Analytical decision making on health & longevity

[mccoy]

Hi all,

I'm starting here a discussion on which dietary(/lifestyle intervention to adopt for healthspan and longevity when faced with discordant opinions, articles, studies.

I beg your pardon if I'm posting this in steps or installments, taking advantage of spare moments of quiet.

So, the scientific papers, even those in credible journals, often tell us things which are one the opposite of another. Also, advocacy, fanaticism, vested interests, academical interests may govern the content of a paper. The sheer size of these papers is such that even specialists sometimes hire teams of people who select the relevant info. Last, if we hear the experts, their opionions on some issues are very discordant. What to do? Is dr. Greger right or perhaps Dr. Mc Dougall is better, or maybe Dr. Fuhrman beats them all but Dr. Ornish is the most authoritative, on the other side people have had very good results heeding Dr. Attia or Dr Rosedale and Dr. fung, the advocates of low carb and keto diets which suggest just about the opposite of the 'vegan doctors'.

One possibility is the adoption of Bayesian inference

 

First step is usually to assign a value or distribution to the a priori function, P(H), our starting point, what we know on the subject.

for example, the subject, not a banal one,  could be beneficial or detrimental effects of the ingestion of fresh fruit (250 to 750 g daily). 

We may at the beginning restrict the subject to apple and oranges. If I eat apples or oranges, am I going to benefit or not or even am I going to get sick?

So, first thing I would hypothize a benefit function (BF) with values ranging from -1 to +1. 

BF= -1→ apples and oranges are going to be 100% detrimental 

BF= +1→ apples and oranges are going to be 100% beneficial

BF=0 → apples and oranges are irrelevant to health, not detrimental nor beneficial

BF= -0.5 → 50% of probability that fruit is detrimental

BF= +0.5 → 50% of probability that fruit is beneficial

And so on....

The BF is just a simple way to assess quantitatively the degree of benefit or detriment that our food or intervention will deliver.

 

[mccoy]

Now, to the question: beneficial or detrimental effects of the ingestion of fresh fruit (oranges or apples 250 to 750 g daily). 

we may assign an a priori value based on the following sources of knowledge:

• Natural medicine: fruit doubtless beneficial
• Nutritional guidelines, traditional: fruit beneficial
• Potential effects of fructose on fatty liver: fruit detrimental
• Potential effects on blood sugar: fruit detrimental
• Effects of fiber on gut fauna: fruit beneficial
• Effects of minerals, vitamins, anthocianides, polyphenols: beneficial
• Effects on caloric balance: neutral
• Subjective effect on myself after 42 years of regular fruit ingestion: beneficial

I introduced my subjective experience because that's part of my prior knowledge, which might change for other guys. I might be forgetting something, but from the above my BF will have an expected positive value, since 5 points are beneficial, 2 detrimental. My estimate of the Prior value of my BF (average or expected value) will be something like +0.5 if I want to be cautious. That is, a prior belief that the cited fruit has a 50% likelihood that its effect is beneficial to my healt & longevity.

P(H)= 0.5

[mccoy]

Now we may obtain our posterior  P(H|E), the final decision-making value after our prior  P(H) has been combined with the data or evidence or likelyhood P(E)

One simple way to do this is to use  bayesian inference for a normal distribution, which uses a normal prior conjugate (the prior and posterior are normal distributions).

With reference to the following link:

https://people.eecs.berkeley.edu/~jordan/courses/260-spring10/lectures/lecture5.pdf

where E(mu|x)  is the posterior, mu0 is the prior, sigma^2 is the variance of the BF distribution and x is our data or evidence.

The subscript  0 = prior distribution.

If we simplify assuming that the variances of the posterior and the prior are equal, then we'll have for the average or expected value:

P(H|E) = posterior BF = 0.5 (prior BF + data BF )

That is, the posterior value will be the arithmetic mean of the prior and the data

 

 

 

[mccoy]

Now, back to our example.

We have to determine a BF from evidence.

BF0 is our prior BF, which is = 0.5

BF will be the evidence or data with which we'll update the prior.

Now, having measured my blood sugar after the ingestion of apples or oranges, I've determined that the spike is not significant. this is a benefit which I'll assign a BF score of 0.9 (good benefit)

So my posterior BF= BF1= 0.5*(0.5+0.9)=0.7 this is the new evidence given the data.

