10 May Case Fatality ration and infection fatality ratio

Sun, 05/10/2020 - 21:13

Dear colleagues,

Kris called my attention for the paper by Streeck et al on the situation in Gangelt, a small town in Germany (12597 inhabitants) , just a few kilometers from the Dutch and Belgian border, so well within the “carnival belt of the Low Lands”.   As most of you may know in the meantime, carnival and corona are almost synonyms for us: think of cities like Tilburg epicenter in NL and Alken-Sint Truiden in BE. (I actually grew up close to Sint-Truiden and was hospitalized in the Trudo Ziekenhuis once.  A few years ago, we spent a few days in Gangelt with our  camper and last year we stayed in Tilburg.  It is strange to see all this happening … ). 

In Gangelt, there was a clear association with carnival (fig 6 p. 28 and p. 11): the chances of being infected rose by 2.5 X as well as the chances of being symptomatic: 16 % ASY carnival vs 36% in the non-carnival infected.  This is tentatively explained (but not proven) to be related with the viral load of the inoculum.

There are  several other aspects, but let’s start with “fatality rate”.  Only 7 people died “COVID-related”. Based on the PCR, it is  bit confusing, because the official number is 340 (2.7%), but the real number 421 (3.3%).  So based on PCR, the case fatality rate (CFR) would be between 1.8 and 2.3.  This is clearly lower than what is reported from China in the 3-6 % range (see second paper). But this Chinese figure does correspond with the mean of Germany: 7549 deaths in 171, 324 PCR cases or 4.4 % CFR. 

But then they also include serology and corrected it for sensitivity and specificity.  So they end up with an infection rate of at least 15 % (which is very high).  Since only 7 people died of COVID, they calculate an infection fatality rate (IFR) of less than 0.4 % !

In Belgium, we have a mortality of 8,656 in 53,081 PCR confirmed cases.  Thus CFR = 16.3 %, about 4 times the German figure and 5-6 times higher than in Gangelt.  Based on the 6% serology prevalence that Pierre Vandamme announced this week, we would not have 53,000 but rather 695,400 infected subjects in Belgium.  Hence the IFR would be 1.25 %, which is still about 3 times higher than in Gangelt. 

These figures will, of course, elicit a discussion on “who was tested” and “what is  the definition of a covid-related death”: is it the same in Belgium and Germany. But knowing the German “Gründlichkeit”, I would be surprised that they underestimate death rate so much, as compared to Belgium.  “We’ll see what happens”, once the European Commission has “normalized” all the figures….

A recent modeling exercise in the US (third attachment) comes up with a IFR for symptomatics of around 1.3 % and 1 % if you correct for a proportion of 25 % asymptomatics. 

But this Streeck paper on Gangelt is very rich and we should compare it also with the data in the paper on Vo, the Italian town in Lombardy, where the first Italian COVID death was reported.  Vo was thoroughly investigated and put under lockdown for 100 days (I heard on VRT news that now there is not a single new case anymore).  I presented this paper already on 29 April (see 4th attachment).

In Vo, they first found a prevalence of PCR (+) of 2.6 % (73 cases)  and 14 days later only 1.2 %, (29 with only 8 new cases) with remarkably 43 % asymptomatics (and no difference in VL between SY and ASY!).  But this was mainly in an older population. Unfortunately, there are no data on mortality.

Other remarkable observations in Gangelt:

  1. Table 2 p. 36 shows a list of symptoms associated with PCR+, such as loss of smell, fever, cough, sore throat, fatigue but NOT shortness of breath, other resp symptoms, nausea, vomiting and stomach pain, which were NOT different between infected and non-infected.  


  1. Households:
  • At the population level, the infection risk was not associated with the number of people in a household cluster (fig 5A), BUT
  • HOWVER in households where at least 1 person was infected, there was a significant association between household cluster size and the per-person infection risk (fig 5B):
    • In a 2 persons household: the second persons risk increased from 15.6 %(population average) to 43 %
    • In a 3 persons household: the 2nd and 3rd person from 15.6 to 35 %
    • In a 4 person household: the 2nd, 3rd, 4th, only 15.6 to 18.3 %
  • Influence of infected children in a household: increased the risk for the other persons
    • In a 3 persons household to 66 % (instead of 35)
    • In a 4 person household to 33 % (instead of 18)


  1. Influence of age, sex, comorbidities: very remarkable: NO influence on chances of infection or symptoms…


Here I add two interesting websites:

  1. Peter Piots testimony about his personal COVID 19 course: he survived, but felt VERY ILL.



  1. The Corona debate (Medische Wereld 9 May) in Dutch was very  interesting.


Best wishes,