Friday night Covid Plotting #9

New Topic
This topic has been archived, and won't accept reply postings.
 wintertree 23 Jan 2021

Now coming to you on a Saturday due to increased lag in the demographic data release and wanting to wait for last Tuesday's case numbers to firm up to reduce bias from weekend effects

Plots 6, 7, 8:

  • Cases in England continue to fall.  
  • Admissions are falling a bit, but there's a lot of random variation in recent points.  
  • Hospital occupancy looks to have turned a corner in to decay
  • Deaths are just about levelled off.  

Plot 9

  • The halving time is continuing to drop on cases.
    • The recent wobble around 01-16 is residual noise from weekend effects and should be reduced when another week of data is available.
  • The halving time for hospital admissions is not falling to such short values as cases; this I think is related to the demographics as covered in a later post - cases went in to decay later for older ages, where hospitalisations are higher.  This is much like the what happened in early December when the doubling time for cases rose much more than for hospitalisations, as there was a demographic shift to younger ages.  Now, the faster decay in younger ages is dominating the total halving time.  

Link to previous thread: https://www.ukhillwalking.com/forums/off_belay/friday_night_covid_plotting_8-7...


OP wintertree 23 Jan 2021
In reply to wintertree:

Clear decay in cases continues everywhere except for Scotland where there's hints of an up-tick in the last day; this isn't yet established enough to have much effect on the halving time measured for plot 9x.  Looking at provisional cases for the most recent days (not used in this analysis) on the government dashboard, this looks like it may in reality be a slackening of the exponential rate combined with a very strong weekend effect biasing the end.  One to keep a close eye on.

Post edited at 18:08

OP wintertree 23 Jan 2021
In reply to wintertree:

Plot 17, 18

  • Cases - Decay continues in all regions (Plot 17, Plots 18 - blue lines).  Halving times appear to be getting shorter (better) in all regions and down to 12 days in the best regions.  A lot of corrugation on this Plot 18 blue lines from residual weekend sampling effects.
  • Hospitalisations - the middle graph on plot 18 is updated to feature the best aligned NHS Region for the English regions.  Not all regions correspond perfectly even when their names are the same.  All regions are now in to decay in hospital admissions.  North East and Yorkshire was the last to tip over to decay.
  • Deaths - these are in decay in some regions and the exponential rate looks to be heading towards negative values for the other (halving times, decay), tracking the hospitalisations and cases as expected.

Plot 22 - hospital occupancy vs ITU occupancy

  • This looks like it's starting to form a loop as admissions decrease but ITU admissions keep rising from people already admitted, before going in to decay soon enough.

Plot 22r - a regional breakdown of plot 22

  • With a bit of optimistic squinting, every NHS region is on the way to forming the loop as admissions then ITU admissions decrease, with this being clearest in the south/east regions. 
  • This plot needs some more work.  The data starts on 2020-09-01.  
  • The aspect ratio of all plots are the same, so that a steeper line means a larger proportion of cases are going in to ITU - London stands out for this.
Post edited at 18:10

mick taylor 23 Jan 2021
In reply to wintertree:

Thanks for these. 
And 478k received their first dose of the vaccine. Which is better than very good

OP wintertree 23 Jan 2021
In reply to wintertree:

This is a rebooted version of plot 16.  

  • The horizontal line markers now show case rates on 2021-01-05 when England went in to Lockdown.
  • UTLA names are still coloured by the late 2020 Tiering level (T1: blue, T2: orange, T3: red).  I'll reset them to black next week.  For now, they make quite a convincing point that T2 did case rates no favours.  There are so few T1 places I don't draw a conclusion there.
  • Annotations are added to hi-light potentially concerning developments:
    • UTLAs that have seen less than a 0.1x drop in case rates since lockdown are annotated in orange.  The annotation shows the fractional drop.
    • UTLAs that have risen more than 0.1x above their minimum level since lockdown are annotated in red.  The annotation shows the minimum level and their fractional rise from that minimum level.
    • Bradford has risen above it's lockdown level and so is annotated by both.  I need to improve the plot to handle that better.
  • This uses my deweekended and mildly filtered data, so it won't produce the exact same results as raw dashboard data.
    • My sense is that which UTLA falls above either of these triggers is a bit noise sensitive for now; nevertheless if noise is tipping them over, they're probably not seeing cases plummet like they should.  
    • I suspect these warning flags will be a bit more meaningful next week with more data and time under the bridge . 
  • There's a map to show where the UTLAs are, matching the colour code.
    • Perhaps there are some geographic clusters 
Post edited at 18:22

OP wintertree 23 Jan 2021
In reply to wintertree:

The last plots for today - demographic breakdowns for London (Plot D1) and England (Plot D2) and a summary of their final values (Plot D3).

Two things stand out to me:

  • Bottom right corner of characteristic time heat map - cases appear to be flipping to growth (orange colour) in ages 0-15, starting first with ages 0-5, then with older ages.  Perhaps this is related to nurseries / daycares reopening after the Christmas break and then schools reopening to some children - and a lot more than in the first lockdown by many qualitative accounts. 
  • Top right corner of the same heat map - the switch from growth to decay (orange to blue in colour) happened later with age over about 75 years of age.  I suspect this is why hospitalisations are falling much more slowly (in exponential rate terms) than cases.  

Next week I'll be looking closely at the exponential rates around ages 25-35 to see if there's possible cross-flow of growth from the < 15 demographic.  

The interpretation and write up of this week's posts was more rushed/chaotic than usual so apologies for the typos etc!  If something doesn't make sense, I probably wrote some nonsense so please to ask!

Post edited at 18:06

 Si dH 23 Jan 2021
In reply to wintertree:

Thanks again. I like the new plot 16, I think that's quite powerful.

 Misha 23 Jan 2021
In reply to mick taylor:

> Thanks for these. 

> And 478k received their first dose of the vaccine. Which is better than very good

I’m starting to wonder which one of my hats I should eat. Although strictly I didn’t promise to eat any. 

Post edited at 19:59
 Michael Hood 23 Jan 2021
In reply to wintertree:

Thanks for all of those.

The exponential rate number (across all the different measures) really does seem to be the key number to see what's happened, what the situation is now, and what it's likely to be in the near future. Your graphs really do show that clearly.

Why are they still banging on about the R number, it's basically irrelevant if they don't include some kind of time element. You could have R=6 but if you only managed to infect someone every 2 months it would be virtually irrelevant in terms of pandemic control.

 RobAJones 23 Jan 2021
In reply to Michael Hood:

I agree, back in March I found more people found it were able to understand that cases were doubling every 4 days and what that might mean if it continued, r being about 3 was a fairly abstract concept. I still think that graph 9e illustrates that we could be in a much better position now, if more people understood that the restrictions (or lack of) around September were not sustainable. The doubling time over 40 days during that time was around 10 days, so cases will have increased by a factor of 16 over that period. Is realistic to have expected a change in policy at the beginning of Sept.? Given the fuss/misinformation  made about the data used for the November lock down, it would have required strong leadership.  On the other hand end Aug/beginning Sept. was a combination of eat out to help out, people returning from holidays and schools opening, but because infection level were low it didn't seem like a major problem to many people.

OP wintertree 23 Jan 2021
In reply to Michael Hood:

Excellent, another convert.

The other big problem with R is that the exponential rate constant is analytically related to the measurable (cases, hospitalisations, deaths) and so can be given to the same precision as the measurable.  To get an R value you have to combine the exponential rate constant with some other stuff that is known to only a poor precision and that doesn't really matter (*) in terms of healthcare overload and death rates. Because of this the R value is given with a wide confidence interval. This reduces its impact to science types looking at it because its seen as so poorly constrained, but all that uncertainty comes from stuff that drops out of the maths of the actual growth, and doesn't go on to affect the measurables that really matter.  

R is still important and has a lot of value to thinking about control measures - clearly allowing anyone to meet with less than R number of people during their infections period is going to reduce transmission, for example.  

I'd rather explain the situation to non-science, leadership types through doubling and halving times than though a week-on-week comparison or even a % change over time comparison.

(*) - this is a gross oversimplification.  The other stuff matters, but the way it matters is fully embodied in the measurable and its exponential rate constant.  There's a degeneracy between R and the period of time over which those infections occur, and there's great uncertainty in both, but that uncertainty appears when we try and estimate both values from the measurable and it's exponential rate constant.

Post edited at 22:57
In reply to Michael Hood:

> Why are they still banging on about the R number, it's basically irrelevant if they don't include some kind of time element

That was my opinion. But then I realised that there is a time element, due to the duration of the infectious period.

But I still think the simpler doubling rate is a much easier way to see what is really happening. And it is not subject to interpretation, unlike the obvious fudging that must be done to get R, resulting in a fairly wide range of R always being quoted.

[edit: should have read wintertree's reply first...]

Post edited at 23:06
OP wintertree 23 Jan 2021
In reply to captain paranoia:

> [edit: should have read wintertree's reply first...]

Yours is more concise.

The other benefit to the exponential rate constant is that it applies similarly to infections, cases, hospital admission and deaths, and you can give a qualitative explanation of the causality between them, and you can visualise it.  I think this can help to bring people along to a much more complete understanding of what's going on, where both the impact of absolute numbers and of policy can have clear focus.

You'd have to really abuse R to apply it to hospital admissions and deaths, and it's difficult even to apply it reasonably to cases.  

I'm more or less convinced that demographical and geographically broken down rate constants should be at the centre of a lot of the presentations and analysis used to inform policy making, and it isn't.  My conviction could just be a consequence of my having spent the last couple of decades going in to other people's fields and telling them they should be doing it differently...

 Si dH 24 Jan 2021
In reply to wintertree and others:

I think for the purposes of driving policy, rate constants are clearly better than r, but it is always better to have more than one source of information or set of analysis. Personally I find it useful to look at your exponential constants together with the dashboard data (both changes in averages and trends apparent in the very latest data) and then to understand the differences where they exist. Neither format should be used in isolation, certainly not by anyone who matters. Both can suffer from issues associated with data lags and calculational uncertainties; having both helps to identify these.

There is a difference though in what should drive policy and what should be communicated to the public. The r rate does have the benefit that it really drives it home to Joe Public that for the epidemic to get small, r has to be less than 1, which means if he or a family get covid, then on average they need to infect less than 1 person, so the aim really is to infect no-one at all. Exponential constants don't give that intuitive understanding (nor halving/doubling times.)

Edit to add: I also think r0 was used at the start of this pandemic because the disease was an unknown and scientists were looking for a way to compare the disease itself, independent of environmental factors as far as possible, with known diseases.

Edit to add, I do also agree exponential constants are a much better tool than r for intelligently discussing a given situation in an epidemic, like we try to do on these forums.

Post edited at 08:49
OP wintertree 24 Jan 2021
In reply to Si dH:

Yes, it's definitely horses for courses, but the rate constant / characteristic time is a horse that I've not seen in any of the briefings or minutes etc.  I think it gives a more intuitive way of presenting the numbers-invariant rate of change than a % week-on-week change, although they're the same information through different transforms.  The other thing I think is useful is to do a time-series presentation of the invariant measurement which could be a rate constant or a % change - heat maps and line plots.

I generally present absolute numbers and invariants together, either as two closely located plots or as a joint plot.  Constrains the choice of heat maps to get appropriate ones for each that don't clash / cause cross-confusion. 

I used to compare the stand-out parts of my rate constant plots with raw dashboard data quite a lot, but I think the raw data starts to loose meaning if started at too much - a sort of Semantic satiation - or perhaps as explained by John Hughes -  youtube.com/watch?v=gAzU-IZXaJk& .  It is important to check any sort of derived analysis against the raw data to understand if it's likely an artefact caused by the sheer scale of weird noise in the raw data.  A more professional approach would be to propagate information on the variance in the raw data through to an estimate of certainty in the measurement, but all the techniques I know for this assume the noise is a lot better behaved than that in cases.  

> Neither format should be used in isolation, certainly not by anyone who matters.

Indeed - absolute case numbers etc are the outcomes that actually break society, not the speed of growth, so they have to be front and centre as well.  The rate constants are the same information put through some mathematical transforms, they illustrate difference aspects of the same information and so help to convey more of the information.  I like the demographic joint heat map plots as a quick-glance that puts a lot of information across. Getting as much of the information presented in one go, as clearly as possible, is the best shot at getting non-science people to grock the bigger picture I think.

Ideally given the increased awareness of high cases producing more transmissible, more deadly and vaccine evading strains, and that examples of all these traits are probably already walking amongst us, we're going to see the maximum acceptable level of cases dialled back considerably.  To achieve this, we have to respond much more rapidly to where things are going, rather than to where they are.  This is where one can use a combination of actuals and exponential rate plots to talk people through a mental forecast of the next few weeks - to quickly build that picture in their heads - and engage them in understanding the effect of the levers they have control over.  Or so I'd like to think.  I've never actually been in a position of advising non science types on science.

I agree that R has a lot of value in both determine and communicating the level of restrictions.  

 RobAJones 24 Jan 2021
In reply to Si dH:

> There is a difference though in what should drive policy and what should be communicated to the public. The r rate does have the benefit that it really drives it home to Joe Public that for the epidemic to get small, r has to be less than 1, which means if he or a family get covid, then on average they need to infect less than 1 person, so the aim really is to infect no-one at all. Exponential constants don't give that intuitive understanding (nor halving/doubling times.)

I agree with your first paragraph but not so sure about this one. I agree that most people probably understand that if r>1 then daily cases will increase, but I think a number of people when informed r is about 1.1/1.2 thought that was near enough to not be a problem, especially if they heard 0.9 mentioned during the odd week.

Also on the personal aspect of risk assessment, it has been raised on here many times, that is usually used by people to try to  justify, bending the rules, "there is virtually no risk of me spreading the virus by doing x. Missing the bigger picture that if thousands or millions of people do x then it is a problem.

 I think more people could answer the question, if 10 people are infected at the moment in your town and the doubling time is 10 days, how many people will be infected by the end of the month, correctly, than the same question with an r value.

 Michael Hood 24 Jan 2021
In reply to wintertree:

> Excellent, another convert.

I was converted some weeks back 😁, it's just that for some reason it was much more obvious this week - me rather than anything in the data I think.

Post edited at 11:49
 Michael Hood 24 Jan 2021
In reply to captain paranoia:

> > Why are they still banging on about the R number, it's basically irrelevant if they don't include some kind of time element

> That was my opinion. But then I realised that there is a time element, due to the duration of the infectious period.

Yes, I'd forgotten about the implicit time period from that.

But using the exponential rate and even more so expressing it as the doubling/halving time, you can lead even dim-witted politicians through scenarios so that they'll understand what difference it makes to act straight away rather than dawdle about (e.g. doubling time = 20 days - keep your eye on it but you've got a bit of time to think about it, doubling time = 3 days - act NOW)

The fact (as mentioned by others) that it allows comparison across cases, hospital stats & deaths is also very powerful.

OP wintertree 24 Jan 2021
In reply to Michael Hood:

> I was converted some weeks back 😁, it's just that for some reason it was much more obvious this week - me rather than anything in the data I think.

It often takes a few weeks of using an approach and thinking about it for it to cross over into new found intuition.  Having things change direction in the last few weeks probably helped that process along.

