered, old face under a plumed bonnet--a face so wasted by age that it seemed only a pair of eyes and a peaked chin. The nose was peaked, too, but between the sharply pointed chin and the great, shining eyes, the small, aquiline nose was hardly visible..'This is Miss Rolleston, Lady Ducayne.'
Claw-like fingers, flashing with jewels, lifted a double eyeglass to Lady Ducayne's shining black eyes, and through the glasses Bella saw those unnaturally bright eyes magnified to a gigantic size, and glaring at her awfully.
'Miss Torpinter has told me all about you,' said the old voice that belonged to the eyes. 'Have you good health? Are you strong and active, able to eat well, sleep well, walk well, able to enjoy all that there is good in life?'
'I have never known what it is to be ill, or idle,' answered Bella.
'Then I think you will do for me.'
'Of course, in the event of references being perfectly satisfactory,' put in the Person.
'I don't want references. The young woman looks frank and innocent. I'll take her on trust.'
'So like you, dear Lady Ducayne,' murmured Miss Torpinter.
'I want a strong young woman whose health will give me no trouble.'"
Many Thanks!
>You don't think it's at all slightly creepy to be treating your own son as a pomme de sang (to borrow a term from the Anita Blake novels)?
>There's guinea pigs, and there's 'holy hell, what???' Organ transplant because family member is dying otherwise, and there's no better match - okay.
Aging _is_ a lethal disease. As things stand, it eventually kills all of us. As I said, and I continue to stand by it:
>Remember that virtually _all_ of medicine looked "creepy" when first introduced - look up the original comments on the first smallpox vaccine, for instance.
To add another example, since it is now part of this thread: There has been (and, to a smaller extent, _continues_ to be) a _lot_ of wringing of hands about transplants. Now, there are reasonable questions about a lot of medicine: efficacy? side effects, short term and long term? limited supplies of various sorts? etc. But 'creepiness' is now, and has always been, a lousy criterion for evaluating medical experiments.
As with any slowly moving malady, attempted treatments for aging _do_ have a very high bar to clear. Ensuring that some novel treatment of e.g. a 50 year old to prevent them dying in 30 years _doesn't_ have a side effect that kills them in 20 years is a very tough problem.
Yet again, over-anxious people with more money than sense turning "Good Lady Ducayne" (published 1896) into a 'how-to' manual rather than a work of supernatural fiction:
https://gutenberg.net.au/ebooks06/0605261h.html#ch03
"She pushed up her loose lawn sleeve, and exhibited a scar, which he scrutinized intently, with a surprised and puzzled look.
'This is no mosquito bite,' he said.
'Oh, yes it is--unless there are snakes or adders at Cap Ferrino.'
'It is not a bite at all. You are trifling with me. Miss Rolleston--you have allowed that wretched Italian quack to bleed you. They killed the greatest man in modern Europe that way, remember. How very foolish of you.'
'I was never bled in my life, Mr Stafford.'
'Nonsense! Let me look at your other arm. Are there any more mosquito bites?'
...Stafford examined both her arms in the broad sunlight, scars new and old.
'You have been very badly bitten, Miss Rolleston,' he said, 'and if ever I find the mosquito I shall make him smart. But, now tell me, my dear girl, on your word of honour, tell me as you would tell a friend who is sincerely anxious for your health and happiness--as you would tell your mother if she were here to question you--have you no knowledge of any cause for these scars except mosquito bites--no suspicion even?'
...'How long do you mean to stop with Lady Ducayne, Miss Rolleston?' Herbert Stafford asked, after a thoughtful silence, breaking suddenly upon the trivial talk of the two girls.
...'Even if you feel your health breaking down in her service?'
...'Then you mean to go on till you break down; to die at your post?'
'Like the other two companions? No! If ever I feel seriously ill--really ill--I shall put myself in a train and go back to Walworth without stopping.'
'What about the other two companions?'
'They both died. It was very unlucky for Lady Ducayne. That's why she engaged me; she chose me because I was ruddy and robust. She must feel rather disgusted at my having grown white and weak. By-the-bye, when I told her about the good your tonic had done me, she said she would like to see you and have a little talk with you about her own case.'
...'You have begun upon my companion, she tells me.'
'I have prescribed for her, certainly, and I am happy to find my prescription has done her good; but I look upon that improvement as temporary. Her case will require more drastic treatment.
'Never mind her case. There is nothing the matter with the girl--absolutely nothing--except girlish nonsense; too much liberty and not enough work.'
'I understand that two of your ladyship's previous companions died of the same disease,' said Stafford, looking first at Lady Ducayne, who gave her tremulous old head an impatient jerk, and then at Parravicini, whose yellow complexion had paled a little under Stafford's scrutiny.
'Don't bother me about my companions, sir,' said Lady Ducayne. 'I sent for you to consult you about myself--not about a parcel of anæmic girls. You are young, and medicine is a progressive science, the newspapers tell me. Where have you studied?'
'In Edinburgh--and in Paris.'
'Two good schools. And you know all the new-fangled theories, the modern discoveries--that remind one of the mediæval witchcraft, of Albertus Magnus, and George Ripley; you have studied hypnotism--electricity?'
'And the transfusion of blood,' said Stafford, very slowly, looking at Parravicini.
'Have you made any discovery that teaches you to prolong human life--any elixir--any mode of treatment? I want my life prolonged, young man. That man there has been my physician for thirty years. He does all he can to keep me alive--after his lights. He studies all the new theories of all the scientists--but he is old; he gets older every day--his brain-power is going--he is bigoted--prejudiced--can't receive new ideas--can't grapple with new systems. He will let me die if I am not on my guard against him.'
"If any real person went to their doctor for an hour daily for years to decrease their risk of cancer by 5%, we would call them an insane hypochondriac!"
The good old-fashioned term for doctor-botherers is valetudinarian, and the likely market for this procedure. Perhaps we should call them Sholtos, in honour of the example of the species as per Conan Doyle's "The Sign of the Four":
“This is Mr. Sherlock Holmes, and this is Dr. Watson.”
