The first question NuSI proposes to address is, what factors drive the body to accumulate excess fat?  That is, what are the fundamental causes of common human obesity?

The conventional wisdom is that obesity is an energy balance disorder, caused merely by the consumption of more calories than are expended. By this logic, combating the obesity epidemic requires that obese and overweight individuals learn to eat less and exercise more, and that our environment change to facilitate those specific behaviors.

“It is in vain to speak of cures, or think of remedies, until such time as we have considered of the causes . . . cures must be imperfect, lame, and to no purpose, wherein the causes have not first been searched.”

- Robert Burton, The Anatomy of Melancholy

An alternative hypothesis is that obesity is a growth disorder, in effect, caused by a hormonal/enzymatic defect that is possibly triggered by specific foods in our diet. By this logic, combating the obesity epidemic requires that obese and overweight individuals restrict consumption of these specific foods, and that our environment change to reduce their abundance and availability.

The initial goal of NuSI is to ascertain which of these two hypotheses is correct and to what degree fat accumulation can be influenced, if any, by the macronutrient content of the diet independent of total caloric intake.

Dive deeper into the studies

Rigorously testing these hypotheses, however, involves considerable challenges and subtleties. One complication is the difficulty of differentiating between the two competing hypotheses under many of the experimental conditions that would seem a priori to be capable of establishing such a distinction. Individuals who consume carbohydrate-restricted diets ad libitum, for instance, often spontaneously reduce their caloric intake, thus conflating reduced carbohydrate consumption with reduced consumption of total calories. Experimental diets that significantly restrict total caloric intake – as with many conventional diets tested — will also restrict carbohydrates significantly, resulting in the same problem: the inability to differentiate between a reduction in fat mass that may have been caused by the reduction in calories consumed or may have been caused by the reduction in carbohydrates alone or by some combination of the two. These issues have yet to be addressed unambiguously, which is why both of these competing hypotheses of obesity can still be considered viable.

Below, we summarize an extensive review of the existing literature on the subject, covering from 1944 onward. Our first criterion was to find experimental trials that involved macronutrient alteration and that were carried out in a setting that allowed for meaningful control over the composition and caloric intake of what the subjects actually ate (e.g., a hospital, metabolic ward, or prison, in particular). We then supplemented these studies with free-living studies of macronutrient alteration. Finally, we looked for overfeeding studies that used a diet significantly reduced in carbohydrate content for at least some subjects. In total, we have summarized nearly 100 studies, which have been aggregated in our appendix. We find that while many of these studies explicitly attempt to address the questions of whether carbohydrate consumption influences fat mass independently of calories consumed and vice versa, none have been able to answer the question conclusively. All of these studies have shortcomings in this context, which include the following:

  1. Lack of control over what the subjects ultimately eat, a problem common to all free-living studies;
  2. Small sample size and/or short duration, both of which diminish statistical power;
  3. Caloric restriction of the control diet, and so effectively carbohydrate-restriction of all the competing dietary interventions;
  4. Caloric restriction of the carbohydrate-restricted diet, thus building into the experiment the assumption to be tested – that the carbohydrate content of the diet can influence weight loss independent of total caloric content;
  5. Testing diets of only marginally different nutrient content — “low carbohydrate” diets, for instance, that are still 30% or more carbohydrate.
  6. Manipulation of fat and protein content in the diet, rather than the carbohydrate content itself;
  7. The use of habitually lean subjects to test the effect of nutrient composition on the accumulation of excess fat.

Let us address these in order:

