MD here. It's "not done" because it could hurt you.
Here's the Bayesian argument:
- Let's assume a lab test (or imaging test) has a 1% false positive rate and 100% sensitivity (that would be pretty good for many tests and imaging modalities!).
- The cancer you screening for has a 1% incidence as well (low-ish, for the sake of argument).
- Therefore the chance that a "positive" scan would be due to a false positive or a true cancer would be equivalent (i.e., the Positive Predictive Value is 50%).
If we did this at population level, we'd be massively overtreating and over-biopsy-ing... not the desired outcome.
Maybe figure out a better way to verify that doesn't require biopsy? Maybe use machine learning in conjunction with better data? I don't know enough but there must be other techniques we could devise that don't require major invasive surgery every single time.
Just sitting on our hands on this information and letting people die is the worst decision. We need to improve our technology to get better results, not throw our hands up in the air and say the problem is too hard.
Well the ways that don't involve surgery are themselves less accurate heuristics with false positives. How can they calibrate without being certain if it is a true or false positive or negative?
Better imaging is an option but that often falls into dangerous in itself or easier said than done like say trying to do detection of cancer by scent.
Machine learning is fundamentally GIGO and tends to learn smartass shortcuts if one isn't careful. "The healthy sample data is less zoomed in while the cancer ones are zoomed in - therefore full chest x-ray is fine but a lung X-ray is cancer."
My limited experience is that for small contained cancers imaging can't predict the future course for the tumor because you need cellar resolution that you can't get without a biopsy and a microscope and or genetic tests.
Even then it's really common to see a biopsy come back as 'indeterminate'. If it's indeterminate then now what? You either punt and do nothing or go for excision.
Consider this tidbit: A lot of slow moving low grade cancers it hardly matters when you find them. Early or later matters little. Whereas aggressive cancers often are untreatable. Find them early, find them late you'll die either way. And fast growing cancers often fully develop between screenings.
This type of attitude is typical, where one likes to dismiss new ideas and keep status quo instead of coming up with new ideas or innovations.
The point is that there is an opportunity for the Steve Jobs of health care to come up with a way, a new innovative novel way, for classification of tumors to be improved without needing biopsy. Once we have that, then we can take tests to our heart's content. Right now, the attitude of "we shouldn't have people take tests and detect cancer early because our current method of determining if they are cancerous would make it too hard" isn't acceptable.
Presumably none of us here are inventors of new medical technology. We are just amateurs talking about this casually on Hacker News. Saying it's "unacceptable" doesn't change anything from either the doctor or patient's point of view. Until the new tech comes along, we need to accept the limitations of what we have now.
I agree with you, but I think the original point was that we may already have technologies that could be employed, but that there is concern they are not being used
”Just sitting on our hands on this information and letting people die is the worst decision.”
If the alternative is saving a few who would otherwise die, but taking away more healthy days from many others or, possibly, even killing healthy individuals because they get overtreated, not spending money on screening _now_, but instead spending money on better understanding is the rational thing to do.
And that’s exactly what we are doing. Technology advances, models get refined. Maybe, some day, statistics (and ethics committees) will decide population screening does make sense.
And it isn’t necessarily about major invasive surgery. Even small procedures, done unnecessarily often enough, may already swing the balance.
”Screening may identify abnormalities that would never cause a problem in a person's lifetime. An example of this is prostate cancer screening; it has been said that "more men die with prostate cancer than of it". Autopsy studies have shown that between 14 and 77% of elderly men who have died of other causes are found to have had prostate cancer.[
Aside from issues with unnecessary treatment (prostate cancer treatment is by no means without risk), overdiagnosis makes a study look good at picking up abnormalities, even though they are sometimes harmless.
Overdiagnosis occurs when all of these people with harmless abnormalities are counted as "lives saved" by the screening, rather than as "healthy people needlessly harmed by overdiagnosis". So it might lead to a endless cycle: the greater the overdiagnosis, the more people will think screening is more effective than it is, which can reinforce people to do more screening tests, leading to even more overdiagnosis.”
Would re-screening everyone positive N weeks after the first positive, and then re-screening the new positives again N weeks later, alleviate this (absent other symptoms)?
I'm assuming here a growing cancer is easier to detect over time.
No, this is the wrong argument. What would happen is, eventually people just wouldn't believe any of these tests and they would become useless, even when they did reach a high level of PPV. Also see the point about biopsies - definitely not non-invasive, often require surgery. Also cost - you're now looking at 5-10k extra charges.
I assume every doctor's visit sees you getting checked for every type of condition every time, or do you want your doctor to play the odds and only look for things you're likely to have?
There's plenty of harmful ways to die, and you'll find they don't check for most of them until they have a decent reason.
Here's the Bayesian argument:
- Let's assume a lab test (or imaging test) has a 1% false positive rate and 100% sensitivity (that would be pretty good for many tests and imaging modalities!).
- The cancer you screening for has a 1% incidence as well (low-ish, for the sake of argument).
- Therefore the chance that a "positive" scan would be due to a false positive or a true cancer would be equivalent (i.e., the Positive Predictive Value is 50%).
If we did this at population level, we'd be massively overtreating and over-biopsy-ing... not the desired outcome.