The big story this week was a new preprint claiming to show that Covid-19 had an unnatural origin. For several days, this was a big story with lots of arguing about it, lots of long threads, lots of people accusing others of bad faith or being idiots or not understanding undergraduate microbiology, and for some reason someone impersonating a virologist to spy on Kelsey Piper.
Then a few days later all discussion of it seemed to vanish. It wasn’t that everyone suddenly came to an agreement to move on. All sides simply decided that this was no longer the Current Thing. See the section for further discussion. In the end I did not update much, so I am mostly fine with this null result.
There’s also more Gain of Function research looking to create a new pandemic. There was a lot of consensus among the comments and those I know that this work must stop, yet little in the way of good ways to stop it. Several people gave versions of ‘have you considered violence or otherwise going outside the law?’ and my answer is no. While the dangers here are real, they are not at anything like the levels that would potentially justify such actions.
Note on Deleted Post from This Week
Finally, I need to address the post that got taken down in a bit more detail. I want to thank Saloni in particular for quickly and efficiently making some of my mistakes clear to me both quickly and clearly, with links, so I could within about an hour realize I’d made a huge mistake and the whole post structure and conclusions no longer made sense, so I took the post down. Please disregard it. Everyone has been great about understanding that mistakes happen, and I want you to know I appreciate it, and hope it helps myself and others similarly address errors in the future.
How did the mistakes happen? Ultimately, it is 100% my fault, on multiple counts, no excuses. What are some of the things I did wrong, so I can hopefully minimize chances they happen again?
My logic was flawed. I wasn’t thinking about the power of the study properly. I let the truly awful takes and absence of good takes defending colonoscopies make me too confident in the lack of available good takes doing so, and let that bias my thinking. I got feedback before posting, but I did not get enough or get it from the right sources. I heard everyone talking about ‘first RCT’ in various forms and failed to notice it was only the first to look at all-cause mortality rather than the first RCT. The authors of this one made the mistake of trying to measure all-cause mortality as primary endpoint despite lacking the power to do so, in a way that my brain didn’t properly process, compounding the errors. I didn’t properly consider the possibility that the main result of a published paper was plausibly highly ‘unlucky’ in part due to training on decades of publication bias. I didn’t fully appreciate the magnitude of the healthy patient bias, which made certain extrapolations sound patently absurd – I’m still super skeptical of those claims but they’re not actually obviously crazy on reflection. And I messed up a few small technical details.
In general, the whole thing is really complicated. There is no question that the study was a disappointing result for the effectiveness of colonoscopies, well below what the researchers expected to find. However, there is a lot of room for ‘disappointing but still worthwhile’ and a lot of additional past data to incorporate. I genuinely don’t know what I am going to think when I am finished thinking about it.
Executive Summary
- New preprint on potential origins of Covid-19, not updating much.
- Gain of Function research continues.
- Please disregard this week’s earlier post until I can properly fix it.
Let’s run the numbers.
The Numbers
Predictions
Predictions from Last Week: 218k cases (-3%) and 2,160 deaths (-7%).
Results: 218k cases (-3%) and 2,262 deaths (-3%).
Predictions for Next Week: 218k cases (+0%) and 2,150 deaths (-5%).
Situation seems straightforward enough. Variants are picking up speed, so we might see cases start to head back up a bit for a time.
Note: I am pretty sure Halloween is not a reporting-relevant holiday.
Deaths
Cases
Case bump in the South seems to purely be a reporting glitch in North Carolina. I am choosing not to manually fix it.
Booster Boosting
A mono-valent booster would have been a better choice. Given that the vast majority of what a booster activates are the existing (original strain) defenses, I don’t see the argument for why we went bi-valent other than The Fear. Bloomberg reports that a new as-yet-unpublished study found the new boosters don’t generate superior antibodies against BA.4 and BA.5 than the old booster did.
HHS to run ads to promote boosters, customized to target minority groups. I do not expect this to be a worthwhile use of taxpayer funds, and I can only imagine what would be said about this if the message was something else instead.
Eric Topol on the benefits of boosting.
You have to love the big font benefits and tiny font negatives, and notice that this was his response to public health Twitter feedback. Here’s the pre-feedback version.
There was a change in the wording on myocarditis to emphasize how one should worry more about Covid-19-induced cases than vaccine-related cases. Most of the change was making the negatives use a smaller font.
Some negatives are rather important absences of desired benefits (duration >4 months, mucosal immunity, better protection from infections, being up to date at the time all lacking).
