Studies may not be what they seem
We need antibody tests to find the number of asymptomatic Covid-19 cases, but the German Heinsberg study was poorly done. Apparently there aren’t many good antibody tests available yet.
The early results of a small study in Germany on the town of Gangelt suggested that as many as 15% of the town might have caught an asymptomatic form of coronavirus and already had antibodies to it. This would mean that death rates to coronavirus were much lower — a mere 0.37%, not 2% (or so), and that aiming for Herd Immunity was a realistic policy. It was picked up in many newspapers and turned into headlines that may have misled a lot of people, including the Prime Minister of North Rhine-Westphalia.
“Coronavirus: Nearly 15% Europeans Now Immune in COVID-19”
“Many people may already have immunity to coronavirus, German study finds”
Five different reasons the results may be spurious
- The test may have detected antibodies to the harmless common cold form of coronavirus instead of the deadly SARS-Cov-2 type. Cross reactivity of antibodies can be ruled out through a time consuming test that seemingly was not done.
- The town infections were started in a cluster by one superspreader at a large carnival. Does that mean the virus spread first through a younger cohort? That would explain both the rapid spread and low fatality rate.
- The data is preliminary and apparently not published in full. No one can analyze it properly. E.g. What were the ages of the people that tested positive?
- It’s not a random test. The researchers selected households, not individuals. People in a household are much more likely to infect each other, and households with children will also be likely to have asymptomatic cases. Family groups may have been overrepresented compared to singles or older couples.
- Even if the antibodies are to Covid-19, that doesn’t mean a person is immune to covid-19, just that they had it.
Other estimates of asymptomatic infections suggest the unknown asymptomatic infections are not 15 times larger, but only 2 or 3 times bigger than known infections of Covid-19. An Austrian team found that only 0.33% of the population was infected about a week after the peak, and estimated just 1% of Austrians were infected. Iceland showed a similar number — around 0.3 -0.8% were infected and that was only twice the number of known infections. These latter tests are RNA tests, which can only find an active infection, not people who have been cleared of the virus. Antibody tests will be more useful (hopefully) but they are barely there, and some are unreliable.
The UK government was urged to abandon one antibody test. They may have another one ready by May. The US NIH announced they have just started antibody testing, and will hopefully release results in a week.
A Stanford Uni group says they have an antibody test working but it won’t be ready for public mass use til May.
Criticism and doubts about study from Heinsberg
Kathrin Zinkant [German language, translated by Google]
…It is probably even methodologically incorrect. Above all, it is unclear whether the antibody test used can show what it is said to have shown, namely immunity to the new virus. The Braunschweig infection epidemiologist Gérard Krause from the Helmholtz Institute for Infection Research had already told the Süddeutsche Zeitung a few days ago that there are currently no widely available tests that could reliably demonstrate immunity to Sars-CoV-2.
On Thursday after the press conference, the virologist Drosten also referred to the problem that some of the tests already available detect antibodies against completely harmless colds, which are also corona viruses. “We are now just a month after the end of the cold season,” said Drosten. According to the Helmholtz Institute, a third of the colds are caused by these four known, harmless corona viruses. After a positive antibody test, only time-consuming neutralization tests have so far shown that the person in question actually went through an infection with the new pandemic pathogen. At the PK in Gangelt, however, there was no question of such neutralization tests.
For example, Streeck’s team selected households for the tests and tested all people in these households. This is legitimate so far – however: “You should not take all the results from these households and convert them into percent, but at most one person per household,” explained epidemiologist Krause on Thursday. The reason: within the household the risk of infection is many times higher than in the general population, a complete count of all family members therefore results in an excessive percentage for immunity, which cannot be transferred to the general public. But that’s exactly what was done in Streeck’s study
Dr Simon Clarke, Associate Professor in Cellular Microbiology, University of Reading
Prof David Heymann, Professor of Infectious Disease Epidemiology
“It is important to understand the sensitivity and specificity of the serological test used in the German studies to be sure the test is not picking up antibody to other corona viruses (4 different common cold causing coronaviruses have been isolated in humans during the past years and 3 more serious ones (SARSCoV1 and 2, and MERSCoV).
Perhaps readers who understand German can help me find the data? The original link to the report has been taken down.