One of the first serosurvey results from Santa Clara county in California reports 48,000 to 81,000 people infected, 50-fold to 85-fold more than the number of confirmed cases.
Living through a pandemic while under lockdown order, with schools and routines and social events and music concerts and movie theaters and sports events all closed means your sense of time has distorted. It was only a week ago I wrote about the first results from widespread serological testing in Germany took place, in Heinsberg Germany, was reported here to be resistance to SARS-CoV-2 at 14% of the population. (Note: The town of Gangelt is considered to be ‘Germany’s Wuhan’ due to the high number of infections and deaths there.)
The importance of serological surveys
For those who do not have a background in biology and immunology, an IgG / IgM test looks for specific classes of antibodies produced by your immune system. (For those looking for more background, I wrote here about the first FDA authorized test under EUA, the Cellex US test.)
And I reproduce the figure below, for illustrative purposes, how the levels of IgM and IgG rise and fade over the time-course of the disease.
These figures are vitally important, to determine how widespread the denominator is in the calculations of IFR (infection fatality ratio or infection fatality rate) and CFR (case fatality ratio or case fatality rate). For the definition of ‘case’, this is confirmed COVID-19 disease: not only rRT-PCR molecular test positive, which are highly sensitive and specific.
Alas as I wrote here these tests may have high false-negative rate due to the virus not being present, not necessarily due to problems in sample collection or handling of the sample, but rather because there is no virus present to detect even though the patient clearly has COVID-19.
By finding out the presence or absence of the antibodies against the SARS-CoV-2 antigen(s), important conclusions can be made about how prevalent the virus has been spreading undetected in the population, masquerading as the common cold or influenza, as well as quantifying the number of individuals in a geography that may have immunity to infection by SARS-CoV-2 as we as a society grapple with re-opening workplaces and get to whatever the ‘new normal’ will look like.
It is important to note at this point, possibility and risk of reinfection in humans is not known definitively, nor details around development of immunity. That said, human monoclonal antibodies from single B-cells of eight SARS-CoV-2 infected individuals have been described (BioRxiv preprint here), and there is animal model data (macaques) that reinfection with SARS-CoV-2 does not occur (BioRxiv preprint here). Also an important note, a positive antibody test result is no substitute for a complete positive diagnosis for SARS-CoV-2 infection, however it is an indication of past exposure to the virus.
Serological surveys underway in the United States
Anyway, getting back to how strange time feels in a time of lockdown, it was only the next day I wrote here about the NIH starting a serological survey , recruiting 1,000 (later told it will be expanded to 10,000). I emailed the address to volunteer right when I saw the press release but understand that the NIH has been overwhelmed with the response and said it could take a week or more to get a reply back. A week later they haven’t replied to my own request to volunteer.
The headline you will be reading today is the result of a previously-announced serological survey (aka ‘serosurvey’) of 2500 volunteers ‘selected to be representative of the county and about 500 of their children’. Santa Clara covers the cities of Palo Alto, San Jose all the way down to Morgan Hill.
The same Stanford group announced a few days ago they would partner with Major League Baseball and 27 of their 30 teams, to test about 10,000 of their employees (everyone from office staff and management to field-keepers to the players themselves). Interestingly, the principal investigator from Stanford, Dr. Jay Bhattacharya, said he would be able to collect data very quickly (the testing group Stanford collaborates with is called Sports Medicine Research and Testing Laboratory (SMRTL).
The Stanford group also has another 2,000 volunteer test in Los Angeles County underway, also through the same testing organization.
In Chicago, antibody testing has also begun, first among healthcare providers, although not part of any academic study. This phlebotomist’s report of ‘30% of those tested have coronavirus antibody‘ is an example of self-selection bias, but interesting nonetheless.
Details about the Sports Medicine and Research Testing Laboratory antibody test
The SMRTL test is a lateral-flow 15 minute test that can use a ‘pinprick’ blood-drop for sample collection, and is very simple. (A video showing the process can be found here.) The MLB was chosen due to their widespread geographic distribution across the country, diverse demographics and access to physicians to help with the administration and logistics, not insignificant in a study of this size, especially with the urgency given the situation.
Interestingly the SMRTL laboratory normally does performance-enhancing drug tests for the MLB organization; the president Dr. Daniel Eichner at SMRTL however had an unused background in viral immunology that he decided to put to use in validating an antibody lateral flow test for detection anti-SARS-CoV-2 IgG / IgM. Their laboratory ended up purchasing 15,000 of the antibody tests from Premier Biotech of Minnesota.