I’d highlight that claim that ‘chasing variants is untenable.’ It is with that attitude. There is nothing physically undoable about ensuring that most people who get a shot get a shot tuned to an updated variant. The reason it is undoable is because our government has decided not to allow the vaccine manufacturers to do this.
ACIP unanimously formally approves new child vaccine schedule that includes Covid. I understand those who think this will further shatter faith in public health. I still don’t see what the alternative was. Not recommending or scheduling the shot would have also been cited the same ways, to the extent that people site such things at all.
This post makes the case that the Covid vaccine shouldn’t be routine for kids. On the merits, I think it absolutely should be exactly that: routine. It shouldn’t be a big deal any more than the others are a big deal. I don’t see how not including it in the schedule makes this easier to achieve. The question then moves to mandates, which again are a pretty dumb idea and I expect to backfire rather hardcore when tried.
Paper identifies strong negative effects to offering vaccination incentives in the range of $8-$125, versus offering the vaccine for free, presumably through some combination of motivation displacement and people wondering why they are being bribed. Given the way things have gone, I am willing to believe that bribing people is no longer a useful opportunity here.
Pfizer to raise price of vaccine four-fold now that deal with US government is expiring, amid complaints that company is not providing infinite supply of its miracle life-saving medicine to those who don’t want to pay for it.
Variants
XBB is dominant in Singapore. Here is how that is going.
As usual, majority of those 15 in the ICU are over 70 years old.
Yes, that is in a population with existing seropositivity that is also highly vaccinated. America cannot expect to match those results. That doesn’t mean you can’t expect to match those results for you and those you know.
Physical World Modeling
Sinovac does not protect against infection for level 5+ Omicron variants. Zero Covid policy continues anyway.
Definitely Not Mario continuing to do P100 erasure.
You can do better than an N95. That doesn’t mean you want to or should do so, but please stop saying that N95s are the maximum possible amount of physical protection.
In Other Covid News
Birth rates seem to have risen (somewhat) during the pandemic, with the decline in childbearing in 2020 due to a lack of birthright-citizenship-claiming travel by foreigners.
Covid inspired the shift to working from home, which has been very good for disabled workers.
Note that the divergence here does not start until 2021. It took a year for some combination of workers and employees to figure this one out.
Can We At Least Ban Gain of Function Research?
Seriously, what is it going to take here?
So not only can we not stop doing it, we cannot even do the mandated cost-benefit analysis testing first.
The very least America can do is follow its own procedures first before causing the next pandemic.
Bloom Lab weighs in on the ‘BU Chimera’ gain of function case from last week. Core point of view is that yes Gain of Function research can cause pandemics, and often we do crazy things we have no business doing, on the other hand Science and this particular one wasn’t too bad.
To repeat, they engineered avian H5N1 flu to be airborne transmissible in mammals.
A policy that says ‘we should allow some such work but not other such work’ would be a lot more credible if we had a track record of not doing the completely insane.
Good news, it was halted.
…and some bad news. Did we mention what’s going on at WIV?
Then come the examples of ‘safe’ research.
That’s Loss of Function research. Definitely a lot better. It still makes me nervous, as it is hard to be sure you are doing what you think you are doing.
My layman reaction is ‘getting something to grow better’ is the kind of thing that changes the value of ‘able to infect humans’ so maybe don’t do that?
What about this particular case?
This is not a standard that works for me. I do not expect it to cause people to not be allowed, in practice, to do dangerous things. When we imagine the scientists doing the research explaining how Nothing Could Go Wrong Here, how much do we believe them? What are their incentives pushing towards?
I do agree with the point that this was less insane than many other actually done Gain of Function research. If you offered me a compromise that said ‘this is the most risky thing we allow, we don’t do risker research than this’ and it was always done with proper safety protocols and prior reviews, I could likely live with that. Alas.
Also, I don’t think the standard here would be unfair?
If you are going to engineer a new version of a deadly virus to answer a scientific question, yes I think you should have to consider if there’s other ways to better address this. We do such things before we approve removing two parking spaces. It sometimes takes years. I don’t think that’s a good policy in that case. In this one? Yeah.
There’s some good news, people are concerned?
Some House Republicans are so outraged they are sending a strongly worded letter.
