How far away is the herd immunity?
During the pandemic, herd immunity has become a hot topic. Critics claim that the Swedish strategy is to allow many to become infected so that the spread of infection stops when enough people eventually become immune. The Swedish Public Health Agency and the Swedish government have rejected this.
However, it is clear that the Swedish Public Health Agency has taken an interest in the population's immunity. Herd immunity has been mentioned at press conferences and in interviews with representatives of the Swedish Public Health Agency. Based on mathematical modeling of the spread of infection, they have tried to predict when it will stop.
March 2020, beginning of pandemic
On March 16, state epidemiologist Anders Tegnell mentioned that six million Swedes could become infected during 6-12 months, of which maybe 1% would get really sick and 5-6% of them would need hospital care. In numbers, that would mean 60,000 "really sick" and 3,000-3,600 in hospital. In this context, it can be mentioned that the fatality rate is in the range 0.5-1%, which was known already in March.
On March 23, Anders Tegnell said that the infection curve had begun to increase and that he expected the culmination to be a month away. However, he did not think it was important to know how many in the country were infected.
"- No, that number is so extremely difficult to verify. Then you should basically test the whole population. And this every three days. I do not know what to use it for. The important thing is to know where we are on the curve so that healthcare has a chance to prepare. "
This was a week after Norway, Denmark and Finland had introduced strict contact restrictions (lockdown) to reduce the spread of infection. But in Sweden, the Swedish Public Health Agency expected the spread of infection to continue to increase, even in regions where it was still very low.
Models by Tom Britton
In March and early April, Tom Britton began modeling the spread of infections in Sweden, especially in Stockholm. Britton is a professor of mathematical statistics at Stockholm University and it later became known that he had contact with FHM in March. According to Britton's calculations on April 1, more than half of Sweden's population would be infected before the end of April, and about two thirds before the end of May. The calculations were based on the assumption that between half a million and one million Swedes were infected at the beginning of April, but he admitted that it could be half as many, without explaining that it would mean fewer infections overall and consequently lower immunity. He later modified his statement and said on April 19 that he expected half of Stockholm, not Sweden, to be infected by mid-May, and that the peak for the spread of infection had passed.
On April 18, Britton published a preprint, which is a manuscript for a scientific study that has not been reviewed. The preprint described Britton's modeling of the spread of infections in Stockholm with different scenarios for different strictness of measures. According to his main scenario, approximately 1,100,000 people in Stockholm would be infected before the beginning of June, when the spread of the infection would stop. The modeling was based on the number of deaths up to 31 March (150) and the assumption that the mortality (infection fatality rate, i.e. the proportion of those infected who die) was 0.3% for covid-19. Based on this, Britton estimated the number of infections three weeks earlier and fed into the model.
However, Britton did not mention in the report or in the media how many deaths could be expected if the modeling was correct. Simple multiplication gives 3,300 deaths in Stockholm based on his assumptions. In addition, the mortality was significantly higher according to the available data. Fergusson and his co-authors at Imperial College London had estimated the mortality rate at 0.9% at the UK population level based on data from the cruise ship Diamond Princess, which would result in three times as many deaths. Mortality is now considered to be between 0.5-1% depending on the age structure of the population and access to healthcare. Britton has later said that he regrets his calculations about herd immunity.
Models by the Swedish Public Health Agency
The Swedish Public Health Agency were also working on models of the spread of infections in Stockholm. On April 21, they presented the results of the modeling at a press conference. However, there were obvious errors such as that 6 million people in Stockholm had already been infected and the Public Health Agency withdrew the report. Two days later, on April 23, an updated report came from the Swedish Public Health Agency. According to the modeling, approximately 26% of the population in the Stockholm region would be or have been infected with covid-19 on 1 May and the peak for the spread of infection would have been sometime 8-11 April.