“A doctor, eh?” cried he, much excited. “Have you your stethoscope? Might I ask you—would you have the kindness? I have grave doubts as to my mitral valve, if you would be so very good. The aortic I may rely upon, but I should value your opinion upon the mitral.”
I listened to his heart, as requested, but was unable to find anything amiss, save indeed that he was in an ecstasy of fear, for he shivered from head to foot. “It appears to be normal,” I said. “You have no cause for uneasiness.”
“You will excuse my anxiety, Miss Morstan,” he remarked, airily. “I am a great sufferer, and I have long had suspicions as to that valve. I am delighted to hear that they are unwarranted. Had your father, Miss Morstan, refrained from throwing a strain upon his heart, he might have been alive now.”
...“That is well! That is well!” said he. “May I offer you a glass of Chianti, Miss Morstan? Or of Tokay? I keep no other wines. Shall I open a flask? No? Well, then, I trust that you have no objection to tobacco-smoke, to the mild balsamic odour of the Eastern tobacco. I am a little nervous, and I find my hookah an invaluable sedative.” He applied a taper to the great bowl, and the smoke bubbled merrily through the rose-water.
...I stammered out some few halting words of congratulation, and then sat downcast, with my head drooped, deaf to the babble of our new acquaintance. He was clearly a confirmed hypochondriac, and I was dreamily conscious that he was pouring forth interminable trains of symptoms, and imploring information as to the composition and action of innumerable quack nostrums, some of which he bore about in a leather case in his pocket. I trust that he may not remember any of the answers which I gave him that night. Holmes declares that he overheard me caution him against the great danger of taking more than two drops of castor oil, while I recommended strychnine in large doses as a sedative."
> So in total, 8 of the 1,000 people have benefited. The average benefit among those 8 people is 4 quality-adjusted life-years, so the total benefit from the scans is ~32 QALYs.
> What’s the total cost?
I don't think this is a good framework in which to view the policy issue. One benefit of doing regular scans of normal people is that you develop a sense of what normal scans look like. (And also a sense of what abnormal scans look like!)
This means that the scans become more valuable over time, because we can interpret them more reliably.
Arguing that they're not very valuable 𝗻𝗼𝘄 misses this point.
> A onetime snapshot of a "could be benign or slow-growing" cancer is unlikely to be better in the future.
This is the opposite of what I'm saying. You're wrong in two ways:
1. A onetime snapshot of a mysterious lump is in fact likely to be better in the future;
2. Cancers are not the only thing we can see on internal scans. Cancers are what motivates the scans now. With better understanding of what we can see, we'll look at more things.
Is QALYs the right measurement for understanding this? Are the doctors who recommend patients not get screening procedures optimizing for QALYs? Like if suggesting screenings gets a doctor 8 5-star reviews and 992 1-star reviews,and suggesting no screenings doesn't, then the ""best"" doctors will suggest no screenings and have plausible reasons why.
How I know I'm not The Exception: routine checkups eventually surfaced that "you probably have prolactinoma, but it's a benign brain tumor, don't worry"...and this freaked me the hell out for years. Every headache that seemed to originate from one particular location, every time vision on one side would go weird, every time "brain fog" would inexplicably descend - it's the brain cancer, I'm dying! And the main reason I never followed up with neurology, despite a recommendation, was indeed cost...attending free city clinics on spotty/no insurance is not a choice a rich person makes. This would have turned out to be a mistake anyway, since things cleared up on their own over time with medication adjustments.
That's the direction I thought the MidScanners section would go - holding the same can opener assumptions steady, let's minimize both the cost and time variables, and also add a further assumption that AI ultrasound magically reaches MRI levels of efficacy. If such a technology had been available back during my scare, would that have actually been a good thing? I don't know! Even if they're all quibblesome numbers, the odds as given don't look especially promising in the median case...so I'm mostly still interested from an R&D Yay Progress standpoint. AI has already done hat tricks with far less, or so they say.
Surprised you didn't address the "if we have squillions more scans, then AI can process them and get vastly more capable at discriminating stuff that matters from the rest" claim. More broadly, it's a mistake to think that the medicine of 5 years ago looks anything like the medicine of 10 years from now.
There’s a Nate Soares quote that breaks my brain ‘there may well come a day when humanity would tear apart a thousand suns to prevent a single death.’ It’s part of his discussion of how the cost of a life isn’t the value of a life.
But while money is a mostly-adequate measure of costs and values, it’s still an artefact to try and measure something else that we actually care about. I think this is the same point as not actually wanting to spend 800 hours in scans for 4 more years of life, or the old joke ‘Exercise doesn’t make you live longer, it just feels like it.’ If building a scanner involves torturing small children to work in heavy metal mines, and going for a scan involves hours of waiting in hospitals and months of anxiety, whatever cost the economics has put on that might not be a good measure of the cost. On the other hand, if getting a scanner involves some bored teenagers and an AI being super happy doing a summer project, and getting scanned involves walking into Starbucks through a gate that snoops your health and never tells you if you don’t need to worry, that’s different.
I think medics have a very good pragmatic feeling for ‘what is worth it’ and a really good sense that we’re all going to go of something anyway (doctors’ approach to end of life care / chemo etc is really interesting) and that’s why they come down more anti whole body scanning than the numbers suggest.
