Hacker News Reader

Comment by dhbradshaw | Hacker News Reader

What exactly does "percent protection" mean in these announcements? Can we use the definition to derive a percent protection associated with having been through the virus? Is there a percent protection associated with your age / gender / race profile?

It feels like the number doesn't mean much without having a similar number for the null cases.

View on HN

It is a reduction of the risk of symptomatic disease by 76% once vaccinated.

But more importantly, there were no hospitalizations or deaths in the vaccinated people. This is a point the media often overlooks.

Because the "no hospitalisations or deaths" claim from the first Oxford paper was comparing n=2 to n=0 and thus was miles from significance. If the new paper has updated data on this happy to be corrected.

Tbh, the reductions in hospitalisations has been super high for almost all the vaccines, which is great news.

It is a constant in any vaccine, including the "lowest" performers from Sinovac and Sinopharm. At this point, I'm inclined to believe that any resonably effective vaccine will have this effect.

It's huge. On the whole, avoiding hospitalizations is far more important than whether people had mild cases.

Not quite when it comes to stopping the epidemic, because transmission seems to be highly correlated to viral load and thus to severity as well.

Indeed. These are some of the best vaccines ever invented[1].

Look at the two charts[2] in this paper from the UK's JCVI. They show the number of severe incidents (severe COVID or death). The linear or exponential growth curves are for people on the placebos. The almost flat lines are for people on the Pfizer-BioNTech and the Oxford-AstraZeneca vaccines, respectively.

Oh, after just the initial dose, not the booster!

[1] https://www.nytimes.com/2021/01/18/briefing/donald-trump-par...

[2] https://www.cas.mhra.gov.uk/ViewandAcknowledgment/ViewAttach...

Is it known what the effect is on so-called "long covid" and other pretty serious long-term covid affects that might be suffered by people who don't get it badly enough to go to hospital?

Anecdotal evidence I’ve been gathering points to the vaccine making long term covid go away. I have no idea how this works, but it is tempting me into signing up for the vaccine

That's the dream. It would be nice to have something better than anecdotal evidence. I hope we do.

I wonder why the evidence of lack of it isn't being publicised.

Out of curiosity, why would you not want to sign up? All the vaccine results have been amazingly positive.

What’s the probability that today’s scientists look as ridiculous in a few decades as the ones from a few decades ago look like today?

Asbestos was considered a great building material.

While science had improved in absolute terms, there is a chance to screw up in new ways.

Sounds like you’re not concerned about all scientific advances, you just want to give them enough to time to be sure there are no adverse side-effects.

Vaccines have been around for a very long time and have proven to be incredibly effective and safe. Nobody is embarrassed about how the vaccines for smallpox, polio and measles have worked out (apart from some misinformation in recent years -- you’ll have to do your own research on that one. If you come to it with an open mind I think you’ll come to agree those vaccines are safe.)

Because the guidance for long haul was to wait. Now that I have anecdotal evidence I will probably sign up sometime this summer when the original antibodies wane

Protection of antibodies from a previous infection is somewhere above 90%.

https://www.medrxiv.org/content/10.1101/2021.01.15.21249731v...

Afaik it means that if you have a control group that isn't vaccinated vs this group (both should be drawn randomly of course), you will see 76% less infections in this group.

That is not what any of the vaccine studies have said so far. I haven't read this one, but I bet it also doesn't say this.

The stated efficacy/protection percentages that have been presented so far have been for the reduction of _symptoms_.

There haven't been any studies on vaccines preventing infections, probably due to lack of resources for testing everyone in a trial multiple times.

Maybe i misunderstood OPs question but i don't think that changes much. It's the difference between $measurement between the control group and the test group.

If that's the reduction of symptoms here, ok.

We don't know how much it changes. We don't have much information on the difference between the ratio of $measurement0 : $measurement1 between the two groups. Once everyone is vaccinated (antivaxxers don't count, they get what they get), symptom rate is all that matters. But in the interim months, transmission rate from vaccinated sources is incredibly important, which is partly determined by infection rate among vaccinated.

IANAD but I'm nearly positive you're incorrect.

I spent last night puzzling over what "efficacy" means vis a vis this video: https://www.technologynetworks.com/immunology/videos/covid-1...

The author explains that the efficacy number is derived from the the number of the number of people in the control/placebo group infected divided by the total number of infections in the study.

76% efficacy means that, for a given population of people who contracted a disease (ignoring all uninfected), 24% were vaccinated (vaccine failures), 76% were not vaccinated. It has nothing to do with reduction of symptoms.

I think this is unintuitive because the "real" solution that laypersons want is, what percentage of vaccinated people exposed to a disease get infected? We can't measure that directly because we can't just expose a bunch of folks to Covid and hope for the best. But if the sample of placebo and vaccinated people is truly random, this would appear to arrive at the same ratio. (re: when I said "ignoring all uninfected"? in a good sample, the ratio of unexposed vaccinated and control units should be about the same)

> We can't measure that directly because we can't just expose a bunch of folks to Covid and hope for the best.

