Energy intake

The macronutrients, i.e. carbohydrate, protein and fat in addition to alcohol, yield energy. The energy content of food was traditionally measured using a bomb calorimeter by calculating total heat liberated under combustion. The result of which is referred to as gross energy value, a value that varies among the macronutrients^{(Reference Schutz, Allen and Prentice15)}. However, not all ingested food is completely absorbed, with approximately 5–10 % of gross energy lost as faecal matter and through urinary excretion. The remaining ‘metabolisable energy’ (ME), expressed per gram of dietary substrate, is available for use by the body^{(Reference Schutz, Allen and Prentice15)}. The metabolisable energy of carbohydrate, protein, fat and alcohol is 17 kJ/g (4 kcal/g), 17 kJ/g (4 kcal/g), 37 kJ/g (9 kcal/g) and 29 kJ/g (7 kcal/g), respectively^{(Reference Atwater and Woods17)}, with an additional energy factor of 8⋅0 kJ/g (2 kcal/g) for dietary fibre⁽¹⁸⁾.

Energy expenditure

Total energy expenditure (TEE) can be split into three conventional components;

1. 1. Resting energy expenditure (REE)

2. 2. Diet-induced thermogenesis (DIT)

3. 3. Physical activity energy expenditure (PAEE) – of which there are two subcategories

   1. a) Exercise activity thermogenesis (EAT)

   2. b) Non-exercise activity thermogenesis (NEAT)

REE refers to the energy required by the body in a resting condition^{(Reference McMurray, Soares and Caspersen19)}, i.e. the ‘metabolic cost of processes such as the maintenance of transmembrane ion gradients and resting cardiopulmonary activity’ ^{(Reference Leibel, Rosenbaum and Hirsch20)}. It is measured under standardised conditions, when the individual is awake, at rest, lying in a supine position and in a thermo-neutral environment^{(Reference Schutz, Allen and Prentice15)}. While REE and BMR are often used interchangeably, REE is more routinely used in research and practice. It is measured exclusively in a post-absorptive state, typically 10–12 h after the last meal and at normal room temperature^{(Reference Schutz, Allen and Prentice15)}. REE represents the largest component of TEE, contributing approximately 60–70 % of TEE^{(Reference Ho16,Reference Leibel, Rosenbaum and Hirsch20)} .

DIT refers to the energy required by the body in the post-prandial period, representing the energy cost of digestion, absorption, transport and storage of dietary nutrients^{(Reference Ho16,Reference Westerterp21)} . It is calculated by dividing the increase in EE above basal fasting level by the energy content of the food ingested^{(Reference Westerterp21)}. Although DIT is a product of EI, it belongs as a component of TEE, equivalent to approximately 5–15 % of total energy consumption, assuming an individual is at or near energy balance^{(Reference Ho16,Reference Westerterp21)} .

Finally, PAEE refers to the additional energy required by the body for movements produced by the skeletal muscle^{(Reference Westerterp22)}. It is subdivided into EAT and NEAT, with EAT representing energy expended through intentional moderate-vigorous exercise, and NEAT representing energy expended as a consequence of daily living and vocation, including low-intensity daily activities above rest (e.g. sitting, standing and walking) and more subtle spontaneous physical activities such as fidgeting^{(Reference Chung, Park and Kim23)}. There is no gold standard for measuring PAEE, with estimates often derived from TEE and REE, or expressed as a factor of BMR or REE, e.g. using physical activity level index^{(Reference Hills, Mokhtar and Byrne24)}. Nevertheless, PAEE is by far the most variable component of TEE, both within and between individuals^{(Reference Chung, Park and Kim23)} typically contributing between 15 and 40 % of TEE^{(Reference Schutz, Allen and Prentice15,Reference Butte, Ekelund and Westerterp25,Reference Westerterp26)} .

