I just posted a new video:
https://youtu.be/KHY5XfMfdig

This post is the text version of that video. I didn’t send my last blog by email (I want to limit emails to one per month). In that last blog post, I linked to 3 videos. In case you missed them, I worked on two vlogs about burning Iron:

Attempt 1: Classic demo and solid oxidizer failure to ignite:
https://youtu.be/46YlW8qjp_c

Attempt 2: Steel wool burning in pure(ish) oxygen:
https://youtu.be/SURy8xRsRp4

And a video essay about robotics and automation:
https://youtu.be/TisXw8zS5r0

This latest work is a bit of a rant about a research rabbit hole. I was trying to ask what seemed like a simple question. Which is more important to longevity: diet or exercise? That seems pretty straightforward. I was surprised to find that it was almost impossible to get a clear comparison between two effects.

And then I found this study by Atefatfar et al. (2023). This study examined all four possible cases of the good/bad diet versus exercise/no exercise. I expected them to have similar, additive effects. But that’s not what they found.

They constructed a graph called a Kaplan-Meyer curve. KM curves show survival over time. So, the best-case scenario is a line that is only slightly sloped and stays high longer. The bad case should show a rapid slide downward. So, we should expect things to look like shallower slopes the more healthy choices you make, right?
Maybe something like this: good diet or exercise (but not both) fall somewhere in between.

But, no. In the real data, diet doesn’t make any difference at all. Why? Why would that be? Diet has to do something, right?! And it does – I found lots of papers on how diet decreases this biomarker or that biomarker or heart attack risk. So, it’s not as if diet does NOTHING, it just doesn’t make us live longer? Except that seems to fly in the face of common sense, too, right?

We can imagine reasons why this might have happened. People tend to be unreliable when they report their own diet. Alternatively, the specific dietary questions asked in this study might not have reflected especially good practices. It might be that diet has a smaller effect than exercise, and so it would take longer to see that in the data. But this was over about 20 years. It also might be that the study size wasn’t large enough, but this did follow more than 6,000 people.

So, I couldn’t believe that this was real. Despite my best efforts, I couldn’t find the study (I was sure must exist) that clearly shows that a good diet increases longevity. That seems like such an obvious statement that it must have been tested. People must have measured this, right? I even thought about picking a fight with the Internet. I knew that if I posted a video with the statement “I don’t think diet actually does much of anything,” someone was going to prove me wrong. Everybody loves proving people wrong on the Internet.

Food is an emotional topic! If diet doesn’t matter, why not subsist exclusively on skittles and rice? I was sure people would respond with passionate, contradictory claims. Like: We must optimize nutritional input to maximize health! And meat is poison! No, meat is medicine! I figured people would go off and find the missing study just to prove me wrong.

I was going to offer a trophy.

But I found the study I was looking for: Ahmad et al 2024. This is the first paper I could find that actually shows clear evidence with large sample sizes of all-cause mortality effects of diet. The meta-analysis of the previous research (Sebastian et al 2024) before that suggested that the effect of a Mediterranean diet was significant for heart attacks, but not significant for overall mortality.

And that’s crazy – health food is a multi-billion-dollar industry! It sure seemed like, when it comes to lifespan, the whole thing was a scam and that twinkies and Big Macs are almost as good as spinach and cashew butter. But there is an effect. My sanity is OK after all.

And even if it’s modest in healthy average adults, diet has a huge effect on some conditions. If you have diabetes, then that dictates that certain diet choices will really hurt your health. If you have PKU, certain diets (especially as a kid) will have a terrible effect on your brain. But when it comes to healthy adults… it seems like the point of diminishing returns is pretty early.
I’m saying that the effort-results curve for diet looks like a slightly sloped plateau, not a hill.

Obviously, we get a big benefit from adequate nutrition. We need certain nutrients in our diet, or we get sick. We can get scurvy, rickets, vitamin A blindness, or beriberi, among other diseases of deficiency. Scurvy can shorten your life by decades – it’s an acute and fatal problem if untreated. Someone who is 25 and has a diet so bad that they have scurvy? They can extend their life by 50 years just by getting out of the danger zone. After that? They might get another 8 years by really optimizing their diet. The effect is actually not that big for most people. Additional nutrients are basically just peed away. Those are extra molecules we can’t use.