With the infinitely rich man, the cost of the scan is trivial, but the value / opportunity cost of his time is more complicated - at the moment we end up with the cost of a QALY being about ‘if a normal person worked normally for a year, they could buy a year of life’, which feels a normal thing for the market to do. But if you are infinitely rich, you’re saying ‘I’d rather be sitting in this scanner and having these tests than be going into low earth orbit or eating grapes peeled by the attractive people of my choice’ - they might value their time now more than their expected time aged 80 (which is where the 4 extra years are likely to go)
I somewhat disagree that "more scans" reduces the marginal benefit. In the extreme (which to be clear I don't believe is still even remotely close to possible) where I can every day just get in the bath and have a whole-body scan done, same as I could step on a scale and weigh myself, then the benefits clearly go back to being positive. There is no "is this zit dangerous?" question any more. I can just wait a few days and monitor its size precisely; its trajectory will say whether I need to worry or not. The anxiety is greatly reduced. For a comparable example of something I *can* already do - all the little defects would become equivalent to the defects I can already see: the moles on my skin. I look at my moles and check them. I know one of them COULD be cancer. Occasionally I get a scare by one that really looks a bit weird and keep a close eye on it for some time, but then I see it doesn't really behave any more weirdly than the others and relax. Very occasionally I go to a doctor to have it checked. And very very occasionally (this has happened twice to me) the mole is annoying, painful, or worrying enough to get it surgically removed. But ultimately, the mundanity of it all means my bar adjusts; and I'm a comparatively rather anxious person.
There probably is a minimum somewhere in there - an amount of scans that is the worst possible amount, enough to cause worry and not enough to deliver the ideal benefits. But at very very high frequency, daily scans are obviously better than no scans.
I place myself in the following category.
I’m a rich person who doesn’t care about time or money, and is immune to anxiety, and won’t make irrational decisions.
The reason I would get a Full Body Scan (FBS) is so that I have a documented reference point for any future health issues that may/or may not be easily identified. The hope is that having a comparative scan at a previous point in time could potentially help identify an issue sooner and track the growth rate. For example detection of an Aortic Root Dilation change could be the difference between living 1 year or 20 years if not diagnosed separately.
I would rather have the opportunity to spend my savings on holidays and memories with my partner than die and leave it to the government to not spend wisely.
Is there no way to have better thresholds for when to escalate care and when not to? Maybe the scan shows a small, likely insignificant finding. Then you don't do followup. Or it shows a really big, bad finding. Then you do followup.
1) This seems like a pretty long-winded way of saying that it's all about the cost and accuracy of the technology. More information is always good, but current tech is too costly. Well... isn't that exactly midjourney's pitch?
2) regarding the negative QALYs from anxiety. I don't buy the model. People are afraid of death, bodily malfunction / decay, etc. Their first encounter with it will always be scary. But this is a *fixed* cost. There's only so much shit to give about death.
Anecdotally, people get more relaxed about it once they've faced it in some form. It follows that the total QALY loss is simply time shifted. It doesn't matter whether you have to work through those feelings based on a medical scan, your neighbor suddenly dying, or what have you. It's unfair to put this cost entirely on the balance sheet of the scan itself.
Furthermore, the culture will eventually adjust to the new reality. An incidentaloma wouldn't be scary if you had literally dozens of people around you treating it as a normal thing that happens. There are many well-understood dangers in the world that don't induce constant anxiety.
A comment on the 'anxiety' aspect, which I think most healthy people will discount as overblown, or paternalistic. You probably feel that in no universe would you ever want to have less information - certainly that was the prevailing opinion on Twitter in most threads I encountered on this debate.
I have multiple sclerosis, and I know something about medical anxiety. I was diagnosed in my late 20s, just after completing the long process of schooling and exams to practice law. I went from being on top of the world, at the height of my optimism and ambition for life, to sitting in a room being told I had lesions on my brain, feeling the walls closing in on me.
MS is indeed a motherfucker of a disease, but three characteristics make it particularly insidious: it is neurodegenerative, progressive, and incurable. The day you are diagnosed, you are told that you will live with it forever, you will almost assuredly get worse over time, and the rate of your worsening is unpredictable.
Most MS patients become obsessed with the future, because we don't know what kind of future we will have. Will we lose our sight? Our ability to walk? We will be exhausted all the time? Or perhaps a mix: we’ll lose 30% of power in one leg, sight in on eye, and be tired most of the time. But even if we accept that these things will happen, it’s difficult to go through life not knowing when. That anxiety over an uncertain future begins to dominate, and become a source of pain all on its own separate from your actual disease course. But I suppose in some deep sense, health anxiety due to the diagnosis is part of the disease too.
An important caveat is that some small percentage of MS sufferers (3-6%) get through life with a benign course that does not affect their lifespan/healthspan. So when you first get scanned, you hope you are one of those lucky few whose initial neurological episode is deemed Clinically Isolated Syndrome - which never develops into full blown MS - or you will continue to get episodes but never progress into real, life-changing disability.
About 6 years ago, I had a stem cell transplant (HSCT) which is the most aggressive treatment for MS, involving high dose chemotherapy and a host of other drugs. It’s risky, it’s very hard on the body, but it resulted in remission for me. For the first 3 years post-transplant, I would regularly have mental breakdowns and periods of depression because I would experience an uptick in some symptoms, and I feared that it must be a sign that the disease had re-activated, and I was back on the descent into hell that was my pre-transplant life.
Eventually, my transplant doctor emailed me back after one of these episodes and just plain told me that I needed to drop my obsession with my condition, and learn to face those fears, because they were doing more damage than anything else. He was right.
It has taken a lot of time to gain some footing against health anxiety. It’s still there, though not as powerful as it once was. HSCT is the best and most effective treatment for my type of MS, but for most people the remission doesn’t last forever. So statistically, I am still expecting a return to active disease. And however much I can tell myself that I have grown, that my psyche is more resilient - I know that deep down, that question mark still hangs heavy over me.
Whatever happens, it is undeniable that I have lost a lot of enjoyment of life that I "should" have had in the last several years because I was worrying about something that had not yet come to pass. And if it does come to pass, then I’ll have just wasted the good years worrying about the thing that came anyway later on anyway.
To Scott's point, if you are a completely rational agent with newly diagnosed MS, you would simply run through the following process: you collect the medical information (how long they think you've had it, how much damage is done), you research the best treatment options available, then pick one, adjust lifestyle factors to minimize severity of disease course, and that's it. That's all you can do. After that, worrying only adds to the level of inflammation your body is already dealing with. But in my experience with fellow MS patients, it's rare to find such a person. There is too much at stake to really just 'let go', even though holding on is not accomplishing anything.
