Tag: Scott Alexander

Book Review: Lifespan

Mary Pat Campbell

Link:https://astralcodexten.substack.com/p/book-review-lifespan

Graphic:

Excerpt:

David Sinclair – Harvard professor, celebrity biologist, and author of Lifespan – thinks solving aging will be easy. “Aging is going to be remarkably easy to tackle. Easier than cancer” are his exact words, which is maybe less encouraging than he thinks.

There are lots of ways that solving aging could be hard. What if humans worked like cars? To restore an old car, you need to fiddle with hundreds of little parts, individually fixing everything from engine parts to chipping paint. Fixing humans to such a standard would be way beyond current technology.

Or what if the DNA damage theory of aging was true? This says that as cells divide (or experience normal wear and tear) they don’t copy their DNA exactly correctly. As you grow older, more and more errors creep in, and your cells become worse and worse at their jobs. If this were true, there’s not much to do either: you’d have to correct the DNA in every cell in the body (using what template? even if you’d saved a copy of your DNA from childhood, how do you get it into all 30 trillion cells?) This is another nonstarter.

Sinclair’s own theory offers a simpler option. He starts with a puzzling observation: babies are very young [citation needed]. If a 70 year old man marries a 40 year old woman and has a baby, that baby will start off at zero years old, just like everyone else. Even more interesting, if you clone a 70 year old man, the clone start at zero years old.

….

So Sinclair thinks aging is epigenetic damage. As time goes on, cells lose or garble the epigenetic markers telling them what cells to be. Kidney cells go from definitely-kidney-cells to mostly kidney cells but also a little lung cell and maybe some heart cell in there too. It’s hard to run a kidney off of cells that aren’t entirely sure whether they’re supposed to be kidney cells or something else, and so your kidneys (and all your other organs) break down as you age. He doesn’t come out and say this is literally 100% of aging. But everyone else thinks aging is probably a combination of many complicated processes, and I think Sinclair thinks it’s mostly epigenetic damage and then a few other odds and ends that matter much less.

Author(s): Scott Alexander

Publication Date: 1 Dec 2021

Publication Site: Astral Codex Ten

Ancient Plagues

Mary Pat Campbell

Link:https://astralcodexten.substack.com/p/ancient-plagues

Excerpt:

But the 1918 Spanish flu has, as far as I know, legitimately died out. Lots of people like saying that in a sense it’s still with us. This NEJM paper (with a celebrity author!) points out that it’s the ancestor of all existing flu strains. But most of these flu strains are less infectious than it was. This didn’t make sense to me the first, second, or third time I asked about it: why would a flu evolve into an inferior flu? Sure, it might evolve into a less deadly flu because it’s perfectly happy being more infectious but less deadly. But I think the Spanish flu was also especially infectious; so why would it evolve away from that?

One possible answer is “because by 1919, everyone had immunity to the 1918 flu, so it evolved away from it – and now nobody has immunity, but it lost the original blueprint.” The 1918 flu was a really optimal point in fluspace, but during all of history up until 1918, the flu’s evolutionary hill-climbing algorithm didn’t manage to find that point, and since flu has no memory it’s not going to be any easier for it to find it the second time, after it evolved away from it. So plausibly, existing flus are strictly worse at their job than Spanish flu was, and digging up an intact copy of the latter would be really bad.

And then there’s smallpox. No mystery why smallpox died out – we killed it. But then we stopped vaccinating people against it, and now if it comes back it would be really bad.

Author(s): Scott Alexander

Publication Date: 14 Dec 2021

Publication Site: Astral Codex Ten

Diseasonality – Why is winter flu season?

Mary Pat Campbell

Link:https://astralcodexten.substack.com/p/diseasonality

Graphic:

Excerpt:

Cold and humidity are definitely important – scientists can make flu spread faster or slower in guinea pigs just by altering the temperature and humidity of their cages. But it can’t just be cold and humidity. But if it was just cold, you would expect flu to track temperature instead of seasonality. Alaska is colder in the summer than Florida in the winter, so you might expect more summer flu in Alaska than winter flu in Florida. But Alaska and Florida both have lots of flu in the winter and little flu in the summer.