We can, but medical ethics hasn't caught up yet. I would participate in a challenge trial, for instance.

> I would participate in a challenge trial, for instance.

Me too. And nearly everyone I know in my risk tier.

We had an entire globe of low-risk people willing to be part of the most acute spread, whether for research, for personal immunity, or to be able to isolate for a defined time to protect loved ones, and we squandered it in favor of pseudoscientific horizontal interdictions like lockdowns. It's really an incredible abandonment of basic principles of public health.

Medical ethicists believe they've solved the trolley problem by turning their back on the lever and letting Jesus take the wheel.

If the sample consists of volunteers, it's hardly random. In particular, given that these volunteers are "low-risk people", it means that the results of this experiment will yield virtually no useful conclusions for the higher-risk groups

Sure, I don't mean to suggest that it can be a replacement for RCTs for vaccines.

But there's a lot of information we left on the table. We had an opportunity to systematically formulate a controlled acute spread, studying the virology in a longitudinal way, studying the immune response in a more intimate and controlled way (from antibodies to the immune response after antibodies fade), and to examine the prevalence and role of innate immunity.

Beyond research, we had a chance to have tens of millions of people with verifiable immunity in Spring or Summer. People who then were in a great position to render aid to the marginalized communities who are instead bearing the brunt of lockdowns.

Instead, we traded all that for zero upside. And we have to deal with the enormous side effects of the horizontal interdiction as well.

Exactly, this is the real issue here unfortunately.

A randomly controlled challenge trial would involve deliberately infecting scores of unvaccinated people, putting them at risk, along with anyone involved in the procedure and anyone they come into contact.

It would mean handling live virus, which requires a Biosafety Level 3 facilities.

And, it's not going to get your vaccine approved any quicker.

See, this is a classic example. You need BSL3 facilities if you want to have a hundred people have it. But the alternative is 27 million people getting it in planes, boats, and buses - none of which are BSL3. That part is okay.

Quite the example of the asymmetry. If you engage with the problem you have to operate absolutely perfectly. Way better not to engage with it and kill a few hundred thousand people. After all, no one can blame you for that part.

And of course it's not going to get the vaccine 'approved' quicker because yeah, 'approval' is also subject to the same people. But it will let us know if it works and whether it hurts.

> You need BSL3 facilities if you want to have a hundred people have it. But the alternative is 27 million people getting it in planes, boats, and buses - none of which are BSL3.

Those things are not true alternatives.

I said "it's not going to get your vaccine approved any quicker" in the hope that you would understand: vaccines have reached the point of global distribution without doing the additional deliberate harm of intentionally infecting people.

Well, we've taken your approach and killed 300k in the process, so I guess we could play at being tankies in the '90s and say "If only the Communism were properly done" or we could revisit the approach.

Of course I am healthy, my family has had the vaccine or the disease and punched through, so I don't really care all that much.

FDA still hasn't authorized the AZ vaccine

Other countries have, though, and they don't have millions of doses waiting for approval or factories waiting to spin up production.

The bottleneck has generally been manufacturing rather than approval so far. It may shift to distribution.

That was a super interesting video, thanks for sharing!

It is so frustrating, trying to figure out at least a bit of whats going on and understanding some numbers, but apparently I cannot do this, without having to dive into papers, too?

There is so much confusion and missinformation flying around, when even here on HN there is discussion about the meaning of basic numbers.

(but please carry on discussing it)

To be clear, this isn't some settled math like basic algebra or science like Newtonian physics - this is cutting edge stuff at the intersection of medicine, public health, and biotechnology. Experts disagree about the nuances of measuring the efficacy of drugs, vaccines, and therapeutics so throw in the glaring ethical concerns into the mix and you have a recipe for a complex field. It's rare (unheard of?) to find a clinical trial where the company (or their CRO) scientists didn't sit down with FDA regulators and negotiate over key features & risk indicators, benchmarking, patient selection, and so on because they bring expertise the FDA is unlikely to have.

It's very difficult to get even a basic grasp of the details because you need to grok some complex statistics, biology, and public health to really dig in. Journalists sure aren't going to get the nuances so of course you have to read the academic literature.

I get it, that I won't really get it, without studying it. But it would be nice, to know what important numbers roughly mean, without that amount of work.

> The author explains that the efficacy number is derived from the the number of the number of people in the control/placebo group infected divided by the total number of infections in the study.

Wouldn't that make a vaccine that does nothing 50% effective, instead of 0% effective (which is what one would expect)?

Well, let's see… say you had 2000 participants in the trial, 1000 with a junk vaccine, and 1000 with placebos.

Furthermore, let's say you had 200 infections, split equally between the junk and the placebo.