The real trick is to get some training material and a good educator together so that come the next pandemic, they can walk in to a room with the cabinet and walk out an hour later with that same intuitive understanding.  You could also, for example, embed this in a cabinet wide pandemic planning exercise as part of a wider effort to insure systems and people are ready for the inevitable.  That seems like a good idea.

Post edited at 12:45
In reply to wintertree:

Whitty, Valance and JVT have mentioned doubling/halving times on occasion. It's almost as if they think in those terms but have been told by the PR machine to use R because the public can't relate.....

In reply to wintertree:

BBC News up-playing the vaccination numbers, but not making much of the 77 cases of the South African variant. 77 detected cases means there will be far more cases in the wild.

So, once again, they have failed to act in time to introduce international travel restriction and quarantine.

2
OP wintertree 24 Jan 2021
In reply to captain paranoia:

The number 77 has been rattling round my head all day.  They did say enhanced contact tracing is being applied - so presumably actual specialists and not the usual phone service.  But yes, I imagine there’s more out there, quite a bit more.  As it doesn’t show in the testing results (SGTF), and as sequencing is applied to a minority of cases, I wouldn’t even have confidence that there aren’t other chains of infection spreading outside the surveillance net of some sequencing and enhanced test/trace.

Let’s hope the current restrictions are enough to prevent it growing; I’m not hopeful especially as it’s likely to evade a lot of vaccine acquired immunity; I don’t know how much better (or worse) naturally acquired immunity is going to fare in terms of degraded efficacy.

In reply to captain paranoia:

It'll be everywhere. What's more cheery to think about is how we're almost guaranteed to be the first country to find the vaccine escape mutations, simply because we'll be the first to create significant selection pressure. So the world should really close its borders to us right now if we're advocating forethought.

2
 Luke90 24 Jan 2021
In reply to Longsufferingropeholder:

> Whitty, Valance and JVT have mentioned doubling/halving times on occasion. It's almost as if they think in those terms but have been told by the PR machine to use R because the public can't relate.....

If that's the case, I think it's the right call. I agree with you all that R is an inferior indicator to be putting in front of the (hopefully) smart and well-briefed people actually making decisions. But I think you underestimate the stupidity and mathematical illiteracy of broad swathes of the general public if you think doubling/halving times would be as easy to communicate as R has been. Simply the fact that higher is better on the way down and worse on the way up confuses matters. And yes, you could quote the underlying number instead because the direction is consistent but then most people won't understand what it actually means.

In reply to Luke90:

I've found it pretty striking when I've looked at some of the other sources of online guff. We're incredibly sheltered here on UKC. It's really brought home the average level of education/research/shit-giving on UKC compared to.... well... choose your example.

When I clicked on some of the Twitter links and that Mumsnet link soneone posted.... Jesus wept. It's night and day. I realise that might sound elitist but I don't know how to take the edge off it. It's a jungle out there. I can understand a lot more of the (lack of) public thought processes having ventured briefly into those dark places.

 Luke90 24 Jan 2021
In reply to Longsufferingropeholder:

Yeah, it's not just online. We all live in bubbles most of the time offline as well. A decade of secondary teaching showed me a much broader range of people than I otherwise encounter in the rest of my life. And most of that range would not comprehend 10% of wintertree's output here.

In reply to Luke90:

I'm reminded of something I read once: "Think of someone you know who is of average intelligence; has an IQ of about 100. Now, remember that half the world is thicker than them".

It's not a new thought, but I don't think climbers, at least the ones I've met, are a representative sample of the wider population.

Raises a loads of inclusivity questions that are well off topic so.... Sorry, will stop there.

Post edited at 18:33
 RobAJones 24 Jan 2021
In reply to Luke90:

> A decade of secondary teaching showed me a much broader range of people than I otherwise encounter in the rest of my life. And most of that range would not comprehend 10% of wintertree's output here.

I was thinking about that when reading your previous post. What do you think? Looking at the cases/admissions/deaths plots and understanding there is a lag, but the numbers are proportionally lower, not sure that kids who were going to get less than a 7 would fully understand that, so only top 20%. Not sure all kids getting 8's would able to understand and talk about plot 9e, so top 5% 

 Michael Hood 24 Jan 2021
In reply to Luke90:

But I think a lot of people could be guided (tutorial/lesson style) through a lot of Wintertree's stuff and thereby gain a better understanding of what's going on.

To just follow it as graphs and texts is more difficult, but then that's the same with most subjects and why most learning has some personal contact rather than just being out of textbooks.

Removed User 24 Jan 2021
In reply to wintertree:

Thanks for this thread as always.

Just seen this: https://www.biorxiv.org/content/10.1101/2021.01.18.427166v1.full.pdf?fbclid...

Not encouraging.

I guess if we are pragmatic about class/process type approval for the vaccines it could be mitigated.

OP wintertree 24 Jan 2021
In reply to Longsufferingropeholder & Luke90:

I think it's got a lot more to do with attitude than aptitude.  

A few people I know with the same letters after their name as me, or more, completely mis-judged this situation back in February and in some cases have struggled to really move on from that.

Generally speaking my experience is that if someone wants to understand something, and is willing to work with someone to do so, they can.  If they don't want to understand it, things are very different.  I'm not trying to reach a very general audience with these threads, but there's a lot that could be done to widen the presentation of the data - talking heads video pointing at things, more egregious use of "good", "bad" and "very bad" on axis labels, things like that.  This applies to anyone's presentation of the data, not just mine.  

The real problem I think is that a lot of people from all walks of life don't want to know, don't want to understand.  Sometimes they have a fixed world view often pegged to political tribal identity, sometimes they just don't want to know (entirely understandable - I saw the future last March [1] and haven't been very happy since.)  Re-engaging those people with the events going on around them is the real challenge; if someone succeeds at that, the rest will follow more naturally.

> When I clicked on some of the Twitter links and that Mumsnet link soneone posted.... Jesus wept.

Twitter horrifies me.  Their business model thrives of giving a platform to dangerous idiots and evil manipulators.  Someone standing up for the truth can rapidly escalate in to offline harassment to a crazy level (I'm thinking of both covid and non covid examples).  

[1] https://www.ukhillwalking.com/forums/off_belay/is_it_worth_it-717284?v=1#x9154...

Edit:  Casting my mind back to the early threads, there was a lot in incredulity at the early estimates of 510,000 dead from Covid in a "do nothing" scenario.  We have about 96,000 dead (by death certificate) and estimates are that about 10% of the population have had Covid.  To get to the point naturally acquired herd immunity wrapped things up, we'd probably require about 70% to 80% of people to catch it, meaning on the order of 680,000 to 780,000 dead.

The is an important lesson here about erring on the side of caution in the face of massive, poorly quantified unknowns and not leaning towards optimistic assumptions early on.

Post edited at 22:06
OP wintertree 24 Jan 2021
In reply to Removed UserBilberry:

I should say "thanks" for sharing that link....  It's a struggle however as I read it first...

Removed User 24 Jan 2021
In reply to wintertree:

> I should say "thanks" for sharing that link....  It's a struggle however as I read it first...


It's ok - I know.

Firstly - I cannot vouch for the bone fides of the article.  Of course, it is probably legitimate, but we live in a warped world.

Taking it on trust, I think there is mounting triangulated evidence (some circumstantial, but it triangulates) that the surface proteins are somewhat to highly polymorphic.  The question then becomes whether the variants suffer a compromise in return for increased transmissability/ pathogenicity.  To-date it appears not; in fact the circumstantial evidence is that the variant with improved transmissability is also more lethal.  Perhaps this is simply required lethal infective dose is more easily met; who knows; but it ain't good.

The only positive is that a pragmatic regulatory approach to the vaccine process, as opposed to the precise sequences involved, will enable a polymorphic vaccine to be delivered in a rapid cycle time - the sequences are known; the bulking process routine.  Seasonal 'flu already enjoys an off-the-radar subtle flexing of the rules to accomplish this.

Bottom line: what is the adjusted risk profile that balances polymorphic propensity to escape vaccines (or other strain dependent treatment such as plasma) with abilty to produce multi-valent vaccines on a cycle time to match.  A classic arms race.

 Misha 24 Jan 2021
In reply to wintertree:

Yes, that’s a good way of putting it and getting back to that 500k number. It’s a bit more complex I suspect as arguably the death toll we’ve had has focused on the most vulnerable people who haven’t been able to shield properly / WFH (as well as those who have simply been unlucky), so a let it rip scenario might not have caused as much death in proportionate terms. On the other hand, the NHS has generally been able to provide adequate care and just about got away from being totally overwhelmed (through a superhuman effort and L3 kicking in just in time) - clearly in a let it rip scenario, with the NHS actually overwhelmed, particularly back in spring when less was known about the disease and there wasn’t enough PPE, the death toll would have been higher. Similarly, the number of infections to date could be more than 10% - then again, more than 70% may be required for herd immunity. So there are some overs and unders but directionally you get to the 500k+ number, plus a wrecked health service.

I think you’re right that some people just don’t want to engage, either due to a lack of interest or due to a certain ‘political’ perspective.

The vaccination number reported today has gone up again, to just shy of 500k. Even if it dips due to reporting delays over the next two days, as long as this momentum can be maintained we might be out of the woods a bit sooner than I had expected. I’m thinking that in order to get back to T1 level restrictions (which would probably need to remain in place until spring next year), we’d need solid herd immunity in the adult population. Let’s say about 10% won’t or can’t have it. That leaves about 50m people. 100m doses. At 3m a week (430k a day) that’s 33 weeks, which is mid September. Sooner if they can maintain or increase the 500k/day. Not bad at all, considering. 

OP wintertree 24 Jan 2021
In reply to Removed UserBilberry:

Thansk for your take on the paper.  I didn’t have any red flags at this pre print - as you say though, it’s a warped world right now.  Generally the warping forces are trying to downplay risk though, and they tend to stand out clearly if you have a science background.

> A classic arms race.

Although perhaps a difficult one.  Some discussion on plotting #8 that the adenovirus carrier used for the AZ vaccine goes in to generate immunity against the virus itself as well as the added antigen/epitopes, meaning that it can’t be reused on the same individual with a different antigen.  This question has also been examined in an official capacity elsewhere going off some private comms.   I assume the synthetic vesicles ones can be reused, but they have more difficult supply issues (especially for the developing world).  I think there will be a few different carrier viruses available to deal with variants, but it’s a limited pool, and unlike changing the antigen I expect changing the carrier virus will require a full approval process rather than a recertification.  This full process was done in astounding record time but it still took long enough to allow a whole bunch of worrying variants to emerge.

I’m hoping that we see a new AZ vaccine ASAP that expresses both “classic” and “South African” spike proteins.

See https://www.ukhillwalking.com/forums/off_belay/friday_night_covid_plotting_8-7...

I’m trying not to be overly negative but I’m also about to go and watch Rick Potion #9.

Post edited at 22:53
Removed User 24 Jan 2021
In reply to wintertree:

Yes - I think we're going to see if RNA-based vaccines' flexibility and responsiveness are what they promised.

What I dread - really, really dread - is the news that patients who are confirmed to have survived current variants, have succumbed to new ones.

 Luke90 24 Jan 2021
In reply to wintertree (and also Michael Hood indirectly):

> I think it's got a lot more to do with attitude than aptitude.  

Absolutely, tribalism and fixed opinions regardless of evidence have a lot to answer for. And I'm sure you could explain almost all of your data very compellingly to someone of almost any intellect provided they were curious and willing to listen.

But I still think there's a more fundamental problem with mathematical literacy and more general skills of critical thinking and assessment of evidence. Because someone curious and willing to listen can be convinced of your valid, well-evidenced perspective, but in someone else's hands they can equally be convinced that lockdown is entirely unnecessary or Brexit leads to the promised land. Because without their own independent ability to assess the evidence and the arguments well, the shysters and the malevolent vested interests are indistinguishable from the genuine experts with good evidence.

And yes, very smart and able people can be duped as well, or wilfully ignorant to serve their own interests. But I think a lot of tribalism comes down to the fact that if you don't have the time or the comprehension to assess a lot of vastly complex issues for yourself, picking a team and sticking with them is the obvious choice really.

Removed User 24 Jan 2021
In reply to wintertree:

> I’m hoping that we see a new AZ vaccine ASAP that expresses both “classic” and “South African” spike proteins.

> I’m trying not to be overly negative but I’m also about to go and watch Rick Potion #9.

Sorry, meant to catch this in my reply.

The issue is not so much in the current variants but in the propensity to produce new variants AND these a) in the absence of any (significant) selective pressure b) that appear to be improving pathogenicity without other fitness detriment and c) the cycle time to address these with the current vaccine technology.

I've no doubt that we can make a vaccine for any variant, given time.  The challenge is the cycle time against the mutation of a binding protein that appears highly fllexible and hence highly variable.

 Luke90 24 Jan 2021
In reply to RobAJones:

> I was thinking about that when reading your previous post. What do you think? Looking at the cases/admissions/deaths plots and understanding there is a lag, but the numbers are proportionally lower, not sure that kids who were going to get less than a 7 would fully understand that, so only top 20%. Not sure all kids getting 8's would able to understand and talk about plot 9e, so top 5% 

Hard to put absolute figures on it but I reckon you're in the ballpark. Though did you mean 9e, that seems like one of the simplest?

In any case, I think dealing with teenagers a lot has probably made me overly cynical about adults, because brains obviously do keep developing and learning past 16 (which is when my contact with the weaker students generally ended). But on the flip side of that, the school environment is one where there's a trusted authority figure to guide them to the right understanding. Out in the big bad world, you've got all the competing interests trying to drag them in different directions. It's so easy to be mislead, or to just throw your hands up and say it's impossible to know the truth so forget it, go with what feels nicest.

OP wintertree 24 Jan 2021
In reply to Removed UserBilberry:

Thanks for that cheerful and prescient addition.

> b) that appear to be improving pathogenicity without other fitness detriment

We’ve always know it has the potential for this, but many people rationalised it as “a bad flu” rather than “a mild SARS-CoV-1” which could also be expressed as “a mild relation of a virus which had the potential to literally decimate society”.  Original SARS was so serious it wasn’t allowed to spread unchecked to the point it could decimate society.  So, the Red Queen looks for a new strategy.  Drop lethality to the point it can spread everywhere first, then pick it up again.  With the modern styles of medical care, increased lethality - up to a sweet spot - isn’t the detriment it used to be but can increase transmission, so long as it’s a slow acting lethality with gradually worsening illness first - which is what we see.

I’m anthropomorphising far too much here, and over-reaching in the comparison, but the selective pressure have changed a lot in recent decades and we know worse is out there.

Post edited at 23:33
Removed User 24 Jan 2021
In reply to wintertree:

> Thanks for that cheerful and prescient addition.

Hmm.  Sorry I guess.

I do think we have a serious problem.  I also think the world is never going to be the same again - we're going to have to get used to C19 as cause of death as much as cancer or dementia or heart failure or stroke.

At the moment we all reject that idea.  The virus must be beaten.  But both history and science tell us differently.  I think an arms race would be a good outcome just now,

OP wintertree 24 Jan 2021
In reply to Removed UserBilberry:

> I think an arms race would be a good outcome just now,

I think it's almost all we're left with at this point.

