Last night we had our first (virtual) meeting of the new season for the Montgomery County Pediatric Society, great to "see" everyone even in the online sense. As I mentioned at the time, I was planning to post a brief comment about a systematic review of PCR and antibody testing for SARS-CoV-2, not necessarily because it is so earth shattering but because it is a nice summary of the current state of the art and a reminder of difficulties in test interpretation.
The article is in a journal probably none of you have ever come across, BMJ Evidence-Based Medicine. However, EBM has been dear to my heart since before the term was invented in the early 1990s. The authors performed a detailed review of publications to try to synthesize evidence on the diagnostic accuracy for all known tests for SARS-CoV-2 and arrive at some conclusions about the clinical effectiveness of this testing. An important caveat, though: the search is current only up to May 4, 2020, so it does not include anything about the newer antigen tests as well as any data published subsequently. Their methodology is sound, though complex, and the basic conclusions are still accurate today.
For detection of the virus, and basically we are just talking here about PCR tests, the authors found that overall sensitivity was 87.8% with 95% confidence interval 81.5% to 92.2%. In the strictly PCR studies, all the patients were diagnosed with COVID-19 so no good way to estimate specificity. The sensitivity might sound good, but it depends on what clinical situation you are dealing with, plus note that it does not include asymptomatic or pre-symptomatic patients for the most part.
A subset of viral detection methods used isothermal amplification assays with PCR as a reference standard, so here it was possible to make a guess about accuracies. The sensitivities and specificities ranged from 74.7% to 100% and 87.7% to 100%, respectively, but because of inconsistencies among the various studies the authors did not feel it was valid to pool the results to provide a single estimate of those numbers.
The results for the antibody studies were more problematic. Of the 10 studies the authors felt had sufficient information to calculate sensitivity and specificity, sensitivity ranged from 18.4% to 96.1% and specificity from 88.9% to 100%. Needless to say, the sensitivity results in particular aren't very encouraging. Antibody testing continues to be solely a research and epidemiologic aid; beware any use for single patient decision-making.
The take-home point of all of this is that we have a long way to go before we are able to make optimal use of SARS-CoV-2 testing for clinical decision making. False negatives in general are uncommon for PCR and related tests, but remember that stage of illness and technique of specimen collection are important determinants of results and need to be considered in interpreting tests for individuals.