Changes in Energy Expenditure with Weight Gain and Weight Loss in Humans

Weight loss and weight gain are associated with declines and increases in energy expenditure (EE), which mainly follow changes in the metabolically active component of the body, i.e. fat-free mass (FFM). Most of these changes are non-adaptive and occur passively. However, weight change does not exactly follow prediction based on calculation of energy imbalance. This is explained by FFM-independent metabolic adaptations, i.e. adaptive thermogenesis (AT). AT limits changes in energy stores in response to varying energy intake and/or EE, e.g. AT explained about 50 % of the less-than-expected weight loss in obese patients [1, 2]. Energy sparing (with weight loss) and energy dissipation (with weight gain) are related to the issues of obesity as well as voluntary (e.g. due to dieting in obese subjects) and unvoluntary weight loss (e.g. during cachexia in cancer patients).

AT refers to (i) the resting (or non-activity-related) component of EE including resting energy expenditure (REE), as well as the diet-induced thermogenesis (DIT), and (ii) non-resting component (i.e. activity-related energy expenditure, AEE, which is further divided into exercise and non-exercise activity thermogenesis, EAT and NEAT) of total energy expenditure (TEE). Metabolically, AT has been explained by the ratio of glycolytic to oxidative enzymes together with an altered efficiency of free fatty acid oxidation in skeletal muscle, “futile” cycles consuming ATP without a net change in products (e.g. hydrolysis of triglycerides and subsequent re-esterification in adipocytes), changes in the ATP-costs per muscle contraction, mitochondrial uncoupling in brown adipose tissue, energy-consuming pathways like lipogenesis, NEAT and/or partitioning of energy to fat mass or FFM [3]. These mechanisms are considered to be under genetic and hormonal control, i.e. by insulin, leptin, thyroid hormones and sympathetic nervous system (SNS) activity.

AT has been related to (i) negative energy balance due to caloric restriction and/or excessive exercise, (ii) chronic overfeeding, (iii) re-feeding after weight loss and (iv) weight maintenance after weight reduction. This will be discussed in detail.

(i) With caloric restriction, negative energy balance and weight loss cause decreases in all the energy expenditure components, i.e. REE, DIT and AEE (for reviews see [3, 4]). Sixty-five years ago, the Minnesota Starvation Experiment was the first quantitative description of AT in humans [5••]. Since then, AT has been reproduced in experimental and clinical studies on weight loss. AT varied between 100 and 500 kcal/day, it is observed in lean as well as overweight subjects. AT is independent of the weight loss strategy.

(ii) In the 1968 Vermont overfeeding study, body weight gain was lower than expected from the excess of energy intake [6]. This had been explained by an increase in EE with overfeeding as a dissipative mechanism to oppose weight gain. The concept was in line with the “gluttony” concept proposed by Miller et al. about 50 years ago [7, 8]. However, throughout the following decades, numerous well-controlled studies could not confirm mass-independent increases in REE and/or DIT during chronic overfeeding (for reviews see [9‐15]). With continuous overfeeding, body energy stored was 60–75 % of excess energy leaving the rest for an increase in EE which was explained by obligatory costs (e.g. for gaining body protein, increased cost of walking, etc. [9‐11]. These calculations left less than 10 % of overfeeding-induced energy expenditure unexplained which was accounted for errors of methods and assumptions to estimate energy balance questioning energy dissipation. However, during controlled overfeeding, the non-resting component of EE, i.e. EAT increases at unchanged NEAT [9, 16, 17]. This effect was independent of weight gain-associated changes in body composition. Using a highly controlled protocol in lean subjects, intermittent periods of 3 weeks overfeeding at +20, +40 and +60 % of energy needs at controlled and low physical activity increased TEE, while again, there was no evidence for mechanisms able to dispose excess energy [18]. Macronutrient composition may add to metabolic changes in response to overfeeding. While there were no differences in increases in EE after either carbohydrate or fat overfeeding [19], a high protein intake may have a greater effect. Hypercaloric protein-rich diets (with a protein content at 25 % of energy intake) increased FFM, TEE and REE [20•]. However, AT was not different between diets differing in protein content. This is also against the idea that overfeeding low protein diets stimulates thermogenesis [21]. Taken together, during overfeeding, an adaptative increase in REE is non-existent and probably more a measurement artefact [15, 18, 20•, 22]. By contrast, AT seems to be related to the non-resting component of EE [9, 16, 17].

(iii) When recovering from starvation, people spontaneously overeat; body weight and fat mass increase. Weight regain may be forced by persistence of the starvation-induced suppression of thermogenesis [5••, 15, 22]. However, during a controlled caloric restriction-refeeding cycle in healthy lean men, mass-independent decreases in REE reversed within 2 weeks of refeeding [5••, 23••]. No energy dissipation occurred with refeeding and weight regain.

