Hacker News Reader

Comment by picture | Hacker News Reader

Routinely scanning people without symptom or otherwise cause for concern is not a good idea. The human body is very messy and variable, and it's hard for anyone or anything to reliably spot thing that are actually worth acting on. If the doctor spots a vague blob - how can they know what it is without any symptoms? The course of action is then "come back in three months for another scan."

Subjecting a population to unnecessary routine full body scans, especially with radiation, will cause harm statistically. There are also countless examples of people undergoing unnecessary operation and suffering complications, like losing a perfectly fine healthy heart and requiring transplant.

View on HN

We need yearly scans or even monthly scans (MRI) for everyone. Privacy concerns aside, I think we would learn an awesome lot. We could then monitor changes over time and machine-learning could probably get really good at recognizing what changes are harmless and which are harmful and need a doctor to look into.

It's a dream scenario, but maybe some day it will be like that. Or it will be like that and be dystopia. :)

MRIs would need to take a fraction of the time for this to be viable.

Or, y'know, we could have 10x more MRI machines, thus incentivizing them to be cost-optimized in ways they absolutely aren't today.

That's not really going to help. For an MRI, you have to lay there being still, and for some you have to hold your breath, all while in a loud little tube. I have to get a couple every year, and I*m lucky not to find them too uncomfortable. Some folks, though, would have serious issues with the cage over your head when you get your brain scanned: Your brain and spinal column will take 45 minutes, as will other areas. You will need to be repositioned often, adding and removing things to get the pictures of all the areas.

In other words, you'll need an entire day off to do this if it is even possible for you to lay there for so long in one event, even if you get breaks. And for some folks it is going to be hours of torture.

More MRI machines won't solve the issue.

Not necessarily, we just need cheaper ones. I interviewed at Hyperfine last year and they seemed to have a promising product

https://hyperfine.io/

No they don't. I would happily take a few hours each year to get an MRI done.

We would need more machines, but even now the prices aren't too exorbitant for something done annually (or- even every few years)

20% of americans havent even seen a doctor in the last year. MRI prices for a fun test are far beyond a huge percentage of the country. https://civicscience.com/20-u-s-adults-not-seen-doctor-past-...

Yeah I think my last MRI was $800, with insurance. And they didn't find anything. I'm not exactly champing at the bit to spend the money to do that every year.

If it were $50-100, maybe then.

Everything's relative. I've had about 8 MRIs in the last few years.

I also wouldn't bat an eye at spending $1k on preventative care each year. Many Americans are paying that annually for premium TV shows.

I've had like 4 MRI's total. The most recent one was I had a scare about a possible aneurism, which is what they didn't find anything for, and then had a couple for looking at my neck and back, where they found two bulging disks, one in the middle of my back and one in my lower back. And they had to check my leg for something once, maybe blood clots? I haven't needed to get one for several years now.

MRI's don't seem that great for an annual preventative check, since they are usually checking a pretty small area (or at least mine were). Like those were 4 different MRI's to check me from head to one of my calves. Getting 4+ of those a year would add up quick. And that's not even taking into account any of a myriad of other tests that could be done every year. I'm sure I could probably take a different test every week of the year, but I'd be bankrupt quick.

Why would you jump to an absurd scenario like one every week?

It would be an enormous benefit to do one MRI a year, rotating between areas of interest, at a relatively negligible cost ($500-1k or so per year - about what it costs for an annual physical).

It's extremely cheap compared to your Dr's hourly billable rate and would get cheaper at scale.

I don't believe it would be an enormous benefit at all. What are you basing that claim on?

Well, I wouldn't call them "fun tests" but otherwise that's true.

However, there is also a chance to save a lot of money by recognizing illnesses early. Sometimes it's possible to prevent a life-long need to take expensive medicamentation or it allows to detect cancer and allow for easy surgery in an early stage allowing for complete cure instead of requiring expensive chemotherapy AND surgery for a long time with a worse outcome. It might also very well prevent people from being unable to work due to illnesses if they are detected early enough.

And that all isn't even considering the changes in quality of life of course.

"20% of americans havent even seen a doctor in the last year"

And 80% have. The needs of the 80% are not determined by the behavior of the bottom 20%

Rich people get better everything, including better medical care. The upper 30% or so can afford better diagnostics and they should be getting those better diagnostics.

Many of us in the software industry could afford a few extra thousand a year for preventative care -- but it is very difficult to obtain care commensurate with one's ability to pay. This is a real problem.

Thats great, but this was the proposal I was responding to: "We need yearly scans or even monthly scans (MRI) for everyone"

As a dream scenario, yes. It's a reasonable goal.

MRIs vary widely in cost and that cost will continue to fall. We'll get monthly scans within our lifetimes.

> especially with radiation, will cause harm statistically.