But I don't think the winning weapon will be vaccination given what we're seeing already.  The key breakthrough will be picking apart how the virus disrupts endothelial function and how it turns the host's immune system on the host, and coming up with pharmacological interventions that disrupt the key mechanisms and pathways.  I think the virus will have fever variations with regards mechanism of lethality than it will with regards immune evasion.

It's going to take some time to get these breakthroughs done and reduced to the point they're proscribed to people when needed; I'd rather we did so in a world with low case numbers and strong control measures, than one subject to endless cycles of healthcare overload, lockdown, death, educational and economic disruption and so on.

We need to identify the target outcome, and then figure out the least disruptive way to get there.  

The big difference I see is that the virus can cycle round through a large set of different configurations in an ouroboros of vaccine evasion, whilst the vaccine endlessly chases it.  But in terms of lethality, there are only so many mechanisms of action within its grasp, and we can learn those and we can pharmacologically diffuse those - or even genetically, one distant day.

Post edited at 23:56
Removed User 24 Jan 2021
In reply to wintertree:

Agreed.  We hoped the surface proteins would be the quick win.  Looks like they're not. 

Christ that's uncomfortable to say.

Removed User 25 Jan 2021
In reply to wintertree:

> So the Red Queen looks for a new strategy.  Drop lethality to the point it can spread everywhere first, then pick it up again.  With the modern styles of medical care, increased lethality - up to a sweet spot - isn’t the detriment it used to be but can increase transmission, so long as it’s a slow acting lethality with gradually worsening illness first - which is what we see.

> I’m anthropomorphising far too much here, and over-reaching in the comparison, but the selective pressure have changed a lot in recent decades and we know worse is out there.

You are a bit (although I agree re. selective pressure).  Killing the host isn't an optimising strategy for any virus.  In fact, frequently we see serious symptoms in man from diseases optimised to other hosts.  Poor adaptation is more frequently the issue.

Which is of no consolation when you're suffocating.

In reply to wintertree:

The nebulised interferon thing is in phase 3 trials now. Loads of -mab drugs are being tested. There's a lot in the pipeline, thankfully.

In reply to Bilberry: I had a cheery water cooler conversation a few months ago that circled around to conclude that we all have a fairly high chance of being killed by a coronavirus. Not necessarily any sooner, but it's challenged all our views on how we might go out. If we thought it'd be the classic flu or heart failure we don't any more.

Jesus this got dark didn't it? Sweet dreams everyone.

In reply to wintertree:

Oh, and one other macabre but possibly good news thought..... Remember we don't need a therapeutic cheap enough to save everyone, it just needs to be good enough for one nation to decide it's ethical to start running challenge trials. That'll open a floodgate of vaccine research that you can decide how you feel about.

 Michael Hood 25 Jan 2021
In reply to all:

My "can't get back to sleep" thread browse is not being helped by the direction this thread's taken. I'm not knowledgeable enough to understand lots of the detail but let me see if I've got this right...

The vaccination strategies (and our antibodies) have so far all been to attack the "spike" on the virus. The SA (and possibly other) variants have mutated in such a way that it looks like such attacks on the spike will be (or will become) ineffective.

Hence we need to be able to develop new vaccines more quickly to cope with any new mutations (that are sufficiently "deadly" and likely to become sufficiently predominant).

So we're likely to end up in an arms race against Coronavirus mutation, but importantly, each step in the arms race is merely a stopgap in the ongoing "game" rather than an endgame.

One possible endgame strategy is instead of fighting the virus directly, work out exactly how this class of viruses work with respect to making us severely ill (and killing us), i.e.how our bodies are affected by the virus, and hence develop an effective defence to this.

Part of this thinking is that there are relatively fewer ways we can be "attacked" compared to the possible number of "avoidance" mutations that the virus can make.

Overall conclusion is that Coronavirus isn't a short/medium term problem. The world is not in a position to stop it becoming a long term global problem which will become part of the "background" to our lives (and deaths).

Overall, that's rather depressing ☹️. I don't know if any of you are old enough to remember the background feeling in the 70's/80's that a completely globally catastrophic nuclear war was actually a significant possibility. If my "summary" above is basically correct, then Coronavirus is going to engender a similar feeling of there being a significant possibility of "life(style) as we know it coming to an end".

Post edited at 06:09
In reply to Michael Hood:

Sort of.

> The vaccination strategies (and our antibodies) have so far all been to attack the "spike" on the virus. The SA (and possibly other) variants have mutated in such a way that it looks like such attacks on the spike will be (or will become) ineffective.

These 'variants of concern', and notably the SA one, have taken the first steps down the road to that situation. Jury's out on how close they are to evading the vaccine but they're closer than any other widespread ones we've found.

> Hence we need to be able to develop new vaccines more quickly to cope with any new mutations (that are sufficiently "deadly" and likely to become sufficiently predominant).

This is true anyway; there will be mutations out there all over the place that happen to be less affected by the vaccines, but since they don't currently have any reason to 'win' they haven't blown up. When you wipe out all the strains the vaccine does work against, that changes. There have been a few papers already that, in vitro at least, demonstrate it's when, not if. 

> So we're likely to end up in an arms race against Coronavirus mutation, but importantly, each step in the arms race is merely a stopgap in the ongoing "game" rather than an endgame.

Depends. People will generate different immune responses to different parts of the spike protein. It probably won't be all or nothing. (Getting to the fringes of my understanding here). Might be that a vaccine can still prevent serious disease without being fully effective.

> One possible endgame strategy is instead of fighting the virus directly, work out exactly how this class of viruses work with respect to making us severely ill (and killing us), i.e.how our bodies are affected by the virus, and hence develop an effective defence to this.

This is an exciting area that's progressing fast. 

> Part of this thinking is that there are relatively fewer ways we can be "attacked" compared to the possible number of "avoidance" mutations that the virus can make.

If you can treat the symptoms or stop it making people sick there's less of a need to prevent the spread.

> Overall conclusion is that Coronavirus isn't a short/medium term problem. The world is not in a position to stop it becoming a long term global problem which will become part of the "background" to our lives (and deaths).

Don't see how it'll ever go away. As had been mentioned though, the optimisation of viruses like these tends to be mild symptoms but epic spread, like a cold. That's the best strategy for it and what we can hope it evolves into.

> Overall, that's rather depressing ☹️. I don't know if any of you are old enough to remember the background feeling in the 70's/80's that a completely globally catastrophic nuclear war was actually a significant possibility. If my "summary" above is basically correct, then Coronavirus is going to engender a similar feeling of there being a significant possibility of "life(style) as we know it coming to an end".

Lifestyle as we knew it won't be back for a while yet. Nobody knows how long. My biggest concern is whether we have to admit it's going to be long enough that we need to reconfigure society. There's only so long we can keep paying everyone to do nothing and if we don't get to a sustainable relationship with this virus quickly, some day we'll have to find new ways to pay the bills.

Things will get better than they are now, both short and long term, but when people say things like 'after covid' or 'when this is over' I'm not sure what they mean.

Sorry.

Post edited at 07:38
 Michael Hood 25 Jan 2021
In reply to Longsufferingropeholder:

Oh well, one positive upside is that some areas of science are going to develop at an incredible speed.

I wonder whether it's possible (or should I say will become possible) for us to force or guide virus mutations down a particular route that's less harmful to us.

OP wintertree 25 Jan 2021
In reply to Michael Hood:

I pretty much agree with how longsufferingropeholder answered.  Except their optimism over it evolving to be less deadly; that’s the long term fate of viruses it seems but we’re in the short term; I don’t think we’re guaranteed an easy road to that long term.

I assume (don’t know) that the coronavirus exposes more surface proteins than just the spike, and that a vaccine using those too would be harder to evade - but the “best” antibodies are those that bind to the tip of the spike protein which is the “receptor binding domain” - the bit that handshakes with the gateway in to the human cell and says “I am permitted to travel, let me in”.  With an antibody stuck to the tip, it can’t do that handshake.  Every time the tip changes enough, these “neutralising” antibodies are invalidated.  Others stick elsewhere but aren’t as effective. 


> of "life(style) as we know it coming to an end".

The future is certainly happening faster than expected.

In reply to wintertree:

I would have added more caveats but I thought it was already gloomy enough. My first cut at the wording was more like "the only future we are likely to experience is the one in which....."
Had to ask, didn't you

Removed User 25 Jan 2021
OP wintertree 25 Jan 2021
In reply to Longsufferingropeholder:

On the other hand, perhaps there are only a limited number of viable configurations for the receptor binding domain, they’ll all be identified soon enough, and a single vaccine can be made covering them all, and they’ll box this problem up once and for all.  

I don’t have any basis other than pessimism to assume the number of variations there will be too many for a single vaccine.  

In reply to wintertree:

Based on literally no knowledge I don't think that can be the case can it? Intuitively you'd think there must be a near infinite number of configurations of that protein. Unless we go for a vaccine that also works against angiotensin-converting enzyme..... which would be..... bad?
No idea. Poking in the dark now. Anyone got real knowledge?

 deepsoup 25 Jan 2021
In reply to wintertree:

It looks like Bilberry beat me to it but..  saw this, thought of you:
https://www.theguardian.com/world/2021/jan/25/the-information-warriors-figh...

 RobAJones 25 Jan 2021
In reply to Luke90:

> Though did you mean 9e, that seems like one of the simplest?

It probably is, but I didn't fancy trying to explain that 55 wasn't "halfway" between 10 and 100 without confusing the majority.

I think your points are spot on, if someone talked 50% (I was thinking 5+?) of students through the plots, and they were interested, they would think they had a degree of understanding at the end. As you say, a big problem arises if that was followed by Gupta/Henagen/Yeadon talking them through their "presentation"    How many will have a good enough understanding to identify which bits are misleading? The principle behind introducing more questions that involved  "problem solving" skills at GCSE, was reasonable, but it highlighted how few students possess these skills and how difficult  it is to teach.

 Dave Garnett 25 Jan 2021
In reply to Longsufferingropeholder:

> Based on literally no knowledge I don't think that can be the case can it? Intuitively you'd think there must be a near infinite number of configurations of that protein. Unless we go for a vaccine that also works against angiotensin-converting enzyme..... which would be..... bad?

Yes, I think that would be very bad, and fortunately quite difficult.  The last thing we need is autoantibodies to our own proteins, especially ones involved in complex homeostatic mechanisms like blood pressure

https://en.wikipedia.org/wiki/Angiotensin-converting_enzyme_2

The SARS-CoV-2 virus has two proteins accessible on its surface, the spike (S) protein and the membrane (M) protein.  The S protein is responsible to binding to ACE2 and the M protein seems to be involved in membrane fusion and internalisation into the host cell.  

https://www.hindawi.com/journals/bmri/2020/4389089/

The current vaccines encode the entire S protein (at least, the 'original version, which I assume was taken from the original Chinese sequence, presumably confirmed here).  As a result, the immune response will be polyclonal - that is different clones of T and B cells will be targeted to different target sites (epitopes) on the S protein.  

As Wintergreen says above, the ideal response would be 'sterilising immunity' where the antibodies generated would bind to the S protein of free virus so as to prevent them binding to their target ACE2 molecules, so neutralising the virus (and also tagging it for destruction).

However, there isn't really any mechanism for the immune system to pick epitopes on the basis of which would produce the best neutralising antibodies.  It's complicated but it depends more on which short sequences of peptides in the S protein happen to bind best to the HLA molecules of antigen presenting cells (and that also depends on which sets of HLA molecules a particular individual inherited).

It's also worth remembering that probably the most important part of the immune response the T cell response the kills virally-infected ACE2 +ve cells.  Nearly all cells constantly display fragments of a selection of their own proteins for inspection by patrolling T cells (which have been selected not to react to them after having tried to inspect them).  However, when they find a foreign protein (whether viral or sometimes a mutated host cell protein) they do react by killing the cell. 

The T-cell epitopes seen by these cells may not be the same ones preferred by antibodies.  They tend to be linear sequences of amino acids (because the HLA molecules present them stretched out in this way).  B-cells and antibodies see surface topology, which depends not only on the amino acid sequence but also the way that the protein is folded (conformational epitopes), and the amino acids they contact may be widely separated in the linear sequence.

This is why it's quite difficult to predict what the effect of mutations in the S-protein will be.  It seems clear the N501Y mutation is in the receptor binding site and that it increases the affinity of the S-protein/ACE2 interaction.  Some other mutations, particularly insertions or deletions of bits of sequence, are likely to slightly changed the shape of the way the proteins fold.

Nice clear article about structure, function and mutations here: 

https://www.nytimes.com/interactive/2021/health/coronavirus-mutations-B117-...

Not sure whether sign-in is required, but I think it's free.

Edit: seems to work OK, and it's really excellent.

Post edited at 11:13
 stp 25 Jan 2021
In reply to Longsufferingropeholder:

> It's not a new thought, but I don't think climbers, at least the ones I've met, are a representative sample of the wider population.

I've often wondered if climbers, as a generalised group, tend to be more analytical because climbing seems to have a higher emphasis on problem solving than many sports.

In reply to Dave Garnett:

Thanks for taking the time to summarise that. Really appreciate it.

OP wintertree 25 Jan 2021
In reply to Dave Garnett:

Thanks for the clear and detailed explanations. 

I'd not thought thought the consequences of the T-cell presentation mechanism being unfolded before; that significantly limits the scope of one substitution or deletion to disrupt that mechanism far from the site of the change - promising.

It seems that a vaccine encoding the fusion protein as well could have given more robustness against variants, and if neutralising antibodies can also block the fusion process...

Thinking about this got me wondering about allostery in the context of immune binding.  It looks like that's an interesting area to be working in right now.  If allostery is important to the binding to the ACE2 receptor, it's also possible that antibodies located far from the binding site could be more effective than naively assumed.

Post edited at 11:50
In reply to Dave Garnett:

One thing I still can't clear up with googling.... because I'm well out of my depth.....
Is the immunogenic "nucleocapsid" that's talked about in the literature the same thing as the "membrane"?

Edit: never mind. Found a paper that explains it.

Post edited at 12:01
OP wintertree 25 Jan 2021
In reply to thread:

Press release from Moderna to go along with a pre-print

https://investors.modernatx.com/news-releases/news-release-details/moderna-...

Their vaccine produces the same level of antibodies against the UK variant and 1/6th the level against the SA variant.  They think - with reason - that will still be enough to be effective against the variant, but that it may mean immunity fades faster.  I need to dig out the pre-print and read it; their logic on this applying is perhaps a stretch if key neutralising antibodies were in the other 5/6ths.

All the vaccines use the same antigen, so hopefully similar behaviour will be seen from the Pfizer/BioNTech and Oxford/AstroZeneca vaccines.

So, the SA variant may serve as the mother of all warning shots, but many not quite launch a fulls-speed arms race.  That would be basically the best possible news at this point if it pans out, as it means we still have a good shot at getting local cases downover the next couple of months and getting enhanced MIQ and surveillance in place for future variants that will drop that 1/6th further.

Post edited at 15:40
In reply to wintertree:

Possibly best news of the day. Hoping we can get on top of it before it makes any other changes.

I thought of another of those questions nobody's asking and possibly shouldn't:
If the AZ adenovirus vector is a one-shot deal, is there likely to be an argument for holding off using it in its current form and modifying it now? Or do we commit a future that relies on mRNA and finding new viral vectors? Or do we update the AZ one and the Sputnik one to produce whatever spike is in fashion next year and then do a swapsie with Russia? Or something in between? Or......
Any of our unpaid immunology tutors about?