(iv) If weight loss-induced AT persists during refeeding, it may carry a long-term risk of “overshooting” pre-starvation body weight and weight loss-weight regain cycles (for a review see [17, 24‐26, 27••, 28]). A low REE is a risk of weight gain, and it impedes weight maintenance [26, 29]. The role of AT in the control of weight loss maintenance has been addressed in a seminal series of controlled experiments performed by Leibel and Rosenbaum (for reviews see [17, 24, 25•]). These authors investigated normal and overweight subjects after a 10 or 20 % weight loss with a subsequently controlled weight stabilisation phase of at least 14 days [17]. In this protocol, AT was about 50 to 140 kcal/day and mainly related to the non-REE component of TEE.

In a long-term observation study on weight-reduced overweight patients, weight regainers had a reduced AT when compared with weight stable patients [26]. By contrast, after massive weight loss of nearly 60 kg in severely obese patients, weight regain was not associated with AT [27••]. The persistence of AT during weight gain after weight loss has been questioned by controlled feeding experiments [5••, 23••] and computational modelling of human energy metabolism [30]. Following a starvation-refeeding cycle, the curve traced by REE on FFM followed a loop with a decrease in the REE-FFM association induced by negative energy balance and a nearly parallel increase during refeeding: At a similar FFM, REE was lower after 12 weeks of starvation when compared with 12 weeks of refeeding [30].

Weight change-associated changes in EE vary among individuals but may be related to one another within individuals responding to weight gain and weight loss [31, 32]. With negative energy balance, a high AT reduces the drive towards weight loss. By contrast, a low AT in response to overfeeding increases the metabolic drive to gain weight. In a given individual, both mechanisms add to efficient energy use to conserve body energy providing a so-called thrifty phenotype [32, 33]. Vice versa, a low AT in response to negative energy balance together with a high AT during positive energy balance favours weight loss during caloric restriction and limits weight gain with overfeeding altogether characterising a so-called spendthrift phenotype [32, 33]. Over long-term, metabolic phenotypes characterised by AT are correlated to subsequent weight changes.

Based on the quantitative physiology of weight change, robust mathematical (so-called “flux-balance”) models of human energy metabolism and AT have been developed and validated [30, 34•, 35]. Crucial points of these models relate to (i) energy partitioning (i.e. energy imbalance is divided between fat mass and FFM) and (ii) possible feedbacks of fat mass and/or FFM on energy intake which are assumed to be constant. Computational modelling is now widely used to predict weight changes in obese patients during dieting or to explain the non-linear function of weight loss.

AT seems to be well-established in the biology of weight loss. Presently, there are three innovative topics related to AT in humans: first, definition and assessment of AT; second, the dynamics of metabolic adaptation related to weight loss and weight loss maintenance; and third, the determinants of AT.

Following Ancel Keyes original definition [5••], AT refers to changes in REE which are independent from FFM and FFM composition. During weight loss, AT reflects decreases in specific metabolic rates of organs and tissues within FFM. It follows that AT can only be assessed based on accurate body composition analysis (BCA) including magnetic resonance imaging (MRI) to quantitate masses of low (i.e. skeletal muscle) and high metabolic rate organs (i.e. brain, liver, heart and kidneys) [3, 36, 37••]. After adjustment of REE for FFM, only AT was calculated to be around 100 kcal/day (or about 50 % of the fall in REE; 23). In that study, FFM decreased by 2.4 kg with concomitant changes in the masses of skeletal muscle (−1.5 kg) and liver (−0.2 kg) [23••]. Adjusting REE for changes in FFM plus the anatomical composition of FFM, “true” AT was calculated to be 70 kcal/day [23••]. It is tempting to speculate that additional adjustments of REE for the molecular composition of organs and tissues (e.g. for their hydration, protein content and density) will further affect the calculation of AT [36].

Normalising REE for FFM or FFM + fat mass has limitations when one addresses changes in REE with weight loss. Normalisation is a statistical approach. It does not take into account physiological changes in adipose tissue-derived FFM (water, protein, minerals in adipocytes), the proportion of the high metabolically active organs and tissues to FFM (see above) and a yet poorly defined thermic effect of adipocytes probably due to their secretory and inflammatory activities [3, 37••]. The characteristics of regressions for FFM and fat mass differ between different degrees of adiposity [37••, 38]. In fact, applying the REE on FFM + FM-relationship before weight loss to data obtained after a 6-kg weight loss led to an error of about 70 kcal/day. The overestimation of AT calculated from the REE vs FFM + FM relationship before weight loss increases after a 50-kg weight loss in severely obese subjects explaining (but also questioning) a magnitude of AT of about 300 kcal/day at a decrease in REE of about 600 kcal/day [39]. By contrast, using a %fat mass-specific regression equation to normalise REE before and after weight loss reduced AT to 120 kcal/day [37••].

