Ag-testing for #COVID19 is much debated. Some believe it will solve almost everything while most clinical microbiologists stay skeptical. I did a non-exhaustive literature search and summarize findings here. Short thread. #testaaminen #TestCovid
My first question: has this been studied? Answer is, yes, a bit. Cochrane did a meta-analysis of point-of-care tests for #COVID19 and could include 5 studies on ag-tests. cochranelibrary.com/cdsr/doi/10.10…
Major finding of the above Cochrane review is that studies are mostly small with huge variation in methodology. I skip specificity altogether and focus on sensitivity since that IMHO is the most important metric when considering epidemic control.
There was big variation of sensitivity: "On average, the sensitivity of antigen tests was relatively poor (56.2%, 95% CI 29.5 to 79.8%)". When they did a subgroup analysis of with samples high-low viral loads, the mean sensitivity for high viral loads was better.
BUT the number of samples with high viral load was small and still there was variation in results. In comparison, PCR-based POC-tests had less variation in their performance. We already know that eg. sampling technique affects ag-tests much more.
And these test were mostly 3/4 done by spesialized laboratory personnel, not random people. So real-life sensitivity most likely would be even more variable. Cochrane concluded that all POC-tests in review need to be confirmed with RT-PCR in a low prevalence setting! 😱
I specifically looked studies that would confirm the claim that ag-tests detect the infectious virus. I found one, there might be more. This was for Biocredit COVID-19 Ag kit. sciencedirect.com/science/articl…
Based on this small and limited study, the ag-test performed worse than viral culture. Ag test was negative for 3 additional viral dilutions when culture was still positive. With clinical samples separated by viral load, high viral samples showed 80% sensitivity with ag-test.
But when they evaluated saliva samples with high viral load, the sensitivity was only 54%. Only 26 samples so needs to be approached with caution. But as a sanity check, this does not convince me. There is no evidence right now that high viral loads based on RT-PCR Ct-values ...
.. correlate with live virus. Viral culture is of course far from actual real-life contagions but it is the best we have. Finally, a study from Belgium where they tested if ag-test could be used to screen samples for RT-PCR. sciencedirect.com/science/articl…
And they conclude, this Respi-Strip does not work for that either. "Of 774 patients tested, 714 negative samples were sent for confirmation, and 159 were found to be positive by qRT-PCR. The median positive percentage agreement was 23.9 % (95 % CI: 14.2 %–38.2 %)."
So like we always say, more research is needed. But if you ask me, if I'd have to choose I would not waste public resources on ag-testing before someone shows me data on a really great ag-test! But this is just me thinking here 😜
unroll @threadreaderapp
I must correct here myself: I wrote it in a misleading way - there is no evidence of ag tests detecting _only_ live virus since it is based on assumption low Ct = culture positive. We see culturable virus even with high Ct values. It is not binary.
