Der Heinsberg-Effekt
Pioneering German blood study: no grounds for celebration.
Knowing what we know now — that the SARS-Cov-2 virus was a secret participant — scenes from the Heinsberg Carnival 2020 unspool with Hollywood horror.
Carnival is a big deal in North Rhine-Westphalia, and has been since the Middle Ages. It typically starts the Thursday before Rose Monday at 11:11 (the time of day of the Armistice on 11/11/1918 for World War I) and lasts to Ash Wednesday.
Carnival is declared open by the three three principle Carnival characters: the maiden, prince and peasant. The Thursday before Shrove Tuesday is ‘Weiberfastnacht,’ Women’s Carnival Day.
The Heinsberg Carnival started Friday 14 February. Most locations hold a dance or ball opening night. In Heinsberg, 350 adults in fancy dress locked arms on long wooden benches and swayed to a live band. People intertwined arms and drank toasts in beer, sometime sipping from the same stein. A waiter with a sniffle rushed food and mugs to the tables. There was much communal singing.
In the street during the day, close-packed crowds watched marching bands, processions and floats, from any number of close-knit carnival societies. There was a lot of kissing. One Rhineland tradition is the Bützchen, a peck on the cheek. Others said to heck with tradition, and kissed on the lips
A few days after Carnival ended, people began getting sick.
In the movie, there’s a period of suspense between the time the first mysterious reports come in, and when the scientists and doctors finally puzzle out what happened.
In Heinsberg this didn’t take long. City Hospital in Heinsberg saw their first Covid patients at the end of February.
On 28 February, a Bonn university student, who works at a school in Poppelsdorf looking after children in the afternoon, was diagnosed. He had celebrated carnival in Heinsberg.
In Düsseldorf, a middle-aged woman came down sick after attending a seminar. The woman sitting next to her at the seminar table had been sweaty, flush and seemed to have a high fever. She had mentioned she had been to Heinsberg for Carnival.
Carnival should have been innocent fun; instead it was a calamity. In the lexicon to which we have sadly grown accustomed, Heinsberg Carnival was a super-spreading event.
By the end of March, Carnival’s toll was clear: among the 250,000 residents of Heinsberg, there were then 1,281 confirmed infections, 34 deaths, and 550 recovered cases.
There have been other such “cluster events” around the world. Most similar to Heinsberg was Mardi Gras in New Orleans, also in February.
Most superspreader events involve large public crowds; loud speech or cheering; sometimes communal singing; sometimes dancing; sometimes kissing; sometimes confined spaces.
Heinsberg was a petri dish in which the virus replicated and spread uncontrolled. Gangelt, whose Old Town had been a center of the celebrations, was put under lock-down 26 February.
It was a unique opportunity for a study.
This was the idea of Prof. Hendrik Streeck, head virologist at the University of Bonn and its University Hospital.
A huge advantage in this study was that the date of exposure, the Carnival, was known. Approvals went quickly. By Tuesday, 31 March, the “Covid-19 Case Cluster Study” was announced and underway.
It was a first for its time in that it would also be a serosurvey — it would study the blood of the participants for antibodies to the SARS-CoV-2 virus.
“Testing” was then a crisis. Public health officials were fixated on discovering currently infected and contagious individuals, and so relying on PCR tests to find live virus. Blood antibody tests were for later.
Residents of Gangelt who had falled sick before the study commenced had, of course, been PCR tested. The peak in confirmed cases occurred around 13 March, three weeks after Carnival.
Prof. Streeck and University Hospital Bonn sent a letter to some 600 households in the area, of which 400 agreed to participate, with parents consenting for children. All got swab tests and had blood dawn.
Six weeks after Carnival, 33 of 1,007 still tested PCR positive. Only 4 had been PCR tested before.
The antibody test was by an ELISA with a validated specificity of 99%.
But the most overlooked tool used by Prof. Streeck and his team was the lowly questionnaire. The researchers asked their subjects about their lives — at home, in the workplace, kindergarten, or hospital —topics in which most number-focused studies have been disinterested.
Months on, the need for “contact tracing” would become a mantra; the Heinsberg study showed one way it could be done.
The final report was published 4 May, 2020. [Universität Bonn press release, in English, here. The report itself, “Infection fatality rate of SARS-CoV-2 infection in a German community with a super-spreading event,” in English, as a PDF.]