Ultrasound is not radiation (as in photons). To the best of our understanding (from a physics/engg perspective) ultrasound imaging capped by appropriate energy/etc limits is expected to be perfectly safe. Likewise with MRI. If we want more pointed study to be sure of this, that's fair, but let's be clear & specific about it and commit to figuring it out one way or another so we don't repeat the same discussion a couple of decades down the line. (We have to think of moving the state of the art forward, instead of festering in unresolved disagreements)

Further, if you're concerned about under-studied possible side-effects of radiation from occasional diagnostic testing, what do you plan to do about being blanketed by mm waves once 5G gets deployed more ubiquitously?

> There are also countless examples of people undergoing unnecessary operation and suffering complications

This sounds far more serious (and fixable) compared to the physics/biology interaction of diagnostic testing. Why do we continue to bury our heads in the sand collectively, instead of trying to fix this with better decision-making tools?

> Likewise with MRI

MRI isn't radiation, but is sometimes used with contrast dye. People can be allergic to the dye. For one person this is a small risk. Across a population we'd be causing harm. We'd balance the risks of harm against the benefits, and so far no-one can find a benefit to routine whole body MRI scans. And if the benefit was there the MRI machine companies probably would have found it by now because it'd massively increase the numbers of machines they could sell.

MRI uses electromagnetic radiation. MRI uses a magnetic field with a radio-frequency pulse and then the protons emit a radio-frequency response - that response location is used to synthetically generate the image. MRI does not use ionising radiation. Nice overview here: https://www.nibib.nih.gov/science-education/science-topics/m...

You've misread their reply. MRI's sometimes use contrast dye. I've personally had these sorts of dyes, both on brain and spinal cord scans (I have MS and get these regularly) and for the first scan of my pancreas (non-symptomatic cyst they are watching to be cautious).

It isn't radiation.

https://www.insideradiology.com.au/gadolinium-contrast-mediu...

Wouldn’t the skill of analyzing scans improve if they were done more frequently?

It seems like you’re promoting a self fulfilling prophecy. Doctors aren’t good at reading scans, so scans will cause false positives.

But doctors are capable of learning and improving their skills. They will surely rapidly learn to screen a lot of false positives, thereby saving lives.

You seem to assume that radiologists currently see patient scans too rarely to deliver accurate diagnosis. That seems very unfounded to me.

"More frequent scans per patient" != "More scans seen per clinician". You would just end up having to hire more clinicians.

(Yes, your family doctor might end up seeing more scans in this scheme, but it is implausible that they would be better at evaluating them than a trained radiologist, which is who this would almost surely be deferred to anyway.)

Is the argument then that scans are intrinsically error prone, as in the noise level is so high that even a team of highly trained radiologists can’t accurately determine if cancer is present from a scan?

It's not the quality of the scan (although that can certainly be a factor). It's that there is a lot of variation in human anatomy, and "these pixels are brighter than the rest" can mean any of many things.

Relatedly, for many types of brain tumors, you simply don't know what exact kind of tumor it is until you actually do a biopsy (= you undergo surgery). This can be important information ("how aggressive is it?"), so in some hospitals it is common practice to send a tissue sample for identification and get back the result while the surgery is still going on.

This is a very naive argument, it assumes the entire problem is lack of skill and that early intervention on positive diagnosis is always a good thing.

Mammography is a really good candidate for random screening and it's still debated because of false positives and the impacts of misdiagnosis.

Yes, the breast cancer screening already has quite questionable results.

the screening guidelines take into account those risks and it's still deemed worthwhile because cancer can grow and become lethal despite being asymptomatic. colon cancer is the mid-life killer...plenty of ppl in their 50s or even 40s get it despite no risk factors. Screening is the only option to catch it early.

This is just silly. You are never disadvantaged by knowing more about what's inside your body. What you don't know can easily kill you, though.

Hopefully in the long term future we will have extremely cheap & frequent MRI and blood testing so you can assemble a high resolution digital history of your body. Doctors are generally not very innovative people though and they will probably resist it to the end.

It's not silly. Look at how the mammogram guidelines are set. If the scans are too frequent, the harm done from false positives (unnecessary surgeries and risks from those surgeries) can become too great. The standards were relaxed to include fewer scans in the last few years because of this.

Apply this to more invasive exploratory surgeries, and you see people dying from internal surgery complications for benign cysts and lumps.

That's because single datapoints are used to trigger intervention. We should be intervening based on growth trends from time-series data. Time-series data which can only be collected by... frequent scans.

Don't discount psychological stress. "You have something growing in your breast which might be cancer, just wait a few months and we'll scan again" is not something that gives people a good nights sleep.

So don't tell them. Like I said, a single datapoint by itself is meaningless — so you wouldn't bring it to the patient's attention. Heck, the software shouldn't flag it for the doctor's attention. (Think: metrics-based alerting thresholds in software systems.)

