Honestly I'm pretty shocked that no one appears to have picked up on this. It's a seriously weak paper that proves very little about anything.
11.03.2026 02:08
👍 0
🔁 0
💬 0
📌 0
Honestly I'm pretty shocked that no one appears to have picked up on this. It's a seriously weak paper that proves very little about anything.
I have now written up the issues with this study.
Turns out it's even worse than I initially thought!
gidmk.substack.com/p/multivitam...
Peak toddler this morning: Little Nerd demanded toast like daddy, then after said toast was toasted wanted porridge, then ate some of my toast anyway 🤣
As my wife would say, just a full on whackadoodle
Some people who think RFK Jr. is crazy wrong on vaccines (he is!) believe he's some kind of north star on food because he says "eat real food" - which EVERYONE in food has always said.
He's mostly crazy wrong about food, too.
@mikegrunwald.bsky.social explains.
www.thebulwark.com/p/robert-ken...
If they had published every analysis they ran and were honest about what they meant, I would have no issue.
Yes, good pick up.
But in addition, this sort of sub-study is a giant waste of time and the fact that it got into Nature Medicine by showing some barely-significant p-values and a Cohen's d of 0.03 is a sign of just how meaningless getting into Nature Medicine really is.
We KNOW that this does not mean that the participants who got a multivitamin are healthier in any way, because they WEREN'T.
They did die less, or get less cancer, etc. They just have some minor differences in their DNA that probably mean nothing.
It's an open question whether 4 years of multivitamin use is worth somewhere in the region of 1 month less of biological aging.
For one thing, there's no strong correlation between these measures and hard outcomes like death or heart disease
Technically, this was a pre-registered analysis and therefore not open to investigator degrees of freedom.
In practice, there are so many ways that the authors could've easily tweaked the data to get a p-value just below 0.05 for the publication that the results are entirely meaningless.
Which brings us to the newest study. Using "biological clocks" derived from measures of DNA methylation, the authors found that there was a barely-detectable benefit for multivitamins in 2/5 of the analyses (p=0.032 and 0.017).
There was also a DETRIMENT for cocoa supplements (p=0.044)
The authors have already published a sub-study looking at cocoa supplementation showing a very small, barely-significant effect on inflammaging markers. As far as I can tell, no word yet on the multivitamin component of the study and whether it had similar effects.
The registration of these secondary outcomes was wonderfully vague. Here's some text from the registration of the current sub-study looking at DNA methylation.
There are *dozens* of ways to analyze the data and meet this very broad description.
COSMOS was a MARS-funded investigation into cocoa supplements and multivitamins for human health. It failed to find a benefit for either supplement for the main primary outcomes, including heart disease, cancer, and all-cause mortality.
However, they registered around 50 secondary outcomes.
The COSMOS study of multivitamins and cocoa supplements continues to be a ridiculous circus that mostly shows how broken science is.
This time, it's DNA methylation.
The problem with health libertarianism a la MAHA and co is that it's fundamentally opposed to the evidence.
You can't easily reconcile the fact that much of health is outside of individual control and the desire to remove all government intervention from public health.
In my opinion it changes nothing about the assessment. The piece was a 25-year-old opinion piece by three people two of whom are dead. It was well-cited, but didn't actually impact the evidence.
Stupid thing to do, of course, and good to see it retracted.
My new piece about glyphosate and why it's probably fine for your health unless you're spraying the stuff in which case the data is mixed.
Also, RFK Jr. got attacked by all of his friends and it was very funny to me.
slate.com/technology/2...
While not a defense, I suspect that graph would look very similar if you were to get US data on the same topic.
What a remarkable statement. Both true and chilling.
"It is clear that the sum of US policy choices amount to mortality worse than a COVID-19 pandemic, ongoing, year after year" @gavinyamey.bsky.social
Sadly that seems to be the case.
I am being yelled at on Threads by a "medical astrologer" who believes that epidemiology is all about correlations and that oncological surgeons fund studies showing that pathological BRCA mutations cause cancer in order to drum up more patients.
The only remarkable thing about this is that the journal editors have not yet retracted every one of these papers.
Making up data - even if it is a case report - is an astonishing breach.
Seems, uh, highly implausible.
❤️❤️❤️
Nothing is more fun than deeply irritating the statistical community online.
Laugh at your peril, they will come for you with their slide rulers and pedantry.
Roses are red,
Violets are blue,
It was statistically significant (p<0.05)
Which means that it's true.
I am the bed of many things
"Coffee is a tea" quintessential Brit well done