Is COVID-19 more dangerous than typical annual coronavirus variants?

Question

I have heard many people claim in private that COVID-19 is not worse than flu and that the reaction is overblown.

As a poster-child example of this, I want to analyze the claims made by the German pulmonologist Dr. Wolfgang Wodarg in this video.

There are other places (Wodarg's personal website, interview, online newspaper) where he repeats and refines his claims but they don't come with English subtitles so I want to focus on the first video.

The overarching claim is that there is no evidence that we are seeing unusual deaths due to COVID-19. It is just that we started measuring things we did not do before, and haven't compared them to a baseline of illness and mortality, and have blown the data we have out of proportion. However, this can be broken down into more easily verifiable subclaims:

1. the lab in China identified a novel corona virus in a small sample of pneumonia patients (<50) of whom there are always plenty in Wuhan (with 11 million people) and put it into "the virus database" (whatever that is)
2. the German virologist Christian Drosten developed a test for this virus which was "rushed to market" in China without proper validation so we don't actually know if it tests what it should
3. we don't know the baseline for coronaviruses in pneumonia related deaths so we can't establish whether COVID-19 actually increased the rates: From the cited Glasgow study we would expect around 5-15% of flu season deaths to test positive for a coronavirus in an average flu season in Germany. That would be 3000 deaths which is more than 100x more than the reported COVID-19 death toll in Germany so far
4. just because someone tests positive for SARS-CoV-2 and died of pneumonia does not mean that they died because of SARS-CoV-2.

Are these claims true?

Answer 1

I'd like to tackle a few sets of claims/arguments Wodarg makes in the two videos:

1. (Unknown) baseline prevalence of corona virus infections.
2. Excess mortality / "we wouldn't see anything special"
3. Comparison with Swine Flu when Wodarg was public health officer

---

(Unknown) baseline prevalence of CoV

• The Glasgow paper he refers to did not look at the general population but at patients showing up in medical practices. They were monitoring 3 particular coronaviruses.
  The 7–15% coronaviruses are presented in an ambiguous way: this is not the prevalence of coronaviruses among the patient population but the relative prevalence, i.e. among those samples where viruses were found. The prevalence among the patient population is only about 1/4 of that since only in ≈ 1/4 of the samples one or more of the monitored viruses were found. So those 3 coronaviruses had a prevalence of ≈ 2–4% among the population that showed up at a medical practice with acute respiratory disease.

• Update Mar 20 This review on the common cold does cite 3 studies saying 10–15% are caused by coronaviruses, though. Looks as if the claim may be true after all, but the data he showed did not support the claim.

• He claims that during "flu season" we monitor influenza viruses, but not others. That is only partially correct, as e.g. in Germany the RKI monitors viruses from 5 groups (including influenza and rhinoviruses). Among them were no corona viruses until they started with SARS-CoV-2 in week 8 this year.

• Website at RKI about monitoring of the flu

• current (11th week) weekly report
  In week 11, they had one sample that was positive for SARS-CoV-2 and also for Influenza.

• Wodarg is right when he says that this year, we have a very mild flu season.

• Wodarg makes an argument that one would expect increasing prevalence for 'coronaviruses in general population' < 'people who show up in medical practice' < 'people who are in the hospital'.
  While this may be true, it also has the implied assumption that: either are coronaviruses associated with severe disease (i.e. existing disease helps coronavirus) or coronaviruses are in fact to be taken seriously in the sense that they can make people so sick that they have to go to the hospital.

Excess mortality / "we wouldn't see anything special"

Wodarg refers to euromomo as a service where we can see whether there's excess mortality, e.g. due to acute respiratory illness. He argues that if we weren't monitoring SARS-CoV-2 now, no-one would realize that there is something unusual (and in the second video, the reporter asks "If we wouldn't test, this would not catch anyones eye?" and he confirms "No, we wouldn't realize it.").