We do have time series data for mammograms. You may think it doesn't meet whatever standard you are setting, but it is objectively false to say it isn't there.

My standard is "doing it for everything, rather than for one thing"; and more importantly, not acting upon data about chronic disease progression, if it isn't a time-series.

In science, you don't call one datapoint "data." You call it an anecdote. Outside of emergency medicine/urgent care, medical practice should be the same.

I honestly can't imagine how you believe breast cancer is diagnosed and treated, but it's clearly inaccurate. So is the cute phrase about "in science".

You didn't understand what I said. Breast cancer screening is the "one thing" I referred to that we're doing right. But who cares if we're doing one thing right? We're doing screening for literally every other cancer wrong. And not getting any less wrong over time.

We — the medical profession in general — are aware of the proven effectiveness of time-series analysis of regular mammograms as a screening technique. However, we have failed utterly to spread this "obvious" knowledge out from the specialty that understands it and has proven its effectiveness, to literally anywhere else. We've been doing regular mammograms for decades now, but that hasn't sparked (wide) adoption of regular scans + time-series-based screening for any other cancer.

There is a certain moral culpability in that. A certain negligence of duty / not seeing the forest for the trees in what it means to be saving lives. (I hypothesize that doctors in at least some specialties, could save more lives over the course of their career by dropping their practice and instead turning into advocates for wide adoption of periodic-scans + time-series-analysis-based screens. Somewhat like would have been true of a doctor in the 1800s quitting doctoring to instead become an advocate for hand-washing before surgery.)

You're right, I had absolutely no idea that you thought we were doing breast cancer screening right. And I don't know if you're now failing to communicate your awareness of regular screening for prostate and cervical cancer, or skin cancer in higher risk locations, or if that doesn't count somehow, or what. I also don't know if you have any evidence for the existence of these other tests that we could be doing on everyone and the risk-benefit profile that shows we should be doing them, or if you're assuming such tests and evidence exists, or complaining that we haven't invented the tests that must be theoretically possible.

It can be hard to predict how aggressive a tumor will be. It isn't uncommon for people to obtain cancers that they "die with" instead of "die of". In those cases, treatment can be worse for the patient than the cancer. But since you don't know which it'll be, people get the treatment and sometimes wind up worse off than they otherwise would have.

And as the GP post cites, the scans themselves can cause health problems down the road. Performing them on massive scale will increase the occurrences of those side effects.

Full body scans, as they currently exist, are not recommended for routine-usage. These scans use large amounts of radiation. This can increase your risk of cancer. If you have more tests, your risk increases.

Further, they are costly, and often times insurance does not reimburse or cover it. They can cost anywhere from $500-1000.

Unless you find a way to make this viable for the general population and lower the radiation exposure, this will never become viable for use.

Knowing that there is something weird in your body sets off a cascade of medical care that easily escalates to surgery, anesthesia, medication to manage complications, side effects of that medication, time in the hospital being at risk for infection, etc. These things absolutely can kill you, or at least put you through hell. When there was actually nothing wrong with you in the first place. That is why the medical establishment is so conservative about it.

Isn't the danger of knowing there's something weird there caused by the rarity of knowing anything at all?

I'd imagine that if there was consistent, good information available, we'd figure out how and when to act on it. The real issue is the massive cost of performing any given scanning procedure to everyone alive.

No. Doctors already think about this. Why do you think they aren't doing prostate scans and mammograms constantly on everyone?

I think modern medical ethics demands a physical treatment for literally everything and anything that's discovered, including quite a few that either aren't physically dangerous (e.g. atypical brains) or are inevitable (a 90+ year old with a laundry list of conditions). And, as other comments have pointed out, a panicked surgical response to a mammogram isn't based on time series data.

I totally understand and agree with trusting that medical professionals "know better", but I also would not conflate current practices with absolute perfection. It's clearly not a perfect profession or practice.

You don't have to conflate current practice with perfection. You just have to bother learning what the current practice is and why - and "it's just so rare to know anything" is not the correct answer. Just to start with basics: yes, actually mammograms are analyzed using time series data of a patients previous scans!

The question is not even what doctors think warrants a follow up. The question is what a malpractice attorney could construe as warranting a follow up, if it turns out later to have been real.

Once a scan exists, doctors are essentially forced to act on it. If an algorithm for deciding which scan results to pay attention to were so solid that it would hold up in court, that might make more scanning reasonable on the margin? I dunno though.

> You are never disadvantaged by knowing more about what's inside your body.

Yes, you are. This has a name - over-testing, over-diagnosis, and over-treatment. It's a significant cause of harm in healthcare.

https://www.bmj.com/content/358/bmj.j4070

https://qualitysafety.bmj.com/content/31/1/54

https://www.choosingwisely.org/patient-resources/whole-body-...

Reply on news.ycombinator.com