COVID-19: The Push for Serological Antibody Testing With a High Degree of Accuracy and Reliability
Novel Coronavirus Antibodies
Two antibodies specific to the novel coronavirus SARS-CoV-2 have been identified as forms of immunoglobulin (Ig), namely IgM and IgG. They tend to appear only slowly during the initial stages of infection, with IgM appearing first, reaching a maximum and then disappearing over the course of a few weeks, and IgG starting to appear slightly later and then gradually reaching a plateau. Their time profiles during the course of infection were first described by Xiao et al. in a letter to the editor of the Journal of Infection and published online on March 21. The authors analyzed antibodies in blood taken from 34 hospitalized COVID-19 patients at different time points. They reported that all patients tested positive for both IgM and IgG by the third week after onset of symptoms, but IgM started to decline rapidly after week 3 while IgG continued to increase, leveling off (but not declining) between weeks 4 and 7. Thus, an IgM-IgG profile may provide information on the course of the disease in a given patient. The authors compared their results to the time profile for the original SARS virus and saw potential similarities; IgM levels with SARS reached a maximum within 4 weeks and were undetectable after 3 months, while IgG was persistently detectable up to 24 months after onset of symptoms.
Because detectable levels of antibodies are slow to appear in the initial stages of an infection, serology tests should not be used as the sole means of diagnosing COVID-19, but only in conjunction with molecular (RT-PCR) diagnostic testing. Serology testing is useful in estimating the extent of an individual’s development of an immune response to infection and, in combination with diagnostic testing, in determining whether they are able to return to work or qualify to donate convalescent plasma that can be used to treat currently sick individuals. It can also be used to identify the percentage of individuals who have been infected but who have never developed symptoms. It should be stressed, however, that the duration of any immunity conferred by infection with COVID-19 is currently unknown, as is the level of antibodies required to prevent (or lessen the severity of) reinfection.
FDA Statements on Serological Antibody Tests
On April 7, FDA Commissioner Stephen Hahn issued a statement regarding serological tests for COVID-19 antibodies. The previous month, FDA had issued a policy allowing developers to market their tests without prior FDA review if certain conditions in FDA’s guidance document were met, in order to allow early patient access. Commissioner Hahn reported that more than 70 test developers had notified FDA they had serological tests available for use, but some firms were falsely claiming their tests were FDA-approved or authorized, or that they could diagnose COVID-19. He said that FDA intended to take action against such firms, or those marketing tests that lacked accuracy and reliability. At that time, only one company had followed the alternative route and developed a test reviewed and authorized by FDA under an Emergency Use Authorization (EUA).
Soon afterwards, on April 18, Commissioner Hahn issued an update, encouraging developers to submit EUA requests for their devices and announcing FDA’s strengthening of the EUA process by collaborating with NIH and CDC to establish a capability at NIH to evaluate serological tests – both those currently on the market and those under development – in support of an EUA and thus to provide greater confidence in test performance. Commissioner Hahn reiterated FDA’s intention to take appropriate action against companies making or distributing non-validated tests or making false claims.
Recent Antibody Testing Surveys
Meanwhile, results of antibody testing surveys of populations in Germany, the Netherlands and the United States (New York City; Los Angeles and Santa Clara Counties, California; Chelsea, Massachusetts; Miami-Dade County, Florida) started to be reported, and they appeared to indicate that the extent of coronavirus infections had been greatly underestimated, with large percentages of each population either asymptomatic or otherwise not included in the official confirmation totals due to only mild symptoms. The surveys reported that between 2% and 30% of these populations had already been infected with the novel coronavirus. A much higher than officially reported infection rate would, of course, indicate a much lower fatality rate and might provide an impetus for the easing of lockdown requirements. However, by April 21, the scientific journals Science and Nature had both published articles questioning the accuracy and reliability of the antibody tests used, inadequate sample sizes, unrepresentative sampling of populations and, by extension, the surveys’ conclusions about the true infection rate.
