Cummins-2020-09-08

About

Back to main page

The post below was written on 18 September 2020. This boxed update was written later (6 December 2020). As serious commentators predicted, the "casedemic" in Europe and the US has since turned into a "deathdemic", with the supposedly "one-peak" northern states in the US being no exceptions.

Intro

This was written on 18 September 2020 in a response to a widely-circulated YouTube video by Ivor Cummins in which he claims that Covid-19 is essentially all over, and that we've been lied to. His manner is persuasive with a kind of hypnotic lilt, carrying the air of the reasonable man wearily dismantling the establishment. His message is seductive because we'd all like to believe the pandemic is nothing to be worried about and we'd also all like to believe we have one over on the establishment who lack the perspective that we naturally possess (or who have some kind of malign ulterior motives that he darkly alludes to). Despite the fact that almost everything in the video is, I believe, completely wrong, I was actually struck by the persuasiveness of the presentation, and given how many views his video has (more than 1.2 million at the time of writing) I thought it was worth spending some time investigating its claims.

Since I'm a mathematician, this will focus on the more mathematical aspects of the video. For a good critique that goes into some medical details, you can read this by Dr Dominic Pimenta.

These are some of the main claims Ivor Cummins makes in his video.

1. In temperate countries/regions you get a universal infection curve - a rapid increase followed by a slower decrease - which he calls a Gompertz curve.
2. In such countries, he says you quickly get 20% of the country infected, with the remaining 80% having prior immunity, so the infection is therefore over after this Gompertz curve has been navigated.
3. He says the number of Covid-19 deaths is nothing special compared to previous years and the extra deaths are just people who were waiting to die anyway (the "dry tinder"), due to fewer than normal deaths occurring in the previous year, thereby leaving the frail behind.
4. He says lockdown and restrictions have little to no effect.
5. He also later says lockdown and restrictions have prevented a useful spread of the virus in summer months, which would have enabled us to build immunity. (Yes, this contradicts 2. and 4.)

Unfortunately all of this is wrong. Almost every country in the world has kept a lid on the disease, to a greater or lesser extent, by lockdowns or distancing and no country has seen a substantial portion of their population infected thus far. There is therefore a large potential for more deaths if distancing measures are relaxed too far.

It's also not true that there has been substantially less than the normal amount of mortality in recent years. His graphs purporting to show this are largely smoke and mirrors. Any such "dry tinder" (mortality displacement) effect is very small at best, and possibly entirely non-existent.

Of course there is a genuine argument to be had about whether the cost of locking down is too high, whether the cure is worse than the disease in terms of economic impact, health and well-being impacts, not educating children etc.. But that discussion can only take place on an honest basis which involves acknowledging the likely full impact of letting the disease have its way.

In detail:

1. The Gompertz Curve

The Gompertz curve ("the classic Gompertz curve" as Cummins is fond of saying) is a fiction of Prof Michael Levitt, the Nobel prize-winning chemist Cummins refers to at 0:50 into his video for his inspiration. To be clear, there is such a thing as a Gompertz curve, but there's no reason to believe that it fits the trajectory of the various Covid-19 epidemics around the world, and every reason to believe it doesn't. It's just numerology. (In fact Cummins mistakes the Gompertz curve, which is always increasing, for its derivative, but let's leave that aside.) In March and April Levitt simply fitted Gompertz curves to the existing epidemic curves around the world and extrapolated them into the future to get a prediction. Of course with three parameters to vary, and the general shape of a steep upslope followed by a slower downslope, you are likely to be able to get a fair visual fit to rather rough data if you fit these parameters afterwards, but this is about as enlightening as the brontosaurus theory of A. Elk (Miss) ("thin at one end, much, much thicker in the middle, and then thin again at the far end").

How have his predictions fared? With China he got lucky and predicted something close to (what China says is) the number of final deaths. With other countries his predictions didn't work out so well. In Israel, Levitt said there would only be 10 deaths in total from Covid-19, based on his Gompertz curve. As of today (19 September 2020) there have been 1196 deaths in Israel, with 27 new deaths announced yesterday. Israel is now fully in the grip of a second wave.

On 25 July, Levitt predicted that the US would have 170k total Covid-19 deaths, which was 42k more deaths than there were on the date of his prediction. He also said the outbreak would be over in 4 weeks in the US. As of 19 Septemeber, 8 weeks after his prediction, there have been 75k more deaths and the disease is unfortunately still going strong.

