If you follow other paleo or low-carb blogs, you’re no doubt aware of the recent study in which the investigators concluded that while low-carb and low-fat diets work equally well for weight loss, the low-carb diet produced greater improvements in HDL. Jimmy Moore wrote about the study, and there were online articles with quotes like these:
Over the long term, a low-carb diet works just as well as a low-fat diet at taking off the pounds – and it might be better for your heart, new research suggests.
Both diets improved cholesterol in a two-year study that included intensive group counseling. But those on the low-carbohydrate diet got a bigger boost in their so-called good cholesterol, nearly twice as much as those on low-fat.
Given all the nonsense over the years along the lines of “Sure, you might lose weight on the Atkins diet, but all that fat will give you heart disease!” I was of course pleased to see this study make the news. But at the same time, something about either the study itself or the reporting of it didn’t feel quite right. We were told that the low-carb group lost more weight initially, but both diets performed equally well for weight loss after two years … and yet this bit of information didn’t make it into the Associated Press version of the story:
The 153 participants in the low-carb group followed guidelines set out in Dr Atkins’ New Diet Revolution. For the first 12 weeks, they were told to limit carbohydrate intake to 20 g per day in the form of low-glycemic-index vegetables. After this, they could gradually increase their intake of carbohydrates by 5 g per day per week by adding more vegetables, a limited amount of fruits, and later small quantities of whole grains and dairy products until a stable and desired weight was achieved.
The 154 participants assigned to the low-fat diet limited their energy intake to 1200 to 1500 kcal per day for women and 1500 to 1800 kcal per day for men, with approximately 55% of calories from carbohydrate, 30% from fat, and 15% from protein.
Are there alarm bells going off in your head? Notice any problems with the comparison here? We’re looking at one group that was told to restrict fat and calories for two years, and another group that was told to strictly limit carbohydrates for just three months, then gradually raise them until a stable and desired weight was achieved!
If you’re consuming 20 grams of carbohydrates per day at three months and then starting adding an extra 5 grams per day each week, at six months into the study you’d be consuming 80 grams of carbohydrates per day. At 12 months, you’d be up to 210 grams per day.
Ask anyone who’s lost weight and kept it off by restricting carbohydrates what happens if the carb count starts drifting up. The answer: first you stop losing, then start gaining again. Most of us find the magic number is somewhere below 100. So at some point in the first year, the low-carb dieters most likely hit their limit and stopped losing weight. From then on, it was merely a maintenance diet.
Meanwhile, I couldn’t find anything suggesting that the low-fat/low-calorie group was told to start eating more once they’d achieved a stable and desired weight. A limit of 1200-1500 calories for women and 1500-1800 calories for men sure sounds like an ongoing weight-loss diet to me, unless we’re talking about people who aren’t very big to begin with.
But as it turns out, the study subjects were big. After digging around and sending out distress signals, I managed to obtain a copy of the full study. According to the researchers, the subjects all had a BMI of over 30. Two-thirds of them were women, and yet the average starting weight was 226 pounds. After 12 months, the average weight in the low-fat/low-calorie group was still over 200 pounds. At that weight, a diet of 1500-1800 calories is a weight-loss diet.
So in theory, we’re comparing a low-calorie weight-loss diet that lasted for two years with a low-carb diet that reached maintenance level within a year. Strange design for a study in which weight loss was listed as the primary outcome. The fact that the low-carb dieters still lost the same amount of weight after two years should have been a major headline — especially since they weren’t told to count calories at any point in the study. But in his press interviews, lead researcher Gary Foster seemed to determined to give low-fat/low-calorie and low-carb diets equal ratings:
Foster, the study leader, said dieters should be less concerned about which diet to use, and focus on finding the support or technique – like writing down what they eat – that keeps them on track. “It doesn’t make a difference for weight loss how you get there,” he said.
When I read the statistical analysis part of the paper, I ended up scratching my head and trying to make sense out of gobbledygook like this: (Don’t give yourself a migraine trying to interpret. I’ll get to that.)
