By Ben Gannon
There are many tests now available for COVID-19, but what are their sensitivity and specificity? First off there are hundreds of different tests in use and in development. These are being collated here. Much of the information on sensitivity (SN) and specificity (SP) is preliminary and based on small sample numbers- really we need thousands of results. Therefore when we see SN and SP data we must be consider it very carefully and remember direct comparison of different tests is only valid if they have used the same samples.
Well-designed PCR should be near to perfect regarding sensitivity and specificity. Some assays are more efficient than others, and have a lower limit of detection, useful when small amounts of virus are present. One theoretical problem is that the nucleic acid sequence the PCR is targeting can mutate - all viruses (especially RNA viruses) are continually mutating at a fairly high rate. Therefore, the PCRs should target stable regions of the target genome that have lower mutation rates. When SARS-CoV-2 emerged, we did not know which parts of the genome were stable and for SARS-CoV-2 one of the tests used early-on targeted a variable region, but this has since been withdrawn.
PCR Example here
Sensitivity 100% Specificity 100% Number of samples tested = 90
For immunoassays – tests based on antigen-antibody interactions - it is harder to get near perfect sensitivity and specificity, due to non-specific and cross reactions. However the more they are used, the more thresholds can be adjusted to increase accuracy. Rapid Diagnostic Tests( RDTs) using antigen detection are normally designed for non-skilled use (e.g. pregnancy tests) and can be performed at the point-of-care. They are very useful to get very quick results but it is very hard to get the sensitivity and specificity as good as laboratory-based tests. Usually laboratory run tests are also needed to confirm results.
ELISA (Antibody detection) Example here
Sensitivity 100% Specificity 100% Number of samples tested = 84
RDT (antigen detection) Example here Sensitivity 88.66% Specificity 90.63% Number of samples tested = 525