The only thing worse than no data is shit data - Ad Libitum:
Quite a discussion over on Woo's place. I want to rant a bit but it's too long and too offensive/off-topic to fit in a comment so I'll just post it here.
I have a few points to make:
Colpo is not a scientist. I can't believe his literature review of metabolic ward studies is being cited as evidence of anything. Here's the thing, Colpo can write whatever he wants but he does not have the years of training which would allow him to understand the problems with what he's reading. No, this is not some dumb argument from authority. It's a simple fact that people who don't have the requisite knowledge to discuss certain matters should not be discussing those matters because they are matters of fact not opinion of some pleb on the street. Oh but Taubes is not a scientist either, I hear you say. Well, Taubes does actually hold undergraduate and postgraduate degrees in science from the world's most prestigious universities and he has an IQ that's a few standard deviations to the right of the curve so that equips him rather well to talk about these matters. Unlike dumbass meatheads from Australia. At least Guyenet is a trained scientist who understands the basics of statistical inference so when he misrepresents studies, it is most likely because he didn't read them carefully enough (as Hyperlipid exposed recently to much hilarity) or because he chooses to misrepresent them for nefarious reasons; in other words, he's lying on purpose to further an agenda which keeps him in a job and keeps his research funded. And because Taubes hurt his delicate feelings. But Colpo, Carbsane and their devoted readership I suspect just don't understand what they're talking about because they are not scientists.
I like to keep this blog anonymous and I am not interested in the slightest in narcissistic posturing or talking about my background because the message is more important that the messenger and ideas should stand on their own merit. But suffice it to say that I am an actual scientist unlike those who teach remedial math and call themselves scientists. I get up every day to go to work which is research and I have undergraduate and postgraduate qualifications in my own area as well as statistics.
But none of this shit matters. What matters is that people who don't know anything read blogs of people who don't know anything and then think they know something.
Example 1: Carbsane displays ignorance of elementary school statistics in her "critique" of the recent JAMA study showing a metabolic advantage for VLC. Dude in comments points it out and is ignored by everyone including our intrepid scientific genius. Blogosphere says the JAMA study is bogus 'cos Carbsane "debunked" it so fuck Taubes and fuck the CIH.
Example 2: Colpo went and read a bunch of shitty old metabolic ward studies which failed to find a metabolic advantage for LC diets and now the blogosphere says there is a mountain of studies that refutes the CIH so fuck NuSI, case closed.
Oh well. Here's the thing. NuSI bursts on the scene with a tremendous piece of work. I would encourage everyone to look at this table. It's AWESOME.
Naive interpretation (Colpo): Read the researchers' discussion which states that there were no statistically significant differences between LC and control diets. Repeat in bold on your blog that Joe Bloggs et al. found no statistically significant differences between LC and the other diet. Repeat procedure for a lot of other metabolic ward studies. Conclusion: fuck you Eades and STFU forever.
Funnily enough, this interpretation is faithful to the studies. It is not a lie to say that metabolic ward studies overall do not support the idea of LC metabolic advantage. But WHY is this so?
Correct interpretation (Attia/Taubes): We don't know. We don't know what these studies are actually showing us because these studies that have been conducted to date are statistically underpowered to detect a difference between the diets or methodologically flawed in other ways and hence all we can say is that they failed to reject the null hypothesis (the null hypothesis being no difference between the diets). We cannot at all say we failed to reject the null because there really are no differences between the diets in reality. We cannot say this for the simple reason: the studies are crap and they have tiny sample sizes. They cannot answer the question being asked.
The status quo merchants, tenured old cunts like George Bray, have it really easy. If you believe in the null hypothesis, i.e., no diet is better than any other diet, "a calorie is a calorie", then essentially all you need to do is not try very hard to reject the null hypothesis which is THE EASIEST THING IN SCIENCE TO DO. In medicine especially where effect sizes for most things are so small, it is actually incredibly difficult to detect a signal amidst all the noise. It's INCREDIBLY EASY to design an experiment that fails to reject the null hypothesis if your job and grant money depend on this outcome. Common shenanigans include but are not limited to:
1) Include only a handful of subjects in each arm of your experiment, making your study grossly underpowered to detect a statistically significant difference between the groups on the studied outcome. Even a cursory glance at Attia's literature review tells us that the majority of these metabolic ward studies have single digit sample sizes, with only a handful that have even as few as 20 subjects per arm. I have never in my life seen laughable, offensive, bullshit studies like this cited in any area of science as evidence of anything except in nutrition research. It is not possible to detect a modest effect, which most real-world effects in medicine are, with those sample sizes. Your effect would need to be HUMUNGOUS in order to detect a difference between the groups with like 9 subjects per arm. What I mean by humungous is that the people in the LC arm would literally need to be dropping dead from emaciation while the subjects in the high carb arm would need to be putting on weight to make it possible for your underpowered study to detect a difference between the diets.