(if it was just humidity, same argument, but change the place names to Arizona and Florida.)

It’s the same story with people being cramped indoors. Common-sensically, this has to be some of the story. But if it were the most important contributor, you would expect to see the opposite pattern in very hot areas, where nobody will go out during the summer but it’s pleasant and balmy in the winter. Yet I have never heard anyone claim that any winter diseases happen in summer in Arizona or Saudi Arabia or terrible places like that.

If it was just vitamin D…look, it’s not vitamin D. Nothing is ever vitamin D. People try so hard to attribute everything to vitamin D, and it never works. The most recent studies show it doesn’t prevent colds or flu, and I think the best available evidence shows it doesn’t prevent coronavirus either. African-Americans, who are all horrendously Vitamin D deficient, don’t get colds at a higher rate than other groups (they do get flu more, but they’re vaccinated less, so whatever).

Author(s): Scott Alexander

Publication Date: 7 Dec 2021

Publication Site: Astral Codex Ten

When Will The FDA Approve Paxlovid?

Mary Pat Campbell

Link: https://astralcodexten.substack.com/p/when-will-the-fda-approve-paxlovid

Graphic:

Excerpt:

For context: a recent study by Pfizer, the pharma company backing the drug, found Paxlovid decreased hospitalizations and deaths from COVID by a factor of ten, with no detectable side effects. It was so good that Pfizer, “in consultation with” the FDA, stopped the trial early because it would be unethical to continue denying Paxlovid to the control group. And on November 16, Pfizer officially submitted an approval request to the FDA, which the FDA is still considering.

As many people including Zvi, Alex, and Kelsey have noted, it’s pretty weird that the FDA agrees Paxlovid is so great that it’s unethical to study it further because it would be unconscionable to design a study with a no-Paxlovid control group – but also, the FDA has not approved Paxlovid, it remains illegal, and nobody is allowed to use it.

One would hope this is because the FDA plans to approve Paxlovid immediately. But the prediction market expects it to take six weeks – during which time we expect about 50,000 more Americans to die of COVID.

Perhaps there’s not enough evidence for the FDA to be sure Paxlovid works yet? But then why did they agree to stop the trial that was gathering the evidence? Or perhaps there’s enough evidence, but it takes a long time to process it? But then how come the prediction markets are already 90% sure what decision they’ll make?

Author(s): Scott Alexander

Publication Date: 22 Nov 2021

Publication Site: Astral Codex Ten

Ivermectin: Much More Than You Wanted To Know

Mary Pat Campbell

Link:https://astralcodexten.substack.com/p/ivermectin-much-more-than-you-wanted

Graphic:

Excerpt:


About ten years ago, when the replication crisis started, we learned a certain set of tools for examining studies.

Check for selection bias. Distrust “adjusting for confounders”. Check for p-hacking and forking paths. Make teams preregister their analyses. Do forest plots to find publication bias. Stop accepting p-values of 0.049. Wait for replications. Trust reviews and meta-analyses, instead of individual small studies.

These were good tools. Having them was infinitely better than not having them. But even in 2014, I was writing about how many bad studies seemed to slip through the cracks even when we pushed this toolbox to its limits. We needed new tools.

I think the methods that Meyerowitz-Katz, Sheldrake, Heathers, Brown, Lawrence and others brought to the limelight this year are some of the new tools we were waiting for.

Part of this new toolset is to check for fraud. About 10 – 15% of the seemingly-good studies on ivermectin ended up extremely suspicious for fraud. Elgazzar, Carvallo, Niaee, Cadegiani, Samaha. There are ways to check for this even when you don’t have the raw data. Like:

The Carlisle-Stouffer-Fisher method: Check some large group of comparisons, usually the Table 1 of an RCT where they compare the demographic characteristics of the control and experimental groups, for reasonable p-values. Real data will have p-values all over the map; one in every ten comparisons will have a p-value of 0.1 or less. Fakers seem bad at this and usually give everything a nice safe p-value like 0.8 or 0.9.