Testing bias can distort not only the public uptake of a situation but also statistics, see e.g. lead time bias.
But with COVID-19, the direction was the other way round: an outbreak of severe pneumonia happened (so people did notice without testing for the virus), and the cause was not known. Now, Wodarg argues that Wuhan is an important virological center in China. Under these circumstances, unknown cause translates to: it wasn't any of the usual suspects (viruses, bacteria, substances). Then people began to search, and found the virus. And we still see severe pneumonia with outbreaks of this virus, see Italy.

I don't think this is the case, though. It is just that SARS-CoV-2 deaths lag behind positive tests.
I'd have preferred to work with seasonal mortality data for Italy and Hubei, but I didn't find any, so I use the European data at euromomo.

• today (Mar 18th) the latest euromomo data is from week 10, i.e. until March 8th, 1.5 weeks ago. In terms SARS-CoV-2 in Europe, in Italy the number of positively tested cases have since increased by a factor of 4.8, in Germany by a factor of almost 11. (numbers from worldometers).

I took the liberty to

• add SARS-CoV-2 positive deaths from Italy in blue: in week 10, 325 SARS-CoV-2 positive patients in Italy died. That's just a bit more than half a pixel in the graph (and a bit fewer than the average weekly traffic deaths in Italy according to the EU traffic death statistic). In week 11, the increase should start to be visible.
  As Wodarg says, this year's flu season was very mild, and the graph shows fewer deaths than usual in this season. (However, we may have to be a bit cautious with this as some other graphs on the web page indicate that numbers may not be complete yet)

• project the Hubei deaths in orange into the euromomo graph as follows: Hubei has 57 mio inhabitants, euromomo monitors 368 mio. (inhabitant numbers from wikipedia), that is a factor of ≈ 6.4. I counted the deaths using worldometers and the COVID-19 daily deaths by region diagram on Wikipedia, multiplied them by 368/57 to "simulate" the same kind of infection intensity as in Hubei for Europe and drew them on top of the red seasonal average. This "bump" sticks out of the +4z line for substantial excess mortality.

  The yellow line is when the quarantine in Wuhan/Hubei started.

  So the mortality bump of the same epidemic as in Hubei happening throughout Europe would have made a noticeable bump (at least unless the epidemic reduces mortality from other causes, such as reduced traffic mortality in quarantine situations), even though it is somewhat smaller than this year's mild flu season. Note though that the death bump is completely behind the quarantine start date – the mode is about 3 weeks later. But this is with severe restrictions. The argument that nothing special is happening and no particular measures are needed is on rather shaky ground.

We know that SARS-CoV-2 has an incubation period of about 5 (1 - 14) days during which it is also already contagious. Median time from onset of symptoms to death was found to be 2 weeks (6 - 41 days). Mortality data will thus lag about 3 weeks after infection.
In the last 3 weeks, the SARS-CoV-2 positive cases went up by a factor of 100 in Italy, x520 in Germany, and x120 in the US. Note that testing is performed in a population that is at high risk of having contracted SARS-CoV-2. Note that is argumentation doesn't even include any time between death until it appears in the official statistics. If public health officers don't have faster access to the mortality data, using euromomo would mean another 1 1/2 week's delay (3 days for the weekly flu report of RKI).

Excess mortality is an important instrument, but it is delayed and therefore IMHO too slow for diseases with the epidemiological characteristics we face here.

Wodarg argues that a typical heavy flu season in Germany causes maybe 20000 deaths, and due to the prevalence of coronaviruses (see above) somewhere around 2000–3000 coronavirus deaths in a season would be normal – which is far more than what we see with SARS-CoV-2.

As I explained above, the paper actually found prevalences that are only a quarter of what Wodarg claims, so the "threshold death toll" should rather be 500–750 deaths. See update above.
Which again, we don't have in Germany.