Factors determining obligatory energy expenditure

Body composition is the primary determinant of REE, explaining 60–90 % of the inter-individual variability^{(Reference Hall and Guo1,Reference Nielsen, Hensrud and Romanski27,Reference Illner, Brinkmann and Heller28)} . Elia^{(Reference Elia, Kinney and Tucker29)} measured the specific REE of different body tissues, referred to as Ki values (expressed as kcal/kg daily). Fat-free mass (FFM) has a significantly higher metabolic rate than fat mass (FM). While metabolic organs and skeletal muscle have Ki values of 200–400 and 13, respectively, adipose tissue has a Ki value of 4⋅5^{(Reference Wang, Ying and Bosy-Westphal30)}. Therefore, an individual with a greater proportion of FFM (comprising muscle and organs) will have a higher REE than height- and weight-matched individuals with a greater proportion of FM. Accordingly, REE is also determined by body size, with a larger body size indicating more metabolically active tissue and higher energy requirements than a smaller body size, despite the same proportional body composition^{(Reference Hills, Mokhtar and Byrne24)}.

Sex differences in REE are mainly attributed to differences in body composition. Females generally have approximately 10–15 % higher body fat^{(Reference Elia, Kinney and Tucker29,Reference Schorr, Dichtel and Gerweck31–Reference Jackson, Stanforth and Gagnon33)} and 5–10 % lower REE than BMI-matched males^{(Reference Hills, Mokhtar and Byrne24)}. Such differences in body composition are suggested to be influenced by sex hormones with oestrogen reducing lipid oxidation and promoting fat deposition in females^{(Reference Brown and Clegg34–Reference O'Sullivan, Crampton and Freund36)} and testosterone promoting muscle protein synthesis in males^{(Reference Griggs, Kingston and Jozefowicz37,Reference Sheffield-Moore38)} .

Nevertheless, in both males and females REE has been reported to decrease by 1–2 % per decade^{(Reference Elia, Ritz and Stubbs39,Reference Molnar and Schutz40)} , due to age-related decreases in FFM^{(Reference Geisler, Braun and Pourhassan41)} and increases in overall adiposity^{(Reference Jura and Kozak42,Reference Karastergiou, Smith and Greenberg43)} . Short et al.^{(Reference Short, Vittone and Bigelow44)} reported that skeletal muscle contributes only 25 % of total weight in 75–80-year-old adults as opposed to 50 % in young adults. This decline in FFM, however, is determined by changes in sex hormones. Males reach peak FFM at 31–40 years, followed by a rapid decrease due to significant declines in testosterone^{(Reference Bredella and Mauvais-Jarvis45)}. In contrast, females start losing FFM 10 years later than males, and to a lesser degree, possibly due to protective anti-inflammatory effects of oestrogen that do not attenuate until the onset of menopause^{(Reference Hills, Mokhtar and Byrne24)}.

While body composition accounts for the most observed variation in REE, research suggests that residual variability may be explained by differences in the amount and distribution of organ tissue, which range in metabolic activity (836⋅8 kJ/kg (200 kcal/kg) daily for liver, 1004⋅16 kJ/kg (240 kcal/kg) daily for brain and 1840⋅96 kJ/kg (440 kcal/kg) daily for heart and kidneys)^{(Reference Elia, Kinney and Tucker29,Reference Gallagher, Belmonte and Deurenberg46,Reference Garby and Lammert47)} . Despite accounting for <6 % of total weight^{(Reference Elia, Kinney and Tucker29)}, metabolic organs contribute 60–80 % of REE^{(Reference Heymsfield, Gallagher and Kotler48,Reference Gallagher, Visser and Wang49)} , meaning even small variations among individuals may influence REE.

Whilst energy content of food is the primary determinant of the obligatory energy cost of DIT, values are specific to each macronutrient due to differing ATP requirements for the initial steps of metabolism and storage^{(Reference Westerterp21)}. As a consequence, macronutrient composition also determines DIT. Fat has the lowest DIT value estimated between 0 and 3 % of ingested intake, followed by carbohydrate with a value of 5–10 %. Protein has the highest DIT value estimated between 20 and 30 %, in addition to alcohol with a value of 10–30 %^{(Reference Westerterp21)}. In healthy-weight individuals, in energy balance and consuming a mixed diet, DIT accounts for approximately 10 % of energy ingested over 24 h^{(Reference Westerterp21)}.

