To the Editor: Dr Ebbeling and colleagues1 conducted a crossover trial to examine energy expenditure during weight-loss maintenance with a low-fat, a low–glycemic index, and a very low-carbohydrate diet. They found decreases in energy expenditure that were greatest with the low-fat diet and least with the very low-carbohydrate diet.
I believe the authors overinterpreted 1 key result in their study. They commented that “C-reactive protein [CRP] also tended to be higher with the very low-carbohydrate diet” and concluded that the very low-carbohydrate diet “may increase cortisol excretion and CRP.” The authors implied that the low–glycemic index diet may be preferable to the very low-carbohydrate diet despite the fact that the latter outperformed the former on most key outcome measures. However, the overall null hypothesis of equal mean CRP values across the 3 diets could not be rejected at the specified .05 significance level, with a P value of .13. Nevertheless, the authors reported the results of a test for linear trend across the 3 diets, arranged from highest to lowest glycemic load and assuming equal spacing, with a corresponding P value of .05. It is this test that the authors used to support the conclusion.