The situation in Italy looks different, though (Wodarg evades answers when the reporter asks about Italy, and what he says is self-contradictory): While Italy has a smaller population than Germany, their annual excess mortality in the flu season is estimated to be in the same magnitude – which, like the SARS-CoV-2 mortality now is attributed to the large proportion of old people.
Worldometer right now (Mar 20th) reports >4000 SARS-CoV-2 positive deaths in Italy so far (and there is no indication that this is over in Italy).

Comparison with Swine Flu when Wodarg was public health officer

Wodarg compares the situation with the pig flu and avian flu when he was working as public health officer where "nothing happened".

• IMHO he is right when he says that the decisions to curb rights and shut down economy must be taken very seriously and that panic hampers good decision taking.

• I also agree with him that a positive-feedback loop in media reporting may increase panic and hinder rational decisions.

• He mentions that the definition of pandemic changed from a disease that causes severe illness spreading over the whole world to a disease spreading over the whole world.
  However, that point is moot here: SARS-CoV-2 causes Severe Acute Respiratory Syndrome (severe pneumonia) so the more restrictive definitions would apply as well.

• Also the comparison with the flu decisions he faced as public health officer is IMHO not quite as easy as he makes it out: Influenza has a much shorter incubation period of about 2 (1–4) days. This means for the public health decision perspective, that dark count of cases in incubation period is much less of a problem for influenza than for SARS-CoV-2. This does make a huge difference for public health decisions.

  Also the duration after onset of symptoms is shorter: usually 1 week for the flu, but 2+ weeks for COVID-19. Again, a longer sick period means a heavier burden to the health system: if we have a capacity to care about n people in hospital, with an average stay of 3 weeks, only n/3 cases per week can be taken care of. With 2 weeks in hospital, that's n/2 per week or 50% more.
  I didn't find statistics on time between onset of symptoms and death for influenza, so I can not comment whether reported mortality rates are better suitable for reacting to flu epidemics due to flu patients dying fast.

---

Update (Mar 19th): I checked euromomo again since it said yesterday that the weekly updates are Thursday around noon. Week 11 also shows considerably lower mortality than expected, but they have a notice about COVID-19 now, including:

[…] there is always a few weeks of delay in death registration and reporting. Hence, the EuroMOMO mortality figures for the most recent weeks must be interpreted with caution. Therefore, although no increased mortality is currently observed in the EuroMOMO figures, it does not rule out that increased mortality occur in some areas or in some age groups, including mortality related to COVID-19.

The per-country graphs are unfortunately barely readable, but as the color changes where they try to correct for delay in reporting of deaths, we can see that actually death notifications from Italy are incomplete for the last 6 weeks or so.

What is correction for delay in reporting? Countries have varying deadlines to report/register deaths at the local inhabitant register, these registers may report only so often to the province register -> country -> euromomo). So this may take a while, and unless the deadlines are synchronized (which AFAIK they aren't), this means that death numbers may "trickle in" at euromomo over a number of weeks. Until this is all done, the reported number of deaths will be too low for the last n weeks. If n isn't negligible, we'd expect the last data points to almost always show fewer deaths than normal. However, over time, euromomo can gain experience what percentage of deaths for a particular country are usually reported to them within how many weeks. That allows to apply a preliminary correction under the assumption that the reporting is as usual (which I doubt at the moment e.g. for Italy). For the question here however, the actual correction is not so important since the absolute numbers do not indicate such a correction, and I used this only as an indicator how long it may take until the final curve is available.

Update Mar 20: The Wodarg videos meanwhile made it into mainstream media in Germany. Tagesschau Spiegel Zeit/dpa merkur (Bild cites Wodargs without any comments)

(While I don't find all the lines of argumentation compelling, I think they are worth reading.)

---

Update Apr 2: Euromomo data for week 13 is out:

• the low after the flu season doesn't go below the expected seasonal average any more. The final downtrend we saw before was thus an indication of reporting delay.

• We now see that in week 12 + 13 the overal mortality across all ages and all participating countries has been very unusually high, approximately +8z from the seasonal average.

