Stanford antibody prevalence study had large rate of false-positives and motivated backers, resulting in unusually low estimate of COVID-19 case fatality rates, says whistleblower: Buzzfeed News

Coronavirus by Engin Akyurt via pixabay.com (open access)

We earlier reported on a Stanford antibody prevalence study performed in early April and published on MedRxiv on April 14.  This study, which included 3,300 people who had blood tests attempting to discern antibody to SARS-COV-2 (the agent which causes COVID-19), claimed that there were as many as 85 times as many infections as nasopharyngeal swab tests had ascertained.  The greater number of infections reduced the case fatality rate (the proportion of infected people who die) to roughly 0.1%, closer to the rate observed with seasonal influenza (which has a case fatality rate of less than 0.1%).

The results of this study are a little hard to swallow because they seem to exaggerate the number of infections somewhat.  Now, a whistleblower complains, as reported by Buzzfeed News on May 15, that the test used to detect the antibodies had a significant false-positive rate, making its estimates suspect.  With the small percentage of subjects found to have positive results, the number of false-positive tests could have been double the number of true positives.  In addition, the people who volunteered for the study likely thought they had been infected and wanted to be tested to confirm this: the recruitment information emphasized this.

The whistleblower states that the study was partially funded by the JetBlue Airways founder David Neeleman, and that he  “sought out the study authors for their congruent policy views” and funded their work accordingly.  They further state that Neeleman was in contact with the authors during the study, in part to encourage them to make conclusions that he supported (not a good way to do science).  The research wound up being cited by those on the right wing (link to NYT opinion piece) who claimed that the severity of the novel coronavirus infection was overstated and that the period of shutdown was unwarranted.  Some of those who used the study for support aired their views on Fox Business News.

If we use the study’s results more carefully, we discover that the actual number of people infected with the new virus in Santa Clara County is probably much smaller than they stated.  The study states that 1.5% of the people studied had positive antibody tests, with a sensitivity of 82.8% and specificity of 99.5%, and after adjustment it calculated that 1.2% of the study’s patients were positive.  After weighting for the population characteristics of the county, it states that approximately 2.8% (between 1.3 and 4.7%) of the people in the county had been infected with the virus.  Since there were only 1,000 confirmed cases at that time and the test implied that there had been between 14,000 and 91,000 cases, the study concluded that there had been a vast undercount of cases.

While there had certainly been a dramatic undercount of virus cases for numerous reasons, including lack of access to tests and asymptomatic infections, the conclusion that was made was an exaggeration.  Those who wanted the new virus to be less lethal jumped to the conclusion that, with the confirmed number of deaths, the death rate was as low as 0.1%.  This is wrong because of the reasons mentioned before, as well as another reason: the confirmed number of deaths was also too low.

Since that study was released, the county coroner made news by indicating that the earliest known COVID-19 death had occurred in early February.  This was based on tests conducted on autopsy cases of people who had died at home, and is, again, “only the tip of the iceberg.”  Only a few of those at-home deaths have even been evaluated for the virus.  Some of those who died were thought to have had “obvious” heart attacks or strokes, but we now know that SARS-COV-2 can actually bring on these terminal events.

A study of the Spanish population was recently released, as reported in El Pais on May 14.  This study (with nearly 70,000 people tested) indicates that, overall, about 5% of the population of Spain, or 2.3 million people, have been infected with SARS-COV-2.  The antibody prevalence ranged from 14.2% in Soria province, to 11.3% in Madrid, down to less than 2% in isolated regions.  Since the confirmed number of cases was only about 230,000 (as of the date of the article), this means that 90% of the cases went uncounted.  Based on the antibody percentage, the actual death rate, with 27,000 deaths, is between 1 and 1.2%.

In truth, we have seriously undercounted both the number of cases (by a factor of ten or more, not a factor of 50-85 times) and the number of deaths (by an unknown factor, but possibly less than double).  Putting all this together, we can estimate that the real death rate is between 0.4 and 1.2%– less than the 6% previously suggested, but certainly far more than the rate we find with seasonal influenza (less than 0.1%).

There’s a lot of uncertainty surrounding the number of people who have had the virus, and some people are exploiting the uncertainty for political reasons.  Those who want the death rate to be low so that we can stop “social distancing” and “safer at home” policies and “jump start” the economy are probably wrong.  Most Americans sense this and are justifiably afraid to venture forth without more information about who has the virus and what we can do to avoid getting sick.  We hope that the delusional optimists (who seem to be on the right-wing end of the political spectrum) do not consolidate control of the country.