There is currently limited evidence suggesting an association between obesity and DIT. Early research by Wang et al.^{(Reference Wang, Strouse and Saunders50)} identified lower DIT in obese subjects compared to lean subjects. This is consistent with findings from a critical review by De Jonge and Bray^{(Reference De Jonge and Bray51)} where 22/29 studies reported a significantly reduced DIT in obese subjects, associated with insulin resistance and reduced postprandial sympathetic response^{(Reference De Jonge and Bray51,Reference Weststrate52)} .

Some studies report that DIT normalises in weight-reduced subjects^{(Reference Schwartz, Halter and Bierman53,Reference Thorne, Naslund and Wahren54)} , suggesting reduced postprandial response is a consequence rather than a cause of obesity. However, others report that DIT remained suppressed in weight-reduced subjects, suggesting that reduced postprandial response contributes to the development of obesity^{(Reference Bessard, Schutz and Jéquier55–Reference Schutz, Golay and Felber57)}. However, several studies report no association between obesity and DIT^{(Reference Granata and Brandon58)}. Such association is further confounded by variation in methodology, energy and macronutrient content of test foods, duration of the postprandial periods and inaccuracy calculating DIT from REE and PAEE^{(Reference Westerterp21,Reference Granata and Brandon58)} . Currently, while reduced postprandial response in obesity seems plausible, further standardisation and validation of experimental protocol is needed to reach a consensus.

For PAEE, both EAT and NEAT are determined by the metabolic cost and the frequency of body movement, both of which are largely influenced by bodyweight. In turn, larger individuals have a higher energy cost of movement compared to smaller individuals, however they can also be behaviourally less active^{(Reference Westerterp22,Reference Westerterp59,Reference Schoeller and Jefford60)} . Other suggested determinants of PAEE include age, exercise training, genetics, EI and disease^{(Reference Westerterp22)}.

Assumption of current weight-loss strategies

Clinical weight-loss prescriptions assume that 14644 kJ (3500 kcal) is equivalent to 1 pound of fat (or about 32⋅5 MJ is equivalent to 1 kg)^{(Reference Thomas, Gonzalez and Pereira61)} translating into advice that a 2092 kJ (500 kcal) deficit daily will result in 1 pound (lb) weight-loss per week. The ‘3500 kcal (14644 kJ) rule’ is based on the findings of researcher Max Wishnofsky who reported that 1 lb of fat stores approximately 3500 kcal (14⋅6 MJ) of energy^{(Reference Wishnofsky62)}. This observation rests on the assumption that weight-loss is composed of 25 % FFM and 75 % FM^{(Reference Burckhardt, Dawson-Hughes and Weaver63)}, a concept based on observations from the Minnesota starvation experiment^{(Reference Keys, Brozek and Henschel64)}. Further simplified assumptions assume FFM consists of about 75 % water (0 kJ/g (0 kcal/g)) and about 25 % protein (16⋅74 kJ/g (4 kcal/g)), meaning 1 g of FFM stores 4⋅184 kJ (1 kcal) of energy, while FM consists of 100 % fat (37⋅66 kJ/g (9 kcal/g)), meaning 1 g of FM stores 37⋅66 kJ (9 kcal) of energy^{(Reference Atwater and Bryant65)}. Based on this assumption, 1 g of total weight-loss is equivalent to 29⋅29 kJ (7 kcal), hence 1 kg is equivalent to 29288 kJ (7000 kcal) and 0⋅5 kg is equivalent to 14644 kJ (3500 kcal).

However, this approach assumes the composition of weight lost as FM and FFM is fixed and remains constant throughout the period of dynamic weight loss. Additionally, it disregards dynamic changes in EE observed when the body is in a negative energy balance, resulting in the significant overprediction of weight-loss^{(Reference Thomas, Gonzalez and Pereira61)}. Despite recognised as over-simplistic, the 3500 kcal (14644 kJ) rule continues to appear in scientific literature and has been cited in over 35 000 educational weight-loss websites^{(Reference Thomas, Martin and Lettieri66)}. It is observed in recommendations by the National Health Services⁽⁶⁷⁾, British Dietetics Association⁽⁶⁸⁾, National Institutes of Health⁽⁶⁹⁾ and American Dietetic Association^{(Reference Seagle, Strain and Makris70)}. By way of illustration, Lin et al.^{(Reference Lin, Smith and Lee71)} demonstrated the bias of the 3500 kcal (14644 kJ) rule in the development of population obesity intervention strategies, where static modelling overestimated weight-loss associated with the sugar-sweetened beverages tax by 63, 346 and 764 % at year one, five and ten, respectively.

