The association between dietary energy density, increased energy intake and weight gain is supported by experimental evidence, but confirmation of an effect in free-living humans is limited. Experimental evidence supports a role of energy density in obesity through changes in food composition, not drinks consumption. The inclusion of drinks in the calculation creates a variable of questionable validity and has a substantive impact on the estimated energy density of the diet. We posit, based on the experimental evidence, that calculating the energy density of diets by excluding drinks and including calories from drinks as a covariate in the analysis is the most valid and reliable method of testing the relationship between energy density and weight gain in free-living humans. We demonstrate, by systematically reviewing existing observational studies of dietary energy density and weight gain in free-living humans, how current variation in the method for calculating energy density hampers the interpretation of these data. Reaching an a priori decision on the appropriate methodology will reduce the error caused by multiple comparisons and facilitate meaningful interpretation of epidemiological evidence to inform the development of effective obesity prevention strategies.