@Vapid-Bobcat
I have no disagreements with you on the questionnaires. When I say single, I mean that there was a single instance of data collection, not that only one type of questionnaire was used. We do not know the real long-term composition of people's diets- we have a single snapshot of their diets from a single filling out of a questionnaire.
You are correct that serum =/= tissue, but if you're going to use that against the study you'd need to provide some evidence that there's a sufficient(model invalidating) and widespread enough(population %) gap in PUFA intake to serum levels.
Adrenic acid is formed from linoleic acid or arachidonic acid(which itself is a successor molecule to linoleic acid)- the increased consumption of either would increase AA in the serum, unless you are someone whose tissues are oxidizing PUFA at such a high rate that it cannot accumulate and is almost immediately taken out of the bloodstream.
Speaking for myself, I have my own thoughts regarding things like PUFA and do not think that it's as much an issue for people who are otherwise healthy- barring excessive intakes, terrible 3/6 ratio, and some other things. For me, this comes from experience with medicine and some studies on the impact of behavioral patterns. About 90% of the reduction in CVD risk is behavioral, not dietary or medical. This is a figure that the study finds in its risk data and it's about what I've seen myself.
We know that observational epidemiology cannot make causal inferences, but one has to wonder why high PUFAs, if they truly are the problem Ray Peat outlines them to be, rarely associate with worse health outcomes.
You are seeing them rarely associate within contexts that are not 1 to 1 of PUFA to negative outcomes.
You can torture the PUFA intake variable to control for a vast litany of confounds(like health behaviors, smoking, etc.) but when you do that you need to ask "how many people are we comparing." It's not enough to say that you've controlled for the variables; if we know that increased risky health behaviors are related to increased saturated fat intake, then the number of people we're doing statistical assessments on is likely very low, when we're looking at high saturated fat intake people who do not engage in risky health behaviors.
You're already taking away 90% of the predictive ability regarding CVD when you do behavior controls, so you're left with a limited predictive variable and and limited sample to run tests on. I am not able to find these figures in your study and I don't want to dismiss it by assuming that the study has these issues, but it's something that I would consider in whether you should draw causal evidence from it.
Even Ray Peat said that someone could have a high PUFA intake(he uses PNW natives and Salmon as an example) and be very healthy due to a high rate of PUFA oxidation- the PUFAs don't store. Controlling for health in saturated fat vs PUFA isn't really going to reveal much except that healthy people are healthy.
You're also removing the possibility that increased PUFA consumption does have negative secondary effects like reduced cognition- making someone more susceptible to increased smoking and drinking(which would lead to reduced health outcomes), but when you control for all of the variables you remove the observation of those secondary effects and are thereby isolating PUFAs impact on health in a way which is not found in the real world.
In short, you are lowering the PUFA effects by removing the analysis of secondary effects from increased PUFA consumption as part of model creation.
We do not live in a world where everyone is healthy- people are under the influence of many metabolic stressors and as such PUFA is going to store, oxidize, and interact with deficiencies in order to create obesity, cancer, diabetes, heart disease, alzheimer's, etc.
If we could live in a world where we've made everyone have strong pro-health behaviors, no metabolic stressors, etc. then we'd be able to have a bigger conversation about the proportion of PUFA in the diet and how it relates to heart disease.
Ray's work is about real people in the real world, not people in a model with a dozen variables controlled for in comparison to others. Outside of the controlled variable models, PUFA consumption is not something that we should encourage people to increase, especially because in people who are already sick it will only make them sicker; the same is true of exercise and sick people- exercise does not help but hinders health(mostly by increasing fatty acid oxidation and stress).
I agree that healthy user bias is probably explaining part or maybe all of the effect of PUFAs on different outcomes. Another possible explanation is over-adjustment for variables such as BMI.
It's lazy to say "healthy user bias". There's unhealthy user bias, as well. There's also underreporting of cigarette and alcohol consumption(especially by current and former smokers and heavy drinkers, respectively) which was an issue in the 90s to 2000s(the time period btw that your study gets its data from). This isn't minor underreporting but on the order of 30-40% for cigarettes.
Underreported alcohol intake is most severe with spirits or hard liquor(one study finding two-thirds underreporting), spirits are also the preferred drink of heavy drinkers which adds significant potential for confounding.
This confounding is only boosted by the relationship between bad health behaviors like smoking and drinking and the consumption of saturated fat. In the US(at least) these behaviors are often closely related; as such, heavy drinkers who have high saturated fat intake may underreport spirit consumption significantly more than heavy drinkers with high PUFA intake, on account of a strong relationship between diet and bad health behaviors.
This is getting long but it's been good mental exercise for me and I hope it's somewhat informative for you, as well.
One last point to consider is the underreporting of energy intake. Obviously, the groups most likely to consume saturated fat, drink, and smoke, are also most likely to underreport their caloric intake- there's a few studies showing this. If there's non-even distribution of calorie underreporting, you're going to have effects in your model that are non-evenly distributed and that would otherwise not exist in a real 1-to-1 calorie comparison. You would see a statistically-significant association with saturated fat and negative outcomes(and the contrary association of PUFA and positive outcomes) within a controlled-variable model when that association is actually non-existent or even reversed but there's a hidden caloric intake on the side of the unhealthy people- an invisible finger pushing down on the scale within the model. Controlling for variables doesn't mean much when you cannot accurately control them because people are giving you bad data.