Popular models of weight-loss

Obligatory changes in energy expenditure

Foremost, the weight-loss-induced decline in REE is primarily due to loss of metabolically active tissue, i.e. skeletal muscle^{(Reference Muller, Enderle and Bosy-Westphal77)}, which expends approximately 54⋅4 kJ/kg (13 kcal/kg) daily^{(Reference Wang, Ying and Bosy-Westphal30)}. This obligatory decrease in EE represents that considered in a settling point model of weight-loss, by which energy deficit decreases as weight decreases, resulting in a new energy balance at a lower bodyweight.

The widely cited quarter FFM rule^{(Reference Heymsfield, Cristina Gonzalez and Shen78)} states that FFM, i.e. glycogen, protein and water accounts for 25 % of total weight-loss while FM accounts for the remaining 75 %. Despite having ‘limited mechanistic basis’^{(Reference Heymsfield, Cristina Gonzalez and Shen78)}, the quarter FFM rule is considered the best approximation of body composition changes in response to underfeeding. However, this rule still incorrectly assumes that the proportion of weight lost as FM and FFM is constant between individuals and during weight-loss.

Early findings from Grande and Henschel^{(Reference Grande and Henschel79)} revealed that the composition of weight-loss differs in the early and late stages of energy restriction, with early weight-loss composed of predominantly water (70 %), some fat (25 %) and little protein (5 %), and later weight-loss composed of predominantly fat (85 %), some protein (15 %) and no water (0 %). The Minnesota starvation experiment^{(Reference Keys, Brozek and Henschel64)} reported similar findings, where weight-loss was composed of about 40 % FM in weeks 1–12, increasing to about 70 % in weeks 12–24.

The rapid rate of early weight-loss is largely attributed to water and glycogen^{(Reference Heymsfield, Thomas and Nguyen80)}. Liver and skeletal muscle glycogen stores are mobilised into circulation by glycogenolysis to provide short-term energy when external energy sources, i.e. food, cannot meet demands^{(Reference Heymsfield, Thomas and Nguyen80–Reference Rui83)}. Glycogen is stored in a hydrated form, with each gram stored with 3–4 g of water^{(Reference Kreitzman, Coxon and Szaz84)}. Once mobilised, the associated water is excreted in urine^{(Reference Murray and Rosenbloom85)}. However, glycogen stores are largely depleted within a week of even moderate energy restriction^{(Reference Heymsfield, Thomas and Nguyen80)}.

Before complete depletion of body glycogen stores, there is a shift from glucose oxidation to fatty acid oxidation. Ketone bodies are used as a glucose substitute through conversion from a surplus of fatty acid-derived acetyl CoA in the liver via ketogenesis^{(Reference Rui83)}. Amino acids are also used as a glucose source for the brain and peripheral tissues through hydrolysis of skeletal muscle and conversion to glucose in the liver, via gluconeogenesis^{(Reference Heymsfield, Thomas and Nguyen80,Reference Rui83,Reference Kerndt, Naughton and Driscoll86,Reference Berg, Tymoczko and Stryer87)} . However, increasing availability of ketone bodies during prolonged and significant underfeeding lessens the demand for amino acids, therefore the proportion of weight lost as metabolically active tissue often assumes this lower stable level for the duration of dieting period^{(Reference Heymsfield, Cristina Gonzalez and Shen78,Reference Forbes and Drenick88)} . The composition of weight lost as FM and FFM can be influenced further by the degree of energy restriction, protein intake, magnitude of weight-loss, baseline adiposity and physical activity level^{(Reference Heymsfield, Thomas and Nguyen80)}.