I think 99% of people would find it very difficult to be handed a report saying “You might have a cancerous growth but it’s probably nothing”, and then just trust that report and sleep soundly at night. You can read all these stats that Scott posted, and if you were told that the finding on your full body scan had a 0.1% chance of being something dangerous, you would insist on investigating it fully and completely. I imagine the same for 0.01%, and perhaps even lower.
The first neurological episode I had began with numbness in my feet and palms, and quickly spread outwards, down my limbs and towards my torso. I went to the ER, waited 7 hours to see a doctor, and was sent home with an appointment to see a neurologist 4 months later.
I had to go to the ER three nights in a row to finally get admitted. Nobody took my symptoms seriously, and even after admission, they wouldn't send me for an MRI because it was Thanksgiving weekend, and i wasn't a pressing enough concern.
I sat in the hospital for 5 days with zero treatment and zero answers, with every doctor telling me it was nothing, until finally the MRI showed those lesions. Then they told me it's likely MS or a handful of other things, none of them any better.
Soon after, when I got assigned an MS specialist doctor, I felt at ease... he suggested to me a treatment course, I followed his advice, and it later turned out he prescribed me a medicine which does almost nothing (copaxone). I spent 2 years getting worse while the medicine silently failed me, and then when I finally just completely broke down with a big relapse, he said we need to go more aggressive and suggested a real drug.
At that point I took it into my own hands, stopped trusting what the doctors told me, and started reading. I found out about the stem cell transplant and all of the other drug options. That same doctor refused to even refer me for the transplant, said it wasn't the right treatment for me... despite the fact that every single journal article/study I read said that this was actually the very best option for young patients who were not already badly disabled....and here I am 8 years after that conversation, still in remission, feeling better than I was in my late 20s.
I've seen both sides of the coin, both trusting and distrusting doctors.
I wonder how the calculation changes for someone who already has a high level of health anxiety.
Thanks for doing the sums for me and justifying my, almost certainly irrational, decsion not to have one. I think the self selection problem is huge here - Old men are notoriously reluctant to go to the doctor (in the UK) unless their leg falls off or they grow a third head.
One thing you are missing is the fact that everyone is going to die anyways.
If humans were generally immortal *unless* they get cancer, the benefit to screening for cancer goes way up.
This is a slightly different point than saying that some slow cancers won’t affect you because you’ll be dead before then.
It’s something I’ve thought a lot about because my grandma had ALS, which is mostly not a heritability risk (family members are more likely to have other brain conditions like depression, autism, and schizophrenia, but not ALS), but for like 10% of cases there is a heritable mutation. The heritable form usually starts earlier and takes longer to kill you. At this point I think it’s unlikely given the age of my dad and my aunts and uncles that our family has the heritable form, but when my dad was younger it certainly weighed on my mind a lot. Would I want the genetic testing to know if I had it?
The answer I came to was “no.” I decided that people aren’t meant to know what kills them that early, that psychologically, it is not a developmentally appropriate thing for someone in their 20s to know. If you knew what was going to kill you, you start thinking about what is going to kill your future kids or if you should even have kids. But nobody is supposed to know what will kill their kids.
Developmentally, your early 30s are the time to have an existential crisis, when you start to realize you aren’t going to live forever and need to do something meaningful. But you can’t create meaning about the future thing that will kill you, because by doing so, you are not living in the moment and you are trying to postpone the inevitable, which kind of ruins the whole existentialism vibe.
Since I will die anyways, I do not want to know how until it starts affecting me.
2K per MRI has got to be a huge underestimation. You're going to get bills from both the ordering physician and the radiologist.
Side story: My kid has a bicuspid aortic valve which is an extremely minor heart defect that about 1% or the population has. You can live into old age with no problem, but they say about 50% of people with it will need a valve replacement (which has become a pretty simple non-invasive surgery) by age 50. But in rare cases you can have an aortic aneurysm related to this that causes instant death.
So the protocol is for my son to get an echocardiogram every 18 months. Each time we get bills for around 4K (usually one from the "outpatient hospital" one from the cardiologist and one from the radiologist).
On the one hand--worth it to prevent a catastrophic outcome for my son. On the other hand, over his childhood that's about 24K that could have gone to his college savings account and we have zero proof that any of this screening benefited him at all, as this defect rarely causes problems in childhood and if anything, may cause him problems in old age. 🤷♀️
Well, I have a 10K deductible so almost never meet it. Some is reimbursed through an FSA.
Among the dozen caveats and assumptions, I would add one more: that measuring this makes sense at all, and that the units used for the measurement are the correct ones. Even within broadly utilitarian framework, I would argue this sort of calculation strongly relies on hedonistic utilitarianism with no discountings. Given preference utilitarianism for example, it doesn't strike me at all as obvious that if individuals assigned 100 utils to a lifetime of 100 years, that they would have to assign 105 utils to a lifetime of 105 years (with healthspan correspondingly elongated). Or, even from hedonistic utilitarianism point of view, I think there are reasons for discounting that I find perfectly sound, but that aren't included in the usual QALY calculations, such as (at least in expectation, a couch potato centenarian *could* in principle pick up extreme sports) reduced amount of novel experiences likely contributing to faster passage of subjective perception of time, which would lead to temporal discounting.
Or, more broadly, while it seems to me that a wise and prudent person would usually follow sound advice from evidence-based medicine, "a life well-lived" (which, as above, even a preference-utilitarian might value) might look less like QALY-maxxing and more like living in the moment and taking a cancer diagnosed too late in your stride. I can see an argument for the opposite ("healthy mind in a healthy body"), but even that would seem to suggest flourishing involving exercise, a healthy diet, etc, not so much delaying the death from preventable disease.
TL;DR - tragedy of the commons. What is arguably rational individual behaviour is a collectively bad outcome.