It is worthwhile to note this antibody test from this provider is not approved under EUA by the US FDA and is considered For Research Use Only (FRUO); it was validated as a Laboratory Developed Test (LDT) by the Utah-based SMRTL laboratory, and in addition to a group at Stanford.
The Stanford preprint study of Santa Clara
This headline, ‘Stanford study suggests far more infections in Santa Clara County than reported‘ is the result of the Stanford 2500 Santa Clara study. A total of 3,330 individuals were tested for antibody, adjusted for geographic representation (zip code), and ethnicity. The preprint itself by Bendavid and Bhattacharya et al is titled “COVID-19 Antibody Seroprevalence in Santa Clara County, California“
The senior author, Dr. Battacharya, is quoted as below in the linked MSN piece:
“I think based on our study that it is less lethal” than has been previously reported, said Bhattacharya, who calculated that it is about equal or slightly more lethal that the flu.Quote from MSN interview
“At the same time, its still quite a lethal virus. There are people getting very sick and dying from it every day.”
Most virologists and disease experts believe the coronavirus is about eight times as deadly as the flu, however.Quote from MSN interview
In Santa Clara County, there are 1700 people diagnosed with COVID-19, 66 fatalies and a population of 1.9 million, a CFR of 3.6%. However, as a result of this study the “population prevalence of COVID-19 in Santa Clara ranged from 2.49% (95CI 1.80-3.17%) to 4.16% (2.58-5.70%)”, or between 48,000 and 81,000 individuals; if the lower estimate is used 66 / 48000 = CFR of 0.14% and at the higher estimate 66 / 81000 = 0.08%. At this point, comparisons to influenza are inevitable; the CFR for influenza is estimated to be in a similar range of .04% to 0.16% (39M to 56M cases, and 24K – 62K deaths from Oct 2019 to April 2020 per CDC).
There are several valid criticisms of this work by epidemiologists, primarily cover the lack of breakdown by age or any other demographic information, as well as using Facebook advertisements for recruitment; thus people with suspected COVID-19 symptoms self-selecting, rather than a true random sampling. Epidemiologic methodology aside (“IANAE”), let’s take a look at the test itself.
Performance characteristics of the serology test
The test performance came from the manufacturer’s data, their own sample of 37 positive and 40 negative controls, and a combination in the preprint. The manufacturer provided performance characteristics of 85 confirmed positive and 371 confirmed negative samples, while the Stanford group used a locally developed ELISA assay to test their 37 positive and 40 negative controls.
Without getting into too many details (you can go ahead and dig into the paper), they derived a sensitivity of 80.3% (95 CI 27.1-87.0%) and specificity of 99.5% (95 CI 98.3-99.9%). There is a large difference between manufacturer’s data on known IgG/IgM positives from ELISA (75/75 100% for IgG known pos, 78/85 92% for IgM known pos), and Stanford’s own ELISA (25/37 only 68% not broken down by IgG or IgM). This explains why their estimate of seroprevalence ranges from 2.49% to 4.16%, or the 48,000 to 81,000 estimated for the entire county.
The specificity was much better (369/371 99.4% from manufacturer, 30/30 100% from Stanford).
While looking around for more information on where the Premier Biotech lateral flow test came from, I discovered a few documents on the McKesson website. One, the CDC Evaluation Report, can be downloaded here (PDF). It has results from 100 ‘definitive diagnosed cases’ and 150 excluded cases, with a clinical sensitivity for IgM at 93.0% and IgG at 100%.
Two new serology test approvals and a list of laboratories
In addition to the Cellex US test, this past week two more vendors of IgG / IgM tests were approved by the FDA under EUA: Chembio Diagnostic Systems of New York (PDF), and Ortho Clinical Diagnostics (with their VITROS ECi/ECiQ/3600 Immunodiagnostic System) (PDF). All the salient details from the FDA can be found on their EUA for 2019-nCoV diagnostics page here.
Also I found a list of 13 existing laboratories at this FDA page who are authorized to conduct serology testing.
Below on the same page, is a list of 97 (as of this writing) manufacturers that have notified the FDA they have validated and are offering serology tests to the market under Section IV.D.
The importance of accurate sampling, and accurate tests
The widespread testing for antibodies against SARS-CoV-2 is of utmost importance to evaluate what the true risks are, as well as understanding how widespread this disease is. With 97 vendors of antibody tests, the FDA certainly has a challenge to determine what level of sensitivity and specificity is acceptable.