An article was recently published in the British Medical Journal that reported on a matched cohort study which compared the risk of dying for those infected with the new British variant (a.k.a. B.1.1.7) and those infected with the older covid variants.
A matched cohort study is a type of observational study where you take a group of people with some condition and then try to find a similar group without the condition to match against. Then you follow the two cohorts over time and see if they differ in some meaningful outcome (like death). Since it is an observational study, it can only show correlation. It can’t prove the existence of a cause and effect relationship, but that doesn’t stop many people acting like it does.
The article has resulted in fear-mongering headlines in news media around the world. Just to take the first example I could find, Al-Jazeera published an article with the headline: “UK variant up to 100% deadlier more deadly, study finds”.
Those darn studies, they’re always finding things. It’s like a never-ending game of whack-a-mole. You knock one down here, and another one pops up over there. Anyway, let’s look in to the study some detail, and see if the claim is true.
There were two criteria that had to be fulfilled for a person to be included in the study. They had to have a PCR-test positive for covid at some point between the beginning of October 2020 and the end of January 2021. And they had to be over 30 years old. The authors don’t provide any reason for the second criterion. The only reason I can see for removing people under the age of 30 is that they pretty much never die when they get covid, and including them would therefore have resulted in less impressive mortality numbers, which would have made it a little bit harder to use the results as part of public fear mongering campaigns.
In the UK, the PCR test currently in use is based on three reading frames. In other words, three separate pieces of viral RNA are sought. The B.1.1.7 variant has some variations in its genetic code that cause one of these reading frames to turn up a negative result. This is useful, because a problem with doing a big study like this and comparing mortality rates for different viral variants is that most people don’t actually get their infections gene-sequenced. So all you have to work with in most cases is a PCR test. But the fact that the B.1.1.7 variant has this oddity, that one of the three reading frames turns up a negative result, means that it can be identified through PCR with pretty good accuracy. No gene sequencing necessary.
So, what the researchers did was to put everyone with a covid diagnosis in which the other two reading frames were positive, but this specific reading frame was negative, in to one cohort, the “B.1.1.7 variant” cohort. Those who had all three reading frames turn up positive were put in the other, “old variants”, cohort.
Now, as mentioned, a cohort study is a type of observational study, and observational studies are rife with confounding factors that mess up the results. In order to minimize this problem as much as possible, the researchers went through and matched each person in the new variant cohort to a similar person in the old variant cohort. The cohorts were matched on date of testing, in order to deal with potential biases caused by one person for example getting tested during the covid peak, when hospitals were overstretched, while the matched person in the other cohort got tested at a time point when nurses actually had time to fluff their pillows. The cohorts were also matched on geographical location, age, gender, and ethnicity.
The endpoint that the researchers chose to look at was death within 28 days. This is a very problematic end point, that will tend to overestimate mortality due to covid. Basically, anyone who had a positive covid test and who then died with the next 28 days was counted as a covid death. Even if they got hit by a bus. Apart from overestimating the covid death rate, this could also muddy the results of the study, making it harder to see a real difference in mortality between the new variant and the older variants, if such a difference does exist. Why they chose to do this rather than actually looking at death certificates, to see whether covid was listed as the cause of death or not, I really don’t understand.
Anyway, let’s get to the results.
54,906 people with the new covid variant were identified, and these were matched with 54,906 people with the older variants. Among those with the new variant, 227 patients died (0,41%). Among those with the old variants, 141 people died (0,26%).
So, the new variant does appear to be a little bit deadlier than the older variants, 0,15% deadlier to be precise. To put this in perspective, for every 700 people who develop covid due to the new variant, you can expect one extra death, as compared with getting covid due to the older variants.
You could of course, like the mass media do, focus on relative risk, and say that the new variant is 61% deadlier, or “up to 100% deadlier” as Al-Jazeera state in their headline (based on looking at the upper end of the confidence interval), but in this instance, looking at the absolute risk gives a much clearer understanding of how deadly the new variant actually is, don’t you think?
One should of course always remember that this is an observational study, and although the researchers have done their best to get rid of confounding factors, it is still possible that the increased mortality rate seen here is due to some unknown confounder, and not due to the new variant itself.
The thing that strikes me most about the results of this study is not the fact that the new variant seems to be a bit more deadly than the old variants, but how un-deadly this study clearly shows that covid is. We have to remember that this study only included people who actually took a PCR test. According to the eminent Dr. Anthony Fauci, 40-45% of covid infections are asymptomatic. Obviously, people who are asymptomatic are for the most part not going to get a PCR test (unless they get caught through contact tracing, but this likely only catches a small proportion of asymptomatic infections). And equally obviously, people with asymptomatic infections aren’t going to die of covid. So, although this study found a fatality rate of 0,41% for the new variant, and 0,26% for the old variants, the real fatality rate is likely considerably lower.
That is especially true if we also factor in that this study excluded people under the age of 30, and counted every death within 28 days of a positive covid test as a covid death. Both of those factors would push the fatality rate down further if factored in. So this study, funnily enough, adds to the existing evidence that the infection fatality rate for covid has been grossly overstated.
To be fair, the proportion of participants over the age of 80 in the study is low, only 0,5%, compared with 3% in the UK population as a whole, which will push the fatality rate in the opposite direction. Whether excluding everyone under the age of 30 (constituting 25% of the UK population) or only having 0,5% of participants be over the age of 80 (when they constitute 3% of the UK population) has the bigger impact on the overall fatality rate in the study, is hard to say. But it raises another interesting point. The mortality rate in the 80+ group in the study is 100-fold higher than it is among the people aged 30-59 (12%, or one in nine people, as compared to 0,12%, or one in 900 people). This is in line with earlier studies that have shown that the risk of dying rises steeply as people reach an advanced age.
As always, the devil is in the details. So, what can we conclude from this study?
The B.1.1.7 variant does appear to be a little deadlier than the older variants, increasing the risk of dying for the average person who gets a symptomatic infection by a marginal amount (0,15% to be precise).
However, the main take-away from this study is that the infection fatality rate, even with the new variant, is very low for most people. I think a more reasonable title for Al-Jazeera’s article about this study would have been “Covid much less deadly than everyone thinks, study finds”.