The headline number was that 14% of the 919 tested were positive for IgG in their blood.
At the time it was first reported, in a preliminary report 9 April, that number seemed shockingly high.
With results in the 20%–30% range since coming in from New York and Boston, it is less so now.
Based on the number of deaths in Heinsberg, Prof. Streeck computed the Infection Fatality Rate (IFR) of Covid-19 there at 0.37%.
No one questioned his math. But whether that rate can be applied elsewhere is controversial. We’ll get to that.
More important findings
The media controversy over the IFR, unfortunately, crowded out many other fascinating findings in the study.
22.2% asymptomatic
Some 22.2% of those who tested antibody positive were asymptomatic for all six weeks. (For nuances of the word asymptomatic, see my Covid With Style: A Writer’s Guide #1 here).
This finding puts a number on a phenomenon that is large in implication. (It also jives with a figure of 17% later reported by Institut Pasteur (summarized here) of a French lysée students and family members. More on this later as well.) But first, and very important:
asymptomatic infected individuals in our study present with substantial antibody titers. [levels]
On the subject of families, the Bonn researchers found the rate of secondary infection among family members was lower than expected. The risk of infection in a two-person household in Gangelt was only 44% if a partner was infected. Still, that means on average of more than four out of ten people become infected in such a two-person scenario — even if both know the virus is present.
Other findings about transmission were less surprising. Participation in Carnival festivities increased both the infection rate (21.3% vs. 9.5%) and the severity of symptoms. Currently well-known, but once unremarked, symptoms such as loss of smell and taste were reported.
Several non-correlations are worth pointing out:
- Sex and age were not associated with the rate of infection.
- Rate of infection not associated with self-reported use of ibuprofen and similar.
- Schools were not a center of infection.
- The researchers carried out an intensive search of the home of a family infected with the coronavirus but found no trace of virus on surfaces. Prof. Streeck is skeptical that it can be transmitted by doorknobs. (But I want to ask him about elevator buttons…)
The IFR debate, encore
The Infection Fatality Rate (IFR) is a number important in modeling the epidemic and in shaping public policy.
Its importance for public policy is easier to see in caricature. To look at the extreme views:
- If the IFR is very low, Covid-19 is “no worse than influenza.” We should no more shut down the economy for Covid than we do for seasonal flu.
- If the IFR is very high, Covid-19 is up there with Marburg, Ebola and Zika, we should drop everything and stop it before everybody dies.
In the U.S, the “low IFR” position has become a center-right position associated with Stanford’s Hoover Institution (a conservative thinktank) and the pro-business editorial pages of Wall Street Journal.
The Stanford researchers in a serosurvey of Santa Clara County, California, were almost certainly guilty of pushing their data too far in support of their views, a species of confirmation bias. To abridge Wikipedia’s definition:
the tendency to search for or interpret information in a way that confirms or strengthens one’s prior personal beliefs or hypotheses.
Opposition to this position by proponents of social distancing and those willing to accept lockdown has been vehement and emotional.
Since I quoted Wikipedia about the Stanford researchers, let’s be fair and quote it about them:
Attitude polarization, also known as belief polarization, is a phenomenon in which a disagreement becomes more extreme as the different parties consider evidence on the issue. … When people encounter ambiguous evidence, this bias can potentially result in each of them interpreting it as in support of their existing attitudes, widening rather than narrowing the disagreement between them.
A denominator at last
The IFR is a denominator static. (Wikipedia’s on a roll: Look here.) The IFR results from a simple division:
IFR = number of deaths / number of infections
In the early days of the pandemic, the denominator was missing, leaving advocates free to guess one that favored their arguments.
The numerator, the number of deaths from Covid-19 in a jurisdiction, is generally known and relatively firm (with all the usual caveats).
But the number of infections has been, and still is, something of a mystery, since we know this disease can be entirely asymptomatic.
A serosurvey, at this date on the calendar, offers a simple operative definition. If you have blood antibodies to SARS-Cov-2, you have had the disease. Whether you knew it or not. You count in total infected.
Knowing the total infected, we now have what we need to compute an IFR.
Serosurveys for antibodies will also pick up people who are still infected and not fully recovered. Note that this changes the IFR only if they subsequently die. One person in the Heinsberg study did in the weeks between the preliminary and final report. The small-numbers effect was such as this single death kicked up the IFR:
The inclusion of this additional death would bring up the IFR from 0.36% to an estimated 0.41% [0.33%; 0.52%].