2) Design both the LC diet and the control diet so that they are both starvation diets (say 800 calories or some such horror) and then pretend that one is LC and the other HC. Newsflash: if both groups are eating starvation diets, then neither diet can be high carb or indeed "high" anything. It is likely that the high carb diet is actually a low carb diet due to the very low number of total calories consumed. What's 60% of nothing? If no difference is shown between the diets, well, no wonder, they're essentially the same diet, the macronutrient ratio means jack shit when you're fed 600 calories a day.
3) Design your study in a way that an additional variable--which is not relevant to the question at hand but hugely problematic due to confounding--is being manipulated (protein) along with % carbs and % fat and just pretend you're testing the effect of carb content of a diet on weight loss and hope that nobody notices that the other variables weren't held constant. This happens all the time where in the LC group they often feed them high protein. No doubt because fat will kill you so we'd rather carry out a scientifically worthless study than feed people fat, or perhaps even, GASP, S-A-T-U-R-A-T-E-D FAT.
4) Call your diet "low carb" even though it's actually high carb and hope no one notices. I've seen 40% carb diets called LC in the literature. I guess they weren't eating 60% carb like the other group so that makes it low carb. LMAOFOREVER
So then these dolts read Colpo and Carbsane and go around saying that there's a mountain of scientific evidence that refutes the CIH. What evidence? Please look at Attia's literature review. Taubes is right, the state of nutrition research is SHOCKING and does not deserve to be called science. If by mountain of scientific evidence you mean sheer number of publications, yeah of course there are lots. But what these dolts don't understand is that if a study is fundamentally flawed or fraudulent, then it doesn't "count" as evidence against anything JUST 'CUZ it's published.
This is why one tightly-controlled methodologically-rigorous adequately-powered metabolic ward study from NuSI will mean more than all those "decades of nutrition research" (dross) put together. I saw this idea being ridiculed recently on WHS (or was it the Asylum?). But it's TRUE. One randomised controlled trial with 100 subjects per arm is more valuable than 10 randomised controlled trials with 10 subjects per arm because the first study is in a position to actually answer the question being asked whereas the other 10 are NOT. This is a FACT. Not opinion. FACT.
Addendum:
For the record: I don't believe there is a clinically meaningful metabolic advantage for LC diets, at least not for non-ketogenic LC diets, over balanced diets. I think Taubes is wrong on this and it is fortunate that he will reach retirement age before any NuSI experiments have been completed, analysed and reported so he won't have to go looking for a shoe store that would have him as a salesman when his "all you can eat eucaloric LC diet" theory of weight loss is disproven. BUT THAT'S NOT THE POINT. What I'm arguing against is people talking about TEH SCIENZ allegedly disproving the CIH when it does not.
Ultimately, I think LC diets work (at least initially after which they seem to stop working for a lot of us) because they lower insulin levels allowing your body to access its own fat stores to a greater extent than starch-based semi-starvation diets, thus increasing satiety and emotional well-being and allowing the person to reduce caloric intake spontaneously without feeling the effects of semi-starvation neurosis aka hacking one's fingers off Ancel Keys starvation experiment style. My suspicion is that initially a lot of ingested fat calories are simply wasted as heat or flushed down the toilet since the body coming to LC from the SAD ran on carbs for so long that it didn't bother producing too many fat burning enzymes. This is why at the beginning of a LC diet you can eat a half-arsed LC and still lose a tonne of weight and why your urine will still be overflowing with ketones even after cheating with crackers and cookies the odd time. Once you're truly keto-adapted and the body has realised you're trying to starve it of calories so it builds the metabolic machinery for efficient fat burning (in addition to just being able to extract calories better from ingested food and dropping the metabolic rate), the fun stops and you stop losing weight. This makes sense from an evolutionary perspective. A ketogenic diet mimics starvation metabolism. The whole point of starvation metabolism is to allow you to stay alive as long as possible hence why every effort is made by the body to "stall" weight loss. This is why the Innuit living in ketosis don't weight 0 lbs and it is also why people who are maintaining on eucaloric ketogenic diets don't continue to lose weight ad infinitum.
