Abstract A holistic statistical modelling procedure for the analysis of energy balance (EB) studies is proposed in this review. We show that a switch from EB-based analysis to heat production (HP)-based analysis has some advantages. The use of augmented EB data with a fasting HP estimate is proposed to correct slope bias due to non-linearity. This is facilitated by joint or parallel analysis of HP and EB data. Increasing levels of metabolizable energy intake (MEI) will lead to an increased HP attributed to greater energy cost of digestion and other metabolic processes following an exponential-like trajectory. This will correspondingly introduce a diminishing-returns-like non-linearity in the EB versus MEI relationship because HP and EB complement each other. This makes the assumption of linearity somewhat erroneous unless the MEI range is very close to the maintenance level. By using non-linear regression models (e.g. diminishing returns curves), we do not need to impose the assumption of linearity of the relationship EB versus MEI. In this study, we show how estimates of FHP, derived from the analysis of HP data (scaled by metabolic weight), can be included in EB data analysis allowing for the estimation of unbiased curvature-based slope parameters.