A parallel longitudinal model structure based on main effects for visit, treatment group, and baseline value and visit-treatment interactions was implemented with logistic regression for binary outcomes. We did estimates by using generalized estimating equations under the logistic regression model for correlated longitudinal binary outcomes implemented in the GENMOD procedure in SAS, version 9. Predicted values for each treatment and visit combination at the mean level of the baseline outcome, with corresponding lower and upper confidence bounds, were produced under each model for the figures. The previously mentioned longitudinal models preclude the use of less robust approaches, such as fixedimputation methods (for example, last observation carried forward or the analysis of participants with complete data [that is, complete case analyses]). These alternative approaches assume that missing data are unrelated to previously observed outcomes or baseline covariates, including treatment (that is, missing completely at random). The longitudinal models implemented for this study relax this missing-completely-at-random assumption in different ways.The generalized estimating equation- based longitudinal logistic models assume that missing data are unrelated to previously observed outcomes but can be related to the treatment because it is a covariate in the model. (that is, covariate-dependent missing completely at random) (18). The likelihood-based mixed-effects models further relax the covariate-dependent missing-completely-at-random assumption by allowing missing data to be dependent on previously observed outcomes and treatment (that is, missing at random). To assess departures from the missing-atrandom assumption under informative withdrawal-that is, the missing weights are informative for which patients chose to withdraw or continue to participate in the study-we present sensitivity analyses. As such, we assume that all participants who withdraw would follow first the maximum and then minimum patient trajectory of weight under the random intercept model.
I thought I knew what they were saying, but to make sure, I asked someone who deals with statistics. My suspicions were confirmed. Here’s what all that Engfish means:
WE INCLUDED THE DROP-OUTS IN OUR NUMBER-CRUNCHING BY ESTIMATING WHAT THEIR DATA POINTS WOULD HAVE BEEN.
Yup … they guessed. So at the end of the day, we’ve got them declaring that low-fat/low-calorie and low-carb diets produce equal weight loss after two years based on estimating the values for the people who quit the study. Nowhere in the paper do we find statistics based solely on the subjects who actually finished. Nowhere in the paper do we find any statistics at all on the actual calories or carbohydrates consumed. Did the low-carb group adhere to the low-carb diet? Did the low-calorie group keep their calories restricted? Since both groups regained some weight between the first and second year, the answer to both questions is almost certainly No.
But the data isn’t there for us to read. That makes me a little suspicious. I couldn’t help but wonder if these people set out to prove the superiority of low-fat diets, didn’t care much for the results, then got creative with their statistical analysis to blur the lines. So I put Google Scholar to work and looked up previous papers by the same researchers. What I found were studies written by people who clearly believe that weight loss is all about calories and that low-fat diets are the best way to keep calories in check. Here are some choice quotes. (I’m not going to italicize them because it’s a lot of text to read.)
“These results show that energy intake increases as dietary fat content increases across the usual range of dietary fat consumed in the United States. Even small reductions in dietary fat could help in lowering total energy intake and reducing weight gain in the population.”
“Our current food supply is high in fat, and high fat diets have been suggested to promote obesity by increasing energy intake, thus increasing the probability of positive energy balance and weight gain.”
“Although there are many environmental factors promoting excess energy intake and discouraging energy expenditure, it is clear that consumption of a high fat diet increases the likelihood of obesity and that the risk of obesity is low in individuals consuming low fat diets. On the basis of the available data, the current public health recommendations to lower dietary fat intake appear to be appropriate.”
“Despite changes in the diet over time, the variables associated with long-term maintenance of weight loss were the same: continued consumption of a low-calorie diet with moderate fat intake, limited fast food, and high levels of physical activity.”
“People definitely lose weight on low-carbohydrate diets. However, there is no evidence that the weight loss can be sustained over time. In addition, there are concerns about the long-term nutritional quality and safety of the low-carbohydrate diet, especially in regards to its effects on cardiovascular health.”