Much higher than expected infection rates indicated by positive antibody tests, if verified by confirmatory testing, might indicate significant progress towards a “herd immunity” that could prevent or suppress future outbreaks of COVID-19. Unfortunately, at this point no one knows how long such immunity might last; as we noted in an earlier Alert, two of the human coronaviruses that can cause the common cold confer a very short period of immunity – about 45 weeks – although the original SARS coronavirus confers a longer immunity of perhaps two years. Also, no one knows for certain what level of antibodies is required to confer a level of immunity sufficient to prevent reinfection; while some governments have suggested that antibody detection could serve as the basis for an “immunity passport” or “risk-free certificate” that would enable previously infected individuals to return to work, the World Health Organization indicates that this suggestion is premature.
The Need for a High Degree of Accuracy
Accurate and reliable serology tests, with very low levels of false positives and false negatives (high “test specificity” and high “test sensitivity,” respectively), are essential for generating results that can be used for making sound policy decisions. So far no test is perfect (although one recently authorized by FDA appears to have come close), but according to the FDA (see below) tests should achieve at least 90% sensitivity and 95% specificity; having no test is preferable to having a bad test, since bad results can lead to bad decisions made from a false sense of certainty. High test specificity (low false positives) is the more important of the two criteria, since a specificity even as high as 95% means that, for example, 50 individuals from a relatively small sampling of 1000 will be incorrectly identified as being or having been asymptomatically infected. One instance where this might have significant repercussions would be in the testing of a population with a low infection rate (“low prevalence”); this might produce as many (or more) false positives as true positives because of the low probability of finding individuals who had actually been infected, thus greatly exaggerating the apparent infection rate. In such cases, FDA has suggested some individuals may need to have two antibody tests performed to generate reliable results.
High test sensitivity (low false negatives) is important if a confirming RT-PCR test is not performed, since false negatives could lead to the clearance of those who might be asymptomatic (but contagious) to return to work after a lockdown, or to the similar clearance of those who had been sick and might still be contagious after quarantine.
Comparison Testing of Antibody Test Kits
Recently, a large group of scientists, mainly from the University of California, San Francisco (UCSF), tested the accuracy and reliability of 11 antibody tests selected from the 70 or so on the market at the time (not through an EUA), plus one in-house test. Their preliminary results were posted online on April 24. The tests consisted of 10 lateral flow assays (LFAs) and two enzyme-linked immunosorbent assays (ELISAs) (one commercial, one in-house). LFAs deliver positive/negative results for the IgM and IgG antibodies with varying positive intensities scored by trained readers from 1 (weak) to 6 (strong), while ELISA provides more quantitative results. In this case, the study set included:
- Samples from symptomatic individuals who had tested positive for the SARS-CoV-2 virus by the RT-PCR molecular test.
- Negative controls from blood samples donated to the American Red Cross pre-July 2018, well before the onset of the COVID-19 pandemic.
Multiple samples taken from some SARS-CoV-2-positive individuals allowed for comparison of test results at different time intervals after the onset of symptoms (1-5, 6-10, 11-15, 16-20, and <20 days).
As expected, the proportion of specimens testing positive for antibodies increased with increasing time from symptom onset, reaching a maximum at time intervals 16-20 and >20 days; thus, correct diagnosis in the early stages of infection still depends on a molecular diagnostic test rather than an antibody test. IgM detection was less consistent than IgG for nearly all assays, but the highest detection rate was achieved by combining the two results (IgM + IgG). Test sensitivity in RT-PCR-positive individuals at >20 days ranged from 81.8% to 100%.
For the negative controls, the data showed test specificity of >95% for 7 of the 12 tests being evaluated (overall range 84.3-100.0%), but was >99% (very low false positives) for only 2 of the 10 LFAs and the in-house ELISA. Four tests gave false negatives in over 10% of samples. However, two of the three tests with high (>99%) specificity (low false positives) were among the lowest performers (high false negatives) in the positive cases, so no test appeared to perform well in terms of both specificity and sensitivity.
Shortly after these preliminary results appeared, on April 28, the chairman of the House Subcommittee on Economic and Consumer Policy, citing the study, wrote to FDA Commissioner Hahn to complain that the Agency had failed to police the coronavirus serological antibody test market, was not conducting a substantive review of the tests it had allowed on the market, had not conveyed any clear policy on serological tests or provided any standards and guidelines, and was unable to validate the accuracy of tests already on the market. According to the letter, the newly released UCSF study had confirmed the Subcommittee’s concern that “FDA's lax policies are permitting a flood of fraudulent tests onto the market.” It asked FDA, among other things, (1) to require serological test makers to apply to FDA for approval to market and to submit validation data to the Agency, and (2) to pull ineffective tests from the market. The first of these requests, at least, was soon to be granted.