OP wintertree 25 Jan 2021
In reply to Longsufferingropeholder:

> Hoping we can get on top of it before it makes any other changes.

It's a moral imperative

> I thought of another of those questions nobody's asking and possibly shouldn't:

That is an area I wasn't going to speculate on, as I don't think doing so can help improve the situation, but I can see how it might worsen it.  I will be happy when production of the synthetic vesicle based vaccines is scaled up and happening robustly from more different locations.

 Dave Garnett 25 Jan 2021
In reply to Longsufferingropeholder:

> I thought of another of those questions nobody's asking and possibly shouldn't:If the AZ adenovirus vector is a one-shot deal, is there likely to be an argument for holding off using it in its current form and modifying it now? Or do we commit a future that relies on mRNA and finding new viral vectors? Or do we update the AZ one and the Sputnik one to produce whatever spike is in fashion next year and then do a swapsie with Russia? Or something in between? Or......

Not completely clear what your question is but, technically, it's pretty easy to slot a modified S-protein sequence into the Oxford ChAdOx1 vector, or to slightly change the sequence of the RNA vaccines.  As to whether we we should switch to one of the new mutants, which one?  The Kent mutant, the South Africa mutant or the Brazilian one?

In principle, you could design a recombinant spike protein that included the interesting bits of all of the above, but I think you'd need quite a lot of both T-cell and B-cell epitope mapping for all of them.  Otherwise you risk not creating a perfect polyvalent immunogen for all variants but one that works less well than the current one for all of them.

   

Post edited at 17:06
 jonny taylor 25 Jan 2021
In reply to Longsufferingropeholder:

> If the AZ adenovirus vector is a one-shot deal, is there likely to be an argument for holding off using it in its current form and modifying it now? [...] Or do we update the AZ one and the Sputnik one to produce whatever spike is in fashion next year and then do a swapsie with Russia?

I haven't been keeping up with all the vaccines, but is this a strong argument against the mix-and-match of 1st and 2nd doses that the UK government was floating at one point, so as not to blow two opportunities in one go? Or is the AZ adenovirus the only problematic one that has been deployed in the west so far?

In reply to Dave Garnett:

What I was getting at was more that since vector immunity is a thing, what's the cost of using the AZ vaccine to deliver an effectively obsolete spike protein? Are there other, similar enough but different enough adenoviruses to call on?

Post edited at 17:15
 Richard J 25 Jan 2021
In reply to wintertree:

>  I will be happy when production of the synthetic vesicle based vaccines is scaled up and happening robustly from more different locations.

I suspect you might be interested (if you haven't seen it already) in this excellent overview of the production, materials requirements and supply chain issues for the mRNA vaccines.  It's complex and transnational!

https://blog.jonasneubert.com/2021/01/10/exploring-the-supply-chain-of-the-...

OP wintertree 25 Jan 2021
In reply to Richard J:

I hadn’t seen that; thanks for the link.  

It’s the fragility and lack of different suppliers for the lipid parts that stands out to me as the risk.   The conventional vaccines look like pretty standard bio lab stuff using standard consumables but these need specialist goops - reading your link, even more specialist than I’d thought.

Which is harder to set up new production for, the lipids or the vesicles?  

Post edited at 18:02
OP wintertree 25 Jan 2021
In reply to Removed UserBilberry:

> A little bit of light 

Yes; it's particularly promising to see a Tory MP joining in, and setting their sites clearly on the leadership at the CEBM.   It's great to see mainstream recognition of the organised web of astroturfing, manipulation and misinformation that's been going on since the start of this pandemic.   

I am pleased to see that Clare Craig now features on their FAQ.

Most of the key people on their FAQ have come up in discussion on UKC at some point; it's surprising how much effect such a small cabal have achieved - they are almost certainly the front people to a network run out of the United States by people with several decades of experience in sowing division in the public perception of science and in scraping the dregs of the academic and media barrels in order to do so.  I hope that they will start to list the various protest groups, their associated companies and links to the pro-Brexit campaign and other sham medical campaigning organisations.  

I feel strongly that this has been and remains an outright attack on the public health and the economy of the United Kingdom, and that the matter should be in the hands of our intelligence services by now.  Also, that the University of Oxford need to prepare three P45s.

 Richard J 25 Jan 2021
In reply to wintertree:

> I hadn’t seen that; thanks for the link.  

> It’s the fragility and lack of different suppliers for the lipid parts that stands out to me as the risk...

> Which is harder to set up new production for, the lipids or the vesicles?  

My guess is that it is the reagents for the RNA synthesis that are likely to be limiting.

I don't think the manufacturing stage for the vesicles is as difficult as that link implies (e.g. I'm pretty sure it doesn't use microfluidic techniques).  The point of it is that it is a bottom up process of self-assembly, so it really is just about controlled mixing in the right sequence.  Moderna contract this out to the Swiss contract manufacturer Lonza, who I guess are good at that sort of thing.  The lipids come from Avanti Polar Lipids, which is (was) a smallish academic spin-out in Alabama (maybe other possible suppliers too but that's the one I know Pfizer are using).  

Avanti have supplied these molecules at scales & pharma grades suitable for clinical trials and drugs for rare diseases, but they presumably have had to scale up an order of magnitude or two now.  Avanti was bought out by that fine Yorkshire company Croda last summer, presumably to help them achieve that scale.  They're fine speciality chemicals, that's the sort of thing they know how to do.  

 bruxist 25 Jan 2021
In reply to wintertree and Bilberry:

Unfortunate news being reported by the Handelsblatt (the German equivalent of the Financial Times): they say the German government is now calculating the Astra-Zeneca vaccine has an effectiveness of 8% in the over-65s. The headline reads, translated roughly, 'A Setback for Corona Vaccine: The Astra-Zeneca vaccine barely works for the elderly'.

https://www.handelsblatt.com/politik/deutschland/pandemie-bekaempfung-rueck...

OP wintertree 25 Jan 2021
In reply to Richard J:

Thanks.  I had assumed the mRNA reagents would be pretty bog standard, but I should really go and talk to some microbiology people as I'm usually wrong when I assume anything there...  I was surprised when you mentioned Croda a few weeks ago as being involved in the supply chain!  

It's fascinating if you step back a bit - it takes a global supply chain to produce something similar to what a single cell can do itself.  A couple of decades ago it wasn't possible at all.  I wonder where we'll be in 20 years time?

Do you know if anyone has looked at cultured exosome production and purification as an alternative approach to somewhat synthetic vaccine production?

 fmck 25 Jan 2021
In reply to wintertree:

I don't understand how the R value works. My wife was positive last week followed by myself. At the same time 3 of my children were showing the same symptoms. I knew the test was pointless as I had the same symptoms as her and the kids. They told me to come in for a test but not the kids. It seemed so pointless and no need to say I was positive as well.

So two positive results for a family of 5! 

 Richard J 25 Jan 2021
In reply to wintertree:

> It's fascinating if you step back a bit - it takes a global supply chain to produce something similar to what a single cell can do itself.  A couple of decades ago it wasn't possible at all.  I wonder where we'll be in 20 years time?

I'm pretty optimistic about the medium term.  I think it took longer to get to this point than people thought 20 years ago (the "stealth liposome" dates to 1990 as a concept).  But having found such a compelling application there'll be a huge impetus to improve and perfect this technology, which I think is very far from being optimised now.  (Don't know about the exosomes).

OP wintertree 25 Jan 2021
In reply to fmck:

> I don't understand how the R value work

It’s great on paper.  Like any mathematical abstraction the more you think about it applied to real situations, the more confusing it gets (or perhaps that’s just because I’m a bit slow...).

Rotten luck on the household infections, I hope it passes without drama.

OP wintertree 25 Jan 2021
In reply to Richard J:

> But having found such a compelling application there'll be a huge impetus to improve and perfect this technology, which I think is very far from being optimised now. 

Interesting; thanks.  

I’ve helped a little bit on some stuff with lab vesicle systems and it’s all so very interesting; you can see the general outline of the roadmap to a fully human created metabolic cell. It’s almost at the point we could try and reconstitute some e.coli grown IP3Rs into synthetic vesicles and try some really quite fun experiments on calcium signalling dynamics, especially if the vesicles could be held in a set of places by optical or electrostatic forces.   The first lockdown prompted me to make some big changes though so I’ll leave that to the next generation.  Various UKC discussions have got me thinking about what I could do with the appropriately functionalised synthetic vesicles however.

Post edited at 21:52
 Dave Garnett 25 Jan 2021
In reply to bruxist:

Seems to be bad news for the Merck vaccine (and maybe J&J too) although it does look slightly like a rookie error of designing a virus relying on the SARS-CoV-2 spike protein to get into cells... and then using an IM administration protocol.  Muscle cells don’t express ACE2...

Post edited at 22:48
 Dave Garnett 26 Jan 2021
In reply to Dave Garnett:

https://www.statnews.com/2021/01/25/in-a-major-setback-merck-to-stop-develo...

Couldn't figure out a way of including the link when I posted last night from my phone.

 joem 26 Jan 2021
In reply to bruxist:

Sounds like this may be complete bollocks, worth keeping an aye on this story though. 

https://www.theguardian.com/business/2021/jan/26/should-we-worry-about-clai...

1
OP wintertree 26 Jan 2021
In reply to Si dH:

>  I like the new plot 16, I think that's quite powerful.

I've updated it a bit; the annotations are now:

▲▲ (red) - case rates are above 1.0× their level at tiering 
▲ (orange) - case rates have risen at least 1.1× their post-lockdown minimum.  A horizontal line shows that minimum.
▼ (blue) - case rates have fallen by less than 0.1× of their lockdown level.

Bradford, Barnsley and Wakefield are the most worrying, all having risen above their level at lockdown in my filtered data, and being geographically clustered.  All three of these also featured in the plot from Saturday.  

Post edited at 15:59

 Si dH 26 Jan 2021
In reply to wintertree:

> >  I like the new plot 16, I think that's quite powerful.

> I've updated it a bit; the annotations are now:

> ▲▲ (red) - case rates are above 1.0× their level at tiering 

Do you mean at point lockdown began?

It definitely looks like South/West Yorkshire down through the midlands might be on edge a bit doesn't it? Bits of Leicestershire at LTLA level also looked worrying in the data last week but maybe when combined at UTLA level they fall just below your cut-offs.

Post edited at 16:35
OP wintertree 26 Jan 2021
In reply to Si dH:

> Do you mean at point lockdown began?

Sorry, yes!  Having a very slow day.  It's all gone very Groundhog Day in my head...

I do indeed mean since the lockdown began on Jan 5th.  

> It definitely looks like South Yorkshire down through the midlands might be on edge a bit doesn't it? Bits of Leicestershire at LTLA level also looked worrying in the data last week but maybe when combined at UTLA level they fall just below your cut-offs.

Things are bobbing over and under the cutoff at the moment.  

There is an LTLA level download although I've never looked at it -trying it now I get an error about a malformed URL parameter...

https://api.coronavirus.data.gov.uk/v2/data?areaType=ltla&metric=newCas... being the link generated for me by the dashboard.  

In reply to wintertree:

Would you like it to be named "The Barnsley variant" or are you going to get in there early and christen it "The Wintertree Strain"?

 fmck 26 Jan 2021
In reply to wintertree:

Thank you all doing fine. Wife gets out of the house tomorrow followed by me on Friday and the kids on Saturday. I understand we all got it at different times so our 10m day stint will have different dates. It does give me the question should we not be isolating until the last person has done 10 days. Doesn't make a lot of sense sometimes.

OP wintertree 26 Jan 2021
In reply to wintertree:

Updates to plot 22 with today's data.

The filtered trendline on the national level data looks to have "turned the corner" to having both occupancy and ITU occupancy decreasing.  

The regional plots shows most regions to have done this or to hopefully be in the process of doing it.  The North West is slightly hard to interpret as currently both occupancy and ITU occupancy shrink over the last few days without a lag between them, so instead of looping round the line doubles back over itself.  There's a lot more statistical noise in the regional level data.

The vaccination data is now out for the most recent weekend; it looks like vaccinations run at about half-rate over a weekend.  Comparing the last two weeks shows it's still increased this weekend, week-on-week:

  • 429,483 people were vaccinated last weekend
  • 500,006 people were vaccinated this weekend. 

Assuming no more growth in numbers from Friday just gone, vaccinations are running at a level of 3 million people per week which is astounding.  As I understand it, more centres have just come on line and so numbers should continue to rise - supply permitting.   Each week that goes by should hopefully remove ~4% of the population from becoming a case statistic or worse - once the vaccine has had time to take effect.   This should add appreciably to the decay rate of cases and follow on measures.  Given the demographics of the vaccination it should have a disproportionately high reduction in the number of people going in to hospital, which can't come soon enough.

Post edited at 17:28

 elsewhere 26 Jan 2021
In reply to wintertree:

> There is an LTLA level download although I've never looked at it -trying it now I get an error about a malformed URL parameter...

> https://api.coronavirus.data.gov.uk/v2/data?areaType=ltla&metric=newCas... being the link generated for me by the dashboard.  

URL works for me, I get a csv download (or json or xml).

Not sure why I get csv/json/xml download - other json opens directly in browser. 

Server response includes some http headers that force download rather than view "inline" within browser. Not seen that before so may cause obscure errors if you are using an HTTP library. 

content-disposition: attachment; filename="ltla_2021-01-26.csv"
content-security-policy: default-src 'none'; style-src 'self' 'unsafe-inline'

mick taylor 26 Jan 2021
In reply to wintertree:

>Given the demographics of the vaccination it should have a disproportionately high reduction in the number of people going in to hospital, which can't come soon enough.

Totally. What I tell my slightly stressed and inquisitive daughter is: there will be a point when the number of people dying will very quickly slow down. Reckon it will start mid Feb and by end Feb we should be in a radically different place. 

OP wintertree 26 Jan 2021
In reply to elsewhere:

Thanks; it’s working for me now - I must have hit it during the update cycle.  I’m busy cooking a batch of Wintertree’s secret recipe chilli now, but I might get to take a look at the data later...

 bruxist 26 Jan 2021
In reply to joem:

Alas, complete bollocks is much more likely to come from the Guardian - UK press standards are woefully low. Handelsblatt have backed up their story with a lot more detail here - https://www.handelsblatt.com/politik/deutschland/pandemiebekaempfung-kontro....

Broadly the new article says: the German Health Ministry held a conference to discuss changing the priority groups in their vaccination programme at the beginning of this week because they already had concerns about the AZ vaccine; two state ministries confirm that this discussion will resume on Saturday; they generally expect approval of the AZ to require a statutory restriction on the over-60s; and all this is because their internal data show an efficacy of under 10% in that age group.

The spiciest bit of the article is this quote from a senior civil servant in the Health Ministry about the idea they've mixed up the percentage of over-60s in the trial with percentage of efficacy: „Eine Verwechselung der Zahlen ist ausgeschlossen. Die Wirksamkeit bei den über 60-Jährigen liegt nach Ansicht der uns bislang zur Verfügung stehenden Daten bei unter zehn Prozent.“ ("A mix-up over the numbers is ruled out. According to the data available to us so far, the efficacy for over-60s is under ten per cent.") A lot can be read from German tone, and the tone of that statement is tabasco-sauce-strong.