Taubes pretty much said this in WWGF, albeit not in those words, it's just that people chose to ignore it because they want to believe that they can achieve their unrealistically low goal weight.
I have a few points to make:
Colpo is not a scientist. I can't believe his literature review of metabolic ward studies is being cited as evidence of anything. Here's the thing, Colpo can write whatever he wants but he does not have the years of training which would allow him to understand the problems with what he's reading. No, this is not some dumb argument from authority. It's a simple fact that people who don't have the requisite knowledge to discuss certain matters should not be discussing those matters because they are matters of fact not opinion of some pleb on the street. Oh but Taubes is not a scientist either, I hear you say. Well, Taubes does actually hold undergraduate and postgraduate degrees in science from the world's most prestigious universities and he has an IQ that's a few standard deviations to the right of the curve so that equips him rather well to talk about these matters. Unlike dumbass meatheads from Australia. At least Guyenet is a trained scientist who understands the basics of statistical inference so when he misrepresents studies, it is most likely because he didn't read them carefully enough (as Hyperlipid exposed recently to much hilarity) or because he chooses to misrepresent them for nefarious reasons; in other words, he's lying on purpose to further an agenda which keeps him in a job and keeps his research funded. And because Taubes hurt his delicate feelings. But Colpo, Carbsane and their devoted readership I suspect just don't understand what they're talking about because they are not scientists.
I like to keep this blog anonymous and I am not interested in the slightest in narcissistic posturing or talking about my background because the message is more important that the messenger and ideas should stand on their own merit. But suffice it to say that I am an actual scientist unlike those who teach remedial math and call themselves scientists. I get up every day to go to work which is research and I have undergraduate and postgraduate qualifications in my own area as well as statistics.
But none of this shit matters. What matters is that people who don't know anything read blogs of people who don't know anything and then think they know something.
Example 1: Carbsane displays ignorance of elementary school statistics in her "critique" of the recent JAMA study showing a metabolic advantage for VLC. Dude in comments points it out and is ignored by everyone including our intrepid scientific genius. Blogosphere says the JAMA study is bogus 'cos Carbsane "debunked" it so fuck Taubes and fuck the CIH.
Example 2: Colpo went and read a bunch of shitty old metabolic ward studies which failed to find a metabolic advantage for LC diets and now the blogosphere says there is a mountain of studies that refutes the CIH so fuck NuSI, case closed.
Oh well. Here's the thing. NuSI bursts on the scene with a tremendous piece of work. I would encourage everyone to look at this table. It's AWESOME.
Naive interpretation (Colpo): Read the researchers' discussion which states that there were no statistically significant differences between LC and control diets. Repeat in bold on your blog that Joe Bloggs et al. found no statistically significant differences between LC and the other diet. Repeat procedure for a lot of other metabolic ward studies. Conclusion: fuck you Eades and STFU forever.
Funnily enough, this interpretation is faithful to the studies. It is not a lie to say that metabolic ward studies overall do not support the idea of LC metabolic advantage. But WHY is this so?
Correct interpretation (Attia/Taubes): We don't know. We don't know what these studies are actually showing us because these studies that have been conducted to date are statistically underpowered to detect a difference between the diets or methodologically flawed in other ways and hence all we can say is that they failed to reject the null hypothesis (the null hypothesis being no difference between the diets). We cannot at all say we failed to reject the null because there really are no differences between the diets in reality. We cannot say this for the simple reason: the studies are crap and they have tiny sample sizes. They cannot answer the question being asked.
The status quo merchants, tenured old cunts like George Bray, have it really easy. If you believe in the null hypothesis, i.e., no diet is better than any other diet, "a calorie is a calorie", then essentially all you need to do is not try very hard to reject the null hypothesis which is THE EASIEST THING IN SCIENCE TO DO. In medicine especially where effect sizes for most things are so small, it is actually incredibly difficult to detect a signal amidst all the noise. It's INCREDIBLY EASY to design an experiment that fails to reject the null hypothesis if your job and grant money depend on this outcome. Common shenanigans include but are not limited to:
1) Include only a handful of subjects in each arm of your experiment, making your study grossly underpowered to detect a statistically significant difference between the groups on the studied outcome. Even a cursory glance at Attia's literature review tells us that the majority of these metabolic ward studies have single digit sample sizes, with only a handful that have even as few as 20 subjects per arm. I have never in my life seen laughable, offensive, bullshit studies like this cited in any area of science as evidence of anything except in nutrition research. It is not possible to detect a modest effect, which most real-world effects in medicine are, with those sample sizes. Your effect would need to be HUMUNGOUS in order to detect a difference between the groups with like 9 subjects per arm. What I mean by humungous is that the people in the LC arm would literally need to be dropping dead from emaciation while the subjects in the high carb arm would need to be putting on weight to make it possible for your underpowered study to detect a difference between the diets.
