I strongly believe that more data is always better. (But more testing is not always better if it's an invasive biopsy.)
If more data results in more unneeded biopsies, then just stop taking biopsies! It seems very obvious that the answer is to perform a series of MRI scans over a period of weeks or months, and track the size of the lump. If it's growing rapidly (or more lumps appear), then you would perform a biopsy. I don't even think you need machine learning for this. It just seems obvious that a quickly growing lump is cancer (almost by definition), so you need multiple scans over time to detect the rate of growth.
Is this a controversial opinion? Why is better to wear a metaphorical blindfold instead of just being smarter about what you do with the data?
Is this because MRIs are expensive and are in high demand, so it's just not practical to do this?
EDIT: I updated my comment to remove "more testing is always better".
> I strongly believe that more data is always better.
At a society level, probably true long term, but in the short term more information results in a 'shit show epoch'. To clarify what I mean there: anytime both in history and in the current world when more information quickly becomes available to the average person there is a 'shit show epoch' which lasts until society adapts. I think the current web / social media explosion is an example of this. Lot more information available, little implanted knowledge on how to navigate it, we're in the midst of a 'shit show epoch'.
To wrap back around to the medical world, cheap MRIs are good long term, but they are going to drive a 'shit show epoch'. The majority of adults have benign growths in their body[1], the ease of access to that information is going to cause a shit show.
Truthfully more testing isn’t always better because the negative consequences of false positives often outweighs the value of tests. It’s the reason you don’t get prostate exams until you’re middle aged. Humans are weird and lumpy and all different. Testing can be bad for you until you until the math works out, and (risk of disease * consequences of missing it) exceeds (risk of false positive * consequences of exploring/treating). It’s just a numbers game.
That makes sense. What do you think about taking regular MRI scans and tracking a growth over time? If biopsies were reserved for fast-growing lumps, then there shouldn't be any harm in collecting a sequence of images. I know MRIs are expensive and take a long time, but I'd gladly jump into an MRI machine for a few hours once a year. And then maybe once every 6 months when I'm in my 60s.
I know that humans are weird and different, and all of us have weird lumps and spots that show up on a scan. My proposal is to just track each spot and keep an eye on them over a period of months or years. We already do this with mole mapping, and MRI scans allow us to do "mole mapping" for our entire body.
From an interview about mole mapping [1]:
> Dr. Curtis: So, generally, what I tell patients is if they have a lot of moles, they should get mole mapped, or if they even have a lot of funny looking moles to them. They might not have a ton, but at the same time, they still might have enough that they are curious about them and they're hard to track, if they're hard to know if they're new or changing.
> Interviewer: Let's talk about the benefits of having this procedure done because there are a lot of them. First and foremost, it can save lives.
> Dr. Curtis: Yes, it can. It can save lives in the way that we can find melanoma earlier and thinner when it is the absolute most survivable.
> Interviewer: And what are some of the other benefits as well?
> Dr. Curtis: So you can catch new moles that could be just benign, not changing, that are not melanoma, which are important to know about. You can also tell if your moles are going away. As you get older some of your moles do go away and it can track that. So it gives you peace of mind. And it can reduce the number of biopsies as well on patients.
While your plan could conceptually make sense, in the US, as soon as anything out of "ordinary" shows up on a scan you have to do everything you can possibly do or be sued out of existence for malpractice. If you follow the recommendations not to test, then you're doing what's expected of you. As soon as you test and see anything then fail to act, you're at huge financial risk.
Time is a big enemy. Your aggressive cancer scenario is ludicrous in a part of the body so near the lymphatic system. Everything is done stat once you have positive scan. Ive known people who have positive imaging one week and full mastectomy the next week and chemo shortly after. Not trying to be rude, but I think you have no clue what you’re talking about and how more biopsies is always going to be better than more cases breast cancer spreading which is what will happen under a wait and see approach
I'm a bit confused by your response. Are you arguing for or against regular MRI scans? I think your example clearly shows why regular scans are very important.