There won't be any error in Handelsblatt's reporting, or in the statements given to them, so I hope that there's some error in the data available to the Ministry. I hope this most of all because we've already started using the AZ vaccine for the over-60s in the UK, though most older groups have had the BioNTech/Pfizer.

3
 Michael Hood 26 Jan 2021
In reply to wintertree:

According to my Dr's practice nurse, they're already onto 65-70 year age group in my area (Bury), but I presume it's a bit of a postcode lottery.

Has anyone seen any stats that shows the proportion vaccinated in each cohort? Nationally, regionally, etc.

 joem 26 Jan 2021
In reply to bruxist:

Why would you assume that there would be no error in one report and that  there Is in another? Links to a source in a language I don’t speak aren’t that helpful really. 
 

unfortunately we’ll find out who’s right. 

 joem 26 Jan 2021
In reply to Longsufferingropeholder:

That one reflects the article I linked above more than bruxist’s summaries 

 bruxist 26 Jan 2021
In reply to joem:

As I said, press standards in the UK are woefully low but even given this, it's worth assuming that a foreign-language report of a report is always going to contain errors. Handelsblatt is authoritative and has direct access to relevant sources in this case; whereas the Guardian is merely parroting with a slant and a linguistically-imperfect understanding of the report.

Press standards in Germany are much higher than in the UK - for example, press libel is a criminal offence, not a civil one - and so one simply doesn't find reckless falsehoods in print. I'm sorry you don't read German: it's a useful language to learn, especially if one is into climbing.

4
 Si dH 26 Jan 2021
In reply to bruxist:

Isn't your argument undermined by the fact that the German government have already refuted what Handelsblatt reported and no-one has any idea where their data came from?

Smells like the Guardian are probably right about this one.

 Misha 27 Jan 2021
In reply to wintertree:

> Bradford, Barnsley and Wakefield are the most worrying, all having risen above their level at lockdown in my filtered data, and being geographically clustered.  All three of these also featured in the plot from Saturday.  

I hope that's due to Yorkshire being relatively low previously and hence it being harder to achieve a reduction, rather than due to a new Yorkshire variant... 

 Misha 27 Jan 2021
In reply to wintertree:

Great to see the loops forming. It's a great graph once you get your head round it. Re vaccinations, the weekend snow may have had some impact as well due to travel disruption. 

 Misha 27 Jan 2021
In reply to bruxist:

Hopefully people will get to the bottom of this AZ confusion soon - whatever the right answer is, we need to know what it is. It strikes me though that by now there should be some early data available on vaccine efficacy based on people already vaccinated. I assume they ask people who get tested or admitted to hospital if they've had the vaccine, which one and when. I get that this is different to a proper randomised control trial over a period of months but you'd think that if AZ really is only 8% effective in over 65s, there would be a fair number of over 65s who've had it testing positive / being admitted. Perhaps the numbers of people who should in theory be protected aren't high enough yet due to the 3 weeks required to develop immunity but I'd have thought any major issues with efficacy (if there are any) should transpire in the next few weeks at the latest.

It also depends what being effective means - if someone gets Covid but doesn't develop severe Covid, that's still a lot better than nothing (eg the Sinovax data showed only 50% efficacy in terms of prevention but near 100% efficacy in preventing serious symptoms).

 Punter_Pro 27 Jan 2021
In reply to bruxist:

Apparently, so are press standards in Germany it would seem.

https://www.reuters.com/article/idUSKBN29V0ZC

https://www.dw.com/en/astrazeneca-german-reports-on-low-efficacy-on-over-65...

Fortunately, my partner works at Addenbrookes Hospital,Cambridge where AstraZeneca's head office is based, she works alongside some of the top immunologists in the country etc, she was talking about this yesterday evening with some of those in the know who are confident that this is the German government trying to pull a ''Wakefield'' all because of the AstraZeneca supplier issues etc.

I just hope the damage they will have now potentially caused isn't too high.

Post edited at 07:21
 Dave Garnett 27 Jan 2021
In reply to Punter_Pro:

> Fortunately, my partner works at Addenbrookes Hospital,Cambridge where AstraZeneca's head office is based, she works alongside some of the top immunologists in the country etc, she was talking about this yesterday evening with some of those in the know who are confident that this is the German government trying to pull a ''Wakefield'' all because of the AstraZeneca supplier issues etc.

Wow.  It seems a weird story to me but if it's a deliberate smear it's an indication of the degree of panic over how slow (relatively) the vaccination project has been to take off in mainland Europe.

 Punter_Pro 27 Jan 2021
In reply to Dave Garnett:

It might be, it might not be.

Emotions are clearly running very high at the moment but It just seems too much of a coincidence seeing that the EU have cocked up their orders big time and are now playing the blame game, using bully boy tactics by threatening to cut supplies of the Pfizer vaccine and you now have this claim on top as well.

I guess it will all be Ironed out over the next few days, the EMA are to set to announce their decision by the end of the week I believe with regards to approval so we shall see.

OP wintertree 27 Jan 2021
In reply to thread:

I'm going to break with my general trend of only analysis current data.  So, take this whole post with a pinch of salt.

Here is a plot to help understand where we are.  It is not a prediction but a way to contextualise the present.  I find this a useful way of thinking.  Others may vehemently disagree.  By comparing progress to this, we can see which direction that progress is going in - better or worse.

To try and "fairly" extrapolate the current situation forwards, I take the average (arithmetic mean) exponential decay rate, for each demographic age bin, over the last 2 days for which there are values, and project this forwards as a constant - this assumes that the exponential decay rate has reached its final value in all age bins, and won't be changing again.  I then use this to propagate cases/day forwards in each bin.  Finally, these are summed over the ages and a plot produced, below.  Annotations show today and the new target date for schools to reopen, March 8th.  I thought it was more appropriate to project with a demographic decay rate as it's clearly not constant with respect to age.

This extrapolation gives us an estimate of ~5,000 cases/day when schools re-open.  That's lower than we've seen since October.  This seems sensible, as the new variant is that much more transmissive that we need cases low to give the government time to respond to any worsening caused by schools re-opening (including the follow on return of more parents and carers to their workplaces or other activities).   It's also going to take some time to return the hospital occupancy levels to something that is sustainable.

It's hard to know what the decay rates are going to do, that would invalidate this prediction.  We haven't observed a decay from peak during lockdown with mass testing before in the UK; Wave 1 didn't have mass-testing, Wave 2 had lockdown fail likely due to the new variant.

  • My hope is that the decay rate will get more negative yet (shorter halving time), as a consequence of the vaccination program, and of non-linear effects (such as spread within hospitals and hospital occupancy).  I don't think it's gong to drop much more as a result of the lockdown, which has now probably been in place for long enough to be fully embedded in the most recent measurements.
  • It could become more positive if you take a "mixed populations" model, with cases halving faster in the population most able to take control measures like WFH, and halving slower in those that can't and those that won't.  When case numbers in the first population drop below those in the second, the halving time moves from being dominated by the first to the second.  In non maths speak, we run out of easy wins.  We're now perhaps passing the time an infections caught just before lockdown would spread a typical household so it's one to watch out fore.

Two ways of measuring progress would be to ask this extrapolation "what is the level on the schools reopening date" and "at what date do we drop below 1000 cases/day".  Currently that is 4950 cases/day and April 25th.


In reply to wintertree:

> It's hard to know what the decay rates are going to do, that would invalidate this prediction.  We haven't observed a decay from peak during lockdown with mass testing before in the UK; Wave 1 didn't have mass-testing, Wave 2 had lockdown fail likely due to the new variant.

We were "counting" deaths in the first wave... do they decay at a similar enough rate? Or at least give some hints at what the rate might do (or not do) next?

OP wintertree 27 Jan 2021
In reply to Longsufferingropeholder:

> We were "counting" deaths in the first wave... do they decay at a similar enough rate? Or at least give some hints at what the rate might do (or not do) next?

I think hospital treatment has changed so much since then, and the demographics were different for reasons including the care homes issues back then.  Throw in the new variant and it really doesn’t add much information I think - other than that the decay was sustained for a long time.

 Michael Hood 27 Jan 2021
In reply to wintertree:

As you're aware, that's a pretty big assumption (decay rate unchanging), but it would be nice to see your "current data" heat map with all that blue in it. Having said that, chances are this assumption will still be in the correct ball park.

Might be an idea to show a range of decay rates; i.e. your extrapolated decay rate +/- 10% or whatever.

Thanks. Good point.

In reply to Michael Hood:

> Might be an idea to show a range of decay rates; i.e. your extrapolated decay rate +/- 10% or whatever.

I'd be interested to see that too. Not that I suspect it changes the conclusion. It's back to looking like it'll be the live unattenuated vaccine for the majority, at the highest rate we can sustain.

OP wintertree 27 Jan 2021
In reply to Longsufferingropeholder:

I’ll work up a plot for Friday with some bounds on the rates.  I’ll also add one that plots the some measures vs cutoff date for the extrapolation - that makes it much easier to see if the future is looking better or worse over time.  No reason I can’t start it over the last 10 days or so - I think that’s a way of reducing all this down to a trustworthy plot of if it’s getting better or worse.

 elsewhere 27 Jan 2021
In reply to wintertree:

> I think hospital treatment has changed so much since then, and the demographics were different for reasons including the care homes issues back then.  

Death rate in hospital has fallen from a third to a fifth due to improved understanding and treatments.

Number 10 briefing yesterday at 708 seconds.

youtube.com/watch?v=HEl6EVErj28&t=708

In reply to wintertree:

Makes me want to play infinite games with what happens if measures ease on day y and we increase all the rates to what they were on (pre-lockdown) day x....
Satisfying curiosity will quickly become intractable without one of them write-your-own-pandemic websites that we saw a year ago, with loads of sliders. Probably already exists actually. I'm off for a google....

OP wintertree 27 Jan 2021
In reply to elsewhere:

Indeed; the problem is that the cases and death rates have a demographic dependence, as do deaths outside of hospitals.  I imagine this improvement has a demographic dependence  You can see this on the various plot 9 variants where the exponential rates never match in value (at peak, or lagged etc) between measures as the consequences of infection depend on age, and the exponential rate is - often very - different with age.  To actually unpick it all would need demographic data for hospitalisations and deaths as well as some serious inference to deal with the coarse age bins on hospitalisations.  By the time you pulled it all together as a model and constrained it to the data, it would need serious peer review to have any credibility, and even then I don’t think I’d have confidence in it.

I think the demographics need to be incorporated for a credible extrapolation, so cases is as far as this can go I think.

Post edited at 16:55
 elsewhere 27 Jan 2021
In reply to wintertree:

> Indeed; the problem is that the cases and death rates have a demographic dependence,

I make the crude assumption that hospital admissions are a uniform demographic of "a doctor thinks this person is the right person to admit at this moment" so the stated blanket average death rate applies.

As vaccination progresses that we should have less pressure on hospitals possibly resulting in the demographic changing to "a doctor thinks this less sick person is the right person to admit at this moment" resulting in an improved survival rate. Fingers crossed for both falling admissions and better survival rates.

Post edited at 17:25
 Toerag 27 Jan 2021
In reply to Longsufferingropeholder:

> Satisfying curiosity will quickly become intractable without one of them write-your-own-pandemic websites that we saw a year ago, with loads of sliders. Probably already exists actually. I'm off for a google....

Here you go. Dunno if it can be modified for the current situation though. 

https://ncase.me/covid-19/

In reply to Toerag:

Not enough sliders

 Toerag 27 Jan 2021
In reply to wintertree:

> I think hospital treatment has changed so much since then, and the demographics were different for reasons including the care homes issues back then.  Throw in the new variant and it really doesn’t add much information I think - other than that the decay was sustained for a long time.


Although we cannot compare death numbers to the first wave surely we can compare the death rate decline, at least from a week or two ago when they found the 'new' anti-inflammatories worked to reduce them by another 10%?  I think I'm right in saying that Dexamethsone, proning and CPAP were found after the main bulk of deaths in the first wave had occurred, so those won't have had an effect.

 Toerag 27 Jan 2021
In reply to Longsufferingropeholder:

Looks like you'll need to write your own - maybe a noddy excel thing? Or ask them to add more sliders to their tool

OP wintertree 27 Jan 2021
In reply to Toerag:

I think so, but that comparison will (hopefully) fail rapidly over the next few weeks, because the vaccination is largely happening in ranked age-decreasing order so the hospitalisation and then death rates should, at a whole-population level, fall much faster than the case rates as the average age of infection starts an ongoing decrease due to the ongoing vaccination program. 

Without a robust demographic model for the relationship between cases, hospitalisations and deaths, that data just isn't there to join all the pieces up usefully.

We also have an open question over the lethality of the new variant which changes what can be extrapolated from the brief period of decay after the 2nd wave.

To actually pull a useful analysis together would approach the amount of works groups like the LSHTM are doing, and would need a robust review to make it useful I think.  Bit much for a hobby project...

 Si dH 27 Jan 2021
In reply to wintertree:

> I'm going to break with my general trend of only analysis current data.  So, take this whole post with a pinch of salt.

> Here is a plot to help understand where we are.  It is not a prediction but a way to contextualise the present.  I find this a useful way of thinking.  Others may vehemently disagree.  By comparing progress to this, we can see which direction that progress is going in - better or worse.

I think it could be useful to compare where we get to in each of your Friday night threads vs this prediction. However I strongly suspect it will turn out to be optimistic. The prime reason for this is that the current exponential decline is not uniform. As you've already picked out some areas are not coming down much at all or are even seeing small recent rises. There are many more areas (some still at high rates) which are dropping, but more slowly than the national average. As rates drop across the country, the cases in the slower reducing or even rising areas will become more and more significant to the national total, reducing the overall average rate of decline. Further, they will bleed cases into adjacent areas , so those adjacent areas will also see a slowing rate of decline as the difference in local prevalence widens. The net effect will be that if nothing changes, the national average rate of reduction will decrease over the next few weeks because of the geographical non-uniformity.

That's my theory anyway I think getting back to the level your model estimates would be an extremely good outcome before schools go back, so despite all the above I think it's still possible an unexpectedly positive position can be achieved come March.

 bruxist 27 Jan 2021
In reply to Si dH:

I get what you mean, but the sequence matters here. Jens Spahn didn't refute or deny the initial report, but said "I don't think much of doing this speculatively in headlines." The Guardian's article was also reporting on the initial report. The one I've posted here is a second article, with what looks like a response from one of the Beamte (the civil servants in the Ministry of Health) to AZ and Spahn's comments. It's one of those situations where there's been an attempt to brush it off, and then more detail has emerged: this is the 'more detail' article.

 bruxist 27 Jan 2021
In reply to Punter_Pro:

Again, what I posted was the response, written and published after the two articles you've posted. The sequence matters: Handelsblatt's initial report came out, there's then an official ministry response from Spahn, and one from AZ, saying that Handelsblatt have mixed up two different 8 per cents (the source of the Guardian, DW, and Reuters articles); and then there is this article, with the response refuting AZ's explanation.