2) Design both the LC diet and the control diet so that they are both starvation diets (say 800 calories or some such horror) and then pretend that one is LC and the other HC. Newsflash: if both groups are eating starvation diets, then neither diet can be high carb or indeed "high" anything. It is likely that the high carb diet is actually a low carb diet due to the very low number of total calories consumed. What's 60% of nothing? If no difference is shown between the diets, well, no wonder, they're essentially the same diet, the macronutrient ratio means jack shit when you're fed 600 calories a day.
3) Design your study in a way that an additional variable--which is not relevant to the question at hand but hugely problematic due to confounding--is being manipulated (protein) along with % carbs and % fat and just pretend you're testing the effect of carb content of a diet on weight loss and hope that nobody notices that the other variables weren't held constant. This happens all the time where in the LC group they often feed them high protein. No doubt because fat will kill you so we'd rather carry out a scientifically worthless study than feed people fat, or perhaps even, GASP, S-A-T-U-R-A-T-E-D FAT.
4) Call your diet "low carb" even though it's actually high carb and hope no one notices. I've seen 40% carb diets called LC in the literature. I guess they weren't eating 60% carb like the other group so that makes it low carb. LMAOFOREVER
So then these dolts read Colpo and Carbsane and go around saying that there's a mountain of scientific evidence that refutes the CIH. What evidence? Please look at Attia's literature review. Taubes is right, the state of nutrition research is SHOCKING and does not deserve to be called science. If by mountain of scientific evidence you mean sheer number of publications, yeah of course there are lots. But what these dolts don't understand is that if a study is fundamentally flawed or fraudulent, then it doesn't "count" as evidence against anything JUST 'CUZ it's published.
This is why one tightly-controlled methodologically-rigorous adequately-powered metabolic ward study from NuSI will mean more than all those "decades of nutrition research" (dross) put together. I saw this idea being ridiculed recently on WHS (or was it the Asylum?). But it's TRUE. One randomised controlled trial with 100 subjects per arm is more valuable than 10 randomised controlled trials with 10 subjects per arm because the first study is in a position to actually answer the question being asked whereas the other 10 are NOT. This is a FACT. Not opinion. FACT.
Addendum:
For the record: I don't believe there is a clinically meaningful metabolic advantage for LC diets, at least not for non-ketogenic LC diets, over balanced diets. I think Taubes is wrong on this and it is fortunate that he will reach retirement age before any NuSI experiments have been completed, analysed and reported so he won't have to go looking for a shoe store that would have him as a salesman when his "all you can eat eucaloric LC diet" theory of weight loss is disproven. BUT THAT'S NOT THE POINT. What I'm arguing against is people talking about TEH SCIENZ allegedly disproving the CIH when it does not.
Ultimately, I think LC diets work (at least initially after which they seem to stop working for a lot of us) because they lower insulin levels allowing your body to access its own fat stores to a greater extent than starch-based semi-starvation diets, thus increasing satiety and emotional well-being and allowing the person to reduce caloric intake spontaneously without feeling the effects of semi-starvation neurosis aka hacking one's fingers off Ancel Keys starvation experiment style. My suspicion is that initially a lot of ingested fat calories are simply wasted as heat or flushed down the toilet since the body coming to LC from the SAD ran on carbs for so long that it didn't bother producing too many fat burning enzymes. This is why at the beginning of a LC diet you can eat a half-arsed LC and still lose a tonne of weight and why your urine will still be overflowing with ketones even after cheating with crackers and cookies the odd time. Once you're truly keto-adapted and the body has realised you're trying to starve it of calories so it builds the metabolic machinery for efficient fat burning (in addition to just being able to extract calories better from ingested food and dropping the metabolic rate), the fun stops and you stop losing weight. This makes sense from an evolutionary perspective. A ketogenic diet mimics starvation metabolism. The whole point of starvation metabolism is to allow you to stay alive as long as possible hence why every effort is made by the body to "stall" weight loss. This is why the Innuit living in ketosis don't weight 0 lbs and it is also why people who are maintaining on eucaloric ketogenic diets don't continue to lose weight ad infinitum.