“Weight loss occurs with low-carbohydrate diets because people restrict caloric intake on these diets, producing a short-term negative energy balance. It is less clear whether they can stay with these diets in the long term and maintain their weight loss. It is also not clear whether the diets are safe to use for years at a time.”
“Control of portion size, consumption of a diet low in fat and energy density, and regular physical activity are behaviors that protect against obesity, but it is becoming difficult to adopt and maintain these behaviors in the current environment.”
“Using data from national surveys, we estimate that affecting energy balance by 100 kilocalories per day (by a combination of reductions in energy intake and increases in physical activity) could prevent weight gain in most of the population.”
Read that last one again. Cutting just 100 per calories per day would do the trick, because it’s all about the calories, right? Now read this one:
“To determine the optimal energy intake of very-low-calorie diets, 76 obese women were randomly assigned, in a double-blind fashion, to one of three liquid-formula diets: 420 kcal/d, 660 kcal/d, or 800 kcal/d. Weight, body composition, symptoms, mood, and acceptability of the diet were assessed throughout the 6-mo study. There were no significant differences in weight losses or changes in body composition among the three dietary conditions at the end of treatment.”
The difference between 420 calories per day and 800 calories per day for six months is 69,000 calories. If you believe every 3500 calories equals a pound of fat, the women in the 420-calorie group should have lost an extra 20 pounds. But they didn’t. So obviously the body can adjust, which means cutting 100 calories per day isn’t the solution to obesity. Those last two conclusions are from two different authors on the Low-Fat vs. Low-Carb study. Perhaps they should get together for a drink and compare previous findings.
I know I’m getting long-winded here, but bear with me. Given this group’s previous enthusiasm for low-fat diets, I doubt they were excited about the new study’s results, which clearly favored low-carb. That’s why I’m suspicious that the low-carb group ended up on what amounted to a maintenance diet within a year, and it’s why I’m doubly suspicious that the two-year results included estimates for dropouts.
So just for fun, let’s compare the results at 6 months. At that point, the dropout rate was still low, both groups were trying to lose weight, and the low-carb group would have been on a true low-carb diet.
Low-fat: 25 pounds
Low-carb: 27 pounds
That result alone should be a wake-up call. We’re talking about people — two- thirds of them women — who weighed an average of 226 pounds. A 1500-calorie diet is pretty restrictive for someone that heavy. And yet the low-carb group, with no calorie restrictions at all, lost a bit more. Did they end up eating less even without a calorie limit? We don’t know, because the investigators didn’t tell us. But if they did eat less, that should tell us something about which diet controls appetite naturally, and if they didn’t eat less, somebody owes the country a huge apology for telling us to count calories.
Low-fat: + 0.89
Low-carb: + 6.21
Those two results should have been the final nail in the coffin of the Lipid Hypothesis and the low-fat diet. Total cholesterol and LDL are lousy predictors of heart disease and always have been. But the ratio of Triglycerides/HDL is actually a pretty decent predictor, most likely because it serves a proxy for large vs. small LDL. The lower the ratio, the larger and fluffier your LDL. Ideally, that ratio should be less than 2.0. (Mine was 1.1 at last checkup.) Here is the change in that ratio at six months for both groups:
Low-fat at baseline: 124 / 45.4 = 2.73
Low-fat at 6 months: 99.7 / 46.3 = 2.15
Low-carb at baseline: 113 / 46.2 = 2.44
Low-carb at 6 months: 73 / 52.4 = 1.39
After six months, the group not counting calories and eating unlimited amounts of “artery-clogging saturated fat” lost more weight and showed a much more dramatic improvement in cardiovascular markers. After 12 months, the results still favored the low-carb group, although both groups backslid a bit.
And then wouldn’t you know it, by the end of two years, results were similar across the board, except for the HDL. And all the investigators had to do to achieve that convergence was instruct the low-carb group to increase their carbs and include estimated results for all the drop-outs.
I can’t help but wonder if they did that for a reason.