Hacker News Reader

Comment by swartkrans | Hacker News Reader

The heart attack detector seems great, but I wonder about the cancer detector. It may lead to a high number of false positives resulting in unnecessary exposures to CT scans, which because it uses radiation can actually lead to cancer, and unnecessary medical procedures and expenses, and unnecessary anxiety. If they get the false positives down it would be very beneficial of course.

View on HN

How many CT scans must someone get before its more detrimental that having the cancer go undetected?

EDIT:

http://www.cancer.gov/cancertopics/factsheet/detection/CT

"In addition, whole-body CT can expose people to relatively large amounts of ionizing radiation—about 12 mSv, or four times the estimated average annual dose received from natural sources of radiation. Most doctors recommend against whole-body CT for people without any signs or symptoms of disease."

I have read that there is no such thing as a harmless amount radiation when it comes to the dosage you get from a CT scan. Every scan increases your risk. I think my impression is that many doctors think there must already be a health problem in the patient that needs to be addressed before someone is subjected to a CT scan. That is use a CT scan when you know it will help, not when you don't know.

> Every scan increases your risk

That is technically true, but it's also true that every banana you eat adds to your risk. At some small percentage increase in risk, it's sensible to stop caring.

I don't know the answer to that, but I know that some people are taking full-body CT scans annually or bi-annually because of fear of cancer.

This is a great way to help increase the risk you get cancer, since it's cumulative, not per-instance.

>This is a great way to help increase the risk you get cancer, since it's cumulative, not per-instance.

Well, I suppose at least in their case it has a high probability of being detected within 6-12 months.

I know, but that proves that cancer scares people to the point of doing silly things. An innocuous screening test, even with a fair amount of false positives could be a huge win.

Are you sure it's CT scans? Why not MRIs, which as far as we know cause much less harm to the body? (They're more expensive, but I'd think people who are that worried about cancer would consider the extra expense worth it.)

CT vs MRI comparison: http://blog.radiology.ucsf.edu/neuroradiology/exploring-the-...

It's brain specific, but I assume it translates to whole body as well.

While a CT scan may be the best known method to confirm later stage cancers now, a better and safer way may yet be discovered that can track the cancer's growth long before a doctor would recommend a scan. Nanoparticles could be a promising way to do that. If we're constantly made aware of many possible risks, the anxiety will (probably) eventually diminish as long as those risks are safely monitored. Overexposure to risks (that were always there) might be just what we need to explore better ways to track medical conditions before it's too late for treatment.

Seems like a straightforward optimization problem. Only alert someone if the expected probability of disease is high enough to be worth the risks of radiation and unnecessary procedures.

Many tests top out at 50% or less and even if the specificity is high you can't necessarily say whether you have a bad form of the disease or not. That's the problem with PSA testing for prostate cancer.

It is until you realize how complex it is to quantify the costs to be balanced (early detection treatment costs, quality-adjusted life years saved, radiation exposure risks, etc.)

Which is anything but straightforward.

There also isn't a straightforward way to tell if a detected cancer/precancer is actually harmful and needs treatment. Ductal carcinoma in situ is almost solely detected through routine mammograms. While 30 or so percent of these lesions will go on to become invasive breast cancer most will cause no symptoms, won't progress to cancer, and be harmless. Since there is no way to tell which cases of DCIS are harmful, usually cases of DCIS are treated as if they are invasive cancer.

Now, this poses an interesting question - what is an acceptable rate of over treatment? When does the harm outweigh the benefits? Those questions i don't have an answer to.

Though the PSA test is no longer recommended by some organizations for similar reasons. Same with the breast self exam.

When does detecting a cancer with no symptoms actually improve quality of life or longevity? These are very complex questions.

Do all or most of the existing cancer detection tests involve radioactive measurements? I would hope there would be a less dangerous 'next step' to confirm before moving on to CT scans.

Do all or most of the existing cancer detection tests involve radioactive measurements?

A follow-up test that involves inserting an instrument into a patient's body to gather a biopsy sample is also invasive. For a disease that has a high risk of early death AND CAN BE TREATED AFTER EARLY DETECTION, a rather invasive test can be justified. If the test result doesn't matter (it doesn't detect an important problem, or the problem it detects is an unsolvable problem), there is less justification for the test.

Good write-ups about the risks and benefits of medical screening[1] can be found at the Science-Based Medicine website, one of my favorite sites for discussion of medical research findings.

[1] http://www.sciencebasedmedicine.org/colonoscopy-and-other-co...

http://www.sciencebasedmedicine.org/a-skeptical-look-at-scre...

http://www.sciencebasedmedicine.org/once-more-into-the-scree...

This is an issue of decision making on what you should do with the information to get the best outcome. There is nothing inherently bad about collecting the data itself.

It will be interesting to see how this progresses. As you said, false positives would certainly be a significant issue, and underlining that would be a question of liability.

If Google's scanner tells you that it has detected the formation of cancer in your body and is incorrect, putting someone through this could well be life altering in the short term.

Similarly, if you believe Google's scanner to be accurate and it fails to detect the formation of cancer, what then? If you get beyond the beneficial stages of early detection of this disease because your Google scanner tells you you're fine, how does that play out and who pays for it?

What if the device is compromised and someone can essentially have what they like reported to you?

Obviously I support any and all efforts to tackle cancer, but putting cancer detection into the hands of the end user has many potentially chaotic scenarios, and this is without mentioning the significant privacy issues that would surround such a device.

"Similarly, if you believe Google's scanner to be accurate and it fails to detect the formation of cancer, what then?"

The same as a CT scan that is a false negative?

" If you get beyond the beneficial stages of early detection of this disease because your Google scanner tells you you're fine, how does that play out and who pays for it?"

It's no different than it is now, where you got a CT scan, it said you were fine, and you weren't? As for who pays for it, this is not a valid tort (medical malpractice is about accepted general standards of the industry. If the scanner meets those, it would not be negligence/malpractice no matter what it says)

"What if the device is compromised and someone can essentially have what they like reported to you?"

Uh, you realize plenty of these scans are remotely read already, right?

People can already report what they like, without any compromise.

Even ignoring that, the hospital machines/storage/etc are all compromisable too. Probably even easier to compromise than something built for end users.

Reply on news.ycombinator.com