It's a decent QALY analysis at micro-level; it's also blind to macro-level constraints and equilibrium effects. I think it relies entirely on three unexamined assumptions:
1) Scan volume is elastic and can scale with no effects/costs/requirements on units, staffing, or quality.
2) The capacity of the broader medical system for the follow-up cascade is elastic.
3) Neither (1) nor (2) above will displace other patients with the medical system, or that there will be zero QALY costs of such displacements.
This argument has been used here in the UK to justify no mass screening for prostate cancer, with the predictable backlash from those who haven't even googled the side-effects of false positives.
Then again, this is early days and when we have enough data AI will be able to catch 90% of the significant cases with no further tests.
The real question we should be asking is: given $x to spend what gives the best health outcomes?
As a pathologist who deals with this kind of discussion regularly, I have mixed feelings about it. I got the Galleri test myself. But one thing is certain: there is a downside to a false positives beyond simple cost and anxiety. People seem to be taking the approach of, maybe this finds something useful (positive), maybe it finds nothing (neutral), maybe it finds something to watch (slight downside). But that paints an overly rosy picture. Whether by bad luck or by incompetence, bad things happen, and I've seen things like a 40 year old man who died on the table getting a second biopsy that probably wasn't necessary. A spleen coming out (you need those!) because someone overthought a report. Once in a while I'll make the call myself after numerous failed biopsies and tell the treating doc to stop chasing something, that they should watch and wait because it's almost assuredly nothing. Sometimes it's like they need permission to stop. These are the kinds of things that are difficult to document in a formal way as part of a study. If people want to do this, I don't really have a problem with it, but "Rare and minor side effects of the testing might contribute -0.005 QALYS per person" is cold comfort to someone in the hospital with a collapsed lung after biopsy misadventure chasing nothing.
I think it's underrated how fucking annoying it is dealing with medical things. As a middle-aged woman with some health issues, my minimum number of appointments per year is 10 (2 dentist, 2 visits to my normal doctor, 2 blood draws, 2 visits with my cardiologist, 1 echocardiogram, 1 mammogram), assuming no follow-ups. I've had a slightly elevated white blood cell count for years and for the past couple of visits, my doctor has been trying to refer me to a hematologist about it in case I had leukemia. So I finally did that. One visit, one blood draw, no leukemia and it ended there.
Was it likely I had leukemia? No. And I got off easy on the testing this time. But there's always a chance that a hematologist will check you for 20 things, one will be out of range (this is similar to the M&Ms xkcd thing, if you think about it), that will lead to further testing, and then in the end it will either be nothing, or you'll turn out to have some syndrome that requires regular checking for the rest of your life but probably wouldn't have done anything to you if you'd never known about it. Of course there's a small chance your life will be saved, which is why I didn't opt out of the whole thing.
Because I got scanned for atherosclerosis, I know I have a small aneurysm, so I have to monitor that annually in case it grows and kills me. I guess I could choose to ignore it, but that seems hazardous. I guess it's good to know about it. But it also sucks to know about it and it sucks to have a whole echocardiogram about it every year and wonder when it's going to kill me. Probably it would never have killed me.
And I think this is all for screenings that are proven beneficial. Do I really want to know my left ovary is 2% larger than normal? Or there's some funny area on my spleen? I kind of don't.
One of my friends is thinking about getting a whole body scan because he wants to have a baseline. I think he might be able to handle hearing about small anomalies.
One item to consider in the US healthcare system is that, for many people on third-party insurance plans, there comes a point at least once per year where their healthcare is "free" (to them) and they are effectively cost-insensitive billionaires. This occurs any time they have exceeded their deductible (for HSA / HRA plans) or their out-of-pocket maximum has eliminated their copays (for more traditional plans). This isn't everyone, but it's tens of millions of people every year that, come June or August or November, don't care at all about healthcare cost.
As a tangential matter, I was extremely interested to hear about actual literature quantifying (or attempting to quantify) QALYs for anxiety and psychological stress, as well as QALY numbers for monetary cost. Of particular note, has anyone used this literature to reassess the Covid response? Because one of the narratives about, e.g., cloth masking in the US was that it was "cost-free" and thus unassailable even though its detractors' argument was that it helped spread panic, anxiety, and psychological trauma (especially for younger kids) and also that it primed adults to perceive extra risk everywhere, leading to added anxiety and also increased openness to harsh response tactics. There was also additional massive stress and anxiety from lockdowns, as people lost their livelihoods and their social connections. If this type of added anxiety were near 0.01 QALYs per person, across the US population that would be over 3 million QALYs. Based on the age profile of decedents, that would suggest NPI anxiety alone would have nearly matched all the QALYs lost to ALL deaths in the pandemic, and many times that of the alleged QALYs "saved" by such NPIs. The lost money would loom even larger.
Also, once you start giving bad advice to calm people down, you lose credibility. Next time, you'll be giving actually useful advice and a bunch of people will ignore you, figuring this is like that nonsense with the cloth masks.
You are missing a term on the benefits side of the equation: personal autonomy
I think there is one little error in attributing doctors some wisdom of erring on the side of caution by not recommending screens.
There are all kinds of situations where doctors over screen, and those are also defended as erring on the side of caution. And if the guidance were flipped, I bet the rationale would remain.
It can’t be both!
Also, I need to read it again, and I’m on mobile so I can’t scroll as I write this, but was the very slightly below the bar QALY per dollar figure a combined number for the four out of ten really severe and also the four out of 300 from follow-ups? If it was, isn’t the natural interpretation that the four out of 300 are well below the bar and shouldn’t actually be followed up upon?