My proposal is that a healthy person will go to get regular MRI scans, and you would compare any new lumps against a baseline scan. For your very first scan, you statistically don't have anything to worry about, since everyone has a few benign lumps or spots. Obviously if they find a massive tumour on your kidney or a lump near your lymphatic system, then they're going to take it very seriously.
However, if something is just "suspicious", then perhaps you would come back in a few weeks or months for another scan, instead of one year. Machine learning could inform the period of time between scans. (Or just a trained radiologist who knows how to determine the risk.)
I'm not a trained doctor or radiologist, so in that sense, I don't have any clue what I'm talking about. I just enjoy discussing ideas and trying to figure out why my ideas wouldn't work, so unfortunately I haven't found your comment to be very helpful. I think we might even be arguing about different things.
There is no baseline. You go to the doctor because you’re concerned about a lump, probably because you had never felt it before. A scan needs to be done. It doesn’t always need to be MRI. But in my non-clinical knowledge of BC survivors, MRI are a must for young/dense breast tissue. It should be required.
I’ll give you specifics on a patient I know (I actually went to all doctors appointments).
32F. Feels something but hard to find next day, I felt it, she couldn’t. I urged her to see doctor. (I actually worked for a pathologist as college job, I know what pretty much every type of tissue looks/feels/smells like... and I knew it was cancer, but maybe benign). Anyways, she goes to doctor. He can’t feel it in exam. But, she has family history so he finds a way to get the MRI done. He explains why it should be standard on young women (denser tissue). Next day, MRI. Confirms lump, doctor is blown away as it was only ~0.5mm in diameter. He’s about 70 and it’s the smallest lump he’s ever identified, surprised it even showed up on MRI. Long story short, she had a full mastectomy ~5 days later and the pathology showed the lump was over 3cm in diameter, super aggressive (triple negative). Started chemo shortly after because the milk duct had opened and cells could have reached her lymph nodes and spread the cancer anywhere in her body.
I know many other people that have similar stories, so this is why you get thrown into full on crisis mode when you get a positive imaging. So I would ask, if she waited a month or more to come back and see if this thing was growing, how do you think that would have played out?
Wow, I'm really glad they caught it in time! That's a very scary situation.
Obviously I wouldn't change anything about the way this was handled. To be honest, my proposal is mostly a concession. It is an attempt to reduce the number of unnecessary tests and biopsies, while still gathering more data about the things that are going on in your body. It doesn't supersede anything that you've described, so if you feel a lump, you would still follow exactly the same course of action that you've outlined.
There must be a way to collect more data and respond appropriately to aggressive and dangerous cancers, while still reducing the number of unnecessary tests and biopsies. This is obviously a very important problem to solve and will save millions of lives.
Interesting debate btw. I think the correlation between high levels of testing as relates to early identification is a big contributor to trend of increasing survival rates (along with advancing treatment options of course).
Also important to note. MRIs are noninvasive. Biopsies, in BC cases, are fairly noninvasive. It’s typically just pulling out some cells with a needle, not to say it’s comfortable, but nothing to really fear either. So, I guess I just don’t see what the big deal is. Except it’s expensive. So, as is always the case with healthcare, the real debate should probably be about cost.
Last time I asked this question, the reasoning was that there are non-rational (i.e. emotional, legal) reasons why an apparent lump on an MRI has to get addressed, so the only acceptable point to say "don't operate" lies at the point where the doctor decides whether or not to test. Apparently, doctors get exactly one chance to say no.
That makes a lot of sense. It also feels like a very solvable problem.
If I had the option to get annual MRI scans, but I had to sign a waiver that prevents me from suing the doctor or the hospital, then I would gladly take that deal. I'd rather have the data and an expert opinion, instead of no data and no opinion. If they miss something and I die, then no problem (mainly because I'm dead, but also because I would have died anyway). If they do a biopsy and it turns out that it was nothing, then that's also no problem. It's what I signed up for, and I understand that false positives can happen. Hopefully I could find a level-headed doctor (or a black-box machine learning algorithm), and just trust them to make the right call.