  A closer inspection of the per-country deviation from the seasonal average shows that e.g. for Italy this peak has a size now like the flu season early 2017 (which, however, went away without lockdown).

• The excess mortality maps now show high mortality for Italy in week 11 (and very high for week 12 + 13), Spain has also very high excess mortality for week 12 + 13 (Belgium, France, England have also shown up in week 12 or 13).

Answer 2

We can be extremely confident that a novel virus is spreading and being correctly detected by our tests

I'm only going to address claim 2, because the other claims are likely correct but essentially irrelevant because if claim 2 is false the chain of reasoning promptly falls apart.

The protocol discussed was tested, and designed, for sensitivity

You can find the testing protocol developed by the Berlin group which includes Christian Drosten here. The first thing to understand about the protocol is that it is based on PCR amplification. PCR amplification is a common molecular genetic technique widely used around the world, it works by amplifying (that is producing vast numbers of copies of) nucleotide sequences that are defined by a sequence before (the 'forward' primer) and a sequence after (the 'reverse' primer) the target stretch of DNA†. PCR will only amplify sequences that have regions that match very closely to these primers. But the protocol uses a more sophisticated and sensitive variant which monitors the amplification process in real time. This real time monitoring is highly sensitive to even single base mismatches in either primer.

The protocol is a three stage process. The first "screening" assay checks for a match in the "E gene" common to many Coronaviruses, the second "confirmatory" assay checks for a match in the "RdRp gene" which will pick up only sequences of coronaviruses very closely related to SARS-CoV-2, and the final "discriminatory" assay will match only SARS-CoV-2 out of all of the Coronaviruses which have been sequenced so far.

The primers chosen for these assay were developed by in silico analysis of 375 virus genome sequences to identify sequences that have highly specific matches to SARS-CoV-2. It is possible for these in silico analyses to be misleading, but for three of them applied consecutively to mislead is extremely unlikely.

To confirm that the in silico results were not misleading, they took bat faeces samples known to contain Coronavirus, cell culture supernatant samples from cells infected with known human coronaviruses, and clinical samples from 75 patients infected with a range of viruses and confirmed that no false positives were recorded.

So it is extremely unlikely that this test is producing a high number of false positives due to the highly specific nature of the testing and the results of their validation steps.

But it is not the only test being used anyway

The WHO lists a range of other diagnostic tests used around the world. Remember that for the claim in the question to be true not just the Berlin test but all others must also be misleading.

And the tests are strongly validated by genome sequencing

Modern sequencing technology has allowed rapid sequencing of samples from around the world, these samples can be compared to generate a phylogenetic tree showing how it has mutated as it has spread. This data provides unequivocal evidence of the spread of a novel virus, with the root of the tree found in those samples collected in China showing that the pandemic has originated there and rapidly spread across the world. This confirms that the spread of cases identified by diagnostic testing reflects a real spread of the virus from its source.

---

† The virus itself is RNA, but the first step of the process produces DNA (referred to as cDNA) from this RNA, and the amplification proceeds with this DNA.

Answer 3

With the benefit of hindsight (i.e. more data), EuroMOMO now (April 7) shows excess scores for several more countries:

The fact that Italy and Spain are in "the top" z-scores should be self-explanatory by now.

Original answer below:

Besides Dr. Aidely excellent points on PCR reliability and the tracing of covid-19 trees, there is another (very) weak point in Dr. Wodarg's arguement. His claim (#3) is that we know almost nothing about the illnesses these other/existing coronaviruses cause. But that's not really the case:

Coronaviruses are enveloped RNA viruses that are distributed broadly among humans, other mammals, and birds and that cause respiratory, enteric, hepatic, and neurologic diseases. Six coronavirus species are known to cause human disease. Four viruses — 229E, OC43, NL63, and HKU1 — are prevalent and typically cause common cold symptoms in immunocompetent individuals. The two other strains — severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV) — are zoonotic in origin and have been linked to sometimes fatal illness.

