13This paper compares predictions of metabolic energy expenditure in gait using 14 seven metabolic energy expenditure models to assess their correlation with 15 experimental data. Ground reaction forces, marker data, and pulmonary gas 16 exchange data were recorded for six walking trials at combinations of two speeds, 17 0.8 m/s and 1.3 m/s, and three inclines, -8% (downhill), level, and 8% (uphill). 18 The metabolic cost, calculated with the metabolic energy models was compared to 19 the metabolic cost from the pulmonary gas exchange rates. A repeated measures 20 correlation showed that the model by Bhargava et al. [7] and the model by 21 Lichtwark and Wilson [21] correlated best with the experimental data. It was 22 found that for all metabolic energy models, the subject's weight affected the 23 calculated metabolic cost. 24 Introduction 25 Humans prefer to walk in energetically optimal ways. Walking speed [1], the ratio 26 between step length and frequency [2], step width [3] and vertical movement of the 27 March 25, 2019 1/25 center of mass [4, 5] are chosen to minimize energy expenditure. Whole-body energy 28 expenditure can be measured using direct calorimetry, by measuring the heat 29 production in the body, or indirect calorimetry, by measuring the volume of oxygen 30 inspired and carbon dioxide expired [6]. However, these measurements are not always 31 available, but an accurate prediction of energy expenditure is often desired. 32 Instead, musculoskeletal modeling can be used to simulate human gait and to 33 calculate the energy expenditure based on a metabolic energy expenditure model (e.g., 34 [7-10]). These calculations are based on variables that are studied in gait analysis, such 35 as joint moments, joint power, or muscle forces, lengths and activations [11]. With a 36 metabolic energy expenditure model, the energy expenditure can be calculated for gait 37 experiments that did not take metabolic energy expenditure measurements, or for 38 predictive gait simulations, where no experimental data is available. Recently, these 39 simulations have been used to analyze 'what-if scenarios' such as the effect of an 40 intervention such as a prosthesis [12], an exoskeleton [13], ankle foot orthosis [14], 41 additional weight [15], loaded and inclined walking [16], or changing the gait pattern to 42 minimize knee reaction force [17] on gait. Energy cost is an important variable in gait, 43 and should be considered whenever a clinical intervention is studied or designed. 44 In literature several energy models were suggested. The Huxley crossbridge 45 model [18] finds both the muscle force and the energy expenditure of a muscle, but 46 requires up to 18 states [19]. Instead, Hill-type muscle models [20] are typically used to 47 simulate muscles, but these do not output metabolic energy expenditure. Therefore, 48 several metabolic energy expenditure models have been proposed that calculate the 49 energy expenditure during walking based on Hill-type muscles [7-10, 21, 22], based on 50 muscle effici...