How not to be fooled by pseudo nutrition research | No, 16:8 fasting won’t cause more heart disease death

Background

Recently, a preview of research on the correlation between time-restricted fasting and cardiovascular death was made popular by the modern news media. It was firstly reported by the American Heart Association (AHA). The SEO friendly title — 8-hour time-restricted eating linked to a 91% higher risk of cardiovascular death — quickly captured the attention of many health-anxious, statistic-illiterate, middle-class, mortgage payers.

I was amazed by the perfect combination of the following:

• The media that frantically reports anything that could capture the audience.
• The degenerated nutrition research community.
• The anxiety driven mass.
• AHA, whose dirty history of rent-seeking is unknown even by GPT-4, acting authoritative and neutral while possibly being none of them.

Anyhow, let me tell you why you shouldn’t be fooled like our contemporaries by pseudo-research like this and avoid even being worried about them in the future — without missing anything important.

Why the 91% higher probability is horseshit

The so-called research is not published at the time of writing, chances are it will never be, but its poster was made available here. Let’s see the two critical tables concerning its data. Can you spot its fault?

The description of the distribution

The correlation between fasting and cardiovascular death

If you can’t find it, let me highlight the part you should look at:

The 16:8 fasting group has 23.2% black people compared to 6.6% in the reference group, a 252% increase; it has 27.1% smokers compared to 16.9% in the reference group, a 60% increase. So, is it quite a finding that the 16:8 fasting group has 91% higher death from heart disease? I think the data is more about the correlation between being black and heart disease death than what it claimed.

Don’t blame yourself if you still don’t get it. Consider reading this book or ask ChatGPT about Confounding. I assure you that it will be worth your time.

Wrong way to debunk

Feel free to skip this part if you don’t like getting too heavy in statistics.

Many on Twitter (now X) claimed the result was meaningless because the fasting group has only 414 samples. Even the mighty Taleb agreed. I think the data size is actually sufficient. I think it’s the most defendable part of this whole study.

The over-hacked P-value is actually a solid metric here. There are many ways to look at it. Let’s use Beta distribution this time. In a nutshell, Beta distribution reliably describes the distribution of Bernoulli’s true P given a H-heads and T-tails coin flip experiment. It addressed the issue of sample size.

In our case, the reference group has 11831 people with 423 deaths. That’s a death rate of 3.6%. Beta distribution told us that in 99.9% of the time, the true death rate is less than 4%. The fasting group has 414 samples with 31 deaths, This gives us a death rate of 7.5%. Its Beta distribution says that in 97% of the time the true death rate is bigger than 5%.

While it’s not slam-dunk, the p-value story is much rosier than the confunder fault above.

The breed of longevity studies

If you have followed the longevity community for some time and have a sound reasoning skill, you don’t need my explanation to ignore most of the newly published results. This research community is a textbook illustration of Why Most Published Research Findings Are False. The researchers are strongly incentivized to publish new findings and are not punished enough by being wrong. Such an incentive structure will always result in what we have today: garbage.

In my opinion, you won’t miss much by ignoring all research since 1900 about nutrition, longevity, etc. Instead, just follow common sense and over value the practices that passed selection (Lindy compatible). You will get 98% of the best possible result. This should be the strategy for 90% of the public, the same way they will be better off just buying SP 500 and holding it until retirement. Those common sense points to

• Diet that exclude toxic stuff (sugar, processed food, white rice/bread, etc) and balanced in protein, fiber, and vegetables.
• Well developed exercise system
• Sleep quality

If you are zealous about such things like me, a research is worth our time only if it meets all the following criteria.

It was published by a highly respectable journal like The Lancet or Nature.

As a generalization of the above, those endorsed it must have much to lose by being wrong. The money used to buy them can’t be close to their punishment. Their skin must be in the game. In this sense, Youtubers with 500K followers are more trustworthy than AHA. Those Youtubers will be ruined if they recommend stupid results either out of their incompetence or for rent-seeking. No individual stands to lose big if the reputation of AHA is ruined, therefore, it’s much cheaper to buy AHA than CDC or FDA.

The impact is more than 80% — risk factor, etc. Ignore ones with 25% impact because those effects, even if true, will be diluted to non existence by other factors in life. To give you some reference, smoking increases risk of lung cancer by 1500% to 2500%; By contrast, eating red meats increases the chance of colorectal cancer by 17%. A sensible person should comfortably eat red meat without fear of colorectal cancer.

The result must be reproduced many times. If the impact is big and the results are true, there must be a strong incentive to repeat it and productionize it. In the end, it will be hard to miss it even for a layman. Were the 91% increase in cardiovascular death were true, many will be incentivized to reproduce the result because intermittent fasting was such a hype in recent years. Same is true for Resveratrol which almost turned into a scandal. To miss such a hot topic by the publication-hungry community is like to miss DTJ by Trump supporters in March 2024.

The math must be sound. Given the above criteria, this is redundancy in practice. I never use my math muscle on any research to just falsify it; I do that for work out: Data analysis is a use-it-or-lose-it valuable skill.

Conclusion

• The research result of heart disease death increase from fasting contains confounding from disproportionate black participants and smokers. Hence the result is junk.
• Health-aware public will do great by following the time tested common sense of diet, exercise, sleep, emotion management.
• They miss little by ignoring newly published research completely.