Previously: Mainstream Nutrition Science on Obesity, Atkins Redux, Did the US Government Give Us Absurd Advice About Sugar?, What Causes Obesity?

If you've been wondering what these posts are doing on LessWrong and you haven't read this comment yet, I urge you to do so. Thanks to commenter FiftyTwo for suggesting I say something like this.

To recap: so taking in more calories than you burn will cause you to gain weight, though calorie intake and expenditure are in turn controlled by a number of mechanisms. This suggests a couple of options for losing weight. You can try to intervene directly in the mechanisms controlling food intake, one of the most well-known examples of this being gastric bypass surgery, admittedly a bit of a drastic option. But intervening at the point of calorie intake is also an option.

Now it turns out that it's relatively easy to lose weight by dieting. That catch is that it's much harder to keep the weight off. A commonly cited rule (for example here) is that most people who lose weight through dieting will regain it all in five years. However, it's important to emphasize that some people do lose weight through dieting and keep it off long-term. An organization called the National Weight Control Registry has made an effort to track those people, and have published quite a few studies based on their work (many of which can be easily found through Google Scholar).

Unfortunately, the NWCR is working with a self-selected sample and asking them what they did after the fact. They're not randomly assigning people to treatments. So for example, a high percentage of the NWCR group reports successful long-term weight loss following low-fat and/or calorie-restricted diets and exercising a lot. And the percentage following low-carb diets was originally small, but it's risen over time. But both of these observations may just reflect the relative popularity of those approaches in the general population.

We may not be able to conclude anything more from the NWCR data than that a significant minority of dieters do succeed at long-term weight loss, some through calorie-restricted diets, some through low-fat diets, and some through low-carb diets. Remember, though, that as discussed in previous posts there's little reason to think low-fat or low-carb diets could cause weight loss except by indirectly affecting energy balance.

And now, one last time, I'm going to talk about what Taubes has to say about this issue. I'm going to quote from Why We Get Fat (pp. 36-38), though Good Calories, Bad Calories contains similar comments, including about the Handbook of Obesity and Joslin's. Taubes begins by citing a review article covering calorie-restricted diets that found that "Typically, nine or ten pounds are lost in the first six months. After a year, much of what was lost has been regained." He also cites a large study that tested a calorie-restricted diet and reached a similar conclusion: participants "lost on average, only nine pounds. And once again... most of the nine pounds came off in the first six months, and most of the participants were gaining weight back after a year."

Based on this, he concludes that "Eating less—that is, undereating—simply doesn't work for more than a few months, if that." Then it's time to really lay in to mainstream nutrition science:

This reality, however, hasn't stopped the authorities from recommending the approach, which makes reading such recommendations an exercise in what psychologists call "cognitive dissonance," the tension that results from trying to hold two incompatible beliefs simultaneously.

Take, for instance, the Handbook of Obesity, a 1998 textbook edited by three of the most prominent authorities in the field—George Bray, Claude Bouchard, and W. P. T. James. "Dietary therapy remains the cornerstone of treatment and the reduction of energy intake continues to be the basis of successful weight reduction programs," the book says. But then it states, a few paragraphs later, that the results of such energy-reduced diets "are known to be poor and not long-lasting." So why is such an ineffective therapy the cornerstone of treatment? The Handbook of Obesity neglects to say.

The latest edition (2005) of Joslin's Diabetes Mellitus, a highly respected textbook for physicians and researchers, is a more recent example of this cognitive dissonance. The chapter on obesity was written by Jeffrey Flier, an obesity researcher who is now dean of Harvard Medical School, and his wife and research colleague, Terry Maratos-Flier. The Fliers also describe "reduction of caloric intake" as "the cornerstone of any therapy for obesity." But then they enumerate all the ways in which this cornerstone fails. After examining approaches from the most subtle reductions in calories (eating, say, one hundred calories less each day with the hope of losing a pound every five weeks) to low-calorie diets of eight hundred to one thousand calories a day to very low-calorie diets (two hundred to six hundred calories) and even total starvation, they conclude that "none of these approaches has any proven merit."

But look at the actual sources and it turns out that, surprise surprise, mainstream experts aren't idiots after all. The second quote from the Handbook of Obesity comes from a paragraph explaining that given how hard obesity is to treat, doctors face a "Shakespearean" dilemma of whether to attempt to treat it at all. The Joslin's article is even clearer (p. 541, emphasis added):

Successful treatment of obesity, defined as treatment that results in sustained attainment of normal body weight and composition without producing unacceptable treatment induced morbidity, is rarely achievable in clinical practice. Many therapeutic approaches can bring about short-term weight loss, but long-term success is infrequent regardless of the approach.

Suppose for a moment that this is true, that long-term weight loss is rare regardless of the approach. If it is, no "cognitive dissonance" is required to recommend treatments that sometimes work. Furthermore, Taubes commits a serious misrepresentation here. Taubes final quote from the Joslin's article, in context, says that, "There are also many programs that recommend specific food combinations or unusual sequences for eating, but none of these approaches has any proven merit." It's pretty obvious in context that the bit Taubes quotes refers only to the programs that recommend specific food combinations or unusual sequences for eating."

It's also worth mentioning that neither of these sources ignore the debate over low-carb diets. The Handbook of Obesity criticizes Atkins-style low carb diets at some length, but also says that, "Moderate restriction of carbohydrates may have real calorie-reducing properties." And the Joslin's article ends up being fairly positive towards low-carb diets in general (p. 542):

Dietary composition may play a role in long-term success in weight loss and weight maintenance. For example, a study comparing a moderate-fat diet consisting of 35% energy from fat and a low-fat diet in which 20% of energy was derived from fat demonstrated enhanced weight loss assessed by total weight loss, BMI change, and decrease in waist circumference in the group on the moderate-fat diet. Retention in the diet study was greater among those actively participating in the weight loss program in this group compared with 20% in the low-fat diet group.

