Rule of thumb: if the study does not mention a control group and a treatment group, it cannot make any predictions of causation. Period. If it doesn't mention a control group and a treatment group, it is an epidemiological (descriptive only) study.
This is an epidemiological study. It means nothing in terms of causation. Epidemiology can NEVER show causation, only correlation. It should only be used to identify factors that might conceivably be studied in a follow-up experimental study, where there's a treatment group that gets some kind of intervention (such as a dietary limitation, or a drug) and a control group that doesn't.
How do I know it's an epidemiological study? Look at the design: "Prospective cohort study." Self-reported data. No interventions or treatments - just one big group of people that get followed for a year or two to see how they do. That's the classic epidemiological study - and it's crap for prediction because you can't predict (prove causation) with epidemiological data. Epidemiological studies are descriptive and exploratory ONLY. If these researchers want to show that saturated fat increase causes Alzheimer's, or heart disease, or anything else, they have to do a study where saturated fat is isolated in the diet and that's all the subject eats.
Certainly, epidemiological studies can talk about association and correlation, and they can give researchers an idea of factors that should be investigated further with experimentation, but as Stabby has already pointed out above, people who eat lots of saturated fat today also tend to consume a whale of a lot of bad things like sugar, grains, and other starchy foods. There's a huge number of uncontrolled and unmeasured variables in this study.
I'd love to get my hands on that data and re-run the correlations to see if there was significance in carbohydrate intake. I'll bet there would be. But that probably won't be possible anytime soon. And in any case, there'd still be a bunch of confounding factors that weren't measured and can't be controlled for, so the data is pretty much worthless for predictive (determination of causation) purposes.
There are other problems with this study, too. They did not isolate their seven factors; note that in their final analysis the "protective effect" factors are only narrowed down to three, not one: Vitamin E, PUFA (shudder!) and folate. Any of those three MIGHT be the causative factor... or it might not. We don't know, because this is study is only good for descriptive and exploratory purposes, not prediction, due to its design.
It drives me crazy that the media thinks anything labeled a "study" is a) automatically trustworthy and b) automatically proves causation. Neither of these things is true. This was a descriptive, exploratory study only. Now, people need to do experimental studies to support or disprove the hypothesis that SFA are causative of AD. They haven't done that yet, so this media hype is frigging irresponsible in the extreme.
Finally, every experimental study that has been conducted to "prove" the lipid hypothesis has come up way short. Check out Gary Taubes' book on that.
I wish we could force everyone in the country to have an adequate education on and understanding of the scientific method. This kind of thing just drives me crazy.
My point is that 'real' science is not even feasible for this kind of study. The best we can ever hope to have is a correlation between what people consume over time and their subsequent change (+/-) in health status over the same period.
Which means that we have to do observational studies like the Eades do - which is halfway between epidemiology and experimentation. Their observations of all their clients have shown remarkable drops in weight, blood pressure, blood sugar and insulin levels while at the same time increasing strength, health, and wellness. They have no control group (which is why it's not experimental) but they have introduced a change/stimulus/intervention: reduction or elimination of dietary carbohydrates. And guess what? It works! Which means that we can have something that is far stronger than a correlation coefficient - we can have results that show an overwhelming response to a stimulus. That's almost as good as a controlled study with a control group and an experimental/treatment group, and far and away better than any epidemiological study. It's far more dependable.
As you say, ethically we can't do the kind of dietary-restrictive experiment I've suggested (although drug companies do it all the time with placebos), but epidemiological studies still should not be trusted as causation-explanatory. They aren't. And freaking out over the results of an epidemiological study is a waste of time and energy. If it's epidemiological, the most we can say about it is "That's interesting. When are they doing the follow-up experimental study?" And that's all we really should be saying until the experimental (or failing that, observatory) study is done.
Griff, I understand that epidemiological studies are only meant to 'tease' out possible causative factors which, in turn, must be subjected to further study. But when we see a list of such factors, as in this study, that are directly contradictory to the map that we have laid out for ourselves, should this not give us pause? Pause for what I'm not sure. I guess we can only go with the science that we have at hand and hope that it pans out in the long run.
My answer is no, because a) those factors are confounded and conflated in any epidemiological study, and b) we have observational proof in this specific case that reduction of certain types of foods and an increase in others creates a much healthier system overall. May I ask if you've read the Eades' books, and Mary Enig and Susan Fallon's "Eat Fat, Lose Fat," and other sources like Gary Taubes? Or are you just depending on Mark's book by itself? I think if you see the plethora of sources out there that support the way we eat and live, you'll feel better about this course of action and more able to dismiss fear-mongering epi studies like the one you've linked to above.
Thanks for the references, Griff. I'll look to build more of a foundation under this whole PB venture. And thanks to all for the discussion! This was my first spin on the forums and it's been good.
No problem, Dave - hope that it helps!
Even if for the sake of argument it was conceded that the study was conducted in such a way that the data were valid, assigning a correlate to each specific food source is metabolic nonsense. The reductionist model of nutrition doesn't take into account the real underlying physiological processes and fails to account for how far a particular food might be from its natural origin and what those processes have done to the food in question. Soliciting participants to self-report on (perceived) food consumption and then rates of some condition will yield nonsense no matter what level of mathematical sophistication is applied to the data. These are lazy studies that are just too many levels of indirection away from actual knowledge to be of any use.
Originally Posted by Dave
Tags for this Thread