New Preprint on Origins of Covid-19
In other humans-causing-pandemics news: New preprint paper claims to show Covid-19 was made in a lab (preprint) by identifying the places it was ‘stitched together,’ this thread has further thoughts. Prof Francois Balloux confirms replication of the findings (which is distinct from them implying what is claimed), finds the argument non-conclusive but compelling, and gets no points for predicting the inevitable explosion of rage and fury.
Matt Shapiro notices that opinions on the paper differ and doesn’t know what to think. Sarah Constantin says they’re doing things the right way, engaging with criticism, publishing the preprint, clear communication. The Economist calls it a ‘first draft of science.’
Those threads:
- Friedemann Weber says (extremely rudely but still counts) that yes you could do it this way but the standard way of doing it would have been easier and wouldn’t match. He also points out that the main preprint author is a fan of artificial origin theories for viruses.
- Alex Crits-Christoph says the preprint is capital-F False, it only exists because the sites examined were cherry-picked and their own metric suggests even higher likelihood of ‘unnatural origin’ for other Covid strains, which sounds more like Not Even False/Wrong than false if true.
- Kristian G. Anderson says the paper is so deeply flawed it wouldn’t pass kindergarten molecular biology. He says he did a similar analysis himself in Jan/Feb 2020 and shares the Github, which is cool.
- Santiago Sanchez applauds the paper for the attempt, then says it can be falsified by recalling freshman molecular biology.
Flo Debarre flat out gives ‘a curated collection of threads and Tweets rebutting’ the preprint, including some not listed above.
Triple Bankshot summarizes the paper as claiming ‘the restriction cites are too regular’ and thinks the approach is novel and interesting whether or not it is right, while also noting several objections that could invalidate the whole paper, he thinks the above objections are real but not slam dunk ‘debunking.’
This is the author updating where they stand on day 2.
Kelsey Piper is investigating things.
(Email, basically saying ‘I am not an expert, help me figure this out.’)
That’s the Andersen who had one of the debunking threads above.
So what do I think about all this?
I find the argument here unconvincing, in either direction. I agree with the authors that this analysis could not rule out non-natural origin, and with the critics that it does not rule out natural origin either. The study seems at core like it was a good idea. It failed to find an impressive or surprising result, either in terms of ‘this clearly indicates man-made’ or ‘this clearly indicates natural origin.’ The likelihood ratios here are not so big. Thus, I am not substantially updating in either direction, and more in the direction of ‘we will never know one way or the other.’
Other Medical and Research News
Pandemics are bad. We should prevent them. They could happen at any time. Still, you know what framing I am pretty sure does not help?
That’s not how this works. That’s not how any of this works.
It is far better than the opposite response of ‘this was a 100-year pandemic so we have 98 years to prepare,’ which is Obvious Nonsense even if I didn’t already think that someone calling an event a ‘100-year X’ meant ‘an X like this is going to happen about once every decade or two’ by default at this point. Seeing a pandemic like Covid-19 makes another arriving soon more likely if we don’t act, not less, as it updates us on frequency of pandemics. They don’t space out evenly. ‘The next pandemic’ will arrive on essentially a Poisson distribution. There’s a chance each day/month/year that it arrives, which we can reduce via various prevention methods, or we can radically increase by doing Gain of Function Research.
Thus, it is like planting a tree. The best time to prevent a pandemic is many years ago. The second best time is right now. If you start next year, yes, ‘it might be too late’ in the sense that this gives a yearlong window for there to be a pandemic you didn’t prepare for. It does not mean that two years from now if we don’t do anything before then, we should say ‘whelp, we missed the window on preventing pandemics, guess we should stop trying.’ Quite the opposite.
Here are Spencer’s proposed solutions:
- Better surveillance, yes, right on.
- Perhaps we should employ more health care professionals and provide more health care. I do not think ‘hire more generic doctors in general’ is a reasonable or efficient response to the worry about pandemics. Perhaps you are using this to push your existing policy and guild preferences.
- Perhaps you are pushing your existing policy and guild preferences.
- Are we going to talk about vaccine research? About streamlining production and approval processes? About plans to impose proper travel restrictions and quarantines? About what you’d actually do to prevent a pandemic if you cared about that, once you discovered it with the surveillance? No. Well, then.
NPR report says that 84% of deaths related to pregnancy are preventable (which tells us little about how many of those we could, by Doing More, prevent), calls for us to Do More. This seems to mostly be referring to psychological issues that result in death, centrally postpartum depression often involving substance abuse and accelerated by kids being taken away. The proposed response is to have the health system check more often for problems (and, presumably, more often take kids away, and subject new mothers to more interrogations and worry). I have no idea if this would be net beneficial or pass a cost-benefit test.