Taubes pretty much said this in WWGF, albeit not in those words, it's just that people chose to ignore it because they want to believe that they can achieve their unrealistically low goal weight.
30 comments:
tess17 September 2012 23:51thank you! :-D just ... thank you!Reply
What? Are you saying you actually slogged through that???
- This shit actually took me three hours to write and now I'm so agitated that I can't concentrate on Hyperlipid's new post. What a waste! LOLReply
- i always have to skim Peter's posts quick before i go back and try to understand them properly, anyway! :-) ...and of course i slogged through! i love it when you, Wooo, EB and Kindke put form to the nebulous discomfort i feel about these bad arguments. i KNOW they're faulty, but i'm not always sure why.Reply
- It's a rather delightful smackdown of Guyenet's high fat diet will rot your brain bogus study.
Yep. Here's the lite version of Petro brilliance:
-Obesity research squad says "palatability and overfeeding" causes VMH damage/fat gain in rodents.
-Scientific evidence PLAINLY SHOWS the VMH damage happens before the over eating or the fat gain; the food itself damages the brain in this specific rodent model which is of no relation to palatability. Over eating is the result, not the cause of the damage.- Stay away from PUFAs, kids. They may cause your brain to urge you to "overeat" what suddenly seems like really "rewarding" food.
- hell, i had to look up "locum" and "rota" in his introductory paragraph.... ;-)
- S, Ironically, the high PUFA (linolenic) diet is the one that protects the rats from obesity because it fails to induce high levels of superoxide in the VMH which then promote death of leptin sensitive neurons, which causes the lack of sympathetic tone to adipocytes and pathological obesity.
The rodents get fat specifically on a high superoxide generating fat like palmitic acid, because of their weirdo rat strain genetics and fragile VMH which appears easily injured by the superoxide normally generated (but this fat is what our body fat is made of, and is entirely functional/healthy for free living human beings.)
From hyperlipid:
"C57BL/6 mice (and Long Evans rats) are specifically bred to get fat on palmitic acid (sometimes plus fructose) based diets. They fail to deal with the absolutely normal levels of superoxide produced in POMC neurons in the VMH which are crucial to energy status sensing. They do not have the luxury of developing insulin resistance as their job is to monitor both glucose and fatty acid levels. They are not allowed to run on lactate with an F:N ratio of 0.2 the way much of the brain does. They take whatever plasma gives them and do their best to cope with it. Or, in the case of rodents bred to become fat on high fat diets, not cope with it."
I haven't fact checked I admit but if what hyperlipid writes is correct, that a high fat PUFA linoleic o6 diet fails to trigger the hyperphagia/obesity in the "high fat diet prone" rodents then this is DAMNING evidence against both the food reward hypothesis as well as the utility of this rodent model in understanding *any form of obesity* except maybe the rare forms of hypothalamic obesity.
The idiot obesity research squad holds up these HFD obese rodents as an example of reward causing overconsumption validating FR hypothesis when in reality they are merely a rodent model for understanding hypothalamic obesity, which is a form of obesity hardly even *relevant* to most people. - I'm watching Hyperlipid's latest series with fascination because either he is wrong or the whole paradigm falls down, there is sort of no middle way. Researchers have been extrapolating from rat studies for so long but I just don't see how these genetically modified rats are relevant to common human obesity at all. Hypothalamic obesity is very rare in humans, no? If saturated fat was so toxic to our brains and bodies, would the body really use this dangerous substance to store energy? I think it's absurd. The amount of PUFA you need to eat is actually minuscule and there is plenty of evidence that high levels of o6 fucks you up (cellular membranes, scary stuff).
- I think the takehome message from all this is fuck the gurus and their supposed supporting scientific studies. Self-experiment until you find what works for you.Reply
- Aye, N=1 is more important than any study youll find on pubmed. Part of the reason for starting my own blog was getting tired of all these so called guru's and I decided to troll pubmed and do my own research.
- [nodding] no one person (or two, or three) ever told me all i need to know, but thanks to six or eight or so, i'm zeroing in on a pretty good action plan for myself! :-)