Alongside obligatory changes in REE due to loss of FFM, obligatory decreases in DIT are observed in response to underfeeding due to reduced EI, as less energy is required in the ingestion, digestion, absorption, metabolism, transport and storage of food and nutrients^{(Reference Trexler, Smith-Ryan and Norton89)}. Assuming a healthy mixed diet, a 2092 kJ (500 kcal) energy deficit would reduce TEE by about 104⋅6–313⋅8 kJ/d (25–75 kcal/d).

Finally, an obligatory decrease in PAEE is also observed in response to underfeeding, in both EAT and NEAT compartments, which is proportional to overall weight-loss^{(Reference Trexler, Smith-Ryan and Norton89)}. This is due to a reduced metabolic cost of movement (i.e. a reduction in ‘ballast’), where 5 % weight-loss has been associated with a 393⋅3 kJ (94 kcal) daily reduction in PAEE^{(Reference Muller, Enderle and Pourhassan90)}.

Adaptive changes in energy expenditure

Energy restriction has been associated with a decline in REE, exceeding that explained by changes in body composition alone^{(Reference Jaime, Balich and Acevedo91)}. Weight-loss studies have shown that the magnitude of fat stores in the body is protected by mechanisms mediated by the central nervous system, which adjust EI and EE through signals from adipose tissue, the gastrointestinal tract and endocrine tissue to maintain homeostasis and resist weight change^{(Reference Rosenbaum and Leibel92)}. The body's protective metabolic mechanism that attempts to preserve energy stores whilst in energy crisis is known as adaptive thermogenesis (AT). AT is defined as the underfeeding-associated fall in REE independent of changes in FFM and FM^{(Reference Muller, Enderle and Pourhassan90,Reference Dulloo and Jacquet93)} . This definition is based on findings from the Minnesota starvation experiment^{(Reference Keys, Brozek and Henschel64)}, where a 50 % energy restriction was associated with a 39 % or about 2510⋅4 kJ/d (600 kcal/d) decline in REE, 35 % (or about 836⋅8 kJ/d (200 kcal)) of which was independent of obligatory FFM loss^{(Reference Muller, Enderle and Pourhassan90)}. AT can be estimated by calculating the decrease in mass-adjusted REE in response to underfeeding, i.e. the difference between measured REE and predicted REE post-intervention^{(Reference Dulloo, Jacquet and Montani94)}. However, some studies have extended this definition to include DIT in response to both underfeeding^{(Reference Dulloo95,Reference Lowell and Spiegelman96)} and overfeeding^{(Reference Lowell and Spiegelman96,Reference Tseng, Cypess and Kahn97)} and cold-induced thermogenesis in response to changes in environmental temperature^{(Reference Lowell and Spiegelman96,Reference Tseng, Cypess and Kahn97)} . The inconsistent definition of AT makes the quantification of metabolic adaptation challenging.

Research to date suggests that AT can explain half of the unsatisfactory weight-loss cases, where weight-loss was significantly less than that predicted by loss of FFM alone^{(Reference Goele, Bosy-Westphal and Rumcker98,Reference Byrne, Wood and Schutz99)} . For example, a 10 % weight reduction has been associated with a 20–25 % reduction in TEE, 10–15 % beyond that predicted by changes in body composition^{(Reference Rosenbaum and Leibel92)}.

Cross-sectional studies have investigated AT by comparing formerly-obese subjects who had lost weight, with BMI-matched subjects who were never obese. A meta-analysis by Astrup et al.^{(Reference Astrup, Gøtzsche and van de Werken100)} reported a 3–5 % lower REE in formerly-obese subjects compared to never-obese controls. However, several cross-sectional studies failed to detect AT^{(Reference Wyatt, Grunwald and Seagle101,Reference Weinsier, Hunter and Zuckerman102)} , likely due to large inter-subject variability in body composition and REE^{(Reference Martins, Roekenes and Salamati103)}.