FDA’s Issues “Umbrella” Emergency Use Authorization for Serology Tests
Also on April 28, FDA issued an Emergency Use Authorization (EUA) for SARS-CoV-19 (novel coronavirus) serological antibody tests that had been evaluated in independent studies performed at the NIH’s National Cancer Institute (NCI) or other FDA-designated government agency, and that had been confirmed by FDA to meet the criteria that it set forth in the Authorization. Tests authorized under this so-called serology “umbrella” EUA are intended to be used in identifying individuals with an adaptive immune response to the virus (indicating recent or prior infection) by detecting one or two antibodies specific to SARS-CoV-19 – IgG, or IgG and IgM, or total (IgG + IgM) – in human plasma or serum. Testing of the serum or plasma samples may only be performed at authorized laboratories certified by FDA under the Clinical Laboratory Improvement Amendments (CLIA) to perform tests of moderate or high complexity. Both LFAs and ELISAs are authorized.
Under the EUA, test manufacturers must submit to FDA validation from NCI or another FDA-designated government agency that consists of the following:
- For determination of test sensitivity: positive antibody tests from at least 30 confirmed SARS-CoV-2-positive serum or plasma samples, for each antibody being identified.
- For determination of test specificity: negative antibody tests from 80 confirmed SARS-CoV-2-negative (or pre-COVID-19) serum or plasma samples.
- For determination of absence of cross-reactivity (i.e., testing positive for another virus in addition to SARS-CoV-2): negative antibody tests from 10 of those 80 samples confirmed to be HIV-positive.
The minimum required overall performance is 90.0% test sensitivity and 95.0% test specificity; for tests that specifically report IgM and IgG results the test sensitivity must be at least 70% for IgM and 90% for IgG. The results should show no detectable cross-reactivity with HIV.
In order to be added to FDA’s list of devices (“Appendix A”) authorized under the EUA, manufacturers must have instructions for use that comply with the labeling requirements of the EUA’s Conditions of Authorization. These include statements that:
- The test is only for use under an EUA
- The test should not be used to diagnose acute SARS-CoV-2 infection.
- Negative results do not necessarily mean an infection is absent; if infection is suspected, it should be confirmed by a molecular test.
- False positives are possible; they may result from cross-reactivity with antibodies from prior infection with a different virus or from other causes, and should be confirmed using a second, different assay.
On May 4, as reported in a separate Alert, FDA issued a new Guidance (superseding its March 16th Guidance) that “drastically changes the requirements for serology/antibody tests for the COVID-19 epidemic” by requiring all manufacturers (but not clinical laboratories) who had been marketing such tests to submit EUA requests and validation data within 10 business days from the date they notified the FDA of their validation data or by May 18, whichever is later.
Also on May 4, FDA Deputy Commissioner for Medical and Scientific Affairs, Anand Shah, and Director Jeff Shuren of FDA’s Center for Devices and Radiological Health, issued a statement describing the main points of the April 28 “umbrella” EUA and the May 4 revised Guidance. In this statement, FDA noted that of the 12 antibody tests that had already been authorized under an individual EUA, most had been authorized within the preceding few days, and that over 200 tests were currently the subject of a pre-EUA or EUA review. (The latest list of authorized tests is available here.) FDA explained that the test performance data being generated by NCI could be used by FDA to authorize a test, to help engage a test developer in providing additional supporting data, or to take action against inadequate tests, including taking them off the US market.
FDA has pledged to continue to work to stop illicit tests from entering the US market, and encourages users – states, hospitals, consumers – to be on high alert and to make informed test purchasing decisions. So there will likely be more to follow, and we are continuing to monitor closely.
 The Appendix to this recent White House document shows how the use of two antibody tests rather than one improves the predictive value of testing in various infection prevalence environments, but dramatically so in areas of low prevalence; in the examples shown, both tests have acceptable sensitivities (95%) and the FDA-required minimum specificities (also 95%).