Recently, increased interest has focused on the possibility that diet content may affect appetite. For example, diets with a low glycemic index may be useful in preventing the development of obesity; subjects given test meals with different glycemic indexes and then allowed free access to food ate less after eating meals with a low glycemic index. Some data suggest that diets with a high glycemic index predispose to increased postprandial hunger, whereas diets focused on glycemic index and information regarding portion control lead to higher rates of success in weight loss, at least among adolescent populations. Low-carbohydrate diets such as the Atkins diet appear to be associated with significant weight loss. However, this diet has not been systematically studied, nor has long-term maintenance of weight loss.

I assume the author of the Joslin's article would say, however, that low-carb diets haven't been shown to completely solve the problem of long-term weight loss being really hard. But would they be right about that? 

To the best of my knowledge, there have been only two randomized, controlled trials of low-carb diets that have covered a period of two years (and none covering a longer period than that). Taubes has cited both in support of his claims. The first, an Israeli study published in 2008, also also included a group assigned to a Mediterranean diet. Here are the results in terms of weight loss: 

So on the one hand, subjects on the low-carb diet did initially lose more weight, about 6.5 kg (14 lbs.) compared to about 4.5 kg (10 lbs.) for the low-calorie diet. On the other hand, both groups started regaining the weight after six months. If, as Taubes claims, data like this shows that low-calorie diets "simply doesn't work for more than a few months," does this data justify saying the same thing about low-carb diets?

Furthermore, if you believe the rule about weight lost to dieting coming back in five years, it seems likely that would happen to both groups. Intriguingly, though, while participants on the Mediterranean diet didn't initially lose as much weight as those on the low-carb diet, the weight regain didn't seem to happen as much on the Mediterranean diet. That makes me wonder what a five-year study of the Mediterranean diet would find.

Note that the Israeli study also found that that participants in all three groups significantly reduced their caloric intake, supporting the hypothesis that even diets that don't explicitly restrict calorie intake work by reducing calorie intake indirectly.

What about the other study, published in 2010, which Taubes has hailed as "the biggest study so far on low-carb diets"? Here are its results (note that the low-fat diet was also a calorie-restricted diet):

That's right, this study found no statistically significant difference between low-fat and low-carb diets in terms of weight loss, and again show the typical pattern of people losing weight in the first six months and then slowly gaining it back. Together, these two studies support the picture painted by Joslin's: low-carb diets may work somewhat better for weight loss, but they don't appear to solve the problem of long-term weight loss being really hard.

One other relevant detail: the second study found that "A significantly greater percentage of participants who consumed the low-carbohydrate than the low-fat diet reported bad breath, hair loss, constipation, and dry mouth." As Taubes' fellow science writer John Horgan has noted, this reveals an apparent inconsistency in how Taubes judges different diets. He goes to great lengths to play up the unpleasantness of calorie-restricted diets, but tells his readers that if they just stick to their low-carb diet theunpleasant side-effects will go away eventually.

So given all this, what should you do if you want to lose weight? I think depends a lot on who you are. I have ethical qualms about consuming animal products, including and in fact especially eggs, which is one strike against low-carb diets for me. Also, while there's some evidence low-carb diets may be better for hunger, my personal experience is that what foods I find filling is kind of random (lentils, black beans, and baguettes all rate highly on the filling-ness measure for me). So maybe just experiment and try to figure out which foods let you personally eat in moderation and not feel hungry. Keep Eliezer's advice in Beware of Other Optimizing in mind, and if one thing doesn't work for you, try something else.

A final point: the truth about weight loss sucks. If your case isn't bad enough to justify something drastic like gastric bypass surgery, your main option is diets which sometimes work but usually don't. Regardless of the approach. Unfortunately, this is not an exciting message to put in a popular book on nutrition. This creates an excellent opportunity for someone like Taubes: imply that if the experts admit they don't have a great solution to the problem, then clearly they don't know what they're talking about, and therefore your solution is sure to work!

Long-time readers of LessWrong, however, will realize that the universe is allowed to throw us problems with no good solution. That's something that may be especially worth keeping in mind when evaluating claims in the vicinity of medicine and nutrition. In a way, Taubes' readers are lucky: following his advice won't kill you, and won't lead to you missing out on any wildly more effective solution. It might have some unpleasant side-effects you could've avoided with another approach, but also might have some advantages. However, I've read enough of the literature on medical quackery to know Taubes' rhetorical tactics can be used for much more dangerous ends.

Just imagine: "It's doctors and pharmaceutical companies that caused your cancer in the first place. That chemotherapy and radiation therapy stuff they're pushing on you is obviously harmful. Don't you now there are all-natural ways you can cure your cancer?" If someone says that to you, then knowing that the universe is unfair, and that sometimes the best solution it gives you to a problem will have serious downsides, well knowing that just might save your life. Or not. Because the universe isn't fair. 

Early on in the process of writing this series, I said when it was over with I'd do a post-mortem to look at how I could have broken it up better. However, Vaniver has given me what seems like good advice on that issue, which I plan to follow in the future. (Unless someone else comes along and persuades me otherwise. You're welcome to try that).

But there are other issues here, the big meta-issue being that downvotes don't help me distinguish between people thinking the posts were completely off-topic for Lesswrong vs. not liking how finely they were broken up vs. me not realizing what a hot-button issue obesity is for some people vs. other things. So suggestions on how I could best solicit anonymous feedback would be especially appreciated.