This seems like an authoritative 25-year research project that the Atkins diet is pretty bad:
http://www.nutritionj.com/content/11/1/40/abstract
Right now my belief is that the Atkins diet is good. It's backed by anecdotal evidence of trying a low-carb diet for 18 months following a 12-month low-fat diet and seemingly getting better results with the low-carb diet.
I'm counting on LWers to tell me how to update my belief in light of this study. Thanks.
If you had not started on Atkins prior to reading this research, then you ought to have revised the odds of it working downwards and probably picked some other diet instead. However, since you had personally tried Atkins and it was working for you, then this first-hand evidence overwhelms any statistical analysis from Northern Sweden. It does not invalidate it in any way for others, but it is pretty much useless for you personally. Similarly, anyone looking to start a diet should discard your anecdotal evidence and favor the research data. So, no update for you, downward revision for others.
I think the reasoning behind this is worth making explicit. even if the expected value of the Atkins diet is negative to a randomly selected member of the population, it may still be reliably positive for some very specific small subset. if you have a strong reason to believe that you're a member of that subset, then information about the average is less relevant to you than it might otherwise be. for example, a reduction in calorie intake, while beneficial for the average member of the population, is not beneficial to people who are severely underweight.
This seems like very little evidence as far as I am concerned.
It claims that eating a higher-fat diet increased cholesterol. This is what I would expect, and I am also entirely unconvinced that this is remotely harmful. They don't even break up "cholesterol" into the relevant subtypes! Was this an increase in HDL or triglycerides? They rely on a section of the paper to claim that the link between higher cholesterol and cardiovascular mortality is well-established... but then why didn't they make a study showing increased CVD or all-cause mortality? They have those data as part of the study! I want the body count.
Also, if you look at the correlations they found with increased cholesterol levels, the ones they don't report in the abstract include "sweet buns and crisp bread rolls, and boiled potato" (as well as boiled coffee and salted fish). So it looks like some kinds of fats and some kinds of carbs correlate with higher cholesterol. That doesn't seem nearly as compelling as the headline. (Let's also note that fat consumption as a % of energy only once again reached 1986 levels in 2010, and yet total cholesterol is still significantly lower.)
The continually-rising BMI is more interesting to me. They lowered fat intake, people got fatter. They lowered carb intake, people got fatter. Hmmm... Oddly enough, they don't report much about total caloric intake - everything is mentioned as a proportion of calories. The shift in fat intake was a fall of 3-4% of calories, then an increase of 3-4% of calories. This would only require a small amount of increased total calories from fat, with no reduction in carb content, to explain the shift as well as the increasing BMI. (Note that they didn't try to draw any correlations with BMI, because of the well-known bias in food frequency questionnaire reporting.)
What other major food shifts did they note in the study? First of all, potatoes were being replaced with rice and pasta. Second of all, alcohol intake rose continuously over the period in question. I would bet hundreds of dollars that the strongest statistical correlation with BMI would be wine intake, based on the figures they report.
Even as far as associational studies go, this is a really bad one. I mean that seriously, this is methodologically one of the worst I've ever seen. I was expecting to actually have my beliefs challenged, a few good associational studies have given me pause, but this is not one of them.
Thanks! I'll go ahead and not update.
Unfortunately, I feel a little smug about not updating, and Eliezer has pointed out that, to avoid confirmation bias, you should only congratulate yourself when you do update your beliefs. So I'll try to have LW pick apart the next pro-Atkins study I see too.
This is also a huge epidemiological study that showed lower cholesterol INCREASED risk for death from cardiovascular disease in women. In men, cholesterol and death followed a U-shaped curve, so men with the highest and lowest cholesterol were at increased risk for death. Of course, with the caveat that this is also not a randomized trial and thus not quality evidence, it does call into question whether the increase in cholesterol seen in the 1st study was a good thing or a bad thing.
We have higher quality data from randomized trials that show Statins, which lower cholesterol, don't reduce mortality when used for primary prevention.
I think you would be able to think more clearly about these topics if you made some additional distinctions among your hypothesis under consideration. Instead of "the Atkins diet is good", try "the Atkins diet allows me to maintain a lower body fat percentage than the SAD" and "the Atkins diet allows the typical person to maintain a lower body fat percentage than the SAD". Based on that study, you should decrease your degree of belief in both of these hypothesis, but to a much greater extent for the latter hypothesis relative to the former.
I've heard so many conflicting things that all I can conclude is that there probably isn't any strong evidence in any particular direction, so I don't worry about it. If the effects were that obvious, there wouldn't be a controversy.
I read Taubes' book on the value of low-carb diets (well worth reading; I learned a lot)-- the evidence seems compelling (but then again, I'm not a nutrician), and I haven't really seen similarly compelling counter-evidence. Atkins/Taubes do seem to be the contrarians here.
