4 ms·

> Worst case with a false positive? You're causing someone anxiety and giving them an extra scan. It is arguable that this as minor an issue as you make it out

by jghn 2y ago

> Worst case with a false positive? You're causing someone anxiety and giving them an extra scan. It is arguable that this as minor an issue as you make it out to be. There has been work to try to assess this (google "cancer overdiagnosis"). The counterargument to what you state is that a false positive can not only lead to stress & unnecessary/more invasive screening, but a treatment plan that's a net negative. For instance: if a cancer were detected and it'd wind up being a cancer that someone dies with instead of from, and the treatment causes worse outcome than the cancer itself, that's not good. The hard issue here is that these things need to be determined at a population scale, and one can always cherry pick personal anecdotes in either direction to tug at heartstrings. We seem to have found ourselves at a point where it's clear there's a balance that needs to be struck, but it's unclear what that balance is yet.

mattmaroon 2y ago
He did say that you tune them to minimize false negatives. I do find it interesting that in the early days of HIV testing even most people who got a "positive" result were in fact negative. The tests have since greatly improved and the number of people taking a test due to hysteria rather than likelihood of infection declined. But I can imagine a world in which we get very many forms of liquid biopsies like this every year, and false positives become a thing we understand and are used to.
- Goronmon 2y ago
  He did say that you tune them to minimize false negatives. That's a bit hand-wavy for something that appears to be the core issue with "test early, test often", isn't it?
  - mattmaroon 2y ago
    Well, it would seem hand-wavy to whichever scientists are tasked with improving it I'm sure! There's probably a lot of work involved. But it's pretty standard for new tests. The first ones are never the most reliable. Obviously the rate of false positives and expense of follow-up testing determine whether it's overall a good idea, but it probably will start out with a positive ROI that becomes a highly positive ROI pretty fast. I mean, yes, some people are going to freak out no matter how many times their doctor tells them about false negatives before they get the test. But overall this will save a lot of lives and as the test improves, a lot of money too.
- ethbr1 2y ago
  Whether one should bias to allow more false positives or false negatives depends on the next steps after a false positive vs the risks after a false negative. False positive: more costly scans False negative: untreatable terminal illness
mapt 2y ago
The existing downsides of a might-be-cancer hit on a test are real, for now - a statistically significant number of people with a breast cancer diagnosis end up killing themselves out of despair or doing radical surgery "to be sure", which detracts from the social benefit of mammograms. But there's no indication that they would persist if false positives were more common because proactive testing was more common. There would be a pipeline of followup testing and risk evaluation, which would be normalized by how common it was and how nonlethal diagnostic hits would become. It would become "Oh, yeah, my cousin got a hit, but followup biopsy said it was a mutation that doesn't metastasize. Guess we'll see next week." rather than "My great-grandmother died of breast cancer, my mother died of breast cancer, my friend's aunt died of breast cancer, this feels like a death sentence", which is the information people who receive hits on their mammogram testing are acculturated with now.
- jghn 2y ago
  Again, this is a well studied topic. Anyone claiming they've identified the one true answer isn't someone who should be believed. The topic is complex, at best. But these are not hypothetical situations. There are data to support the point to which you're countering. You are correct that there exists a world where the problem is minimized and at that point obviously the math changes. But we're not there yet.
  - mapt 2y ago
    The argument is over whether we should seek to get there.

by jghn 2y ago

mattmaroon 2y ago

He did say that you tune them to minimize false negatives. I do find it interesting that in the early days of HIV testing even most people who got a "positive" result were in fact negative. The tests have since greatly improved and the number of people taking a test due to hysteria rather than likelihood of infection declined. But I can imagine a world in which we get very many forms of liquid biopsies like this every year, and false positives become a thing we understand and are used to.

Goronmon 2y ago
He did say that you tune them to minimize false negatives. That's a bit hand-wavy for something that appears to be the core issue with "test early, test often", isn't it?
- mattmaroon 2y ago
  Well, it would seem hand-wavy to whichever scientists are tasked with improving it I'm sure! There's probably a lot of work involved. But it's pretty standard for new tests. The first ones are never the most reliable. Obviously the rate of false positives and expense of follow-up testing determine whether it's overall a good idea, but it probably will start out with a positive ROI that becomes a highly positive ROI pretty fast. I mean, yes, some people are going to freak out no matter how many times their doctor tells them about false negatives before they get the test. But overall this will save a lot of lives and as the test improves, a lot of money too.
ethbr1 2y ago
Whether one should bias to allow more false positives or false negatives depends on the next steps after a false positive vs the risks after a false negative. False positive: more costly scans False negative: untreatable terminal illness

mapt 2y ago

The existing downsides of a might-be-cancer hit on a test are real, for now - a statistically significant number of people with a breast cancer diagnosis end up killing themselves out of despair or doing radical surgery "to be sure", which detracts from the social benefit of mammograms. But there's no indication that they would persist if false positives were more common because proactive testing was more common. There would be a pipeline of followup testing and risk evaluation, which would be normalized by how common it was and how nonlethal diagnostic hits would become. It would become "Oh, yeah, my cousin got a hit, but followup biopsy said it was a mutation that doesn't metastasize. Guess we'll see next week." rather than "My great-grandmother died of breast cancer, my mother died of breast cancer, my friend's aunt died of breast cancer, this feels like a death sentence", which is the information people who receive hits on their mammogram testing are acculturated with now.

jghn 2y ago
Again, this is a well studied topic. Anyone claiming they've identified the one true answer isn't someone who should be believed. The topic is complex, at best. But these are not hypothetical situations. There are data to support the point to which you're countering. You are correct that there exists a world where the problem is minimized and at that point obviously the math changes. But we're not there yet.
- mapt 2y ago
  The argument is over whether we should seek to get there.