So yeah, 100 / 200 = 50% effective. A coin toss, which is to say, no better or worse than not getting the vaccine.

Again, this isn't measuring "what one would expect" (how many vaccinated people, exposed to the virus, get the disease), it's "what proportion of vaccinated people get sick relative to unvaccinated people." (again IANAD)

Kind of seems odd, given that you can never have 0% efficacy (unless your vaccine somehow caused every vaccinated person to get the disease instead of a non-vaccinated person). It seems like you're wasting half your "spread", basically.

If you measured efficacy as "1 - (vaccinated infections / unvaccinated infections)", it would be much more intuitive, and you could even get a bit of negative efficacy in the case where your vaccine did nothing.

No, 50 % means the infection rate (or whatever you're actually measuring) in the vaccine group is half that of the control group. In your example the infection rate are exactly the same, giving an efficacy of 0 % (± whatever confidence interval the size of your trial and placebo groups gives you). See e.g. https://blogs.sciencemag.org/pipeline/archives/2020/09/21/th...

> There haven't been any studies on vaccines preventing infections

This study has looked at that. From the press release: "Analyses of PCR positive swabs in UK population suggests vaccine may have substantial effect on transmission of the virus with 67% reduction in positive swabs among those vaccinated"

In the Oxford trial, participants take weekly PCR swabs (https://www.bbc.com/future/article/20200721-coronavirus-vacc...)

This was posited (but has not been confirmed) as one of the reasons it was posting lower numbers - it was catching some asymptomatic cases

The vaccines which are administered intramuscularly aren't expected to prevent infection at all.

You won't have neutralizing antibodies in the mucosal surfaces in the upper respiratory tract which would be required to completely prevent infection.

You will have NAbs in the bloodstream which should very effectively isolate the infection to the upper respiratory tract. Immune responses will also be primed so that the infection in the upper respiratory tract will be dramatically reduced. In many cases though the infection will be detectable via rtPCR but that won't be significant since the person won't be symptomatic, won't "feel sick" (or very sick) and will be very unlikely to transmit the virus. It should also entirely cut out the multi-organ involvement of covid since the neutralizing antibodies in the blood will prevent spread to the lungs/kidney/heart/brain/etc.

Testing via rtPCR for "infection" would be an expensive waste of time in a vaccine which isn't expected to confer that level of protection. And when it failed to confer that level of protection you're left trying to explain why that low number doesn't matter to a population expecting some kind of medical miracle.

Instead they watch for symptoms and then they confirm that is a SARS-CoV-2 infection via rtPCR, and this is the typical "efficacy number" which is reported (I think the initial Oxford trial results did report the results of randomized rtPCR testing so their numbers were dramatically lower, which has caused all kinds of confusion since its not apples-to-apples with the mRNA vaccine results).

Reduction in transmissibility is more difficult to determine which is why those numbers aren't reported, it is assumed that reduction in symptoms correlates with reduction in transmissibility. So none of the vaccine trials can conclusively rule out asymptomatic spread in vaccinated individuals. However since the people are being followed over time what is caught is any infection which subsequently produces symptoms. So the vaccinated asymptomatic individuals would be truly asymptomatic and not just presymptomatic. Dramatically cutting down symptoms is expected to be a pretty good proxy to dramatically cutting transmission.

Relevant quote from a recent BMJ article on truly asymptomatic transmission:

> The transmission rates to contacts within a specific group (secondary attack rate) may be 3-25 times lower for people who are asymptomatic than for those with symptoms.1121415 A city-wide prevalence study of almost 10 million people in Wuhan found no evidence of asymptomatic transmission.16 Coughing, which is a prominent symptom of covid-19, may result in far more viral particles being shed than talking and breathing, so people with symptomatic infections are more contagious, irrespective of close contact.17 On the other hand, asymptomatic and presymptomatic people may have more contacts than symptomatic people (who are isolating), underlining the importance of hand washing and social distancing measures for everyone.

https://www.bmj.com/content/371/bmj.m4851

The title study here on HN on the Oxford vaccine is pretty clear that they're looking for "primary symptomatic COVID-19" which means they're not screening for asymptomatic infection, they're waiting for symptoms, then confirming.

They found that one shot conferred good protection (76%) for at least 90 days post vaccination (after the first 2 weeks) with little waning and some evidence that it gets better. They also found that the BEST boosting interval was 12+ weeks or more, reaching 82% efficacy. This matches what I've heard some virologists yap about informally online that longer boosting intervals are generally what is recommended (more or less letting the learning immune response bake longer while letting the immediate response to the vaccine vector wane) but the early studies used short intervals due to pandemic driven concerns around giving health providers and the elderly the best immunity in the shortest period possible. Now that we're transitioning to needing to vaccinate literally everyone the message may unfortuntely shift in ways that make people suspicious.

Reply on news.ycombinator.com