Study on All-Cause Deaths by Vaccination Status Makes Absurd Claim That a Fourth Dose Cuts Total Deaths by Two Thirds

I was pointed recently to a study from Sweden which looks at all-cause death rates by vaccination status – something very few studies do. Published in July, it compares all-cause mortality in those who had three and those who had four vaccine doses among both over-80s and care home residents in Sweden.

The results are, to be blunt, incredible, which is to say, they’re not credible. They purport to show that compared to three doses, a fourth vaccine dose reduces all-cause mortality by a massive 71% in the over-80s and 39% in care home residents during the first two months after inoculation. Let me say that again: the study is claiming that in the over-80s a fourth vaccine dose cuts deaths from all causes (heart disease, dementia, cancer etc., as well as Covid) by 71%, over two thirds. If true, the vaccine is truly a wonder drug.

Let’s take a closer look at why these are not credible figures. (If you want to skip the maths, go straight to the paragraph that starts “To repeat”.)

According to the latest report from the UKHSA, vaccine effectiveness (VE) in over-50s against death with Omicron is estimated as:

• Dose 2 at 40+ weeks: 52%
• Dose 3 at 2-4 weeks: 85%
• Dose 3 at 15-19 weeks: 75%
• Dose 3 at 25-39 weeks: 63%

For our calculations we need to make a couple of assumptions; I’ll explain why I think they’re fair. First, we need to assume that these figures for the third dose will be very similar for the fourth dose. This is likely to be correct as we can see that the third dose is declining to a similar level as the second dose over a similar time period, so the next dose will likely behave similarly. Second, since in Sweden fourth doses were stipulated to be given at least four months after the third dose, we will assume that three-dosers during the study period are in the 15-19 week category, i.e., they have a VE against Covid death of 75%. We will also assume the VE for four-dosers against death in the immediate post-jab period is 85%, since that is what it was for the third dose.

This allows us to calculate that the fourth dose cuts Covid deaths in the over-50s by 40% compared to dose 3. (This is because dose 3 cuts deaths to 0.25 of their unvaccinated level, dose 4 cuts them to 0.15 of that level, and 0.15 is 40% less than 0.25.) I note that UKHSA actually gives direct VE estimates for dose 4 versus dose 3 against hospitalisation, and these indicate our estimates against Covid death look about right. Of course, this all assumes that UKHSA VE estimates are accurate; in reality, as they’re from a Government agency they’re probably skewed towards being supportive of the vaccines, so we can assume our calculations here are conservative estimates which grant a higher VE than may be the case in reality.

Next, ONS data show that in the pre-vaccine and pre-Omicron era (when there was little natural immunity) around 40% of total deaths during Covid waves were Covid deaths (e.g. the proportion was 37% Covid deaths in April 2020 and 46% in January 2021). I’m using figures for deaths with Covid rather than from Covid for consistency with the UKHSA VE estimates.

We’ve seen that three doses are 75% effective against Covid death, so this cuts the number of Covid deaths in the three-dose population down to 10% of the original number of deaths. If we assume Omicron is half as lethal and taking into account the reduction to the overall number of deaths, this leaves Covid deaths making up around 8% of all deaths in three-dosers during Omicron waves. (This is not far off the observed figure of 12% of deaths in January 2022 in England being Covid deaths, which also includes the unvaccinated.)

Since we’ve calculated that dose 4 prevents 40% of these deaths in three-dosers, we can now calculate that dose 4 reduces the total number of deaths (from all causes) in four-dosers by around 3% (40% of 8%). Assuming it has no effect on the rest of the deaths from other causes, this is a VE against all-cause mortality of 3%.

How then can the Swedish study claim that a fourth dose (compared to three doses) reduces all-cause mortality by 71% in the over-80s and 39% in care home residents during the first two months after vaccination?

To repeat: on UKHSA data, a booster dose during Omicron reduced Covid deaths in the over-50s by 40%, which translates to a reduction in all-cause deaths of just 3% (because Covid deaths make up just 8% of the overall deaths in the three-dosers). Yet the Swedish study finds a reduction in all-cause deaths in the over-80s of 71% and in care home residents of 39%. How can 3% in over-50s in England become 71% in over-80s in Sweden?

There’s clearly something wrong here. A vaccine booster which reduces Covid deaths by 40% at most cannot reduce all-cause mortality by 71% when Covid deaths make up such a small portion of overall deaths. It would imply that the vaccine booster is massively cutting deaths from all causes, including heart deaths, dementia deaths, cancer deaths and so on, even more than it’s cutting Covid deaths. There’s no evidence of this at all. The opposite in fact: Covid waves since Delta have been accompanied by significant numbers of excess non-Covid deaths, many of which in England at least are in the over-85s age group, the cohort which received a fourth vaccine dose in spring. (It’s true that Sweden didn’t have excess deaths during the study period, but neither did it have a Covid wave – it skipped the BA.2 wave somehow – and like other countries it has had excess non-Covid deaths during recent Covid waves.)

So the study’s claim that deaths among fourth-dosers are lower by over two thirds doesn’t stack up at all. The authors note: “Although third-dose recipients had similar baseline characteristics as fourth-dose recipients, some third-dose recipients likely did not receive the fourth dose because of deteriorating health that was not captured by the baseline characteristics.” Could this ‘healthy vaccinee’ effect explain it? It’s hard to tell; it’s an awfully large effect. Have the researchers somehow excluded all those most likely to die in the four-dose group and included all those most likely to die in the three-dose group?

Either way, this study clearly cannot be relied on to tell us anything about the effect of vaccination on all-cause mortality. Which is a shame, as I had hoped that once researchers turned their attention to this question we would quickly get some answers on the overall benefits of the vaccines. Instead, it appears that the answers we get make no sense at all.

It only goes to show that what we need is the full data – mortality data broken down by cause, age group, vaccination status, prior infection status and underlying conditions, individualised as far as is consistent with anonymity – released so that it can be analysed properly, and obviously anomalous results like this Swedish study avoided.

Stop Press: An Emeritus Professor of epidemiology from the U.S. has got in touch to say he fully agrees with this critique, and adds:

1. Although they estimated VE at days 7-60, the K-M curves [the chart above] are shown since day 1. They diverge at the very beginning! We don’t expect any beneficial effect so quickly. That’s clear evidence for the ‘healthy vaccinee’ effect, i.e., clear evidence of confounding.

2. K-M curves are unadjusted, of course, but the data in the tables allow us to see what adjustment did. Comparing unadjusted VE (which can be computed by hand from the rates) to partially adjusted and then to ‘fully adjusted’, it seems that we observe some strengthening of the ‘beneficial effect’ following adjustment. That implies negative confounding i.e., fourth dose recipients were sicker to begin with, not healthier, so we didn’t see the ‘true benefit’ before adjustment. But I don’t think there was anywhere preferential vaccination of elderly or care home residents in worse mental or physical state. If anything, it was the opposite.

So, yes. Another useless study and another absurd claim.