  1. The overwhelming majority of dietary trials are free-living studies that provide subjects with diet guides – often with popular diet books themselves – nutritional counseling on the dietary approach, and then send the subjects back to their lives with the hope that they do indeed continue follow the prescribed dietary intervention. Knowledge of what the subjects actually eat is attained through food diaries or questionnaires, both of which have demonstrated to be incapable of accurately assessing either diet quality or quantity [Brinkworth et al. 2009; Dansinger et al. 2005; Gardner et al. 2007; Johnstone et al. 2008; Meckling et al. 2004; Miller et al. 2003; Noakes et al. 2006; Petersen et al. 2006; Sacks et al. 2009; Shai et al. 2008; Tay et al. 2008; Volek et al. 2002; Yancy et al. 2004]. Free-living studies – even the ones that find a benefit to low-carbohydrate consumption – can offer no guarantee that the true or even correct effect of a dietary intervention has been determined.
  2. This first issue can be effectively solved by isolating subjects in metabolic wards but this has historically compromised sample size and duration of trial for control of the dietary intervention. Many of the trials cited in the older literature as addressing the question of calories vs. macronutrients are small [Kekwick and Pawan 1956 (6 subjects); Kinsell et al. 1964 (5 subjects); Olesen and Quaade 1960 (8 subjects); Werner et al. 1955 (6 subjects); Pilkington et al. 1960 (9 subjects); Young et al. 1971 (8 subjects)] and/or were conducted for a short time period [Hill et al. 1991 (8 subjects, 7 days); Lewis et al. 1977 (10 subjects, 14 days); Yang and Van Itallie 1976 (5 subjects, 10 days per diet); Krehl et al. 1967 (4 subjects in one protocol and 7 subjects in another, one month per diet)]. Because of the lack of statistical power with such a small sample size, and because of the inability to measure fat loss specifically over such short durations, these studies were unable to detect the modest differences in fat loss that may have been present during the study, or attenuated with a longer study or a greater sample.]
  3. Both free-living and metabolic ward studies typically restrict the caloric intake of the control diet and the carbohydrate-restricted diet. Metabolic ward studies typically used diets so hypocaloric that the control group was severely carbohydrate-restricted compared to a usual diet [Bogardus et al. 1981 (830 calories); Pilkington et al. 1960 (800-1000 calories); Golay 1994 (1000 calories); Kekwick and Pawan 1956 (second treatment, 1000 calories); Vazquez and Adibi 1992 (590 calories); Vazquez and Kazi 1994 (615 calories); Piatti et al. 1994 (800 calories) Miyashita et al. 2004 (1000 calories); Yang and Van Itallie 1976 (800 calories)]. Consider an intervention that compares a 1000 kcal/d “calorie restricted” diet to a 1000 kcal/d carbohydrate-restricted diet of 10 % carbohydrate. If the usual diet of the obese subjects was 3000 kcal/d, of which 50% was carbohydrate, the diet that restricts all calories equally will be restricting carbohydrates from 1500 kcal/d to 500 kcal/d. The diet that restricts primarily carbohydrates will be restricting carbohydrates from 1500 kcal/d to 100 kcal/d. Such a study, therefore, can be perceived as comparing a diet that restricts carbohydrates by 1000 kcal/d to a diet that restricts carbohydrates by 1400 kcal/d. The question would then be whether or not this difference is significant enough to induce an observable different in fat loss over the period of the study. Under these circumstances, any differences might be too modest to detect.
  4. Restricting the total calories consumed for the carbohydrate-restricted diet assumes that such a restriction is either necessary for weight loss, which is a hypothesis being tested, or tries to control for total caloric intake, in which case it does so by sacrificing a valid test of the hypothesis that such a restriction is necessary. Allowing the subjects to eat ad libitum, though, allows for intervention effects: specifically subjects might restrict calories – even when advised not to – with the belief that this will maximize the effect of the diet, and so increase the difficulty of differentiating between the effects of carbohydrate restriction and of total calorie restriction.
  5. Some studies do not limit carbohydrates sufficiently to establish a significant difference between the two interventions, and one that would be expected to induce a statistically significant effect on fat mass. This problem is exacerbated in free-living trials in which a large proportion of the subjects apparently fail to maintain the intervention. It is exacerbated further when both dietary interventions are prescribed hypocaloric, as previously discussed [Grey and Kipnis 1971 (at least 25% carbs); Hill et al. 1991 (at least 20% carbs); Torbay et al. 2002 (25% carbs vs. 30% carbs); Sacks et al. 2009 (at least 35% carbs)].
  6. A few studies are well designed and implemented, but do not establish complete macronutrient compositional changes. For instance, Bray et al. (2012) studied hypercaloric diets (140% of baseline) of varying protein and fat proportions, while maintaining carbohydrate content at approximately 40 percent in both diets.
  7. Several trials cited in the literature as establishing the role of nutrient composition on fat accumulation, or the lack thereof [Bray et al. 2012, Leibel et al. 1992] either overfed lean subjects, building in the assumption that obesity is caused merely by the consumption of more calories than are consumed — a hypothesis to be tested — or study the effect of macronutrient composition on lean subjects without overfeeding. It’s quite possible, though, that lean individuals are those who are resistant to the dietary factors that cause overweight or obesity in others and so are not suitable subjects to answer the relevant questions. As Leibel et al. 1992 phrased this problem, “Similar results might not have been obtained in a group of obese individuals or lean individuals susceptible to obesity.”

In short, the state of the scholarly literature is such that we need well-designed, well-controlled, large n, lengthy studies directly addressing the question of whether the carbohydrate content of the diet influences fat mass independent of total caloric intake and whether total caloric intake has an influence independent of carbohydrate content. Such studies would also ultimately address the effect of nutrient composition on biomarkers of disease and other relevant outcomes.

  • Download - summary of studies Download - list of all citations