The marketing world makes it easy to think that more effort will have proportionate results. “Spend more to get more.” But that doesn’t seem to be the case.
The best evidence for a benefit suggests the following 9-point scale. The higher you score, the better your diet. A score of 6 or higher was correlated with a significant reduction in death rates (in other words, you live longer). The study has limitations: it was conducted exclusively in women, and the effects were still not very big, and still had big error bars. But it at least shows the effect size of a good diet: 23% +/- 7%.
(i) high intake of vegetables
(ii) high intake of fruits
(iii) high intake of nuts
(iv) high intake of whole grains
(v) high intake of legumes
(vi) high intake of fish
(vii) high proportion of unsaturated fats
(viii) low red/processed meat
(ix) low alcohol consumption

And, clearly, this is good advice. It’s rooted in measurable effects of a good diet. Other studies show that it lowers the odds of cancer, diabetes, heart disease, and degenerative neurological conditions. It improves biomarkers of health like blood cholesterol. These are good guidelines.

But these guidelines are… boring. They undermine massive industries, countless marketing campaigns for “superfoods,” and a cultural perception of some foods as sacred while others are sinful.
Look at saturated fat.
Look at trans fats.
Look at antioxidants.
Or low-fat substitute products.
Or low-carb substitute products.

All of these have effects on the same order of magnitude. They change mortality by 20-30%. I want to believe that my spinach smoothies will help me live longer – that they have more impact than pleasant poops. But the effect size is just not that big.

The problem is that studying small effects is hard. It gets complicated by real world problems I talked about above: self-reporting is not very accurate, and it takes a long time. And there is at least one more wrinkle: different diets affect people differently. The Mediterranean diet is probably good for most people. But maybe there’s an ideal diet for each person! But given a small effect size, it will be really hard to measure what diet is right for each person.

So, we don’t actually get a huge health benefit from a lot of effort and expense on food. Qualitatively, a good diet is good. Quantitatively, you can maybe get about 8 years lifespan extension by good adherence to a great diet over your whole life. That’s bigger than a rounding error, and it’s worth trying for, but it’s not radical life extension. I take nutrition hype with a grain of salt.

And it’s similar to the effect of exercise (though exercise might be a little bigger, see Sheehan et al 2020).

Further reading:
Atefatfar, Ayda, Siavash Babajafari, Noushin Mohammadifard, Fatemeh Nouri, Maryam Boshtam, Masoumeh Sadeghi, Mohammad Hossein Sharifi, Asma Kazemi, and Nizal Sarrafzadegan. “A Healthy Diet, Physical Activity, or Either in Relation to Cardiovascular and All-Cause Mortality: A Prospective Cohort Study.” Nutrition 116 (December 1, 2023): 112186. https://doi.org/10.1016/j.nut.2023.112186

Ahmad, Shafqat, M. Vinayaga Moorthy, I-Min Lee, Paul M Ridker, JoAnn E. Manson, Julie E. Buring, Olga V. Demler, and Samia Mora. “Mediterranean Diet Adherence and Risk of All-Cause Mortality in Women.” JAMA Network Open 7, no. 5 (May 31, 2024): e2414322. https://doi.org/10.1001/jamanetworkopen.2024.14322

Sebastian, Sneha Annie, Inderbir Padda, and Gurpreet Johal. “Long-Term Impact of Mediterranean Diet on Cardiovascular Disease Prevention: A Systematic Review and Meta-Analysis of Randomized Controlled Trials.” Current Problems in Cardiology 49, no. 5 (May 2024): 102509. https://doi.org/10.1016/j.cpcardiol.2024.102509

Fadnes, Lars T., Carlos Celis-Morales, Jan-Magnus Økland, Solange Parra-Soto, Katherine M. Livingstone, Frederick K. Ho, Jill P. Pell, et al. “Life Expectancy Can Increase by up to 10 Years Following Sustained Shifts towards Healthier Diets in the United Kingdom.” Nature Food 4, no. 11 (November 2023): 961–65. https://doi.org/10.1038/s43016-023-00868-w

Caprara, Greta. “Diet and Longevity: The Effects of Traditional Eating Habits on Human Lifespan Extension.” Mediterranean Journal of Nutrition and Metabolism 11, no. 3 (September 15, 2018): 261–94. https://doi.org/10.3233/MNM-180225

Sheehan, Connor M., and Longfeng Li. “Associations of Exercise Types with All-Cause Mortality among U.S. Adults.” Medicine & Science in Sports & Exercise 52, no. 12 (December 2020): 2554. https://doi.org/10.1249/MSS.0000000000002406