*Geek Box: Adjusting for Total Energy Intake in Epidemiology
Total energy intake is a critical factor to account for in any epidemiological analysis of diet and disease for three main reasons. First, the level of energy intake itself may be the primary factor influencing disease risk. Secondly, individuals vary in their body size and physical activity levels, and consequently in their total energy intake; this means absolute levels of nutrient intake will vary from person to person, which could introduce random error into the analysis. Finally, if a nutrient is a more direct cause of disease than total energy intake, then the effect of a nutrient of interest would be distorted by total energy intake [because as energy increases or decreases, the intake of the nutrient would increase or decrease].
The best analogy for adjusting for total energy in epidemiology is that it is seeking to achieve the same effect as having isocaloric diets in an intervention. Let’s say you want to compare two diets, one high in fat and lower in carbohydrate vs. the opposite: if one diet had less energy, we would say the effects were likely due to this difference in total energy intake. This is the same for epidemiology: the exposure of interest is the composition of diet, independent of total energy intake.
For epidemiology, other methods have been proposed to more assess total energy intake, in particular adjusting for body weight and adjusting for physical activity. However, these methods do not cancel out measurement errors because of body weight and physical activity are independently related to energy intake.
However, because both total energy and individual nutrients are calculated from the same foods, the errors for both are strongly correlated. Therefore, by carefully adjusting the intake of a nutrient for total energy intake, these correlated errors cancel each other out and the validity of the measure of a nutrient is improved. There are a number of methods of adjusting for total energy, including the nutrient-density method, the energy-adjusted residual method, the energy partition method, and multivariate methods. Each has certain advantages and limitations, and it is important to consider what the variable of interest is, how that variable relates to other factors, and the implications for biological plausibility of the chosen method.