I'm less inclined than your partner's colleagues to leap to conspiracy theory, I guess; and I think that we've got used to such strange, abnormal behaviour from our own government that we start to think all governments behave in the same way. It's potentially a significant report. We'll find out this weekend if the EMA do or don't approve the AZ vaccine for the over-60s and, if they don't, the pressure will be on them to share their reasoning and data. There's obviously something amiss, and I hope it's with their data.

 bruxist 27 Jan 2021
In reply to Misha:

Yes, absolutely: the disquiet needs an answer sooner rather than later. Either there's a problem or there isn't; I guess we'll find out this weekend. I'd like to know sooner than that, as we've used the Pfizer vaccine for the NHS priority group strand of vaccines, but have been using AZ for frontline care workers, gravediggers, social workers, dentists, registrars et al since Monday. If there is actually a problem as asserted in the over-60s, not the over-65s, then I'd like to be lining up the 60-67-year-old workers for Pfizer rather than AZ.

OP wintertree 27 Jan 2021
In reply to Si dH:

A falling but large exponential masking a rising but small exponential - right until the moment it doesn't - is a common trope of this pandemic and you're right that we have some areas already showing the potential to grow enough to eventually dominate cases, as with the failed November lockdown.

The big question is "why"?  Local factors, or a new variant?  NZ have just confirmed two more community cases that trace back to their MIQ facilities; all are the SA variant and the latest ones trace back to an earlier case in the facility.  Hopefully they will soon announce if any breaches of protocol were discovered at the facility, or if this is a red flag that the variant is more transmissible, or has a different timeline in carriers than previous variants that will need extensions to the protocol.

If the developing hotspots are down to "local" issues, I don't think it casts a great pessimistic cloud over these extrapolations as that's more than balanced by the baked in pessimism of the vaccination not yet showing in the data used, and not being added in.  If the hotspots are down to the SA variant or similar, we'll know soon enough.  

Hopefully PHE are throwing everything they have at our trouble spots.  It's making the local papers now that they're becoming hotspots.

I've put something together to track how the extrapolations develop.  To be consistent, it needs to be run against the same week day for all weeks - due to residual weekend effect wobble in the cases data.  I'll re-set it to run against Mondays when the next set of updates comes out - for now there isn't enough data to get two data points on Mondays.

Post edited at 20:03

 Offwidth 28 Jan 2021
In reply to wintertree:

Some real concerns with the vaccination programme struggling to reach some of the most vulnerable.... 

https://www.theguardian.com/world/2021/jan/27/doctors-call-better-targeting...

 Offwidth 28 Jan 2021
In reply to Offwidth:

Some really good analysis on restriction compliance showing it is mostly very good despite all the hot air  The sub thread looks at where there are problem areas and by far the most serious reason is low paid workers who can't afford to self isolate. 

 https://twitter.com/rowlsmanthorpe/status/1349459843833352192

More worryingly the NHS gossip grapevine is showing more indications of heavy handed and arguably dangerous PHE intervention. Where Pfizer needs using up or be wasted it is said some vaccination centres have been explicitly told it is not allowed to use this as a second dose for NHS frontline staff. Public Health England seemingly putting politics in front of maximising NHS employee safety.

Post edited at 10:48
 Toerag 28 Jan 2021
In reply to Offwidth:

It's ridiculous to waste vaccine, as long as it can be recorded properly people should be vaccinated.

mattmurphy 28 Jan 2021
In reply to Offwidth:

> Some real concerns with the vaccination programme struggling to reach some of the most vulnerable.... 

I’ve never understood why the BMA gets treated as a similar authority to the JCVI.

A doctors trade union run by GPs doesn’t have the same level of knowledge as a group of immunisation experts.

The article in question is calling for the JCVI’s guidance to be deviated from to target other vulnerable groups. Yes they need vaccinating, but  the additional complexity of changing the schedule would cause issues. Speed and volume are key here.

Why the Guardian publishes this rubbish is beyond me and in general The Guardian has been shocking in terms of its Covid coverage.

Publishing claims about the first does of Pfizer only having 33% efficacy without the claims being validated and prominently reporting on the German nonsense that the AZ vaccine has only 8% efficacy in over 65s.

Really dangerous stuff.

3
 Punter_Pro 28 Jan 2021
In reply to topic:

So the Germans are sticking to their guns and have not recommended the Vaccine to the over 65's

https://news.trust.org/item/20210128115400-59715/

The data shows an estimated efficacy of 6.3% for the over 65's

https://mobile.twitter.com/olivernmoody/status/1354782232876085249

However, I still say that the Hasselblatt report was misleading as there simply isn't enough data to draw a conclusion to it, it doesn't mean it is less effective for that age range and for them to flat out say it is definitely only 8% is just dangerous and the perfect fuel for the anti vaxers etc.

I guess we are now the guinea pigs for the AZ vaccine and whether it is effective or not for the elderly, hopefully we will have enough data soon.

Post edited at 14:34
1
In reply to Punter_Pro:

n=1. Yes! Science!


I actually LOL'd

 jonny taylor 28 Jan 2021
In reply to Punter_Pro:

> The data shows a confidence interval of 6.3% for the over 65's
> https://mobile.twitter.com/olivernmoody/status/1354782232876085249 
> However, I still say that the Hasselblad report was misleading as there simply isn't enough data to draw a conclusion to it

That might have been a typo by you, but if I'm understanding correctly it's not the *confidence interval* that is 6.3%.

If I am understanding that table, it is saying that the nominal effectiveness is 6.3% but that the true figure could be anywhere between 94% and minus 1400% (i.e the spreadsheet is going nuts unable to calculate anything meaningful from the small numbers). But I don't understand where the 8% figure ever came from - is the table you linked the source for the 8% article in the first place?

If I'm understanding this correctly, it's beyond irresponsible that anybody has used the 6.3 (or 8?) percent figure for anything, ever.

Post edited at 14:29
OP wintertree 28 Jan 2021
In reply to Punter_Pro:

Good commentary on that Twitter link.

I'm commenting in a quiet corner of the internet that doesn't influence much, and I have tried to be very mindful of not producing "technically correct" discussion that is not suitably qualified and that could be twisted in to fuel for the pro-disease fire. 

To state an efficacy from that data without including the CI is a mockery.   Entirely in keeping for a tabloid like Bild, but their broadsheet should hold themselves in contempt of proper reporting.

If the Germans want to make a decision based purely on that data alone and not look at other pertinent data, that is their decision to make.

As the UK presses ahead with rapid immunisation coincident with very high infection rates, that confidence interval will soon shrink to something that is appreciably smaller than the possible range of values and meaningful data will emerge.  I hope for everyones sakes that the data will allow them to change their minds soon enough.

 Punter_Pro 28 Jan 2021
In reply to jonny taylor:

Yes apologies, a typo, I have just corrected it. But yes, that would be my understanding of it also....

 Offwidth 28 Jan 2021
In reply to mattmurphy:

I'd agree what the BMA said on 12 weeks vs 3 weeks was wrong . They could have worded it in a way that was acceptable but chose not to. They simply don't know the government evidence on which the decision was made (as the government is refusing to be transparent on that detail). What they are saying about ethnic minorities in that Guardian article simply isn't what you claim and is following JCVI advice. They highlight real issues with the programme compared to who is actually most vulnerable and look for some local flexibility and increased focus on the issue. It was obvious from the beginning that poor and older BAME citizens with health problems and crowded living conditions, made them especially vulnerable and really needed vaccination before middle class octogenarians living at home with no serious health problems. The same applied to why those in care homes should really have been vaccinated before healthy octogenarians (as Scotland did).

Interestingly on the subject of transparency Nichola said today although the UK government forced them to remove information on vaccine logistics they are going to put it up again in the public interest. Scotland are catching up fast on the first phase and look due to overtake and meet their first target before England. 

Post edited at 15:04
 neilh 28 Jan 2021
In reply to wintertree:

Did not the authorities in the Uk and USA also want more data from AZ before approving.But my memory is hazy on these issues.

In reply to Offwidth:

Last night this caught my eye in the BMJ:
https://www.bmj.com/content/372/bmj.n86

That was published on the 12th of Jan.

Edit: if it's too long to read, just search for the sentence beginning "Absolutely fascinatingly" and go from there.

Post edited at 15:17
OP wintertree 28 Jan 2021
In reply to jonny taylor:

>  (i.e the spreadsheet is going nuts unable to calculate anything meaningful from the small numbers)

You know what really bothered me?  All the CIs were given to 1 decimal place, except for this one which was given to zero.  How duo they ever expect me to take them seriously if they can't be consistent.  I mean the actual value could be anywhere from -1405.4% to -1404.6%... 

Edit: Well, that and giving a CI that is clearly the gibberish result of an algorithm failing, as you say...

Post edited at 15:22
In reply to wintertree:

You probably have noticed, but I'm not sure you've noticed........
They had one case in the vaccine arm, and one case in the control arm.

That's one case in each group. One.
They've then taken the inexplicable step of trying to turn that into an efficacy figure and then someone has written it down, presumably as some kind of joke that got way out of hand.
I'd love to see the maths that they used to turn a one-all scoreline into a number.

Post edited at 15:33
Removed User 28 Jan 2021
In reply to wintertree:

This is an interesting thread where by a mathematician who has been looking at infections in the over 80s. His curve fitting suggests that the vaccination program may be starting to have an effect.

https://twitter.com/JamesWard73/status/1354787659441385473?s=19

OP wintertree 28 Jan 2021
In reply to Longsufferingropeholder:

Indeed.  

Going as far as calculating a value and giving it for that case is not "good science", and clearly any reasonable person would a * in the table cell, with a footnote reading "* insufficient numbers for a determination.

But.... the nonsense confidence interval bothers me a lot more though, as it implies that they are using incorrect mathematics to analyse the data and they didn't notice.  No statistically valid approach can give a CI that exceeds the range of physically possible data.  If an algorithm is incorrect on a corner case, is it correct on better data, or just less incorrect?  Failure in this sort of maths is not normally stepwise by progressive.

OP wintertree 28 Jan 2021
In reply to Removed User:

> This is an interesting thread where by a mathematician who has been looking at infections in the over 80s. His curve fitting suggests that the vaccination program may be starting to have an effect.

Thanks; I'll have a read tonight.  I've been looking at my demographic breakdowns carefully.  I think it might show through clearly by this Friday's update (now due on a Saturday...).  Let's hope so!

Post edited at 15:43
In reply to wintertree:

They must have noticed.
They don't approve the AZ vaccine for over 65's. For the next weeks/months they're only vaccinating over 65s. Noise about supply of AZ vaccine - dealt with.

 Michael Hood 28 Jan 2021
In reply to Longsufferingropeholder:

Not sure of the basic maths but you'll notice that they're comparing 1/341 (vaccine) to 1/319 (control). The %age difference between those is 6.45% or 6.9% depending on which one you use as the denominator. Rather close to 6.3% so I suspect it comes from some kind of sophisticated statistical variation on that simple calculation 😁

As everyone's saying, 1 +ve in each group means "not enough evidence" - somebody should have noticed that rather than allowing the modelling whatever to follow through.

It's the usual problem where a sophisticated model/calculation is done and nobody's bothered to do the simple back of an envelope calc to make sure it's at least sensibly in the ball park.

 Michael Hood 28 Jan 2021
In reply to wintertree:

The maths/algorithm may "technically" be correct if they didn't have "cutoff" limit checks on the confidence limits because they assumed they would always be narrow enough to be sensibly in range.

But the big thing is as you say "and they didn't notice"

So a confidence limit of -1405 is probably about correct as an indication of how much confidence to have in their work 😁

Post edited at 16:33
In reply to Michael Hood:

Yeah I did start reading about how they get from the numbers to an efficacy a while ago. What I learned was that it's really boring. It's slightly more complicated than what you've done there, but mostly really dull stats. Either way, they've used one out of 341 and one out of 319 to get to their number so yeah, it's not great.

 minimike 28 Jan 2021
In reply to Michael Hood:

The fact they’re quoting CIs at negative values rather than just zero is ridiculous. Unless the vaccine GIVES you covid.. (for clarity, it doesn’t and can’t, ok?)

In reply to minimike:

> The fact they’re quoting CIs at negative values rather than just zero is ridiculous. Unless the vaccine GIVES you covid.. (for clarity, it doesn’t and can’t, ok?)

From the data they have, they are unable to conclude that. Yes it is ridiculous.

Edit: If they hadn't seen that one case in the control arm, and went ahead blindly with the algorithm in the same way, they would have presumably had to conclude, with a huge CI, that the vaccine gives you covid. And that's no more or less ridiculous than what they've done.

Post edited at 17:35
OP wintertree 28 Jan 2021
In reply to Michael Hood:

> The maths/algorithm may "technically" be correct if they didn't have "cutoff" limit checks on the confidence limits because they assumed they would always be narrow enough to be sensibly in range.

A rigorous approach derived from correct precepts (e.g. as Minimike says, that the vaccine can't give you Covid) would not be able to produce a -ve value for a CI; it suggests to me that they've used some pre-canned formula that isn't actually applicable for n~1.

> So a confidence limit of -1405 is probably about correct as an indication of how much confidence to have in their work

Or at least in their ability to communicate the pertinent findings of their work...

> It's the usual problem where a sophisticated model/calculation is done and nobody's bothered to do the simple back of an envelope calc to make sure it's at least sensibly in the ball park.

I don't think the modelling can actually have been valid to have produced this, but I totally agree with your comment on the need to cross-check modelling; preferably against ballpark estimates and other models.

Post edited at 17:56
 elsewhere 28 Jan 2021
In reply to Longsufferingropeholder:

The must have noticed and just chucked it into the spreadsheet thinking "it's obvious, everybody knows it's insufficient data".

Using the logic for not vaccinating the over 65s, you'll never offer centenarians any vaccine or drug because there aren't enough* of them to do a big enough clinical trial. 

*13,000 or 0.02% of the population in the UK. If you had to test every drug and vaccine on them they'd be like pin cushions!

At some point you have to say "we have no clinical trail data for centenarians, diabetic men who have had prostate surgery, a specific ethnic minority or some other cohort, but we will try it anyway".

Post edited at 18:16

771710 tests. That's a good number.

 Si dH 28 Jan 2021
In reply to Longsufferingropeholder:

> 771710 tests. That's a good number.

Yes, although interestingly not far off half of those were LFT. PCR tests totalled 416k. There has been a fairly rapid, approximately linear increase in LFTS taking place in all regions since new year, the usage is now fairly evenly spread around the country (England at least.) I'm not sure what the main purposes they are being used for are.

Post edited at 18:36
In reply to Si dH:

It's not clear to me either. Still, I'll take an LFT over what we had in March.
In other good news, this is potentially exciting:
https://www.bbc.co.uk/news/health-55840547

OP wintertree 28 Jan 2021
In reply to Longsufferingropeholder:

If that includes the sequencing component that’s fantastic assuming the information is being used.

OP wintertree 28 Jan 2021
In reply to Michael Hood:

> Might be an idea to show a range of decay rates; i.e. your extrapolated decay rate +/- 10% or whatever.

Ta-da.

This multiplies the exponential rate constants (for each demographic bin) by 0.9x and 1.1x to do ±10% to give some idea of the expected variance.  It's not a representation of any sort of uncertainty in the data, purely indicative.

One notable effect is that cases were about level around Jan 5th, meaning a rate constant of ~0.  Making this proportionally larger or smaller keeps it at 0, so the blue shading is very tight.  Perhaps I should fuzz that data by doing ± a specific amount on the rate constant instead.