Longitudinal weight-loss studies have provided a more accurate method of investigating metabolic adaptation, where AT of clinical significance has been detected in both lean^{(Reference Keys, Brozek and Henschel64,Reference Muller, Enderle and Pourhassan90,Reference Weyer, Walford and Harper104)} and overweight/obese subjects^{(Reference Leibel, Rosenbaum and Hirsch20,Reference Heilbronn, de Jonge and Frisard105,Reference Bosy-Westphal, Kossel and Goele106)} . In most cases, 10–20 % weight-loss is associated with AT equivalent to 418⋅4–1255⋅2 kJ/d (100–300 kcal/d)^{(Reference Leibel, Rosenbaum and Hirsch20,Reference Keys, Brozek and Henschel64,Reference Muller and Bosy-Westphal107,Reference Doucet, St-Pierre and Alméras108)} . Based on such evidence, a formerly-obese individual will theoretically require 418⋅4–1255⋅2 kJ (100–300 kcal) fewer daily for weight maintenance compared to never-obese individual of the same weight and body composition.

However, reductions as much as 2092 kJ/d (500 kcal/d) have been detected, suggesting large inter-individual variability. Such a case was observed in a weight-loss study at Laval University^{(Reference Tremblay, Major and Doucet109)} where a woman adhering to a 2092 kJ/d (500 kcal/d) energy deficit for 15 weeks had a resultant weight gain of 2⋅1 kg, despite strict compliance and close nutritional support. This clinical paradox can be largely explained by indirect calorimetry measurements, which revealed a 552 kcal daily decrease in REE at the end of the weight-loss phase.

Nevertheless, there is inconsistent evidence regarding the onset of metabolic adaptation. Heinitz et al.^{(Reference Heinitz, Hollstein and Ando110)} detected AT within a week of energy restriction, associated with the rapid declines in insulin secretion, depletion of glycogen stores and loss of intra- and extracellular fluid. This aligns to resultant alterations in glycolytic and oxidative activity reported to induce metabolic slowing, with the primary aim of ensuring the brain's energy needs are met^{(Reference Muller, Enderle and Bosy-Westphal77)}. Muller et al.^{(Reference Muller, Enderle and Pourhassan90)} reported similar findings with metabolic adaptation detected after 3 d of energy restriction. The magnitude of the observed AT closely correlated with reductions in insulin secretion, changes in glucose oxidation, fluid balance and free water clearance rate^{(Reference Muller, Enderle and Pourhassan90)}.

In contrast, substantial evidence suggests that underfeeding-associated AT takes weeks to develop^{(Reference Dulloo, Seydoux and Jaquet111,Reference Cannon and Nedergaard112)} and is associated with lower sympathetic nervous system activity, triiodothyronine and leptin^{(Reference Rosenbaum and Leibel92,Reference Muller and Bosy-Westphal107,Reference Rosenbaum, Goldsmith and Bloomfield113)} . This delayed onset of metabolic adaptation is reported to be triggered by signals from depleted adipocytes with the primary aim of preserving TAG stores and preventing loss of basic biological function, e.g. reproduction^{(Reference Muller, Enderle and Bosy-Westphal77)}. Such findings support the possible existence of two components of AT, an immediate metabolic adaptation associated with decreased insulin and carbohydrate availability, and a delayed adaptation associated with decreased leptin secretion from depleted adipose tissue stores.

There is, however, conflicting evidence regarding the persistence of metabolic adaptation. While some research suggests that underfeeding-associated AT can be reversed within 2 weeks of refeeding or 4 weeks of weight stability at energy balance^{(Reference Leibel, Rosenbaum and Hirsch20,Reference Martins, Roekenes and Salamati103)} , others report that the effects of AT are long-term, being still detectable 6 months to 1-year post-surgical^{(Reference Butte, Brandt and Wong114–Reference van Gemert, Westerterp and van Acker116)} and diet-induced weight-loss^{(Reference Keys, Brozek and Henschel64,Reference Rosenbaum, Hirsch and Gallagher117)} and even up to 6 years post weight-loss^{(Reference Fothergill, Guo and Howard118)}.

Mechanisms of adaptive thermogenesis

Skeletal muscle and brown adipose tissue have been identified as important sites of thermogenesis regulation^{(Reference Dulloo and Samec119)} using uncoupling proteins, proton leakage and substrate cycling^{(Reference Norwack, Giroud and Arnold120)} to alter EE in response to changes in the external environment.

