Hacker News Reader

Comment by spion | Hacker News Reader

I know that seems like a large number but in the vaccine arm there were 20x more subjects and only 15 cases. So with the 1359 previously infected subjects, assuming identical behavior to vaccinated subjects, we would expect around 0.9 cases if vaccines were approximately equal to infection.

So this study doesn't have enough statistical power to say how vaccines compare to previous infection. Could be better, could be up to 3 times worse within a high confidence interval.

(disclamer: back-of-the-napkin math, did not double check everything)

View on HN

Here are the numbers:

Total participants: 52238.

Previously infected: 2579, of whom 1220 vaccinated, 1359 not vaccinated. No subsequent infections in either of these groups.

Not previously infected: 49659, of whom 29461 vaccinated (15 subsequent infections), 20198 not vaccinated (2139 subsequent infections).

Comparing previous infection with vaccination, we have 0/1359 infections in the previously infected and 15/29461 infections otherwise. According to Fisher's exact test, the two-tailed P-value for this is 1.0, which is very much not significant. You're correct to say that the study doesn't show how vaccines compare to previous infection. The study is way underpowered to show that.

For the unvaccinated, it gets more interesting. We have 0/1359 infections versus 2139/20198. According to Fisher, that has a P-value of 2 * 10^-64, which is most definitely significant. The study is not underpowered to show this at all. A previous infection is definitely protective.

With the caveat that the participants didn't have asymptomatic screening, so there is quite possibly bias due to people with a previous infection getting a second milder case and not being tested.

Using the exact method, 95% confidence intervals for infection rate are:

Previously infected, not vaccinated: 0% to 0.3%

Previously infected, vaccinated: 0% to 0.3%

Not previously infected, not vaccinated: 10.1% to 11.0%

Not previously infected, vaccinated: 0.03% to 0.08%

Thanks for this! Yes there is no doubt that there is an effect, of course, but it doesn't rule out the booster shot being possibly very helpful in strenghtening immunity

>> So this study doesn't have enough statistical power to say how vaccines compare to previous infection. Could be better, could be up to 3 times worse within a high confidence interval.

Could be. But it would be great if they'd just say having had a confirmed case is equivalent to getting the vaccine because several studies have now shown that to be roughly the case. It's also conventional medical wisdom although it varies by virus.

Unfortunately, we have reason to believe that over time, the immune defenses from COVID-19 infection seem to decline, particularly in older folks. It'd be great if we had some definitive data on secondary infection rates, but until a lot more people get infected, we just won't have it.

What is the evidence that “we have reason to believe that over time, immune defenses from COVID-19 infection seems to decline”?

I think this is another case where improved surveillance is turning up things that, while they look bad, happen with other vaccines and never turn out to be a problem.

The same goes for variants, which a lot of people expect to break through vaccines because that's how the flu works, but it doesn't actually seem like they will.

Yes, the extreme focus of attention is turning routine and expected medical observations into scary anecdotal news cycles.

I keep hearing this, but have yet to see evidence that the elderly are catching second rounds at a high rate.

Because there isn’t any. But that is how the aging immune system responds to most other diseases, so we assume the same will be true here.

No kidding. It's clear to me that protection at some level is provided following infection - which simply makes sense.

You'll notice they say - we don't have enough data. Then data is incomplete. Then we can't really say because study was X / Y / Z.

They ignore the MASSIVE difference in your chance of being infected if you are unvaccinated and have not been infected vs many other combos.

Just treat folks who've been infected as vaccinated, they can get booster / variant shots with everyone else when the time comes.

I'd like to know the death rate of infected unvaccinated people who get covid again against vaccinated people who get covid and get sick.

According to this study it could be a long time before we know that - I think cov-19 in the presence of a non overwhelmed medical system armed with steroids and oxygen kills less than 2/100 infected folk so getting enough incidents in either of those arms of a trial is going to be hard work.

The death rate for certain demographics (especially say over 70s) is far higher than 2-in-100. While some countries have high levels of vaccination - especially in those demographics - most don't.

The UK is enterring a third wave, and has a high number of elderly people having been fully (2 shot, mainly AZ) vaccinated. Early indications are that those enterring hospitals now on the whole have not been fully vaccinated.

And for most people under 70, it's way under that, FWIW.

It does suggest that people with 2 shots are unlikely to end up in hospital. I haven't seen enough data about the hospitalisations when adjusted for age+vaccines to draw any more conclusions.

There are still many over 70s who haven't had any shots though. My (estranged) uncle and his mother-in-law haven't left the house for 18 months, he disinfects his food deliveries and leaves the post in a box for 3 days. He refuses to get the vaccine because he's worried about catching it. Chances are he won't end up in hospital either, but people like that could skew the cases as more data comes out to allow spurious claims from anti-vaxers (They might claim that 'before 2 in every 100 over 80s were going into hospital) , now it's only 1 in every 100 unvaccinated over 80s going in to hospital), because the chance of catching covid from the population still unvaccinated dwindles because of continued isolation

The infection fatality rate varies exponentially with age. A good rule-of-thumb is that for a given age group, the IFR of catching Covid is about the same as the all-cause mortality risk over one year of life. It just so happens that these two numbers are about the same, when looking at age groups.

Edit: nicely presented here: https://www.bmj.com/content/370/bmj.m3259

I'd like to know the disability rate, which seems much higher than the death rate.

Yes, this and "long COVID" are quite underreported, and it's the much bigger risk. However, there is some evidence that getting vaccinated improves/cures long COVID.

>> You'll notice they say - we don't have enough data. Then data is incomplete. Then we can't really say because study was X / Y / Z.

This is a weird trend these days. Medical science refuses to take an engineering approach to anything and only accepts statements based on controlled studies. That includes rejecting previous understanding of how immunity works in general. Early on when claims were made about HCQ, Zinc, or whatever, they were shot down as "not approved treatment" even though there were no approved treatments! The utter failure to recommend vitamin D supplements is another failure - we know deficiency weakens immune response, we know most (Americans) are deficient in the winter, it's a no-brainer but in their mind it was not "proven" so not even mentioned or recommended for anyone.

The only pattern I see considering all cases like that is the underlying notion that people must not do anything for themselves.

The rough consensus is that an infection is equivalent to a single dose of vaccine. Some countries do not give the second dose to previously infected people.

Reply on news.ycombinator.com