Further evidence may be hearing Lustig that, in the interview with Peter Attia, explains that the fructose in fruit is absorbable by gut bacteria, so that's another benefit which may be valued as a 0.5

Updated posterior: BF= BF1= 0.5*(0.7+0.5)=0.6 this is the new evidence given the data.

Further, I noticed that fruit in those quantitites is easily digested without any gastric distress. That's a valuable benefit to me hence BF=1

 New BF1= 0.5*(0.6+1)=0.8 this is the new evidence given the data.

More evidence: the article in which it is described that, in lean people with no energy surplus, the effect of fructose on a fatty liver condition is nihil. That's a very valuable evidence, hence BF= 0.9

ew BF1= 0.5*(0.8+0.9)=0.85 this is the new evidence given the data.

At this point, I have a Bayesian belief that fruit or oranges in the quantity of 250 to 750 gr is beneficial to me with an 85% assigned degree of probability.

Bottom line, Bayesian inference suggests to me that I can keep eating oranges or apples and I'll receive a most probable benefit from it.

The result is objective because derived from an objective mathematical procedure, whereas there is an amount of subjectivity given the assignation of values to the BF and the personal measurements or observations.

In an individual with impaired glucose tolerance, blood glucose might be spiked and that may constitute a marked detriment with BF=-1

We may also assign a variability to the BF values according to the certainty or uncertainty of data.

Any comments, suggestions, critiques are welcome

 

 

 

[Sibiriak]

Quote

... the question: beneficial or detrimental effects of the ingestion of fresh fruit (oranges or apples 250 to 750 g daily). 

Might a better question be something like-- would I benefit from substituting   X portion of my diet (keeping total calories constant) with  250-750g of oranges and apples?--  where X might be, say an isocaloric portion of leafy greens or  vegetables, for example.     I.e. doesn't the benefit of ingesting a certain food type depend to a significant degree on the food type that ingestion displaces, and how that displacement alters the overall nutritional profile of the diet (macrotnutrient  ratio, amino acid balance, vitamin, mineral, micronutrient/ polyphenol content etc.)?

[mccoy]

4 hours ago, Sibiriak said:

Might a better question be something like-- would I benefit from substituting   X portion of my diet (keeping total calories constant) with  250-750g of oranges and apples?--  where X might be, say an isocaloric portion of leafy greens or  vegetables, for example.     I.e. doesn't the benefit of ingesting a certain food type depend to a significant degree on the food type that ingestion displaces, and how that displacement alters the overall nutritional profile of the diet (macrotnutrient  ratio, amino acid balance, vitamin, mineral, micronutrient/ polyphenol content etc.)?

Hi Sibiriak, those are all reasonables doubts, some of which I have been thinking about. The example is of course generic and in my case there was no displacement, since I already eat that amount of fruit.

 the assumptions are : isocaloric conditions, minimum requirements of AA and other nutrients respected, macronutrients ratio not significantly altered. Of course I won't diplace protein with carbs, unless I clearly include such situation in the analysis, and then the BF might be even negative if I go below the minimumrequirement of protein after such displacement.

The micronutrient/polyphenols content to an extent is already considered in the analysis, since a positive benefit function BF was assigned because of vitamins, minerals and phytochemicals content.

Of course we can impose further constraints to the problem, according to our personal situations.

You may try and propose an analysis, like for example, 'what's going to be the BF if I displace an isocaloric amount of cereals with X gr of apples with skin?'

Or the same, displacing vegetables (which ones?) with X grams of apples, or other fruit.

In the previous example, if I included bananas, these are hi-carb fruits whose effects on my blood sugar i did not measure, so I would have less evidence of benefit and probably the BF, until further update,would be less beneficial than apples or oranges.

 

[mikeccolella]

I think we need to be careful about “isolating” something like fruits. Foods need to be considered in the context of meals and overall dietary choices.

[mccoy]

3 hours ago, mikeccolella said:

I think we need to be careful about “isolating” something like fruits. Foods need to be considered in the context of meals and overall dietary choices.

We may add another decision on whether to consume fruit within meals or by themselves, as commented in the other thread (the mousetrap).

To me, eating  fruit within meals, in significant quantities, would mean unacceptable bloating and fermentation, hence a totally negative benefit function.

Others may enjoy a modest amount of fruit within a meal without any discomfort, and that would constitute a positive benefit function.

Overall dietary choices: probably eating fresh raw fruit, in the absence of impaired glucose tolerance and caloric surplus, is always beneficial, whatever the dietary context.