FDA Delenda Est, did you know that all available-in-America IUDs are too big for a large percentage of women because the FDA decided to regulate IUDs as if they were drugs which has resulted in it taking clinical trials and on average 12 years to get approval? So we only have 5 choices available instead of 22, often all too big. And then there are the condoms, which for 5%-10% of men are required to be too small due to an explicit width limit, increasing chance of breakage.
With that in mind, the FDA is planning on ramping up its War on Knowledge. Diagnostic tests are its latest enemy. They want Congress to pass a law explicitly allowing them to prevent such tests, and are threatening to do it themselves anyway if Congress decides not to cooperate. They will simply use ‘rulemaking.’
“The other option besides the law is federal rulemaking – which is always an option for the FDA – but it is a very lengthy and contentious process,” he added. “It would be much better for the lawmakers to come to an agreement … but we also can’t stand by.”
It is a real shame that the FDA wants to do something it thinks is vital to people’s health, and it is currently unable to do so without going through a very lengthy and contentious process. One can understand why they would want the law to change so they did not have to do that.
Meanwhile the director is taking victory laps talking about how they ‘foster an environment to ensure the US remains the world’s greatest innovators and improves life expectancy’ when the FDA’s job is to prevent this and also partly due to their handiwork life expectancy is declining.
There is also a welcome explicit call for more telemedicine.
Although I do not understand how the solution to telemedicine is distinct from ‘make it legal (and reimbursable?) for doctors to do telemedicine.’
Claim that OTC medicines combined with IT tech are disrupting healthcare and greatly lowering costs, estimates 5%-10% drop in medical visits already from drugs being moved from prescription to OTC even under FDA’s rules. Imagine what we could do if the rules were more reasonable. The post frames this for any given situation as an either-or, which seems to miss further opportunity – if you have an AI that can diagnose pneumonia, you don’t have to then skip the doctor visit if it comes back positive, whereas if it is negative you are in the clear (assuming the system is sufficiently accurate).
Also worth noting is that this is a claim that ~10% of doctor visits used to be about getting a prescription to use something like Advil that is now over the counter. How often do we run into problems because such drugs are now OTC? This seems like the kind of thing we would hear about, and I never hear about it. Not gatekeeping Advil and friends was a pure win. This strongly implies that we are not doing this for enough of its friends.
Claim that NIH is denying access to its genetic data to those it believes will use that data to look for differences between groups or find a link between genetics and intelligence. They do not want people to look.
Periodic reminder in honor of the latest ‘this more active activity may prevent dimension whereas this less active activity may cause it’ study that correlation is not causation. It could well be that there is a real effect here, but studies structured like this tell us nothing.
New review system adapted by eLife whereby there won’t be acceptances or rejections, only a compilation of peer reviews. Then you can, if desired, filter papers by various thresholds to approximate the prior system. I am cautiously excited here. I do notice the temptation to submit something to see the thing in action.
FIRE says that University of Minnesota is issuing a Hippocratic Oath that, among other things, promises to “honor all Indigenous ways of healing that have been historically marginalized by Western medicine.” That means either your doctor will violate that oath, or they won’t.
America’s Finest News Sources
I sent out a Tweet this week asking if anyone had sources they thought should be included in this week’s updates. The only response I got was ‘what do you think of Taylor Swifts’ new album?’
(It seems… fine? Lesser work. One note, little impression. Cursing felt jarring.)
Getting more efficient information filtering systems, and figuring out how to scale by incorporating the efforts of others, will be key to the success of Balsa, so I’m not about to give up that easily. This is an explicit call to help me compile the worthwhile information sources in an efficient form. On sufficiently extreme success I will attempt to work with you.
Thus, your mission, should you choose to accept it, is to provide one of the following:
- Leave comments at the Substack version of this post suggesting potential information I may have missed, in response to the comment threads I will start for that purpose.
- Create a Twitter account for this purpose, so you can retweet things and tweet out links to things with brief descriptions. If it is worthy, I can add it to my lists. Note that I have to place a large premium on the signal-to-noise ratio, if there is too much irrelevant stuff then a source becomes non-useful.
- Suggest an existing source I should be looking at that I seem not to be (it can be you). Once again, signal-to-noise is important here.
- Compile a Google Doc of such links and things and share it with me some time around Tuesday or Wednesday.
- Suggest another new method of improving my information filtering.