The connection between information and stress, and the connection between stress and a wide range of physiological effects effectively proves that more data is NOT always better. It's clear that there are cases where being given more data will reduce the general well-being of the patient.
MRIs are expensive and in high demand so it's not practical to do that for the majority of patients. We have limited dollars available to spend on the healthcare system as a whole and so buying more MRI machines and training more technicians and radiologists is probably not the best use of resources from a cost effectiveness standpoint.
Multiple repeated imaging procedures are also disruptive for patients with busy lives or healthcare access issues. At lot of patients just don't show up.
This comment could only be made by a person sitting in front of their computer with no experience in healthcare, no understanding of how screening works, and no knowledge of breast cancer biology. Sorry to be blunt.
Yes, I'm always surprised to see so much negativity from people in the healthcare industry whenever this topic comes up. Maybe I'm completely wrong, or maybe the status quo needs to change.
Which part do you mainly disagree with? Can you not see a future where regular scans and machine learning might facilitate the early detection of cancer for millions of people?
Wow, that was an amazing article! Thanks very much!
The mammography and PSA sections were very eye-opening. I did know that most men die with prostate cancer, but I didn't know that treatment often leads to impotence and incontinence. That's very scary.
> He was diagnosed with low-grade localized cancer, the kind that can be observed without treating. But he couldn’t face living with the knowledge that he was harboring an untreated cancer. He was afraid of surgery and opted for radiation treatment. He developed radiation proctitis and had rectal pain and bleeding for years.
I can understand the anxiety of living with an untreated cancer, and I'm starting to see why this might be a bad idea for the general population as a whole. "A person is smart. People are dumb, panicky, dangerous animals."
It doesn't change anything for me personally, and I would prefer to have more tests and more information. If the doctor says that I don't need to worry about it, then I'm not going to insist on surgery or radiation treatment. It's hard to know what it would feel like in the moment, but I don't think I would struggle with too much anxiety. I have already accepted that I probably have a lot of benign lumps and cancers in my body, because I know this is very normal. On the other hand, you also have to be careful about the other extreme, where you delay surgery or try to treat it with natural remedies.
> If a prostate cancer is localized and low grade, it is reasonable to observe the patient and not treat unless he develops signs of progression.
This is what I have been suggesting, but on a larger scale. People will just need to control their anxiety and realize that these cancers are normal, and they usually don't cause any harm. This is a tough problem, but we need to start educating people better.
They also talked about whole body CT scans, which is similar to what I am suggesting:
> Whole body CT scans were popular for a while; they were offered in free-standing CT facilities on request, without a doctor’s orders. They exposed patients to the same amount of radiation as 923 chest x-rays. They found abnormalities in 37-86% of patients, creating anxiety and leading to unnecessary follow-up tests that were sometimes invasive and potentially life-threatening. Fortunately they seem to have gone out of style.
Regular MRIs would be much better than CT scans. I think it would also be much better to have a series of scans, where you can look at the differences over time (and even use machine learning for this), instead of just having a single snapshot. I think this is similar to the blood test for PSA, where you look at the rate of rise over time. If these scans are leading to unnecessary follow-up tests, then the problem isn't the scans - the problem is the unnecessary tests. If this became a common practice, then it would be much easier to figure out the correct frequency of tests. It seems very foolish to throw away all the data instead of solving this problem.
I find it interesting that the USPSTF recommends against regular EKGs. I've heard lots of stories about the Apple Watch detecting AFib [1] [2]. I also read some articles that talk about the risks [3] [4]. My conclusion is that I'm going to buy an Apple Watch 4.
The "What makes a good screening test?" section was very helpful. Instead of using this list as a filter, I like to see this as a list of problems to solve:
1. Disease has serious consequences
Yes, 609,640 people die from cancer every year.
2. Screening population has high prevalence of the condition
Yes, and it would make sense to screen for the most common types of cancer. Melanoma screening can already be done with a phone camera and machine learning [5].
3. Not too many false positive or false negative results
I believe the results can be vastly improved with more data and better machine learning algorithms.