How did Levitt respond to these failures of his predictions? He did at least admit he was wrong, though he was pretty ungracious and minimised it as you can see here, though of course that article was from the 28 August when the true scale of his wrongness was yet to make itself clear. See how he makes definite statements about what will happen without a hint of uncertainty. Such (rather unscientific) statements from such a high profile scientists have the potential to be used by Ivor Cumminses the world over and cause carnage ("Who are you to disagree with a Nobel Prize winner?"). Yet when Levitt is proved wrong he just shrugs and says "oops", or to give you the exact quote "It was probably wrong of me to say that lockdown was a mistake. If people have done it, they should be told that it was probably a good idea when they did it, but it’s no longer necessary and there’s nothing left to be frightened of." a half-hearted apology with more disinformation tacked on the end.

The fundamental point is that to the extent that the Gompertz curve fits the data at all, it is fitting the effect of the social response to the outbreak, not the natural trajectory of an unchecked outbreak. Ivor Cummins is making exactly the same mistake. This curve works to an extent so long as the the distancing (etc) measures play ball with it. But it's completely wrong to assume the infection is done when we're at the bottom of the curve. If distancing is relaxed enough the infection will certainly be back.

2. Prior immunity, infection rates, and second waves

From their antibody survey, the ONS estimates about 6% of the UK has had Covid-19. It's possible the true number is a bit higher for various reasons, but there's no evidence that 20% of the country has been infected. And there is also no evidence that 80% of the population have some kind of prior immunity. (For some evidence against this idea, there are examples where the virus has circulated in a closed communities and infected most people, e.g., 85% on this fishing vessel or a probable 87% in this choir, which obviously wouldn't be possible if 80% of them were immune.)

Of course if everyone were immune from the first wave, there could be no second wave. How then does he explain the second waves we're now seeing in Europe, US and Israel? He uses a variety of evasions.

In the case of Europe he denies there is a second wave. It's all a "casedemic" - just lots of false positive tests with no-one actually getting ill. That's partly true, in that we had much lower testing capacity in March-April, so now testing is finding a much larger proportion (though by no means all) of the cases. But the reason this isn't translating into a large number of deaths right now is because the new cases are almost all in young people who are mostly exempt from serious Covid-19 effects. Of course the danger is that eventually the cases will move from young people to older people and we'll start seeing serious cases and ultimately more deaths. This has started to happen to some extent: Spain is currently suffering more than 100 new deaths each day. Not yet as bad as March/April, but still significant and inexplicable by a Gompertz curve (or rather its derivative) which only has one peak.

At 32:00 he talks again about Spain. Since his video was made a week and a bit ago, the Spanish mortality curve has grown to about twice what it was. But this is the nasty bit: he draws a large second hump and says it's not a proper second wave until you get that kind of curve. He doesn't mention that that second curve would represent about 30,000 more deaths, and that by the time the daily death numbers - a lagging indicator - show the signs of starting to form such a curve, it would be too late to prevent it. He is arguing for never taking countermeasures in time.

In the case of the US, he introduces a new idea that some regions of the world should naturally expect two peaks. Apparently temperate regions only get one peak - "the classic Gompertz curve" - while tropical regions see "the classic double curve". The mechanism for this is not explained. In this theory the southern US states count as tropical and are expected to have two peaks in their infection curve, or just rumble on at a constant level. This is meant to explain the two peaks in the US infection curve, because the two-peak southern states are mixing their infection curve with the one-peak northern states. This is partly true but only by retro-fitting. If it's true, it's because of the progression of infection around the US, the distancing counter-measures, population density and a whole range of other things, not because the climate of Kansas is so different from Nebraska (or whatever - I don't know where his cutoff is meant to be). And in fact there is a serious resurgence right now in northern states like Ohio and Pennsylvania. I don't know how he'd explain these.

(From 28:20 in the video, supposedly the Southern and Western regions are "virome-triggered", which makes them different from the other US states. "Virome-triggered" is a meaningless gobbledegook phrase used to sound impressive.)

As for Israel which has maybe the most pronounced second wave in the world right now, Cummins simply ignores it. I don't see how he can fit it into his model, but no doubt there would be a new idea to explain it.

3. Mortality displacement ("dry tinder")

The "dry tinder" theory. Around 2:45 into his video Cummins uses Euromomo data to compare with 2018. This is a somewhat cherry-picked year because winter 2017/18 was a bad flu season. But it's also slightly misleading for another reason: there wasn't a particular shortage in deaths in the winter of 2020 (prior to Covid). In fact it was roughly average, perhaps even a little worse than average for that time of year (depending on how you calculate the average). The trick from the Cummins video is that you are invited to believe that because the period Dec 2019-Feb 2020 wasn't as bad as Dec 2017-Feb 2018, there must have been some missing deaths. But there weren't: an average number of people died for that time of year. Another trick is that he is using the Euromomo set of countries, some of which haven't suffered particularly badly from Covid-19. This has the effect of diluting the Covid spike.

