To the Editor,
It is widely believed that return to normalcy after the COVID-19 pandemic will include widespread testing for both active (presence of viral nucleic acids) and recovered (presence of anti-SARS CoV2 immunoglobulins) infections . The latter can inform if individuals were exposed to the virus and if they presumably developed immunity. The required analytical performance characteristics of COVID-19 antibody (Ab) tests have not been discussed in detail. Suboptimal analytical performance can create confusion and may lead to false reassurances, especially when carried out on large populations that have yet to be exposed to the virus and in the absence of a gold standard comparative method . Here, we analyze the importance of assay specificity and calculate the required threshold for the tests to provide clinically relevant information.
Most discrete clinical tests provide two types of results: positive or negative for a specific parameter of interest. Positive results can be true positives (TP; positive with assayed parameter present) or false positives (FP; positive without assayed parameter present). Negative results can be true negatives (TN; negative without assayed parameter present) or false negatives (FN; negative with assayed parameter present). The analytical requirements for the adoption of a clinical test depend on its intended use. It is widely claimed that COVID-19 serological Ab tests could be used for the safe return of presumably immune people to normalcy. In this case, FN Ab results are almost inconsequential as the tested individuals will be told that they are negative and thus have no presumed immunity, when, in fact, they do, leading at most to some unnecessary extension of their isolation practices, but not viral spread. But FP results are consequential as these individuals may be reassured of presumed immunity that they do not have, rendering them vulnerable to infection and viral spread. The rate of FP results depends on the analytical specificity of the assay.
The analytical parameter that reflects the required analytical specificity in order to adopt such a test or not (based on the FP/FN ratio) is the positive predictive value (PPV). The % PPV reflects somebody’s chances of having the disease if the test is positive and equals: PPV=TP/(TP+FP)×100. The required PPV value depends on the intended use of the test and the presumed risk/benefit value of its application. The PPV is dependent not only on the sensitivity and specificity of the test but also on the disease prevalence. Below, we present simulation data for COVID-19 infection based on two US prevalence values. The one at 0.2% (one infected per 500 individuals) and another one at 1% (one infected per 100 individuals) are based on preliminary estimations of prevalence in some countries (we remind that prevalence varies widely among countries, regions, cities, etc. and is continuously changing). We also fixed the sensitivity of the test to 100% (no FN, as a best case scenario) and varied the specificity from 90% onward, calculating the PPV in each case (Table 1). These data show that in order to achieve a minimum of 90% PPV (allowing one FP out of every nine TP), the analytical specificity of the COVID-19 Ab test must exceed 99.9% in both scenarios. With most clinical tests, such exquisite specificities are quite rare and particularly challenging to achieve.
Positive predictive value of a test with a 100% sensitivity and variable specificity, at two disease prevalences.
|% Specificity at 100% sensitivity||Cases per 100,000 people||True positives (TP)||True negatives (TN)||False positives (FP)||False negatives (FN)||Positive predictive value (PPV), %|
We urge COVID-19 Ab kit manufacturers to check thoroughly the specificity of their COVID-19 Ab assays by using sera from patients with various autoimmune conditions (about 5% prevalence in the US population)  and patients positive for heterophilic antibodies  (about 1% prevalence in the general population; these patients have serum antibodies with broad and unpredictable specificities against human and animal antibodies) and other infections with diverse viruses (optimally including SARS-CoV and other ‘common-cold’ CoVs), to make sure of no cross-reactivity . Universal standardizations and assay harmonizations would contribute tremendously to minimal analytical variability between kits .
The high urgency for new SARS CoV-2 serological tests should not compromise assay quality standards. FP results could lead to vital errors both in prevalence prediction studies and in misleadingly reassuring patients of presumed immunity, thus allowing them to return to societal activities at a great risk of getting infected and further spreading the virus.
Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.
Trevethan R. Sensitivity, specificity, and predictive values: foundations, pliabilities, and pitfalls in research and practice. Front Public Health 2017;5:307–12.
Preissner CM, Dodge LA, O’Kane DJ, Singh RJ, Grebe SK. Prevalence of heterophilic antibody interference in eight automated tumor marker immunoassays. Clin Chem 2005;51: 208–10.