Another point about how many people will get these scans and how they will be distributed, that I think buttresses the anti-scan argument, is that the US system tends to massively balloon recipients of any initially niche intervention. If the scans are OK for some subgroup (e.g. "rational billionaires"), the political impetus is that "this is a life-saving technology, it should be available to everyone." Then there is political pressure to add it to Medicare coverage and mandate coverage by private insurance. Enterprising politicians, eager medical groups, and aggrieved patient organizations will highlight the 4 saved out of 1000, then contrast with a child or young mother whose cancer "could have been caught," to push for mandatory widespread adoption. So even if in isolation the "rational billionaire" could use the tech, at the societal level all of the political incentives will push and push until it suddenly becomes "life saving care" that "shouldn't be restricted" and usage balloons to tens of millions because of coverage mandates. So normalizing it even for niche cases has predictable consequences for the public fisc and public health. I say this as someone who favors letting boutique health and patients do what they want, but I understand why the opponents are saying "niche = public" because that is the pattern of pressure for US health care interventions.
Cost aside, part of the problem is current medical ethics. If I search for "headache" as a symptom online, I'm told that it *could* be cancer. But if I got to a doctor they don't give me a scan or a biopsy, they don't tell me it *could* be cancer. They tell me I'm fine and prescribe an OTC pain killer.
It seems like this set of ethics isn't impossible for scanning. Tell the 300 people they are perfectly fine. (I think the risk with current ethics is that they can sue, or they can claim moral outrage that doctors don't care, etc.)
The dystopian version of this is forced yearly scans where a large corporation collects a ton of data but doesn't show patients the results, reassures them they are fine, and carefully/surreptitiously gathers more data in select individuals. Direct communication to the patient only happens in extreme cases. This is probably not outside the realm of possibility for the brand on a large healthcare company in the right regulatory/cultural environment.
I don't think "We’ll assume that the extra cost of treating the treatable cancers is zero. Why assume this? Because if we didn’t detect the cancer early, we would have detected it later, and still had to treat it, and the cost of treating small early cancers is probably no worse - and likely better" holds at all.
This may make sense for the costs of immediate treatment, both financial (e.g. surgery costs) or non-financial (e.g. post-surgical pain). But there are also long-term costs from treatment that add up over time if the patient remains alive. If someone gets erectile dysfunction from prostate surgery at age 50, this is not the same as getting it at 70s, and you should definitely subtract some QALYs for earlier tretament in that case. Similarly, if someone starts on hormonal blockers for prostate cancer a decade earlier, this will increase the cost of treatment rather than decrease it.
The other case in which the assumption doesn't hold is in the case of cancers that will be treated because of screening but would otherwise never manifest because the patient ended up dying from something else before it caused any trouble (a situation usually known as overdiagnosis). Again, prostate cancer is the classic example where a significant fraction of cancers detected by screening will end up being overdiagnosed, but this is probably non-negligible for other cancers such as thyroid, breast or kidney as well.
You'd have to look up the numbers in overdiagnosis of individual cancers to make a good estimate of how much treatment is directly attributable to screening and would not have happened in its absence. But it is certainly not zero, so "zero is an upper bound for the extra costs of treatment" is probably an incorrect assumption.
I think the analysis under-emphasises how much the benefit would vary depending on the individual (apart from financial situation).
"The average benefit among those 8 people is 4 quality-adjusted life-years"
An otherwise fatal cancer which is detected early in a 50 year old should result in an additional 30 quality-adjusted years. It would be a very different calculation for an 80 year old.
For a 50 year old, It is not unreasonable to pay a fairly high cost for even a small possibility of such a large benefit. Even if the expected value may be zero, or slightly negative, to save someone from "ruin" it may be a highly rational choice.
I don't think you can substract from both sides AND then look at their ratio to figure out the cost.
Suppose you can get $101 for if you give me $100 and one apple. This indeed looks as if I am offering $ per apple. Having ten apples you might expect to gain $10. But what if you are short on money, and don't have $1000 dolars to pair with 10 apples? In real world you might borrow them for a moment.
But now, in your analysis you use time.
You can't borrow time
This is a lot for a 'patient' to digest and I confess I have not read the entire article word-by-word. Is it reasonable to assume that in the absence of a FBS, the responsibility for discovering negative health findings is left to individual physicians, typically during routine wellness checks? (That seems to be the state-of-the-art presently.) As an apparently healthy 72-year-old patient, I do not have confidence that today's crop of physicians is particularly adept at diagnosing or effectively treating medical problems. Some are not even capable of making appropriate referrals. I speak from the personal experiences of my late wife and experiences of friend and associates. In today's medical marketplace, it next to impossible to find physicians who are interested, competent, and inquisitive to the point of effectively delving in to a patient's concerns. With the rise of PAs and NPs, it is difficult to even obtain an appointment with an MD. Most are now employed by hospital conglomerates, who, in turn, rely on insurance company payments for their financial viability. In today's medical world, it's easy to see why FBS would be attractive, and prospective customers would not be inclined to view the potential personal benefits through the lens of an analysis such as presented in this article.
Always, always, always ask: Qui bono?
I think a potential problem here might be that the incentives of doctors, due to malpractice, the insurance system and everything, is going to be to drive towards investigating absolutely everything to death. Defensive medicine. The doctors know this. They don't want to waste their time being compelled to investigate a bunch of stuff that is low-likelihood of being an issue. But they don't want to say "if you do this, I'm going to be forced to investigate every little thing". They don't want to acknowledge how much medicine is driven by risk aversion due to malpractice risk. And they also don't want scarce resources to be consumed and diverted away from people who are higher risk. So the safer route is to recommend against doing it. Because in a lot of ways, it's just throwing up more noise.
Doctors often prescribe a benign medicine just to make the patient feel better. We need something similar for slow growing ignorable cancers. What comes to mind for me is one of the prostate cancer treatments, which is to insert a small radioactive ball into the interior of the cancer to shrink it. Something similar for other cancers might be a small ball that emits something/anything for 10 years that might have a mild physical benefit but at least does no harm. The real benefit would be emotional tho, that at least the doctor did something, so the patient can relax and be happy.
Thoughtful, thank you.
New favorite phrase for many such situations:
"I worry that the set of people who think they fall into this category is bigger than the set of people who really do."