The results of the study in Sweden do not seem conclusive to me:
After 2004 fat intake increased sharply for both genders, which coincided with introduction of a positive media support for low carbohydrate-high-fat (LCHF) diet.
From this preliminary result, it's hard to conclude that low carbs are the problem here. The first step would be to show that, well, the subjects actually followed a low-carb diet, not just a high-fat diet. Hopefully the study, once completed, clarifies this.
Data from epidemiological studies is so weak as to be almost worthless.
According to Dr. John Ioannidis the conclusions reached from observational studies are wrong 80% of the time. So unless there is a randomized trial that also supports the hypothesis it is more likely than not that the conclusions of the authors of this study are wrong.
In fact, you shouldn't conclude anything from epidemiological studies. They are hypothesis generating studies that should give you an idea of what kinds of interventions you should investigate in a real scientific study.
Here are some randomized trials:
Compared with a low-fat diet, a low-carbohydrate diet program had better participant retention and greater weight loss. During active weight loss, serum triglyceride levels decreased more and high-density lipoprotein cholesterol level increased more with the low-carbohydrate diet than with the low-fat diet.
In addition, the low-carbohydrate group lost less lean body mass:
Patients in both groups lost substantially more fat mass (change, -9.4 kg with the low-carbohydrate diet vs. -4.8 kg with the low-fat diet) than fat-free mass (change, -3.3 kg vs. -2.4 kg, respectively).
Mediterranean and low-carbohydrate diets may be effective alternatives to low-fat diets. The more favorable effects on lipids (with the low-carbohydrate diet) and on glycemic control (with the Mediterranean diet) suggest that personal preferences and metabolic considerations might inform individualized tailoring of dietary interventions.
The patients on the low-carb diet lost more weight & better improved HDL:
The mean weight loss was 2.9 kg for the low-fat group, 4.4 kg for the Mediterranean-diet group, and 4.7 kg for the low-carbohydrate group (P<0.001 for the interaction between diet group and time); among the 272 participants who completed the intervention, the mean weight losses were 3.3 kg, 4.6 kg, and 5.5 kg, respectively. The relative reduction in the ratio of total cholesterol to high-density lipoprotein cholesterol was 20% in the low-carbohydrate group and 12% in the low-fat group (P=0.01).
CONCLUSION: Successful weight loss can be achieved with either a low-fat or low-carbohydrate diet when coupled with behavioral treatment. A low-carbohydrate diet is associated with favorable changes in cardiovascular disease risk factors at 2 years.
They also saw greater improvement in HDL ("good") cholesterol in the low-carb group:
Increases in HDL cholesterol levels were significantly greater in the low-carbohydrate than in the low-fat group at 3, 6, 12 and 24 months.
They also mention another reason that LDL increases in the low-carb group may not increase cardiovascular risk:
[...]assessment of LDL cholesterol concentration without information on LDL particle size has limitations as an indicator of coronary heart disease risk because small, dense LDL particles are more atherogenic than large LDL particles (24). Data from carefully controlled studies demonstrated that isocaloric replacement of dietary carbohydrate with fat increases plasma LDL cholesterol concentration but shifts LDL particle size from smaller to larger and less atherogenic LDL (25).
CONCLUSIONS: The low-carbohydrate diet produced a greater weight loss (absolute difference, approximately 4 percent) than did the conventional diet for the first six months, but the differences were not significant at one year. The low-carbohydrate diet was associated with a greater improvement in some risk factors for coronary heart disease.
Again, the low-carb diet better increased HDL: +18.2 for the low-carb group, +3.1 for the conventional group (P=0.04).
It also better decreased Triglycerides: -28.1 for the low-carb group, +1.4 for the conventional group (P=0.04).
Experimental evidence is very clear that low-carb diets allow large weight loss in 6 months. But by 12 months, other diets are catching up. Thus it is not surprising if the long-term effect difference nothing, which is what this study suggests. But that's hardly "pretty bad."
I was going to reply to you about the feasibility of weight loss in general (you haven't said you're interested in weight loss, but that's what people usually do Atkins for), but my comment really wasn't answering your question, so I posted in the open thread instead. Here it is if you're interested.
Some (implicit) counter-evidence to the study was provided in response to a question I asked on Paleo Hacks.
Is having a more accurate belief more important to you or having an effective diet? Also, what other diets have you tried?
I don't really feel qualified to tell you how you should update quantitatively, especially after vague descriptions such as "the Atkins diet is good", and "seemingly getting better results". But obviously, that study should not be something that strengthens your belief in the Adkins diet.
Is having a more accurate belief more important to you or having an effective diet? Also, what other diets have you tried?
I don't really feel qualified to tell you how you should update quantitatively, especially after vague descriptions such as "the Atkins diet is good", and "seemingly getting better results". But obviously, that study should not be something that strengthens your belief in the Adkins diet.