The primary outcome of the study was the change in body weight over a period of 2 years, and the secondary outcome was the change in waist circumference. Data were pooled from the diets for the two factorial comparisons: low fat versus high fat and average protein versus high protein. The analysis also included a comparison of two of the four diets, the diet with the lowest carbohydrate content and the diet with the highest carbohydrate content, and included a test for trend across the four levels of carbohydrates. The effects of protein, fat, and carbohydrate levels were evaluated independently with the use of two-sample t-tests at a two-sided significance level of 0.05. Exploratory post hoc analyses were performed with threshold amounts of weight loss as outcomes, with Bonferroni’s adjustment for multiple comparisons. Associations between adherence to the fat and protein goals and weight loss were also explored in post hoc analyses (see Methods in the Supplementary Appendix).
We performed an intention-to-treat analysis in which long-term weight loss for persons who withdrew from the study early (after at least 6 months of participation) was imputed on the basis of a rate of 0.3 kg per month of regained weight34 (link) and a rate of 0.3 cm per month of regained waist circumference after withdrawal (see Methods in the Supplementary Appendix). Risk factors for cardiovascular disease and diabetes were also analyzed according to the intention-to-treat principle, with zero change from baseline imputed for missing data. The study was powered to detect a 1.67-kg weight loss as an effect of the level of protein or fat in the diet over the 2-year period, assuming a withdrawal rate of 40%.