The number of women who think they are above average drivers: 55%. The number of men who think they are above average drivers: 80%. No source, stat told to me by a (female) friend because they thought it was funny. Almost certainly falls into the 88% of statistics made up on the spot. But somehow holds a bigger truth. I haven’t looked it up - finding it false would destroy the pleasure I take in it.
Slight mindbender: people have different conceptions of driver quality, such that all 55%/80% could conceivably be correct (relative to their idiosyncratic metric).
What about full body scans actually causing cancer? I guess MRIs are safe, but definitely not CT scans: https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2832778
And CT scans are better at detecting some cancers.
MRIs with contrast agent can also do bad things to you, for example if you have an allergic reaction to the contrast agent or something.
It is also important to consider the potentially substantial medicolegal costs associated with widespread whole-body screening. Diagnostic errors by radiologists, complications arising from biopsies, delays in obtaining timely follow-up appointments, and inadequate communication of a finding’s clinical significance can all expose physicians and healthcare systems to litigation. These legal risks contribute to rising malpractice insurance premiums, which are ultimately passed on to patients through higher healthcare costs.
The real trick pulled by Midjourney is to get people discussing all these random numbers and questions like “are doctors evil actually?” instead of just asking “is this Midjourney thingy a Theranos like pivot?”
Like I get it, there’s lots of low hanging fruit out there, but why tf is a random AI image generation company doing a supposedly amazing, novel, medical breakthrough? Could it just be that… they’re running out of cash due to OpenAI/Gemini eating their lunch and now they’re trying to pivot into a hot area to get a bunch of VC money? Why isn’t **that** the main question vs all these (useful!) Bayesian calculations?
Something that few people are mentioning - lawsuits. It's all well and good to say we'll just raise our threshold for intervention to avoid harm. But even if a physician correctly calculated that something is 99% chance of being inconsequential, then for every 100 calls they'll miss one finding that leads to harm which they can get sued for.
This is mostly an American problem. Other countries either make it very difficult to sue doctors, or the payout is small, or the payout comes from a government fund.
Left out of this analysis is the value of time. It's the one thing you can never get back. All kinds of time: time spent making ( & unmaking) schedules, time lost in transportation, time sitting in offices, time in follow-up appointments, time worrying & etc. I' m sure everone can add to the list.
>But sometimes harmless things have the same size, shape, density, and location as dangerous things, and even after you’ve wrung as much information as possible out of those parameters, you’re still not sure. Mr. X is 5’8, overweight, and lives in Chicago. Is he a terrorist? Would it help if I told you his height to five decimal places?
>(“But if you get multiple scans at different times, you can see how fast it’s growing!” Okay, Mr. X gained three pounds since last year, ready to answer now?)
Agreed. Which is why I find Yudkowsky's "That Alien Message" essay https://www.lesswrong.com/posts/5wMcKNAwB6X4mp9og/that-alien-message so irritating. We build specialized instruments _for good reasons_ , and additional precision on the _wrong_ measurement plus even unbounded thinking time on the wrong measurement, is not going to make up the deficit.
( One minor quibble with:
>So, converting everything to the same units, whole-body screening costs $108,000 to save one quality-adjusted life year. Usually in health economics, $100,000 per year of healthy life saved is considered the bar for a good cost-effective intervention (though other sources say $150,000). So this is somewhere around the bar, not unambiguously good or bad.
Since the dominant cost is financial, large scale screening MRIs might come out more favorably. The learning curve for going from millions of MRI screenings to hundereds of millions. Quoth Google/Gemini:
>When cumulative MRI manufacturing volume scales from10^7 to 10^9 units, learning curve theory predicts a cost reduction of approximately 14% to 27%
which isn't huge, but enough to considerably shift the decision. )
There is a second half of this equation that is also difficult for me to approximate: every hour spent chasing incedentaloma is an hour that is not spent on another patient. I don't know how I would begin to quantify this, but it has a nonzero cost in time and attention.
Interestingly, the main arguments given against whole body scanning (anxiety and unnecessary time spent) don't move the needle sufficiently to argue against it. With 32 QALY benefit, and only 7 QALY time cost, those arguments are only a relatively small factor and the financial cost is the main cost.
So decreasing sensitivity would plausibly be a bad thing. You could lose some of the 32 QALY benefit, without reducing the main cost.
I think your analogy with the person getting an hour of scans every day for 3 years misleads our intuitions to judge it less rational than it is.
First, for scans in the actual world, there is very little benefit from each scan after the first, when you're getting them daily. So we rightly judge someone doing this to be crazy.
Even if we imagine that each of the 899 hours were necessary, there is also the further problem that when you calculated the cost effectiveness, you priced in stuff like driving to and from appointments. But if you get 1 hour of scans a day, you will have to add that back in again, double counting.
Also, if you're getting 1 hour a day, that will likely interfere with your life a lot more than the raw time, since it'll probably consistently make you unable from participating in activities certain times a day.
A more accurate test would probably have to imagine that the hours were more concentrated in bigger lumps. Or even better that the cost were, say, staring at a wall 1 hour a day, at *some* point throughout the day, on average over 3 years (don't ask me how that would work). If we do that, then I don't think it's nearly as intuitively crazy.
Add to that that we're bad at probabilities so that 5% of 4 years doesn't intuitively register as 900 waking hours.
All that is to say that I don't think that particular analogy was a particularly good gauge of how rational the scan is for a rich person.
But what you say after that point might still be right
Very informative, thank you.
Aren't there a lot of non-catastrophic benefits to the scan? Seems like you could get meaningful data on things like muscle/fat composition, tendon/joint health, and many others. This would be particularly helpful for people who are trying to optimize their health through different supplements and practices.
I have nothing to add here except that this device should be called A Scanner Wholly.
Perhaps this is too far afield for your analysis, but I wonder whether the multi-cancer screening blood tests (e.g., Galleri (GRAIL)) have better QULYs and make more sense than whole-body imaging?