4. Test detects disease before critical point
Yes.
5. Test is safe – causes little morbidity
Yes, this would just be an MRI scan.
6. Test is affordable and available
This is the main problem that needs to be solved. This might be more feasible in a country with universal healthcare, so that the government can fund this program.
7. Treatment exists and is not too risky or toxic
No, surgery and radiation treatment is often very risky and toxic. I think this is most difficult part to get right. The general population will need to have a mental shift, where we understand that it can be ok to live with an untreated cancer. Especially for prostate cancer, since it's extremely common. Another solution might be to keep people ignorant and just send the data through a black-box machine learning algorithm.
8. Treatment is more effective when started earlier
I'm not quite sure how to respond to this because it just seems like a ridiculous statement to me. Of course I tell patients things in a way to reduce their anxiety. It often doesn't help. Your last sentence is just plain wrong. Checkups don't keep you healthy. Good diet and exercise does.
If you tell someone "you have cancer but it's growth rate isn't fast enough yet for it to be worth treating" that will have a very different impact than "you're fine but we need to keep an eye on you to make sure you're in optimal health". But really they describe the same state of the world.
The alternative being proposed is to lump the people who are fine in with the people who are going to die unless they get treatment urgently and effectively forget or destroy knowledge about their state so that the amount of anxiety avoided in the first group exceeds the amount of horrible death and pain in the second.
It seems pretty clear that we have a ... problem ..(?) somewhere if we're even considering this.
I'm going to guess that the problem is that we can't deal with hazardous information very well. The knowledge that a person has a specific not-worth-treating cancer somewhere is hazardous to them and they need to be insulted from it. But it is not hazardous to the medical system, rather it is useful since with further monitoring it can turn into something actionable (a successful cancer treatment instead of patient death). But people are expected to be managers of their health (which makes sense), so that means they kind of have to know this hazardous information.
Be academic all you like. However, I live in the real world, with real patients. I live in a world where I can tell a patient that a mole is completely benign and it still makes them anxious and they want me to cut it out.
Some people are anxious types. Few people can comprehend medical statistics. Even an intelligent person such as yourself doesn't seem to comprehend the danger that can come from a "check up".
After you've diagnosed thousands of patients and discovered hundreds of incidentalomas, perhaps then you can tell me how simple this is.
> However, I live in the real world, with real patients. I live in a world where I can tell a patient that a mole is completely benign and it still makes them anxious and they want me to cut it out.
What I am saying is that things need to change, not that this is a trivially fixable problem. Exactly how you would change things is not obvious to me, but I would say that not doing tests because of the information hazard issue is a symptom that something is wrong.
Perhaps the more general version of this problem is "don't set up systems that give billions of people nonconstructive negative feedback". It's applicable beyond medicine, I would point at unrealistic celebrity beauty ideals as another manifestation.
A related issue is thyroid cancer. Diagnosis rates started increasing dramatically almost everywhere in the world in the mid 80's. Diagnosis rates in the US for example have tripled since 1980, in South Korea they have gone up 13x since 1993. Meanwhile the mortality rate has been completely flat. The most likely cause of this is increase in both screening and the sensitivity of testing [1]. There is also a current debate about if we should be screening / treating thyroid cancer the way we are, i.e. it seems like we're detecting a lot of innocuous thyroid cancers and then operating without need resulting in unnecessary life long complications for the person.
From the article it sounds like most of the harm of false positives from the MRI is the extra stress of the biopsy and/or waiting for biopsy results. As such it feels rational to try to treat the stress/anxiety with well know methods like talk therapy. Prescheduling an appointment for after the MRI/Biopsy might greatly reduce the stress of patients. Of course this runs into the cost "problem".
The rest is pure speculation based on how it appears these procedures are portrayed.
Patients think of the MRI as yes/no for cancer and when comes back as yes they get understandably upset. If instead the MRI biopsy are presented as a single procedure (ideally done together, but money) the MRI is just getting targeting information needed for the biopsy not providing a yes/no decision. If it turns out there is nothing to biopsy the patient gets a happy surprise no-one is ever upset about this. (Again problem of money)
It's the same results for starting mammograms at 40 vs 50. They're now recommending to start at 50, unless you have a family history, because they found a lot of false positives in women who started getting mammograms at 40 who didn't have high risk.