Furthermore, we also know how similar this newly detected coronavirus is, relative to the known ones:

Although 2019-nCoV is similar to some betacoronaviruses detected in bats (Figure 4), it is distinct from SARS-CoV and MERS-CoV. The three 2019-nCoV coronaviruses from Wuhan, together with two bat-derived SARS-like strains, ZC45 and ZXC21, form a distinct clade. SARS-CoV strains from humans and genetically similar SARS-like coronaviruses from bats collected from southwestern China formed another clade within the subgenus sarbecovirus. Since the sequence identity in conserved replicase domains (ORF 1ab) is less than 90% between 2019-nCoV and other members of betacoronavirus, the 2019-nCoV — the likely causative agent of the viral pneumonia in Wuhan — is a novel betacoronavirus belonging to the sarbecovirus subgenus of Coronaviridae family.

Simply the fact that COVID-19's virus is more related to SARS than to other coronaviruses we knew about (hence the official name SARS-CoV-2 given by ICTV--this is the committee that maintains the "database" from Dr. Wodarg's point #1) should be convincing enough that we might have a clinical problem with SARS-CoV-2 that's going to be more serious (in the direction of SARS) than the mere common cold from the other families of coronaviruses we knew about. The ICTV paper that decided this name has a fairly helpful figure in this regard:

[c], Shown is an IQ‑TREE maximum-likelihood tree of single virus representatives of thirteen species and five representatives of the species Severe acute respiratory syndrome-related coronavirus of the genus Betacoronavirus. The tree is rooted with HCoV-NL63 and HCoV-229E, representing two species of the genus Alphacoronavirus. Purple text highlights zoonotic viruses with varying pathogenicity in humans; orange text highlights common respiratory viruses that circulate in humans. Asterisks indicate two coronavirus species whose demarcations and names are pending approval from the ICTV and, thus, these names are not italicized.

This is of course not definitive proof what COVID-19 will actually do to its host, without further clinical evidence. (In fact, the WHO objected to adopting SARS-CoV-2 as name in its mass communication materials, precisely because the mortality from SARS-CoV-2 infections [COVID-19] is actually less than that of SARS.)

I find it interesting however that Dr. Wodarg never seems to mention SARS in his video, although I admit I didn't watch it very closely. He might mention it in passing, but he certainly harps on the "usual" coronaviruses a lot more... If I'm allowed an analogy, this is a bit like saying: we just discovered a new 1.5-meter reptile, but I'm pretty sure it's as harmless to people as geckos are. Oh yeah, I might have heard of crocodiles, but I'm [somehow] sure they're not relevant to the dangerousness potential of a 1.5-meter reptile. And of course, the reality is different...

Estimates vary, but one top CDC official (Fauci)

said COVID-19 is at least 10 times “more lethal” than the seasonal flu.

So whether COVID-19's overall effects will be noticeable or not in a public health perspective, relative to the seasonal flu, now depends on how many cases of COVID-19 you're going to have. (cbeleites answer covers this latter issue quite well, so I'm not going to try and compete with that here.)

Answer 4

A recent (May 14, 2020), open-access peer reviewed paper came out in JAMA Internal Medicine.

In the paper they asses the deaths from COVID-19 and seasonal influenza in the United States. It would appear that COVID-19 is 2-2.6 times worse than the seasonal influenza.

However, the authors took great pains to try to create a new metric which, when used, rather than common practices, appears to place COVID-19 as being from 9.5 to 44.1 times more deadly than the seasonal influenza.

The ratios we present are more clinically consistent with frontline conditions than ratios that compare COVID-19 fatality counts and estimated seasonal influenza deaths. Based on the figure of approximately 60 000 COVID-19 deaths in the US as of the end of April 2020, this ratio suggests only a 1.0-fold to 2.6-fold change from the CDC-estimated seasonal influenza deaths calculated during the previous 7 full seasons.3 From our analysis, we infer that either the CDC’s annual estimates substantially overstate the actual number of deaths caused by influenza or that the current number of COVID-19 counted deaths substantially understates the actual number of deaths caused by SARS-CoV-2, or both.