Yes, it's about threshold values, but what values? Threshold values are conceptual, not only clinical, if ever. Shouldn't we be more cautious about how our very own language is used? I would like to see more intense scrutiny in science about what gradients constitute something being allowed to be something, e.g., a word, a meaning and so on. In other words, an immensely thorough operationalization when it comes to our own narrative. Otherwise, we just might end up scanning threshold values for our own narratives rather than anomalies/diseases, etc.
Gosh. I’ve been a doctor for 40 years, trying to help my patients live longer and better, and you, with no experience, are so smart. It just seems obvious, just plain obvious
The shape of our problem is that there is a set of data that could improve decision making, but knowing the data by the party most affected has costs. We currently lack a technology to a) generate the data b) make decisions on it c) dont share it with the party that suffers.
In the good ol days I think doctors would hide data like this from patients. That was determined to be bad for a number of reasons.
An AI third party could be a technology for solving this problem? Midjourney would not be a third party but it could make the introduction of one easier. The third party needs to be able to consent on your behalf to follow ups without disclosing specifics to you. You need to be conditioned to not worry about what your AIs actions suggest about the data.
Actually, the only problem I care about is what extent the rest of us are paying additional costs due to unnecessary medical procedures. This seems to me that it could be a much more serious issue, with society wide implications for gov't healthcare policy and the average cost of insurance premiums. What fraction of the rise in medical costs are due to such procedures? In what ways are the rest of us paying for it? That's the conversation I would like to see.
If it was cheap enough, it would be pretty cool to use it for body composition tests.
Great piece, thank you for digging into this. I am a cancer survivor who benefited from super aggressive intervention following a surprise scan finding. Amazing, wish more people benefited in this way. On the other hand, I am now closely monitored by constant scans of different types and despite frankly being one tough as nails mfer, it is often psychologically torturous. I consider myself pretty immune to anxiety. Work and interpersonal matters do not faze me. But having an ambiguous finding (as you often do! the human body is confusing!) and waiting weeks or months to discern whether it’s anything serious is NOT for the faint of heart. It is IN your body, invisible to you and entirely beyond your control. People vastly underestimate how hard this is to face, psychologically.
All that is to say, given the collective mental health levels of our population I’m not particularly optimistic people will handle frequent scanning well.
Scanning good. Waiting for scan or biopsy results… bad.
The most important reason for having a whole body screen is to get a baseline for evaluating further changes. After I was hospitalized for a serious, life-threatening condition, I was subjected to every test imaginable over 3 days.
Most came out normal, but in some cases, the correct interpretation depended on whether I was just an outlier, or whether something had gotten worse.
If they had an earlier baseline, doctor could have seen if something had actually CHANGED or if I was just an outlier on some reading.
I notice you've been using AI estimates more frequently for your research. Have you tried using them to estimate values that you're already knowledgeable about, to see how reliable they are?
I notice you've been using AI estimates more frequently for your research. Have you tried using them to estimate values that you're already knowledgeable about, to see how reliable they are?
I'm grateful for this analysis, which pretty much confirms my preexisting views about overzealous screening.
What a lot of people don't seem to recognize is that interfacing with the medical system in any way is dangerous. Many of the other comments here illustrate this fact very well. When it's worth the risk, it's worth the risk, but when it's not, it's not, so the decision to subject yourself to it should not be made lightly. You definitely shouldn't just assume that more is better.
My recommended heuristic for most people is: don't overthink it. Go to the doctor if you have troubling symptoms, but as far as screenings go, just get the normal ones recommended for someone with your risk profile. I'm sure the official recommendations aren't calibrated perfectly, but I think it's very unlikely that the average layperson could beat them with confidence.
In addition, consider spontaneous remission rates: https://pubmed.ncbi.nlm.nih.gov/19029493/ . This study estimates a 22% higher incidence in those under active screening.
In addition, consider spontaneous remission rates: https://pubmed.ncbi.nlm.nih.gov/19029493/ . This study estimates a 22% higher incidence in those under active screening.
The most interesting part of this is doctors asking people not to bother, waste unnecessary money or time and the people behaving they know better!!
As an Indian I can say with absolute certainty NO DOCTOR in India will tell people not to overtest and spend money on shit they dont need. This technology is god send for Indian docs.
Rethinking about this this morning, in a perfect market, which is what your QALY sums are sort of assuming, people will choose to do a thing when it gives them more extra life than it takes them to do (whether that is physically doing the thing, or working to get the money to pay for the thing.)
But the reductio ad absurdum of that is that people ought to be happy spending an entire year working to get an extra year and a second of life, but not to get an entire year less a second of life.
Imagine a hypothetical where the cost of buying extra life was sitting in a doctors waiting room, (and all the actual health saving happened by the pure magic of being in the room, not any more inconvenient tests or scans or treatments.) This is a just-better-than-marginal intervention, sitting in the room gains you a as-much-life-plus-a-minute as you spent sitting in the room.
Which would mean over the years of life you end up with, half of them (or at least half of the awake hours, or the productive hours) would be spent sitting in a doctors waiting room.
If you frame it as ‘do you want to live for a year doing whatever you want, or live for two years but for 8 hours a day you have to sit in this one spot thinking about your health’ it feels a much starker thought experiment to me.
We already have various DNA tests with exactly the same market niche and “more data will make it so much better” mantra. I still don’t know a single person who actually get anything useful from unprompted genetic test. You are more likely to find criminal among your relatives than unnoticed disease. All medically significant results is some form of “3 times higher risk of X, follow healthy diet and exercise” and only useful for such a thin slice of people I doubt there is more than 100 of them in the world
One might factor in that different hours may be worth more or less than others. I might yield an hour every day to gain an additional calendar year.
Only gaining 4 years for early detection seems incredibly pessimistic?
If one has a small, early stage cancer, should one not tend to gain the entire rest of one's life back (decades?)
I just hope by the time there's an actual functioning procedure they still only need water. Imagine the cleanup required after hanging out naked in a vat of ultrasound gel for a while!