This data is very noisy and the most recent 8 days are subject to change as the estimates of exponential rate constant for those days will improve with more data - I suspect that it will pull the RHS of this plot down a bit more as it comes in, currently the RHS is in a period of elevated reporting from the "weekend effect".  

The vertical grid lines shown days which are a Monday, so that like can be compared with like, and because I consider Mondays to the best point in the data to measure things. 

Some other observations:

  • Deaths are now in decay by the week-on-week measure used for the government dashboard, meaning that the whole dashboard is now green.
  • Looking at the data in the provisional window that will resolve in time for my Saturday update, things are looking promising for cases.
  • I've updated Plot 22, it looks like ITU occupancy has now turned a corner as well as general occupancy.    It really does take a long time to start falling after cases.  I can't imagine how the staff feel, particularly from ITUs.  If there's any staff reading from the County Durham area, drop me a PM over the summer and there's a round on me...  (If there's any pubs by then).
  • The vaccination rate is down, hopefully that's just the logistics machine continuing to come together.  It's still high enough to exceed 2m people per week.
  • The deaths by death certificate data is now available from the dashboard with daily resolution as well as weekly. 
Post edited at 20:36

 bruxist 28 Jan 2021
In reply to wintertree:

I'm really disappointed in that Twitter thread, and especially in Oliver Moody, who has clipped the table out of the RKI report and pretended that it's the whole of the data. It's actually one table in the section on the AZ vaccine, the main part of which does give the CI and explicitly states that the efficacy rate is statistically insignificant. A copy is here; relevant section is on p.35 (yes, it's in German, sorry!: https://www.handelsblatt.com/downloads/26863416/2/stiko-entwurf-28-01.) When Moody protests that "the data presented here is completely inadequate", he's right. He's cropped it.

What's seemed clear to me reading the German reporting since Tuesday is that both Spahn and the RKI are simply unhappy with the lack of data, and aren't going to approve AZ on emergency grounds without supporting data. The one thing that really doesn't bode well about this situation is the beginning of hostile reporting and media twisting. That in this case it's coming from Moody, Berlin correspondent with the Times, just makes me rather despondent.

3
OP wintertree 28 Jan 2021
In reply to bruxist:

Interesting, thanks.  One of those times I really wish I'd picked up languages.

Skimming the document

  • They also have a -ve CI value for the BioNTech vaccine for > 75s  - Tabelle 9 the vaccination effectiveness value is 75% with a CI of  [-13.1% to 100%] 
  • They have no CI for the Moderna vaccine (k.A. which I assume is like our N.A?) for over 75s.

It seems all the studies unsurprisingly suffer from weak statistical power in the oldest ages - although the AZ suffers this worse.  It is interesting that the Pfizer / BioNTech is approved for over 75s on the basis of this data, as the CI still renders the finding uncertain - but less so than AZ.

> the main part of which does give the CI and explicitly states that the efficacy rate is statistically insignificant. 

Good. I still think there's something wrong with having a negative value in the CI; it is inconceivable that the full report wouldn't recognise this, and a prominent response on the Moody thread notes this.

So, the question to me is "do the German newspaper reports convey the statistical insignificance of this number" - although given the misunderstanding with 8% it's all just out of kilter.

My understanding of Moody's twitter thread - and why I think I disagree with your assessment - is that he is explaining the problem with "the story" - as in the newspaper reports, and when he says "But what the story missed is that this figure is completely useless. They just don't have enough data to say anything meaningful.", he is echoing what you have said.  

I think there's some crossed wires somewhere on one of our behalves (me, you, Moody, or more than one).  Probably what comes from trying to tackle something nuanced through a medium like Twitter!  He could have said “as recognised in the wider document, this figure has no statistical power” but that’s not exactly Twitter friendly.  There’s a reason I avoid Twitter...

> That in this case it's coming from Moody, Berlin correspondent with the Times, just makes me rather despondent.

I thought he was trying to un-pick where the German media reports misrepresented (by omission) the decision making process embodied in the German government report that you linked, but perhaps I'm missing something.  Like being able to read German for a start...

Post edited at 22:14
 Richard J 28 Jan 2021
In reply to wintertree:

Excellent news on the phase III results from Novavax just out, 90% efficacy.  Doesn't sound like they've got as far with manufacturing at scale as AZ, though.  An engineered protein nanoparticle, to be brewed up on Teeside, it seems, so perhaps a less complicated supply chain than the mRNA vaccines (but not sure about the somewhat black magic sounding "plant based adjuvant").

 Si dH 28 Jan 2021
In reply to Richard J:

Yep, good news, I was just reading this article:

https://ir.novavax.com/news-releases/news-release-details/novavax-covid-19-...

The South African trial results are of interest as well as the UK ones.

OP wintertree 28 Jan 2021
In reply to Richard J:

Great news.  Is this associated with CPI’s Teeside presence?

Is that a phage derived protein nanoparticle?   [Edit: A quick google image search later] Their marketing graphics don’t look like a phage.

It’s astounding how many different research based delivery vehicles are getting a boost from this pandemic. 

> but not sure about the somewhat black magic sounding "plant based adjuvant").

It seems like there’s a lot more diversity of delivery vehicles and adjuvants out there than of antigens.  It does make me wonder how much of the secret sauce to the efficacy is on the adjuvants, and - once data from widespread immunisation is analysed - if there is mileage in pooling adjuvant technology to improve efficacy down the line.

In terms of plant based stuff; it certainly makes tight QC combined with high throughput more difficult than stuff made by chemists...  That’s somewhere on the list of causes for my greying hair.   

Post edited at 22:26
OP wintertree 28 Jan 2021
In reply to Si dH:

> The South African trial results are of interest as well as the UK ones.

Fate protects fools, little children, ships named Enterprise and global public health.

Another sign that the SA variant is hopefully the mother of all warning shots and not a total reset of the situation.

 Richard J 28 Jan 2021
In reply to wintertree:

Fujifilm's biotech CMO in Billingham, I believe.  Not directly dependent on CPI, though they certainly work with them on various projects.

In reply to wintertree:

'Extrapolated cases on 2021-03-08'...?

Have I missed the time machine...?

 bruxist 28 Jan 2021
In reply to wintertree:

I don't quite know how to explain those negative values and it isn't my field - I can only guess that German literalism demands them - but I do know that in presentation of such data in Germany, the 'Tabellen' are subordinate to the summary. One reads the summary first, then uses the table/chart/picture as an aide-mémoire. You know how, when you see old British newspapers, they have a wall of text and a tiny illustration or picture? British papers have it the other way round nowadays - a huge picture with the text almost as an adorning frame - but German presentation still expects the reader to read first and look at the pictures & graphs afterwards. I really would have expected Moody to know this, and not to clip just the bit no-one looks at.

The much weirder bit on his Twitter feed is the bit you identify, where he says "But what the story missed is that this figure is completely useless. They just don't have enough data to say anything meaningful". It's weird because that point he's making is made explicitly and repeatedly throughout Handelsblatt's article. It's like he's repeating their reporting whilst relying on most readers' inability to check whether what he's saying is true or not. The Handelsblatt report repeatedly say things like "the number of older participants was still too low to provide statistically meaningful data on the effectiveness of the Covid vaccine".

Where you say "I thought he was trying to un-pick where the German media reports misrepresented (by omission) the decision making process embodied in the German government report that you linked" - I'm not sure. The papers on Monday and Tuesday were basically printing quotes from members of the relevant ministries leaking it in advance, after which this RKI report with the decision-making has only been released today. Had I been reading German media and the RKI reports only, this would have all made pretty straightforward, logical sense, even if I disagreed with the decision. And if I look at other non-UK reporting on the issue - say, CNN: https://edition.cnn.com/2021/01/28/europe/germany-astrazeneca-vaccine-coron... - that seems pretty straight too.

Coming back down to basics, I would reckon this is what we should have expected anyway. There wasn't much chance of the EU or the US regulators approving the AZ vaccine for over-60s on slim evidence; we on the other hand have gambled on theoretical but unproven efficacy and, if it succeeds, the EU and US will look like they've been fruitlessly cautious.

OP wintertree 28 Jan 2021
In reply to captain paranoia:

> 'Extrapolated cases on 2021-03-08'...?

> Have I missed the time machine...?

See here https://www.ukhillwalking.com/forums/off_belay/friday_night_covid_plotting_9-7...

It’s using the England wide demographic breakdown of cases data to extrapolate forwards from the x-axis date to 2021-03-08, by taking the cases/dayand exponential rate constant from the x-axis date and running that forwards to 2021-03-08.  It’s not a forecast as the rate constants will change (hopefully for the better from vaccination and warming weather; possibly for the worse from social or virus variant factors).  What it does to is give us a way to see if the situation is improving or worsening, in a way that will adapt to the demographic dependences of absolute case numbers and vaccinations as they change with time.

 Richard J 28 Jan 2021
In reply to wintertree:

> It seems like there’s a lot more diversity of delivery vehicles and adjuvants out there than of antigens.  It does make me wonder how much of the secret sauce to the efficacy is on the adjuvants, and - once data from widespread immunisation is analysed - if there is mileage in pooling adjuvant technology to improve efficacy down the line.

Secret sauce pretty much describes it, I think.  I think the Sanofi/GSK vaccine is built on similar lines to the Novavax, but it gave disappointing results, despite adjuvant technology being GSK's core USP in vaccines.  If you haven't read it, Dan Davies's book "The Beautiful Cure" gives a good introduction to the weird history of adjuvants and what is and isn't known about how they work.

In reply to wintertree:

Ah; what confused me was that the graph didn't extend anywhere near 03-08, or, in fact, anywhere into the future.

 Dave Garnett 28 Jan 2021
In reply to Richard J:

> (but not sure about the somewhat black magic sounding "plant based adjuvant").

All adjuvants are black magic, I don’t think there’s anything particularly spooky about this one!

Edit: Ah, just seen your secret sauce comment - and the book recommendation.  It was definitely more voodoo than molecular biology when I was doing it!

Post edited at 23:04
OP wintertree 28 Jan 2021
In reply to bruxist:

Thanks for taking the time for that detailed chronological summary with context from the foreign language papers - it's much appreciated and well beyond my abilities.

> It's weird because that point he's making is made explicitly and repeatedly throughout Handelsblatt's article. 

Okay; that's the crux and it changes significantly my interpretation towards yours vs the Twitter thread.

> but I do know that in presentation of such data in Germany, the 'Tabellen' are subordinate to the summary. One reads the summary first, then uses the table/chart/picture as an aide-mémoire.

That is really insightful, thanks.  Largely I think from severe dyslexia I am a very visual person; I try and read a lot of papers to keep abreast of several areas and I do this by reading the abstract and conclusions, then skimming through the figures and captions, then tables, then hunting out key bits of text for anything not so conveyed.  I hadn't realised that the division in approach was so strong with UK/USA vs Germany - but then the only German output I have read is Albert Einstein's Brownian motion paper (a masterpiece, the reading of which I highly recommend even if you're not as interesting in Brownian motion as me).

As you said:

> I really would have expected Moody to know this, and not to clip just the bit no-one looks at.

I now understand the basics for you saying this.  That the newspaper article does clearly cover the insignificance of that critical row makes picking on the table alone highly misrepresentative.

We don't need this right now.

>  There wasn't much chance of the EU or the US regulators approving the AZ vaccine for over-60s on slim evidence; we on the other hand have gambled on theoretical but unproven efficacy and, if it succeeds, the EU and US will look like they've been fruitlessly cautious.

Sometimes in a crisis you have to roll the hard six.  It can be better to make a decision and do everything in your power to make it work, than to agonise over the right decision.  So long as you continuously review things, you can never go very wrong.  So, I hope we are scrutinising the longitudinal data from the vaccination program daily.  

OP wintertree 28 Jan 2021
In reply to captain paranoia:

> Ah; what confused me was that the graph didn't extend anywhere near 03-08, or, in fact, anywhere into the future.

Thanks for the explanation of the confusion.  I shall have to think more about the formatting of that plot.

 Richard J 28 Jan 2021
In reply to Dave Garnett:

The inner bark of the Chilean soapbark tree, Quillaja saponaria, also used as a foaming agent in root beer...

I'm sure you're right, you have to forgive those of us who naively hope that someone knows how some of this stuff actually works!

In reply to wintertree:

> I shall have to think more about the formatting of that plot.

Should I point out that plot 22 update 2 has a US format date...?

😁

OP wintertree 28 Jan 2021
In reply to Richard J:

> Dan Davies's book "The Beautiful Cure" gives a good introduction to the weird history of adjuvants and what is and isn't known about how they work.

Excellent; ordered now before I forget it.  Thanks.

(Hijacked from another one of your replies to Dave)

> I'm sure you're right, you have to forgive those of us who naively hope that someone knows how some of this stuff actually works!

It's been fascinating moving in to the biosciences to see the reverence with which basically everyone views the immunologists.  I'm hoping I can cash out my current endeavours in 5 years or so, and the question arrises "what to do next?".  Can you enter immunology around age 50 and make a positive difference?  I suspect that the field is going to change a lot between now and then.  It's also possible that in 5 years time I'll be exhausted and destitute...

> Fujifilm's biotech CMO in Billingham,

Interesting; I wrote up my analysis of the spatiotemporal spread of the failure of lockdown on Dec 16th as a context for the (pre)discovery of the new variant.  I put it on here and in more details on another more professional platform; that other site tracks viewers and one of the key organisations for views was Fujifilm Diosynth.  Another sign that perhaps I'm not just shouting in to the void.

Post edited at 23:36
OP wintertree 28 Jan 2021
In reply to captain paranoia:

> Should I point out that plot 22 update 2 has a US format date...?

You absolutely should; thanks.  It wouldn't be a problem in itself it formats weren't mixed within the figure - the epitome of failure.

F*****g python should throw an exception when I format a datetime object with "%s" rather than assume it knows how I want to handle it.  As Wintertree, Sr. always said "A bad workman always blames his tools"...

 Misha 29 Jan 2021
In reply to wintertree:

It would be great to open schools again but I suspect 8 March is wishful thinking. If your prediction of 5,000 detected cases a day is correct (and 5,000 sounds pretty optimistic 5 weeks out, compared to 20k+ currently), presumably that's around 10,000 infections. Which is a lot more than T&T could reasonably cope with. If we could keep it at that level with schools open, that should be ok in terms of NHS pressure. So the question is whether it will start creeping up again. It certainly doesn't leave much slack for opening up elsewhere any time soon.

What might help your decay rate based prediction is the increasing impact of the vaccines. However I suspect the impact on cases will be modest. The impact on deaths would be more significant, though how much of that will have 'fed through' by 8 March is another question.

 Misha 29 Jan 2021
In reply to Longsufferingropeholder:

> They must have noticed.

> They don't approve the AZ vaccine for over 65's. For the next weeks/months they're only vaccinating over 65s. Noise about supply of AZ vaccine - dealt with.

That strikes me as a plausible explanation...

 Misha 29 Jan 2021
In reply to minimike:

> The fact they’re quoting CIs at negative values rather than just zero is ridiculous. Unless the vaccine GIVES you covid.. (for clarity, it doesn’t and can’t, ok?)

I blame Bill Gates! 

