No medical condition has generated as many dietary remedies as obesity. All diets have their followers, but hard data on the efficacy of the diets are scarce. In this issue of the Journal, Sacks et al. report the results of a large, long-term trial that tested the efficacy of weight-loss diets that were high or low in carbohydrates, protein, or fat. Highcarbohydrate, low-fat diets became popular approximately 20 years ago, when it was thought that calories from carbohydrates were less fattening than the same number of calories from fat. A high-fat, low-carbohydrate diet was popularized by Dr. Robert Atkins in the 1970s2 and recently enjoyed a revival. The appeal of high-protein diets is that protein is thought to provide more satiation per calorie than fat or carbohydrates.
The trial by Sacks et al. lasted longer than most, the dropout rate was low, treatment was intensive, and compliance was assessed with objective biomarkers. Unfortunately, the dietary goals were only partly achieved. Protein intake was intended to differ by 10% of energy between the high-protein-diet group and the average-proteindiet group, but the actual difference, as assessed by the measurement of urinary nitrogen excretion, was 1 to 2% of energy (according to my calculations, which were based on a diet that provided 1700 kcal per day). Extreme carbohydrate intakes also proved hard to achieve. When fat is replaced isocalorically by carbohydrate, high-density lipoprotein (HDL) cholesterol decreases in a predictable fashion. The authors used the difference in the change in HDL cholesterol levels between the lowest- and highest-carbohydrate groups to calculate the difference in carbohydrate content between those diets. That difference turned out to be 6% of energy instead of the planned 30%.
The reduction in caloric intake was also not sustained. Weight loss averaged 6 kg at 6 months, which fits reasonably with the planned daily deficit of 750 kcal. However, after 12 months, subjects started to regain weight, which suggests that they were eating more than planned. Final weight losses averaged 3 to 4 kg after 2 years. This weight loss is similar to the weight loss that can be achieved with pharmacotherapy, and it is a clinically relevant effect that will slow the onset of type 2 diabetes. To that extent, all the diets were successful. But the weight regain during the second year, although slow, suggests that in the end many participants might have regained their original weight even if treatment had continued.
Within each diet group, some participants achieved much better weight loss than others. Participants who lost more weight attended more counseling sessions and adhered more closely to the prescribed dietary composition. These observations led Sacks et al. to conclude that behavioral factors rather than macronutrient composition are the main influences on weight loss. That is a plausible hypothesis, and it has been observed before, but the present data do not allow a firm conclusion to be reached, because differences in macronutrient intake were too small.
Even if the planned differences in macronutrient intake had been achieved, the absence of blinding would have made it difficult to ascribe the effect of a particular diet to protein, fat, or carbohydrate molecules. Weight-loss studies are behavioral studies; they require participants to eat less. Cognition and feelings have a huge impact on such behavior. Participants may eat less not because of the protein or carbohydrate content of a diet but because of the diet’s reputation or nov-elty or because of the taste of particular foods in the diet. Specific effects of fat, protein, and carbohydrates on food intake and body weight can be determined only when all diets look and taste the same. Studies that have accomplished this goal with the use of porridges (similar to oatmeal) and standardized snacks or with covertly manipulated foods have been carried out for short periods, but few subjects would be willing to eat those foods for the several years that would be needed to examine long-term effects. Therefore, this issue is unlikely to be settled soon. If behavior rather than diet composition is the key to weight loss, macronutrient composition may be of secondary importance anyway.
The trial by Sacks et al. lasted longer than most, the dropout rate was low, treatment was intensive, and compliance was assessed with objective biomarkers. Unfortunately, the dietary goals were only partly achieved. Protein intake was intended to differ by 10% of energy between the high-protein-diet group and the average-proteindiet group, but the actual difference, as assessed by the measurement of urinary nitrogen excretion, was 1 to 2% of energy (according to my calculations, which were based on a diet that provided 1700 kcal per day). Extreme carbohydrate intakes also proved hard to achieve. When fat is replaced isocalorically by carbohydrate, high-density lipoprotein (HDL) cholesterol decreases in a predictable fashion. The authors used the difference in the change in HDL cholesterol levels between the lowest- and highest-carbohydrate groups to calculate the difference in carbohydrate content between those diets. That difference turned out to be 6% of energy instead of the planned 30%.
The reduction in caloric intake was also not sustained. Weight loss averaged 6 kg at 6 months, which fits reasonably with the planned daily deficit of 750 kcal. However, after 12 months, subjects started to regain weight, which suggests that they were eating more than planned. Final weight losses averaged 3 to 4 kg after 2 years. This weight loss is similar to the weight loss that can be achieved with pharmacotherapy, and it is a clinically relevant effect that will slow the onset of type 2 diabetes. To that extent, all the diets were successful. But the weight regain during the second year, although slow, suggests that in the end many participants might have regained their original weight even if treatment had continued.
Within each diet group, some participants achieved much better weight loss than others. Participants who lost more weight attended more counseling sessions and adhered more closely to the prescribed dietary composition. These observations led Sacks et al. to conclude that behavioral factors rather than macronutrient composition are the main influences on weight loss. That is a plausible hypothesis, and it has been observed before, but the present data do not allow a firm conclusion to be reached, because differences in macronutrient intake were too small.
Even if the planned differences in macronutrient intake had been achieved, the absence of blinding would have made it difficult to ascribe the effect of a particular diet to protein, fat, or carbohydrate molecules. Weight-loss studies are behavioral studies; they require participants to eat less. Cognition and feelings have a huge impact on such behavior. Participants may eat less not because of the protein or carbohydrate content of a diet but because of the diet’s reputation or nov-elty or because of the taste of particular foods in the diet. Specific effects of fat, protein, and carbohydrates on food intake and body weight can be determined only when all diets look and taste the same. Studies that have accomplished this goal with the use of porridges (similar to oatmeal) and standardized snacks or with covertly manipulated foods have been carried out for short periods, but few subjects would be willing to eat those foods for the several years that would be needed to examine long-term effects. Therefore, this issue is unlikely to be settled soon. If behavior rather than diet composition is the key to weight loss, macronutrient composition may be of secondary importance anyway.