GRIM – make sure means are possible given the number of numbers involved. For example, if a paper reports analyzing 10 patients and finding that 27% of them recovered, something has gone wrong. One possible thing that could have gone wrong is that the data are made up. Another possible thing is that they’re not giving the full story about how many patients dropped out when. But something is wrong.

But having the raw data is much better, and lets you notice if, for example, there are just ten patients who have been copy-pasted over and over again to make a hundred patients. Or if the distribution of values in a certain variable is unrealistic, like the Ariely study where cars drove a number of miles that was perfectly evenly distributed from 0 to 50,000 and then never above 50,000.

Author(s): Scott Alexander

Publication Date: 17 Nov 2021

Publication Site: Astral Codex Ten at substack

Mantic Monday: Judging April COVID Predictions

Mary Pat Campbell

Link: https://astralcodexten.substack.com/p/mantic-monday-judging-april-covid

Graphic:

Excerpt:

In April 2020, I made my yearly predictions, and many of them were about the (then new) coronavirus pandemic.

Two other people on Less Wrong, Zvi and Bucky, decided to test themselves against me by trying to predict the same questions. Zvi saw my answers beforehand; Bucky didn’t. Here’s how we did (except where otherwise stated, all predictions are for 12/31/20):

Author(s): Scott Alexander

Publication Date: 22 February 2021

Publication Site: Astral Codex Ten

Coronavirus: Links, Discussion, Open Thread

Mary Pat Campbell

Link: https://astralcodexten.substack.com/p/coronavirus-links-discussion-open

Graphic:

Excerpt:

R in most US states right now is closely clustered around 1. Mutant strains are more contagious, enough to bring the R0 up to 1.5 or so. But having a lot of the population vaccinated will bring it back down again. Also, I’m acting like there’s some complex-yet-illuminating calculation we can do here, but realistically none of this matters. It’s not a coincidence that all US states are closely clustered around 1. It’s the control system again – whenever things look good, we relax restrictions (both legally and in terms of personal behavior) until they look bad again, then backpedal and tighten restrictions. So we oscillate between like 0.8 and 1.2 (I made those numbers up, I don’t know the real ones). If vaccines made R0 go to 0.5 or whatever, we would loosen some restrictions until it was back at 1 again. So unless we overwhelm the control system, R0 will hover around 1 in the summer too, and the only question is how strict our lockdowns will be.

In autumn, if we haven’t already vaccinated everyone there’s a risk things will get worse again because of the seasonal effect. Also, for all we know maybe the virus will have mutated even further and become even more vaccine resistant. Now what?

Author(s): Scott Alexander

Publication Date: 15 February 2021

Publication Site: Astral Codex Ten at Substack

COVID/Vitamin D: Much More Than You Wanted To Know

Mary Pat Campbell

Link: https://astralcodexten.substack.com/p/covidvitamin-d-much-more-than-you

Excerpt:

Are there real randomized controlled trials? Yes – three of them. One from Spain (n = 76) tried randomizing hospital patients to get or not get Vitamin D; the patients in the experimental group did much much much better (25x lower odds of having to go to the ICU!) than the control group. Another from India (n = 40) tested asymptomatic people with mild cases; everyone stayed mild but the patients treated with Vitamin D were three times more likely to clear the infection quickly. The last, from Brazil (n = 240) was a large multicenter RCT that tested whether Vitamin D affected the outcomes of hospital patients. It found no effect whatsoever, not even a hint of a trend.

I could take or leave the Indian trial; nobody is worrying very much about seroconversion in mild cases. The Spanish and Brazilian ones are a pretty jarring contrast. In the Spanish one, the Vitamin D treated patients did 25x better; in the Brazilian one, there was no benefit at all.

I side with Brazil. It’s bigger, more professionally-done, and has fewer minor statistical flaws that really shouldn’t matter this much but make me nervous. Also, in the past I have learned to side with negative RCTs rather than astoundingly massively positive ones when the two conflict. There were some early astoundingly massively positive RCTs for hydroxychloroquine, and then the bigger and more-professionally-done ones that came later showed no effect.

Author(s): Scott Alexander

Publication Date: 16 February 2021

Publication Site: Astral Codex Ten at Substack