The survey probably under-sampled children. It did not include any people in retirement centers or nursing homes, where death rates have been extremely high.
The IFR scale
Unlike the Case Fatality Rate (CFR), which depends on the state of testing, treatment, and hospital infrastructure, the IFR is considered by Prof. Streeck to be more a property of the virus. He believes it can be used to back-calculate the number of infected elsewhere.
Fatalities do depended on demographics, so the IFR in Italy, where the population is older, will be higher than where the population is younger.
Prof. Streeck got into media trouble when he announced the Heinsberg IFR, Critics felt the 0.39% fatality rate was too low, or at least could not be applied to all of Germany. If it were, it implied over 1.8 million Germans were infected, or had been in the recent past.
When estimates are all over the place, I like to arrange them low to high (then throw a dart). The real problem is that there is no single, global valid IFR. But here is the range:
- Typical seasonal flu (Source, CDC). Deaths per 100,000 population 2.0 = 0.02.
- Stanford Santa Clara County study, between 0.1%–0.2%.
- Oxford CEBM, between 0.1%–0.26%, with many caveats pertaining thereto.
- Prof. Streeck’s Heinsberg study, 0.41%, with a possible range between 0.33% and 0.52%.
- University of California, Berkeley, and Lawrence Berkeley National Laboratory data scientists, looking at New York City and Santa Clara County data, “no less than 0.5%”
- Number currently favored by Neil Ferguson, of the now-infamous Oxford Model: 0.8–0.9%
- My quick calculation for New York City (all five boroughs), based on New York State’s 15,000 person semi-random serosurvey: 1.15%
- And now we get to some real caveats. Gianluca Rinaldi, Matteo Paradisi in “An empirical estimate of the infection fatality rate of COVID-19 from the first Italian outbreak,” here:
We estimate an overall infection fatality rate of 1.29% as well as large differences by age, with a low infection fatality rate of 0.05% for under 60 year old and a substantially higher 4.25% for people above 60 years of age.
COVID-19 kills the weakest segments of the population.
Prof. Streeck has wisely decided to step back from the public policy debate. He gave us the IFR number his study came up with. “What you make out this of policy-wise, that’s a question for the politicians,” he told Freddie Sayers of unherd last Monday.
Not an optimist
Prof. Streeck is a virologist, not a professional optimist. He believes it most likely that SARC-Cov-2 will become endemic, that is, always be with us, and the human species will need to learn how to live with it.
Vaccine: maybe not
Prof. Streeck has been involved in HIV research for years. Which has, infamously, been unable to come up with a vaccine, despite four decades of research. He questions the faith that a vaccine will be developed for SARS-Cov-2 at all, let alone in 12 months.
No other coronavirus has a proven vaccine. This list of viruses that have resisted vaccines despite decades of research:
- HIV. In 1984, the US Secretary of Health and Human Services Margaret Heckler predicted that a preventative vaccine would be ready for testing in two years. Four decades and 32 million deaths later, the world is still waiting.
- Dengue fever, which infects as many as 400,000 people a year. For dengue, feline coronaviruses, and HIV, vaccines can actually make things worse, through a process known as antibody-dependent enhancement (ADE). In this, the viruses take advantage of anti-viral humoral immune responses.
- rhinoviruses and adenoviruses . The common cold awaits its cure.
- porcine respiratory CoV (PRCV) and avian infectious bronchitis CoV (IBV). The vaccine news from the animal kingdom on coronaviruses is not great.
Plan B
Prof. Streeck has a hypothesis about a possible Plan B, based on his observations in Heinsberg, that we might learn to live with the virus in a weakened form.
He is very careful to call this a hypothesis, which needs testing, not a policy prescription.
It’s based on the 22% who developed high levels of antibodies for SARS-Cov-2 and never realized they had the disease.
If requires several things to be true. One is that these people developed asymptomatic cases because they received low doses of the virus, or received weakened strains that had passed through other people.
The second is that traditional hygiene measures, such as hand-washing, will reduce the virulence of the virus — thereby throttling down its most severe outcomes — without completely eliminating its infectiveness.
Under these circumstances, we would in effect be living a natural experiment in variolation, where we were likely to get mildly infected, but that would be okay.
Monday’s interview with Prof. Streeck is well worth watching, if you have time.
Follow this blog on Twitter @Will_Bates_sci