Despite being in a very high impact journal, I think I detect a lot of bias on the side of the authors since they try very hard to make the numbers swing towards COVID-19 being worse.

However, from this paper, it would appear that COVID-19 is at least twice as deadly, likely up to 2.6 times as deadly, and possibly up to ~44 times more deadly, than the seasonal influenza.

Answer 5

https://www.bloomberg.com/opinion/articles/2020-04-01/coronavirus-italy-shows-we-may-be-underestimating-death-toll

Mortality in parts of Italy is 5x normal. While most of the cases aren't identified as Covid-19 it would be quite a coincidence if most of them weren't. (I would expect some increase in other causes of death because the hospitals are swamped with Covid-19 cases.)

Answer 6

Now, 6th May, it becomes clear that YES. COVID-19 increases total deaths but only in few countries in Europe according to recent EUROMOMO data. https://www.euromomo.eu/graphs-and-maps/

Despite proven facts that virus was present months before to be reported by WHO at least in Europe (France) and USA since end of 2019 total death tolls are down till beginning of March. https://www.cdc.gov/coronavirus/2019-ncov/covid-data/covidview/04102020/nchs-data.html Thus, logical answer based on numbers is: Yes, it could be more dangerous, depends where you live.

Answer 7

Dr Wodarg’s videos and articles show strong evidence of confirmation bias and fundamental attribution error. For example, in his widely shared original video, he distorts the facts, makes correct but not that important generalisations, attributes commercial motivations (without evidence), and avoids key issues to support his hypothesis. Hidden behind the confirmation bias are some fundamental flaws.

An example of distorting facts, and fundamental attribution error, is his implication (in the video) that the Chinese government made a meal of the infection to support their ambitions to introduce more surveillance. The opposite is the case because the government initially tried to suppress the reports of the infection, first observed by the late Dr Li Wenliang. The Chinese government have only just issued an apology and paid some compensation to his family.

Examples of correct generalisations, which aren’t that important, are his observation that coronaviruses appear every year, that introducing a new test can distort one’s understanding, and that we are currently working on incomplete and imperfect data. Although all are true, the key question is whether the best-available data shows a more-than-reasonable likelihood that this particular strain is represents a significant new threat. Dr Wodarg only looks at overall, historical coronavirus trends. He does not properly consider whether the current trends are hiding the emergence of a dangerously atypical coronavirus.

The fundamental flaws in his argument can be found in the issues that he avoids, particularly previous pandemics with high mortality rates (such as the 1918 Spanish flu), and the role of R0. He refers only to SARS and MERS, which did not result in mass fatalities. He thereby fails to identify how to distinguish between a devastating pandemic such as Spanish flu and the normal winter progression of infections.

This is where the estimated R0 of Covid19 is important. For normal flu, R0 is around 1.3, but it is brought below the critical value of one through vaccinations, a degree of natural immunity, and the natural social distancing behaviour of people who are infected. SARS R0 was brought under one through containment.

For Covid19, however, the R0 is typically estimated to be around 2.5, with no current vaccination or natural immunity (ibid.), and for many who are infected the symptoms are so mild that they may not engage in any social distancing even after the onset of symptoms. Although these are provisional facts, they present an urgent existential threat to very large numbers of people. There is nothing in Dr Wodarg’s argument that addresses that particular threat.

Answer 8

I'm looking for data sets from additional countries, but as of the week ending 2020-03-07, all-cause mortality in England is tracking well under recent years such as 2015 and 2018 which had harsher than average influenza seasons. (Note: 2015 and 2018 were harsh enough to be classified as influenza pandemics.)

Data source

Edited to add: A month later, these apparent trends in the provisional data have not held up:

Graph Source