The Public Health Agency's modeling was based on confirmed covid-19 cases in combination with a random survey of the number of infections at the turn of March-April. The survey showed that 2.5% of Stockholm's residents were infected at the time, i.e. had a positive PCR test. The Swedish Public Health Agency used that figure to estimate the number of new infections each day. To do so, they needed to know how long PCR tests showed positive, which they assumed was 5 days. This was soon questioned by Adam Kucharski, a mathematician and epidemiologist at the London School of Hygiene & Tropical Medicine, who pointed out that PCR tests give positive answer on average for 2 weeks. This was later confirmed by Jan Albert, professor of microbiology and infection control and Joakim Dillner, professor of infection epidemiology. With longer time, the number of infected people would spread over more days, and the number of daily infections would decrease. This in turn would mean that fewer people had been infected than what the Public Health Agency had concluded.
The Swedish Public Health Agency seems to have trusted its results. On April 16, Anders Tegnell told the Norwegian TV channel NRK.
"- Our mathematical models indicate that there may be herd immunity in Stockholm during May."
"- We had expected an upturn after the Easter weekend, but we do not see it. We are still at the level we did a couple of weeks ago."
However, Tegnell also said that the positive development with a decreasing spread of infection was dependent on the population continuing to follow the authorities' recommendations.
In another interview on April 16, Anders Tegnell said that the spread of infections was at a level where herd immunity began to have an effect. He explained that the effect begins already at 10-20% immunity but that around 50% of the population must be immune before the spread of the infection would change to a steady state where each infected person infects one person on average (R = 1). He also said he thought it would happen before a vaccine was available.
"-I would probably bet a penny that we will reach that level of immunity before a vaccine is available."
However, he did not want to comment on how many could be expected to die. At that time, 1,100 people had died of covid-19 in Stockholm.
"- We have deliberately not counted on the number of deaths because we hope that our measures will reduce that group as much as possible."
Also on April 19, Tegnell said that the Public Health Agency saw a certain slowing of severe cases and number of deaths.
"- It is likely that herd immunity has an increasing effect."
Lower herd immunity threshold or social distancing?
On May 9, Tom Britton went public with results from a new modeling study that at that time had not been peer-reviewed. The new study was about the spread of infections in a scenario when parts of the population are more isolated and other parts have more contact with each other. From these assumptions, Britton concluded that the spread of infections could stop at a lower rate of immunity than if everyone had equal contact with each other. Britton and his colleagues later published the study as a peer-reviewed scientific article. Based on the study, Britton said that this would mean that the spread of infection in Stockholm would stop in mid-June. However, this was not included in the published article.
Antibody results
On May 20, the Swedish Public Health Agency presented at a press conference its first results for antibodies to covid-19 virus. The samples came from outpatient care and showed that a lower proportion than previously predicted had antibodies.
- 7.3% in Stockholm
- 4.2% in Scania
- 3.7% in Västra Götaland.
However, the public health authority said that the result was in line with their modeling with the explanation that it would take 3-4 weeks before the body's immune system develops antibodies. This has later been questioned by the immunologist and science writer Henrik Brändén who said that according to basic immunology it takes a week to ten days from symptoms of an infection until large amounts of antibodies flood into the blood, and pointed out that the Public Health Agency also gave the same information in a compilation of current knowledge about serology and covid-19. Based on that, 7.3% with antibodies in Stockholm week 18 would correspond to those who had been infected in week 16 (April 13-19).
The Swedish Public Health Agency had difficulties explaining how those levels of antibodies in Stockholm could have any effect on the spread of infection. On May 25, they said at the press conference that the sample (patients in outpatient care) was not representative. The public health agency would therefore try different types of selection and present new figures. However, no such figures have ever been presented and according to the Swedish Public Health Agency's latest report, the figures were on a par with those presented at the first press conference. The Swedish Public Health Agency's antibody tests among blood donors have also yielded similar results.
Immunity from t-cells
At the end of June, researchers at Karolinska Institutet published results from a study of t-cell immunity, which included 200 people. The study was published as a preprint on June 29 and as a peer-reviewed scientific article on August 17. The researchers said that
"- Our results indicate that the immunity to covid-19 is likely to be significantly greater in society than antibody tests have shown. If this is the case, it is of course very good from a contagion perspective."
