Determination of energy density of freely selected diets: methodological issues and implications

BACKGROUND: There is increasing evidence that dietary energy density (ED, kJ/g) may be an important dietary characteristic, particularly in respect to control of energy intake; however, there are no agreed methods for deriving the ED of freely selected diets, and ED values may be markedly affected by the inclusion or exclusion of specific dietary items, particularly beverages. OBJECTIVE: To highlight the consequences of using six different methods of ED calculation, and their implications for characterizing differences between weight status groups and identifying associations of ED with macronutrient intakes. DESIGN: ED was calculated using six defined methods: (1) all food and beverages; (2) all food and energy beverages; (3) food, milk and alcohol; (4) food only; (5) all dry matter; (6) protein, carbohydrate and fat only, of varying exclusions of different beverages and water. For illustrative purposes, data from 41 lean (LE, body mass index (BMI) 20 - 25 kg/m(2)) and 34 obese (OB, BMI greater than or equal to 30 kg/m(2)) adults who kept 4-day weighed dietary intake records are described. RESULTS: ED values (and coefficient of variation, CV) differed substantially by methods of calculation. OB reported significantly greater mean ED compared with LE by one method (all food, milk and alcohol, excluding other nonalcoholic beverages); however, the opposite was found using another method (dry weight). For most calculation methods, ED was negatively associated with percentage energy from carbohydrate for LE, in contrast to OB. All methods found positive correlations for ED and fat (g) among LE, but only one method found such a correlation among OB. Similarly, three methods produced positive correlations between ED and percentage energy fat amongst LE; however, this was only observed amongst OB with one method. CONCLUSIONS: Methods of calculating ED of freely selected diets must be carefully defined, and can markedly influence apparent relationships of ED with other dietary measures and subject characteristics.