To check this, I made a graph (click to enlarge)

using what I think are reasonable methods/assumptions. Starting with the England+Wales weekly death figures (from www.mortality.org) from 2010, for each week of the year (1-52) construct the 10-year average for that week. Then plot the actual deaths and the 10-year average deaths on the same graph. You can see that from 2019 the number of deaths is just about average for the corresponding time of year, and in the run-up to Covid it's actually slightly above average, so I see no evidence of "dry tinder" here.

If you change the assumptions, e.g. using a 5-yr average, you will get slightly different results, but in any case I don't believe you will see evidence of dry tinder accounting for more than a very small proportion of the Covid deaths.

Note also that just because a frail person evades a chance death in one year, there is no law of nature that says they have to die the next year. They might live another 5 or 10 years. In other words, even if there were some kind of mortality deficit (which I don't believe there is) it wouldn't automatically translate into an expectation of the same number of people dying in the following year.

What of the alternative "dry tinder" graphs he presents 8:30-10:20? The solid blue lines in the graphs are correct in the sense of representing true numbers, but his interpretation of them is highly questionable. For one thing, it relies entirely on comparison with the dotted lines, but these have not been constructed in any principled way. E.g., looking at the Netherlands graph at 8:45, there is no reason to believe that mortality has to increase year on year like that. In fact mortality rates have generally been decreasing in recent years.

He might say that the dotted line is a baseline to take into account of population increase, but the problem with that is that that's not how it is chosen. E.g., at 9:19 you see the Sweden dotted line going down, despite Sweden's population increasing over that period. And even if there is a population increase over the period, that doesn't necessarily translate into a proportionate number of deaths because immigrants tend to be young.

So in reality, the dotted line has probably been chosen to go through the blue curve to create humps and troughs on either side of it and create the impression of a mortality deficit where there isn't one. (If you make the dotted line horizontal, which I think is more reasonable, then you get an entirely different interpretation with no mortality deficit in the run up to Covid. But really I think this kind of graph is not the way to settle this question.)

Another misdirection (or maybe mistake) is that you are encouraged to look at the humps and troughs of the blue graph despite the fact that it is already a 52 week trailing average. Looking at the areas created between the blue graph and the dotted line is effectively integrating it, and integrating something that has already been convolved with a year-long kernel has the effect of convolving with a trapezium kernel which is not what you want. And other misleading impression from these graphs arises because they don't go far enough into the future. Even if the main mortality from Covid-19 is limited to March/April, the 52-week trailing average will continue to show this up to April 2021. To make the effect of Covid-19 visually equivalent to other mortality effects, you ought to be extending the graph to April 2021.

In other words, this is another lot of smoke and mirrors and does not prove what he thinks it proves.

4. "Lockdown has no effect in preventing spread"

Saying lockdowns have no effect is ridiculous. Apart from the mountain of experimental evidence in the case of Covid-19, we've known for well over a hundred years how diseases spread by person-to-person contact (directly or via surfaces or through the air) and reducing contact has to reduce the rate of spread. There's no plausible mechanism by which isolating people wouldn't prevent the disease spreading. It doesn't spread by magic. We know that $$R_0$$ for Covid-19 in the UK is about 3 in the absence of distancing measures. Broadly speaking if contacts are reduced by a factor of 3 then $$R$$ should go below 1, which is what we need. We can measure $$R$$ and see how it responds to distancing measures and recently it has been going up in response to relaxation of such measures. This rise is entirely inexplicable in Cummins' view of the world where the disease is already played out and distancing makes no difference.

At 12:33 he asks how can the number of deaths continue to fall even though we've been relaxing distancing. The answer is that relaxing distancing increases $$R$$, but at the end of June $$R$$ was low enough that you could increase $$R$$ and still maintain it below 1. In other words, relaxing distancing is compatible with reducing infection rates. As it is now, we are starting to see what happens when $$R$$ goes above 1: increased number of infections, which will in turn lead to an increased number of deaths.

5. "Lockdown has too much effect in preventing spread"

Fast forwarding to 33:58. This is weird, isn't it? Now he's arguing that we should have let people infect each other during the summer in order to build immunity and thereby reduce winter infections. He argues it's better for people to get infected during the summer to build "T-cell, mucosal population community immunity" and we've avoiding developing the "normal, ancestral, evolutionary summer period community immunity". Ignoring all the other dubious aspects of this idea, notice how he has completely flipped from pretending that lockdown and distancing has no effect on the spread of the virus.