AT can be calculated from the difference between REE adjusted for FFM + fat mass before and after weight loss. Alternatively, the difference between measured REE and REE calculated from organ and tissue masses times their specific metabolic rates can be used [23••, 37••, 40, 41]. This difference is a measure of the mass-independent changes in energy expenditure, and it increases after weight loss. However, the approach depends on some assumptions, e.g. regarding the molecular composition of organs and tissues and their specific metabolic rates. Since 3 days of caloric restriction may result in losses of up to 2–3-l water with an accompanying 3.6 % change in FFM hydration [36], a weight loss-associated change in FFM does not necessarily reflect a decrease in metabolically active FFM. Overestimating the loss of metabolically active FFM results in an underestimation of AT [22, 36]. In addition, specific metabolic rates of individual organs and tissues are not constant but vary with adolescence, age above 55 years, obesity, weight change and work load [41, 42].

There is need of more sophisticated concepts to assess AT. Presently, detailed BCA at the organ-tissue level (using whole body MRI to assess organ and tissue masses) together with BCA at the molecular level (using balance and dilution techniques to assess changes in tissue hydration and protein content) is needed for proper adjustments of REE. In addition, direct estimates of specific metabolic rates of organs and tissues by magnetic resonance spectroscopy and/or positron emission tomography (PET) will add to assess functional body composition and AT. Faced with these conceptual and methodological caveats, AT cannot be considered as a biological entity. In addition, scientists should be honest about the limits of detection using state-of-the-art technologies to assess EE [22, 43]. The precision differs between assessments of REE by indirect calorimetry (between 1 and 3 %) and measurements of TEE with doubly labelled water (DLW, above 5 %) where the intra-individual variance may exceed the inter-individual variance [22].

Changes in energy expenditure are functions of time with fluctuations within minutes, hours, days, weeks and months. In the Minnesota Starvation Experiment, REE and body composition were measured after 1, 3 and 6 months of caloric restriction [5••]. This time scale is in line with more recent clinical studies on AT which vary in observation periods between 3 weeks and 6 years ([1, 27••, 44‐49]; for reviews see [3, 4, 50]). Weight loss results from a negative energy balance and changes in body composition; it is not continuous but curve-linear ending when a new steady state and, thus, a new equilibrium between energy intake and energy expenditure are reached [51].

During caloric restriction, the first phase of weight loss is rapid and lasts less than a week (phase 1); this is followed by a second phase characterised by a slower weight loss (=late phase) [51]. With total starvation, phase 2 continues until fat mass is nearly completely depleted. Weight loss becomes deleterious to the subject when body protein is the only endogenous energy source left. During controlled 3-week semi-starvation, healthy young men lost about 3.2 kg during week 1 and 1.3 to 1.4 kg/week during week 2 and 3 [23••, 36, 40]. Decay constants in weight were calculated as K ₁ = −0.78 kg/day (week 1 = phase 1) and K ₂ = −0.19 kg/day (week 2 and 3 = phase 2). As to body composition, the corresponding decreases in FFM were 313, 90 and 66 g/day, respectively [23••]. Concomitantly, fat mass decreased by 168, 109 and 142 g/day.

Weight loss is characterised by defined changes in body composition. These changes relate to AT. During phase 1, decreases in FFM exceed the decrease in fat mass; they are due to losses in intracellular (due to glycogen and protein mobilisation) and extracellular water (due to an increase in natriuresis) rather than to losses in cell mass per se [23••, 36]. By contrast, phase 2 relates more to the steady and ongoing loss of fat mass. Since the energy content per kilogram change differs between body fat (i.e. 9434 kcal) and FFM (i.e. 1815 kcal), the composition of weight lost is the main determinant of weight changes in response to a given energy balance [34•, 35, 51]. While decreases in fat mass explained 34 % of weight loss during phase 1, this proportion increased to 64 and 68 % during phase 2. Thus, the energy content per kilogram weight change increased from 4409 kcal during phase 1 to 5209 and 7105 kcal during phase 2 [23••].

Obviously, assessment of AT must follow the dynamics of weight loss. EE should be measured longitudinally before and during the two different phases of weight loss. Unfortunately, this idea has not been addressed in nearly all of the above-mentioned clinical studies on AT where EE had been measured before and after weeks or even after months or after years of weight loss only. During controlled caloric restriction, AT becomes manifest within the first 3 days with no further changes during phase 2 [23••, 40]. This is evidence for the idea that regulation of AT occurs during phase 1. By contrast, the decrease in REE closely followed weight loss during phase 2 (i.e. there was no further mass-independent adaptation in REE). Faced with these dynamics clinical investigations cited above did not address the regulation of AT. All the worse, many clinical studies on weight loss are metabolically uncontrolled, i.e. it remains unclear whether weight-reduced patients have reached a weight stable state or whether they are still losing or already regaining body weight.

