The FDA published results from the first batch of COVID-19 antibody diagnostics to have their accuracy independently evaluated by federal laboratories—starting with the 12 blood tests the agency has already authorized for emergency use against the pandemic.
The data illustrates each test’s approximate ability to avoid false-positive and false-negative results, known as specificity and sensitivity, as well as their overall predictive value—essentially, how much a clinician could trust the likelihood of a correct reading when discussing the results with a patient, based on the test’s performance as well as the estimated prevalence of the novel coronavirus throughout the community.
The validation work is part of a project launched last month in collaboration with the Centers for Disease Control and Prevention and the National Institutes of Health, and the results come just days after the agency clamped down on the broad marketing of antibody tests across the country.
Conducted by the National Cancer Institute’s Frederick National Laboratory for Cancer Research, the test data was also used by the FDA as it decided whether to grant the products a green light. In addition, the FDA last month issued a broad authorization for antibody tests that are voluntarily submitted and pass review by the program.
RELATED: FDA, CDC, NIH to begin validating COVID-19 antibody tests as more enter the market
It is currently unknown how many people in the U.S. have been exposed to the novel coronavirus and have developed antibodies against the disease. In addition, this percentage varies among the groups of people being tested—healthcare workers, for example, are at greater risk and show higher rates of infections compared to the broader population.
To calculate the tests’ positive and negative predictive values, the FDA assumed a baseline 5% prevalence rate—which, taken nationwide, would amount to over 16.4 million people. However, the agency said that any single antibody test, when put up against a largely asymptomatic general population, “is not likely to be sufficiently accurate to make an informed decision,” and that additional tests focusing on different aspects of the virus would be needed.
RELATED: FDA increases oversight, requirements for COVID-19 antibody tests
The FDA also posted a simple calculator that shows predictive values based on different levels of prevalence, which may change over time or across different locations.
The first COVID-19 serology test granted an Emergency Use Authorization by the FDA—Cellex’s lateral flow rapid test, using drops of blood and a test strip similar to a pregnancy test—showed a combined sensitivity of 93.8% and a specificity of 96.0% when searching for two different antibodies linked to the coronavirus.
At 5% prevalence, that equates to a positive predictive value of 55.2%. At 10% prevalence, that estimate of performance increases to 72.3%, while at 2.5%, it drops to 37.6%. In short, the more likely it is that a person has been exposed to the disease, the more likely the test is giving a correct positive result. Cellex’s test was more accurate at ruling out cases, showing a 99.7% negative predictive value at 5% prevalence.
RELATED: Current COVID-19 antibody tests aren't accurate enough for mass screening, say Oxford researchers
Other products use different testing technologies and examine larger blood samples—such as the high-throughput ELISA tests developed by Abbott, Roche and other companies.
Roche’s recently authorized Elecsys test showed 100% sensitivity and 99.8% specificity across multiple antibody types, with a positive predictive value of 96.5% and a negative predictive value of 100% at 5% prevalence. Similarly, Abbott’s Architect test for IgG antibodies showed a sensitivity of 100%, specificity of 99.6%, and positive and negative predictive values of 92.9% and 100%, respectively.
Meanwhile, a separate, two-step ELISA test developed by clinical laboratories at the Mount Sinai Health System in New York showed a sensitivity and specificity of 92.5% and 100%, resulting in positive and negative predictive values of 100% and 99.6%.
The full list of test results from all 12 FDA-authorized tests is available here.