Such mechanisms increase the body's capacity to dissipate energy, with an established role of brown adipose tissue and skeletal muscle in non-shivering^{(Reference Busiello, Savarese and Lombardi121,Reference Roesler and Kazak122)} and shivering thermogenesis^{(Reference Roesler and Kazak122,Reference Haman and Blondin123)} under conditions of chronic cold exposure.

Animal studies have also supported a role of brown adipose tissue thermoregulation as a means of energy dissipation in response to chronic overfeeding^{(Reference Rothwell and Stock124–Reference Feldmann, Golozoubova and Cannon126)}. However, these animal observations are not consistent with findings from short-term^{(Reference Schlogl, Piaggi and Thiyyagura127–Reference Vosselman, Brans and van der Lans129)} or long-term human studies^{(Reference Peterson, Orooji and Johnson130)}, where no change in brown adipose tissue activity was observed after overfeeding, despite a greater-than-predicted increase in REE.

In human subjects, it has been proposed that underfeeding-associated AT is predominantly mediated by thrifty mechanisms specific to skeletal muscle which downregulate thermogenesis, particularly in response to signals from adipose tissue (Fig. 1)^{(Reference Dulloo, Jacquet and Seydoux131)}.

Fig. 1. Schematic diagram illustrating direct (—) and indirect (− −) pathways for adaptive thermogenesis, triggering thrifty mechanisms specific to the skeletal muscle. During periods of energy restriction, leptin secretion in the adipose tissue decreases due to reduced TAG stores. Also, a reduction in plasma insulin is observed secondary to restricted dietary intake. Such hormones directly downregulate substrate cycling in the skeletal muscle. Additionally, both leptin and insulin indirectly reduces skeletal muscle thermogenesis through suppression of the sympathetic nervous system (SNS) and thyroid gland, and subsequent triiodothyronine (T3), and norepinephrine (NE) production.

The skeletal muscle is the primary site of a thermogenic effector system^{(Reference Dulloo, Jacquet and Seydoux131)}. This system is orchestrated by substrate cycling between lipid oxidation and lipogenesis, and regulated by hormones including insulin, leptin, triiodothyronine and norepinephrine^{(Reference Dulloo, Jacquet and Seydoux131)}.

Leptin is secreted by adipocytes in adipose tissue in proportion to existing FM^{(Reference Rosenbaum, Nicolson and Hirsch132)}. During periods of energy restriction, depleted TAG stores result in reduced leptin production, which directly downregulates substrate cycling in the skeletal muscle. Additionally, leptin indirectly downregulates skeletal muscle thermogenesis through suppression of the sympathetic-thyroid axis, with reduced norepinephrine and triiodothyronine production having similar regulatory effects on substrate cycling^{(Reference Dulloo, Jacquet and Seydoux131)}.

Insulin is secreted by the pancreas in response to elevated blood glucose concentrations, hence during periods of energy restriction, lower dietary carbohydrate intake results in reduced insulin production, which is reported to have similar direct and indirect effects on substrate cycling and thermogenesis in the skeletal muscle^{(Reference Dulloo, Jacquet and Seydoux131)}. This suggests that early alterations in glycolytic activity are one explanation for the proposed immediate onset of AT.

However, the skeletal muscle is a major glucose consumer and a primary site for glucose metabolism, meaning suppressed thermogenesis will result in reduced glucose utilisation during periods of refeeding. The resulting hyperinsulinemia will cause spared glucose to be redistributed for lipogenesis (tri-acylglyceride storage) in adipose tissue. This phenomenon, referred to as ‘catch-up fat’, is characterised by a disproportionate rate of FM recovery relative to FFM^{(Reference Dulloo, Jacquet and Seydoux131)}. This preferential restoration of FM has been observed in several influential weight-loss studies^{(Reference Weyer, Walford and Harper104,Reference Debray, Zarakovitch and Ranson133)} including the Minnesota starvation experiment^{(Reference Keys, Brozek and Henschel64)}, where FM exceeded prestarvation values by over 75 % after refeeding^{(Reference Dulloo, Jacquet and Girardier134)}.

Factors determining adaptive thermogenesis