AT is considered as an outcome of autoregulatory control that operates to limit weight loss and to restore body composition [3, 14, 22, 52, 53]. During weight loss, there is a considerable inter-individual variance in AT [3, 23••, 44, 54•] (Fig. 1). Taking into account the precision of indirect calorimetry, up to 60 % of subjects experienced a greater-than-expected decline in REE after weight loss [3, 23••]. On the other hand, AT was found to be reproducible [23••]. Thus, AT is considered as an individualised trait.

AT was independent of macronutrient composition of the diet [44‐47]. It was thought to be proportional to the degree of weight loss with an additional effect of baseline weight [47]. However, analysing data on obese patients before and after weight loss [1, 55, 56], the association between weight change and AT was weak (see Fig. 2). This was in line with the results of a controlled clinical study on obese patients using total starvation or different hypocaloric diets for weight loss [48]. By contrast, the starvation-induced fall in REE was closely related to REE before weight loss: The higher baseline REE was, the higher was AT [23••].

In addition, there is an association between AT of the non-REE component of EE and maintenance of reduced body weight [24, 25•, 52]. Most of AT occurred during weight maintenance after a moderate weight loss of −10 % with smaller additional effects induced by a weight loss of −20 % [17, 52]. Recently, Rosenbaum and Leibel proposed three models of AT during weight maintenance [25•]: Model 1, no AT, i.e. a “mechanical model” related to settling of body weight; model 2, fixed AT due to a threshold response (model 2) and model 3, AT is proportional to weight loss. With initial weight loss, both REE and AEE decreased by a fixed amount of kilocalories. This was independent of changes in body composition. With further weight loss, only AEE continued to decrease. Thus, AT related to the resting compartment of TEE followed the threshold model, whereas AT in the non-resting compartment of TEE is proportional to weight loss [25•].

Determinants of AT have to be discussed in the contexts of (i) weight loss and (ii) maintenance of reduced weight.

With weight loss, AT has to be differentiated in relation to (i) the individual component of TEE, (ii) different phases of weight loss and (iii) weight loss maintenance. The first issue is about resting and non-resting EE. The second is about hepatic glycogen mobilisation and mass-independent sparing energy with negative energy balance. The third issue relates to a possible “reset” of the biological defence of body weight after weight loss and thus the risk of weight regain. Figure 3 surveys metabolic adaptation to weight loss and weight maintenance and their determinants with time. We propose to differentiate between three phases of weight loss addressing two different regulatory systems. The first is the immediate control of metabolism in response to negative energy balance (phase 1), whereas the second regulatory system impacts weight maintenance after weight reduction (phase 3). Between the two regulations (i.e. during phase 2), changes in EE are proportional to weight change, and there is no further increase in AT with ongoing weight loss.

During negative energy balance, mitochondrial carbon load decreases and ATP demand is met by the mobilisation of endogenous sources, i.e. glycogen and triglycerides. Losses in functional body mass and AT add to reduce ATP demand. Metabolically, the paramount characteristic of phase 1 is the depletion of glycogen stores, whereas in phase 3, AT is triggered by the loss in body fat. As far as set points (i.e. feedback systems designed to match a target) and settling points (i.e. control systems without a set point) in body weight regulation are concerned [76, 77], the data give rise to the idea that (i) AT is part of body weight regulation and (ii) there are two set points with a settling between them. The first set point is about the decrease in glycogen stores only. We assume that in early starvation, lower thresholds of (i) liver glycogen or (ii) negative fluid balance associated with glycogen depletion trigger AT. It is tempting to speculate that this early response is related to energy needs of the brain (i.e. brains metabolism requires 80 to 100 g glucose per day). By contrast, AT does not occur after glycogen depletion in response to an isocaloric ketogenic diet and moderate weight loss [78]. This may be explained by concomitant increases in plasma ketones providing an alternative fuel to the brain and, thus, put the glycogen threshold into perspective. During weight maintenance, a lower margin of the leptin-fat mass-association is the second set point keeping energy expenditure low to prevent triglyceride stores getting too low which again would limit basic biological functions (e.g. reproduction).

Future biomedical research on the physiology, genetics, cellular and/or molecular aspects of body weight regulation and AT needs robust concepts taking into account conceptual issues, advanced methodological approaches, the dynamics of weight change and in depth physical and metabolic phenotyping. We still lack an integrated understanding of AT. It is also worthwhile to remember that our present knowledge about AT in humans is based on a small number of controlled experimental studies (including the nearly 70 years old Minnesota Starvation Experiment; 5) as well as a greater number of more or less uncontrolled clinical observations. Both approaches do not address the “real” physiology of weight changes which may take several years to obtain a new steady body weight [79].