 John Kelly 29 Jan 2021
In reply to wintertree:

think this is interesting

Manaus - signs of a resurgence despite 'herd' immunity - lot of possible reasons are discussed, seems to be to early for conclusions

https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(21)00183-5/...

Post edited at 08:27
OP wintertree 29 Jan 2021
In reply to Misha:

> It would be great to open schools again but I suspect 8 March is wishful thinking. If your prediction of 5,000 detected cases a day is correct (and 5,000 sounds pretty optimistic 5 weeks out, compared to 20k+ currently), presumably that's around 10,000 infections.

It’s an extrapolation of the status quo rather than a prediction.  The hope is that the vaccination program starts to being these numbers way down over the next few weeks - although it will do so in older ages than those involved in household<>school spreading.  

> Which is a lot more than T&T could reasonably cope with.

Not being sarcastic, but is there a case number greater than zero that our current T&T can usefully contain.  If didn’t manage too late last summer; my feeling is that it has

  • too poor an engagement with testing (poor employer and state support for isolation and low paid or insecure jobs)
  • too much latency to be effective (systematic structural issues not purely related to total load)
  • too little engagement in terms of people answering the 0300 number

I think we need locally based priority t&t teams with boots on the ground, and that we need rapid response teams to “trace up” around super spreader events to find the source and other branches of infection from them.

> If we could keep it at that level with schools open, that should be ok in terms of NHS pressure. So the question is whether it will start creeping up again. It certainly doesn't leave much slack for opening up elsewhere any time soon.

No indeed, and in terms of NHS pressure people are going to be utterly exhausted by then.

> What might help your decay rate based prediction is the increasing impact of the vaccines. However I suspect the impact on cases will be modest. The impact on deaths would be more significant, though how much of that will have 'fed through' by 8 March is another question.

Yup; that’s why this is done demographically.  The vaccination won’t stop when schools open so the demographic trend in rate constants come March 8th will be interesting.

Ome possibility would be to shift 3 weeks of the summer holiday from June/July to the remainder of March.  This would trade open school time for when vaccination will be higher, cases lower and the weather in favour of more ventilation and outdoor time.  I doubt it’s going to interrupt many foreign holiday plans...

Also, school busses I think need some serious attention - mask wearing, social distancing and ventilation on the upstairs of double deckers. 

Post edited at 08:51
 Dave Garnett 29 Jan 2021
In reply to Richard J:

> I'm sure you're right, you have to forgive those of us who naively hope that someone knows how some of this stuff actually works!

I could make up something credible and some of it might even be right but I think I might also get the book!  In my day it was mostly all about whether use complete or incomplete Freund's Adjuvant (some unholy oil emulsion with or without mashed up dead TB bugs in it). 

What some adjuvants do is to cause inflammation and ginger up the local immune response in a non-specific way.  You can get the same effect by poking around with a needle (even better if you manage to get it infected) but for some reason this has fallen out of favour.   

OP wintertree 29 Jan 2021
In reply to John Kelly:

Yes, worrying isn’t it.  I hope some conclusions can be drawn soon, but what ever they are I doubt we will be happier for them.

I still find it unbelievable how hard some people were arguing for naturally acquired herd immunity early on in this pandemic.  

1
 Richard J 29 Jan 2021
In reply to Dave Garnett:

I thought the book was very good (recommended to me by my boss, who is herself a very eminent biologist).  It does seem that things have got a bit further than mashing up bugs - looking up a bit more about the Novavax adjuvant they seem to think it's important that it is two separate saponin fractions from the bark each wrapped up in 40 nm phospholipid nanoparticles.  

OP wintertree 29 Jan 2021
In reply to Richard J:

I’ve been teaching Jr to do saponin extraction from conkers to make a shampoo.  We’ve only got as far as drying, crushing, leaching, settling and taking the relevant coarse fraction but I’m hoping we can try something more scientific when the pandemic is over and I can reclaim some WFH space for a small lab, and when they’re interested in finer work than smashing, stirring and pouring.  I’ve been eying up the cost of a basic, used gas chromatograph...  Detectors have come on a lot since I used to do GC in the 90s.  The council knocked down our favourite conker trees for the new county hall building so our haul was much reduced this year

I had no idea saponins had adjuvant properties.  Interestingly, as they’re natural they could presumably find their way in to topical neutracuticals.

Post edited at 09:43
 Offwidth 29 Jan 2021
In reply to wintertree:

ONS infection rates reported on BBC news to be broadly unchanged in England this week. Worrying.

Edit not quite as bad looking at the details as it's compared to two weeks before.  https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/con...

Post edited at 13:02
 groovejunkie 29 Jan 2021
In reply to Offwidth:

> ONS infection rates reported on BBC news to be broadly unchanged in England this week. Worrying.

I'm struggling to understand any of the stats. If a month ago community prevalence was approx 1 in 50 with recorded daily infections of 60 000. How is it now 1 in 55 with 28 000 recorded daily infections? And also, all the published charts of rates per 100 000 vary between 200 and 700 (wish!). If we take an average of 500, then that is 1 in 200 is it not? I appreciate the real world application is much more complicated than this but if we are basically saying the situation is not changing despite all the measures being taken that is utterly utterly depressing. 

OP wintertree 29 Jan 2021
In reply to Offwidth:

React was at odds (with a giant CI) to the P1+P2 testing data as well.  They’re not measuring quite the same thing as one is detecting some new infections and the other “live” infections by some definition of “live”.  

For now, hospitalisations look to me to be doing the expected sort of thing following cases.  Given the lag in the ONS data and the near immediacy of the hospitalisations, that seems in conflict.  I wonder if the potential higher viral load of the new strain is changing how long a new infection persists in the ONS data?  They have a small decrease in the mean over the last two datapoints - less than the errorbars.  You’d see something like that convolving new infections (P2) with a distribution function for how long they’re “live” to get the ONS data.

We’ll see what another week brings.

In reply to groovejunkie:

Hospitalisations are going the right way.  That’s a key litmus test that restrictions a few weeks ago were working.  I think there’s an effective “lag” in the ONS data compared to P1+P2 as it measures any age of infection, and infections are long lived.  P1+P2 gets new infections.  Just as the decay in hospitalisations lags P1+P2 I would expect the decay in live cases to lag - after all you’re not normally hospitalised without a live case!

ONS used to estimate daily infections in their reports; they stopped with various reasons given I didn’t really understand. Potentially related to the new variant breaking some assumptions used to estimate this rate?

I could be wrong but that’s my hunch on what we’re seeing.  For all I dislike ZOE, it reflects new cases and looks closer to P1/P2 than the ONS data.

Post edited at 13:08
 Offwidth 29 Jan 2021
In reply to groovejunkie:

Positive tests went up in the week after the week ending Jan 2nd and then back down again for the week ending Jan 16th. So it's not as bad as the BBC report implied.

 groovejunkie 29 Jan 2021
In reply to Offwidth & Wintertree:

Ah some good news, thankyou. And also thank you Wintertree for all your charts/analysis etc. I don't understand a lot of it but massively appreciate the amount of work you put in. 

 Si dH 29 Jan 2021
In reply to wintertree:

I had thought quite a bit about these discrepancies and come to the same tentative conclusions as you. It's needs investigating properly though I think - I hope someone in PHE is doing this. If people are remaining test-positive for longer after infection it could have an implication for isolation times.

Post edited at 16:27
 Punter_Pro 29 Jan 2021
In reply to topic:

A decision has been made and the EMA has recommended an approval for the AZ vaccine for all ages, with a reported efficacy of 60%. 

https://www.ema.europa.eu/en/news/ema-recommends-covid-19-vaccine-astrazene...

 

Post edited at 16:37
OP wintertree 29 Jan 2021
In reply to Punter_Pro:

Thanks.  

"In addition, as the vaccine is to be given as two standard doses, and the second dose should be given between 4 and 12 weeks after the first, the Agency concentrated on results involving people who received this standard regimen."

So, it seems that the EMA do not have a problem with the 12-week gap.

 (Edit: As applied to the AZ vaccine, not the Pfizer one which is a different discussion).

Post edited at 16:45
OP wintertree 29 Jan 2021
In reply to Si dH:

> It's needs investigating properly though I think - I hope someone in PHE is doing this.

Indeed.  

> If people are remaining test-positive for longer after infection it could have an implication for isolation times.

NZ are now adding a home quarantine period on to the end of release from MIQ after several community cases traced back to MIQ.  It's not clear to me how much is because they thing the infectious period of the new variant (SA in their case, shares a convergent substitution with the Kent variant) and how much is because they think increased transmissibility or protocol breaks are leading to increased transmission during the quarantine period.

Edit: Looking at the provisional window on cases, they look to be falling really quite well.  

Post edited at 17:03
 Toerag 29 Jan 2021
In reply to wintertree:

> > It's needs investigating properly though I think - I hope someone in PHE is doing this.

> Indeed.  

> NZ are now adding a home quarantine period on to the end of release from MIQ after several community cases traced back to MIQ.

Are they doing 'test to exit'?  We introduced that after they found people testing positive on day 13 of self-isolation...

 Punter_Pro 29 Jan 2021
In reply to wintertree:

> So, it seems that the EMA do not have a problem with the 12-week gap.

>  (Edit: As applied to the AZ vaccine, not the Pfizer one which is a different discussion).

Apparently not, they are inline with the MHRA report which is reassuring.

 bruxist 29 Jan 2021
In reply to wintertree:

> Thanks for taking the time for that detailed chronological summary with context from the foreign language papers - it's much appreciated and well beyond my abilities.

Glad to be of help: likewise your analyses are well beyond my abilities. The UKC forums continually amaze me in being a place where really rather sophisticated scientific work can be presented and then explained and discussed in layman's terms. Not what I was expecting when I joined!

Thanks for the tip about Einstein's Brownian motion paper. I don't think I've ever read anything he wrote - which seems quite astonishing, now I think about it.

> Sometimes in a crisis you have to roll the hard six.  It can be better to make a decision and do everything in your power to make it work, than to agonise over the right decision.  So long as you continuously review things, you can never go very wrong.  So, I hope we are scrutinising the longitudinal data from the vaccination program daily.  

It does seem as though there's a lot of attention being directed at the Israeli data on the Pzifer/BioNTech roll-out. I'd fully expect the UK to be producing the same level of data on the AZ roll-out. In my local area we've just ended the first full week of AZ vaccination, which was solely for care workers; that changes to mass vaccination for the public as of Monday, so we'll be generating the data we need soon.

OP wintertree 29 Jan 2021
In reply to bruxist:

> Thanks for the tip about Einstein's Brownian motion paper. I don't think I've ever read anything he wrote - which seems quite astonishing, now I think about it.

Clear, precise and - interestingly to me - written in the first person, something strongly beaten out of researchers in the UK.  It subtly changes how one reads the paper I think - and for the better.  He read some wooly reports of what people saw down microscopes ("It is possible that the movements to be discussed here are identical with the so-called Brownian molecular motion however, the information available to me regarding the latter is so lacking in precision, that I can form no judgment in the matter" - not a lot has changed in a century in may ways), sat down and thought about it and realised that the equipartition theorem applies to everything not just atoms and molecules, and out pops the rules of diffusion that are used the world over today.

I don't think he thought much of the experimentalists of the time - "I will not attempt here a comparison the slender experimental material at my disposal".  It turns out capturing an accurate recording of Brownian motion down a microscope is really quite hard with convective currents and evaporation driving all sorts of false signals, and with the effects of vibration being hard to decouple.  I'm not sure when it was first possible to record sufficient data on the motion of a single particle down a microscope to show that it was Brownian, but I doubt it was in Einstein's lifetime.  Bulk diffusion is a different issue though and much easier to test.

There's an English translation out there although I envy you if you can read the original  - http://www.damtp.cam.ac.uk/user/gold/pdfs/teaching/old_literature/Einstein1...

 bruxist 29 Jan 2021
In reply to wintertree:

I'll have a go on this first and report back: https://www.zbp.univie.ac.at/dokumente/einstein2.pdf (which is surely not a copy of how it was originally printed - it would have been printed in Fraktur, what we think of as German Gothic font, but getting my hands on a copy of the original paper as printed might take more time than just reading it.)

 Richard J 30 Jan 2021
In reply to bruxist:

To join in with this appreciation of Einstein's lesser known 3rd paper from his miracle year, it's hard to remember now that in 1905, to not accept the reality of atoms was a completely reasonable position, as the evidence for them was equivocal at best.  But Einstein's paper on Brownian motion - and the careful experimental work by Jean Perrin that confirmed the theory - really was decisive.

 Richard J 30 Jan 2021
In reply to wintertree:

> I’ve been teaching Jr to do saponin extraction from conkers to make a shampoo.  ... I had no idea saponins had adjuvant properties.  Interestingly, as they’re natural they could presumably find their way in to topical neutracuticals.

Well, maybe if conker saponins work that could take some pressure off the unfortunate Chilean soapbark tree.

My favourite story about substitutions for rare natural products I learnt about about several decades ago, when I was working a lot with an Anglo-Dutch food conglomerate.  They were (probably still are) very keen to learn how to control ice-crystal growth, to prolong the shelf-life of ice cream in countries with dodgy cold-chain distribution networks.  Somebody in their R&D team thought that it would be a great idea to add the anti-freeze peptides from arctic cod (which suppress ice crystal growth by a specific binding mechanism at the ice-water interface).  Even a company with their resources pulled back from the idea of large scale blood harvesting from polar fishes, but they'd got as far as genetically modifying soy to express the peptide.  And then someone even smarter in the R&D team thought, wait a minute, arctic cod can't be the only organism that needs to protect itself against frost - and found a similar anti-freeze peptide in the grass clippings from the lawn outside their lab.  That got them a nice paper - but then someone from closer to the D end of R&D managed to get nearly as good a result using locust bean gum.  

OP wintertree 30 Jan 2021
In reply to Richard J:

> Well, maybe if conker saponins work that could take some pressure off the unfortunate Chilean soapbark tree.

It's all I've used on my hair for 3 years now - just filtered leachate rather than fractionated.  It's almost odourless compared to commercial shampoo, and it literally grows on trees.  The soapbark business is not good - a trend in hair washing in the UK shouldn't be able to disrupt a way of life half a planet away.

Interesting food story; I've been amazed by the money spent by the food industry to improve the quality of mass produced ice-cream; the level of leading physics and instrumentation brought to bear on ice formation is staggering. X-ray diffraction from synchrotrons being one bananas example of what's going on now.  I quite like it really; a lot of people rally against mass produced food but I see their quest for improved quality on a budget as the gradual democratisation of gourmet cuisine.  Long way to go, like...

Edit: Ice formation is fascinating stuff though.  

Post edited at 13:53
 Richard J 30 Jan 2021
In reply to wintertree:

I'm with you on the democratising effect of high quality mass produced food.  There's a lot of nostalgia about an imagined past of artisanal produce in a rural Arcadia.  But (to take a personal example) for my grandmother, who grew up in rural North Wales, the food of her childhood was bread, buttermilk, potatoes and occasionally fatty lamb, so she thought Kwiksave was marvellous.

There were people doing synchrotron x-ray studies of chocolate at least 20 years ago - ice formation is indeed fascinating, but fat crystallisation is even more complex!  And, famously, Margaret Thatcher's first job after getting her chemistry degree was working on ice cream with Walls. 


New Topic
This topic has been archived, and won't accept reply postings.
Loading Notifications...