4 ms·

There's always the chance that the test itself is not particularly good and just has a high false positive rate. Although it is difficult for a PCR test to prod

There's always the chance that the test itself is not particularly good and just has a high false positive rate. Although it is difficult for a PCR test to produce a match if there is no genetic material present, it may amplify trace signals, depending on the cutoffs chosen.

JackFr 7y ago
This seems to me the most obvious implication. That when you only test symptomatic people, the test might be accurate, but in testing everyone the false positive rate is much higher.
ImaCake 7y ago
I looked into this a few weeks ago. The PCR test has an incredibly low false positive rate, below 1%. It is low enough that it is feasible to test a 100,000 people and expect basically zero false positives. The false negative rate is somewhere around 30%. So the real danger with PCR is that you will miss cases. The opposite is true for the antibody tests. The false negative rate is around 10% and the false positive rate is also around 10%. So in 100,000 tests, you would expect to miss about 1,000 true positives. But, and this is critical, you would also mistakenly get 1,000 false positives - even if only 1/100,000 people are actually infected! Next time you see someone claiming 50% asymptomatics. First thing to ask is whether they used PCR. Or if they used an assay with a terrible false positive rate.
- danieltillett 7y ago
  If not done well PCR has an incredibly high false positive rate. If you contaminate your set up area (or equipment) with past PCR amplification material you will get a very high rate of PCR positivity. Of course if you do things correctly this shouldn’t happen, but if the average pathology lab technician is as skilled as my ex-students and colleagues (and myself) then I would be not be so keen to rule out false positives. Yes there is a major issue with false negatives, especially when the swab was not performed or stored correctly. RNA is not very stable and it is very easy to destroy all the viral RNA if you are not careful.
  - ImaCake 7y ago
    I am familiar with these drawbacks, having worked as a research assistant doing RT-PCR and qPCR. My understanding from a brief stint in industry is that there is a lot more quality control to be found in industrial and diagnostic labs than is typically found in a research lab. So I imagine there is still an appreciable amount of false positives even in a diagnostic lab, but they do have a lot more quality control in place to reduce it.
    - danieltillett 7y ago
      Yes normally this is the case, but when they are under immense pressure to pump through the samples we should expect the normal high standards drop.
- jobigoud 7y ago
  > The false negative rate is around 10% and the false positive rate is also around 10%. So in 100,000 tests, you would expect to miss about 1,000 true positives. > you would also mistakenly get 1,000 false positives I think you mean missing 10,000 true positives and getting 10,000 false positives.
- andrewla 7y ago
  > The PCR test has an incredibly low false positive rate, below 1%. Do you have a source, offhand? The problem with asserting any sort of false positive rate is that you need to benchmark against a source of truth. What source of truth are we talking about? A second PCR test? Clinical diagnosis? If we benchmark against the clinical diagnosis, for example, then clearly this is a huge false positive rate -- most of the patients are clinically diagnosed as negative, which is all this article is saying. Benchmarking against a second test doesn't really tell us much if trace signals are being amplified, or if the laboratory is in any way compromised because of the urgency of the testing.

JackFr 7y ago

This seems to me the most obvious implication. That when you only test symptomatic people, the test might be accurate, but in testing everyone the false positive rate is much higher.

ImaCake 7y ago

I looked into this a few weeks ago. The PCR test has an incredibly low false positive rate, below 1%. It is low enough that it is feasible to test a 100,000 people and expect basically zero false positives. The false negative rate is somewhere around 30%. So the real danger with PCR is that you will miss cases. The opposite is true for the antibody tests. The false negative rate is around 10% and the false positive rate is also around 10%. So in 100,000 tests, you would expect to miss about 1,000 true positives. But, and this is critical, you would also mistakenly get 1,000 false positives - even if only 1/100,000 people are actually infected! Next time you see someone claiming 50% asymptomatics. First thing to ask is whether they used PCR. Or if they used an assay with a terrible false positive rate.

danieltillett 7y ago
If not done well PCR has an incredibly high false positive rate. If you contaminate your set up area (or equipment) with past PCR amplification material you will get a very high rate of PCR positivity. Of course if you do things correctly this shouldn’t happen, but if the average pathology lab technician is as skilled as my ex-students and colleagues (and myself) then I would be not be so keen to rule out false positives. Yes there is a major issue with false negatives, especially when the swab was not performed or stored correctly. RNA is not very stable and it is very easy to destroy all the viral RNA if you are not careful.
- ImaCake 7y ago
  I am familiar with these drawbacks, having worked as a research assistant doing RT-PCR and qPCR. My understanding from a brief stint in industry is that there is a lot more quality control to be found in industrial and diagnostic labs than is typically found in a research lab. So I imagine there is still an appreciable amount of false positives even in a diagnostic lab, but they do have a lot more quality control in place to reduce it.
  - danieltillett 7y ago
    Yes normally this is the case, but when they are under immense pressure to pump through the samples we should expect the normal high standards drop.
jobigoud 7y ago
> The false negative rate is around 10% and the false positive rate is also around 10%. So in 100,000 tests, you would expect to miss about 1,000 true positives. > you would also mistakenly get 1,000 false positives I think you mean missing 10,000 true positives and getting 10,000 false positives.
andrewla 7y ago
> The PCR test has an incredibly low false positive rate, below 1%. Do you have a source, offhand? The problem with asserting any sort of false positive rate is that you need to benchmark against a source of truth. What source of truth are we talking about? A second PCR test? Clinical diagnosis? If we benchmark against the clinical diagnosis, for example, then clearly this is a huge false positive rate -- most of the patients are clinically diagnosed as negative, which is all this article is saying. Benchmarking against a second test doesn't really tell us much if trace signals are being amplified, or if the laboratory is in any way compromised because of the urgency of the testing.