The Swedish Public Health Agency viewed their results positively and Director General Johan Carlson said on June 30.
"- It fits in well with the forecasts and modeling we have done about how much spread there should be in the Stockholm area at this time of year, so in that way it has been frustrating that the antibody levels have been so low. But this gives a complementary results in a good way. "
Increasing immunity, at least in FHM's heads
At the Public Health Agency's press conference on 16 July, Johan Carlson said that up to 40% of Stockholmers could be immune to covid-19. According to Carlson, 17.5-20% of Stockholm's residents had antibodies and about 20% t-cell immunity. It is unclear where the figure for antibodies came from. Week 24 (8-14 June), the proportion with antibodies was just under 12% both among outpatients and blood donors. There are no other reports of antibody tests in the whole of Stockholm, and the spread of infection had decreased since the spring, so a doubling is unrealistic. Carlson's statement was questioned by Gunnveig Grødeland, a researcher in influenza and adaptive immunity at the University of Oslo, who said that the proportion with antibodies could be correct with regard to antibody tests in New York and Bergamo, but that the figure for T cells was uncertain and that it was not clarified whether T cells induce immunity. Carlson maintained, however, that 17-20% had antibodies and that at best twice as many could be immune.
At the end of July, Anders Tegnell said that he hoped that the worst period was over in Sweden.
"- Considering what we have seen so far and think we know about how many are immune in society, it must be said that it is very unlikely that we would return to the situation we had in the spring."
The Swedish Public Health Agency had then developed three scenarios for the spread of infections in the coming year, and Tegnell obviously considered that the immunity of the population was high enough that the spread of the infection would not increase again.
"- What we see are these local outbreaks in workplaces and in regions that have not been so widespread before."
FHM has still not let go of the idea that many have been infected. At the end of August, they said that 20% of Sweden's population had been infected. At the press conference on October 1, they said that they stand by their assessment that 20% in Sweden have been infected (at 23:15), but they have no evidence to show that.
What did FHM know and when?
But what did the Public Health Agency's modeling really show? As mentioned above, they were questioned from several parties because the estimate of the number infected from the sample surveys was based on the assumption that PCR testing is positive for 5 days.
The Swedish Public Health Agency also became aware of this. In June, they published a report on covid-19 fatality. According to the report, the infection fatality rate in Stockholm was 0.6%. The estimate was based, among other things, on the sample survey of infected people at the turn of the month March-April, but with the assumption that PCR tests show a positive 10 days in median. The report also showed estimated mortality in a 5-day positive PCR test, which would be around half as high as the number of infected would be much greater.
The report was published on 16 June, but the Swedish Public Health Agency informed the media as early as 15 May about the same estimate of fatality. Thus, they had already changed to assume that the PCR tests show positive for 10 days. In a report on 1 July, the Swedish Public Health Agency presented new modeling of the spread of infections in Dalarna, Skåne, Stockholm and Västra Götaland based on a 10-day positive PCR test. With the new modeling, the proportion of those infected in Stockholm was estimated to be higher
- 13% May 1
- 19% July 1
- 20% September 1
This is more in line with the fact that about 7% in Stockholm had been infected on 19 April according to the antibody tests. Recent studies of antibodies have not shown any major increase, however, and it is unclear whether this is due to the models overestimating the number of infections or whether the antibodies decrease over time.
Summary
My overall impression is that the Swedish Public Health Agency has chosen to emphasize things that have supported their view that the immunity of the population has been high and that it would soon reach levels where the spread of infections would stop. It has been based on assumptions that have not been in line with current knowledge, partly about how long PCR tests show positive and partly about the fatality rate of covid-19. Tom Britton's first modeling was based on a fatality rate of 0.3%. It has later emerged that Britton had contact with the Public Health Agency in March when he made his first modeling and discussed a mortality rate between 0.1-0.3%. It did not match the data that existed then and it does not match what is known now. The Swedish Public Health Agency already knew when they presented their first antibody studies that their modeling was incorrect. Still, they maintained that the immunity of the population was high, even after updating their own modeling.
