Why the centenarians and Blue Zones narratives on longevity are potentially very dangerous

Being a person with a complicated health history from a family that hasn't exactly been dealt the most favorable genetics, I do my best to try and improve my personal odds of survival, that compared to the general pulation are not good at all, at least not at my current age of 47. I run experiments with my own data using my background in engineering and control feedback system design to try to find out if the stuff I find through pubmed searches and social media links could be valid 'for me'. For me, risk is all about hedging my bets. Unlike many people in nutrition will try to tell you the science is clear and settled, it truly is not. I do a lot of data engineering, including a lot of simulation-based hypothesis quantification and I can assure you that nutritional science is not settled at all. You may hear a lot of food X is the new cigarettes but rest assured, nutrition, at least the part of nutrition that is extensively studied has no indiscriminately bad stuff. Not anywhere close to tobacco. There are a lot of viable hypotheses out there for the experts to base relatively solid educated guesses on, but given that much of the theories require multiple layers of inductive reasoning to arrive at a model, however good the model, the science is not at all settled enough for anyone at high risk not to hedge their bets. Most nutritional scientists don't have much skin in the game themselves so they should have no issue with betting on the fastest horse or the second fastest horse who has a jockey with a track record better than that of the fastest horse. For me and others like me though, betting on just one horse would increase our chances of being worm food long before the science ends up being settled.

In some cases it is possible to hedge your bets, put a small amount of money on the long shot slow horse with a 80:1 payout. For example, I don't tolerate statins, and while there is very little reason to believe LDL reduction by non-statin drugs does in any way reduce my risk of an early death, I do take high dosage Niacin as a way to hedge my bets. Look at it as the slow 80:1 horse that I put a five euro note on.

Orthogonal to the idea of hedging our bets between different models of disease is the increasingly popular yet horribly misunderstood concept of longevity. Currently, there are multiple popular narratives in nutrition around the idea of longevity diets. In this blog post, I want to outline why the approach taken by the people who support and believe these narratives is horribly flawed and likely to lead to lower, not higher life expectancy.

To understand the problem, we need to realize what we are looking at when we decide to study for example regions with a relatively large population of centenarians and we need to consider the potential validity of the old adage "what doesn't kill me makes me stronger" and what it means to the tail of the population statistics.

My own family history has a mortality curve that would make my chance to live to be a hundred quit a bit above average. While the place I am now at 47 is jokingly called the valley of death in our family, and the probability of me biting the bucket before I am 60 years old is many many times higher than for the general population, if I manage it to 70, the mortality in the 70 .. 90 year old range is surprisingly low. If you were to purely look at me and my family from a blue zones type of longevity perspective, you would be fooled into thinking my life expectancy was above average while in fact it is very much below average.

There are many such examples where the life expectancy of one group might have an early modus yet a long and fat tail and another group might have a relatively late modus but a less favorable tail distribution if looked at from a simplistic longevity focus.

Given my personal risk profile, I don't have the luxury to even consider focusing on longevity fact, my risk attenuation strategy has so far been on trying to counteract the 'favorable' longevity curve my family is cursed with.

I would consider a longevity curve like this a Keith Richards longevity curve. While Keith Richard, at 74, might end up living to 90 or maybe even a hundred, most people from his generation with his lifestyle were buried before reaching 40. So even if Keith Richards would manage to live to a 120 years old, describing his lifestyle as a longevity lifestyle would be folly. We are looking at the far tail of a distribution and if we discard the rest of the life expectancy curve, doing so makes us jump to horribly wrong conclusions.

When we actually go hunting for places where longevity tails exist, as in the Blue Zones and more recently but less outspoken, the Pioppi diet, we are making things worse as the probability that we end up hunting either fat and long tails or simply just spurious zones that won't be here a decade from now becomes astronomical. In fact, if we look at the blue zones, we see some of these already disappeared, but that doesn't stop the believers. A picture of a Mcdonald's on some stay corner of a street somewhere in the region is enough to maintain the narrative. Doesn't matter that most visitors of the restaurant are tourists, the spurious results aren't invalidated because, well, McDonald.

As said, it is always good to hedge your bets based on the available evidence. But it is just as important to recognise models that are based on highly ambiguous data that should never actually have been used for model creation or confirmation.

First of all, numbers you should actually look at when exploring diets in different regions should depend on your age, health status and family history. If you are healthy and your family history gives you nothing much to worry about in the next decade, go for the remaining life expectancy for a person of your age and gender. If like me, that isn't the case: go for the probability of surviving the next decade.

Secondly, it is never, and I mean absolutely never a good idea to go hunting for outliers in data sets. Don't select on outliers. Don't select. Anyone who has ever done anything with statistics, data science or data engineering knows the place most likely to yield interesting and useful results is the central mass of the data points. In fact, using statistics that avoid putting too much weight on outliers has become the norm in most of science. Blue Zones and similar longevity centered approaches however sadly seem to actually embrace what outside of nutrition is frowned upon. So when trying to hedge your bets, evaluate all available data but weigh the data according to proper science without making silly assumptions about your own risk based on tail extreme outliers such as presented in much of longevity geared nutritional arguments.