1) From the POV of the patient, how bad / invasive is getting a biopsy? Wouldn't you always want the low-risk biopsy over having discovered a problem too late? From a public health / insurance perspective, I get that you're trying to lower costs and time wasted, which are sadly conflicting goals :-(
2) It's always been so shocking to me that the reaction to "we get false positives" is not "we need to pour money for research into much better diagnostic techniques yesterday" and instead it's "well I guess we'll just not look then".
These are interesting questions. If biopsy was totally benign you'd be right. But there are complications and rarely serious complications of breast biopsy.
"How bad / invasive is getting a biopsy?" perhaps you are willing to accept those risks, but you have to also accept the risk that you are getting treated for cancer you don't have. Pathology isn't perfect [0].
Should we invest more in getting better results? Yes, and we are. But, we also have to come up with a risk/benefits strategy for the technology we have now for the patients we have now.
Everything in medicine is risk/benefit because our data and our interpretation of that data isn't perfect.
Considering that this was funded by the Patient-Centered Outcomes Research Institute[1], which was started as part of the Affordable Care Act, my guess is that this was indeed done to help reduce overall health care costs from unneeded procedures.
About the same as getting a prostate biopsy, except you can lie on your back instead of on your side while they shove several large hollow needles into your flesh to cut out pieces of a very sensitive gland.
I don't think comparing it to prostate biopsy helps most people. I assume they anesthetize you so you don't actually feel pain just really weird, possibly upsetting tugging right? Afterwards you take some regular pain killers for a few days. This doesn't feel like a big deal to me but I'm likely wrong. Maybe it is only a big deal for some people like how how people who are claustrophobic freak out in the MRI.
Depends on where, I have had kidney biopsys its painless but does take a team of 3 and you spend most of a day flat on your back (taking up a hospital bed) being monitored every 15 mins
> It's always been so shocking to me that the reaction to "we get false positives" is not "we need to pour money for research into much better diagnostic techniques yesterday" and instead it's "well I guess we'll just not look then".
I think it is both. I think we tell people not to get bad diagnostics while also putting a lot of research into making the diagnostics better so we can tell them to get the diagnostics again.
> Breast MRI had a cancer detection rate (CDR) of 10.8 per 1,000 exams compared to mammography’s CDR of 8.2 per 1,000 examinations. However, breast MRI had a biopsy rate of 10.1%, double that of mammography’s 4% rate. [emphasis added]
What's going on with the biopsy rate? Is it something as simple/nefarious as recouping the cost of the equipment?
I have not read the article but I believe that usually routine exams are done with X Ray, because it is quicker and cheaper. However if something is a bit worrying, doctor may ask for a different type of exam, MRI or Ultrasound, to have some confirmation.
But since this is rarer people who have MRI are already people at risk, so it makes sense to have a higher rate of biopsy in the end.
It’s not a broken system. It’s a limitation of the current state of human knowledge. Perfect information about X without the state of knowledge to interpret that information prompts flawed action.
You are not living at the end of time, at the peak of human knowledge. There are things we know enough about to interpret; there are things we don’t. The former we test; the latter prompts scientific research.
If more data results in more unneeded biopsies, then just stop taking biopsies! It seems very obvious that the answer is to perform a series of MRI scans over a period of weeks or months, and track the size of the lump. If it's growing rapidly (or more lumps appear), then you would perform a biopsy. I don't even think you need machine learning for this. It just seems obvious that a quickly growing lump is cancer (almost by definition), so you need multiple scans over time to detect the rate of growth.
Is this a controversial opinion? Why is better to wear a metaphorical blindfold instead of just being smarter about what you do with the data?
Is this because MRIs are expensive and are in high demand, so it's just not practical to do this?
EDIT: I updated my comment to remove "more testing is always better".