We are currently in the midst of a global obesity epidemic, with its prevalence more than doubling since 1980( 1 ). The development of obesity is closely associated with a number of metabolic complications including insulin resistance and dyslipidaemia, which can in turn increase an individual's risk of developing type 2 diabetes mellitus (T2DM) and CVD( Reference Leiter, Fitchett and Gilbert 2 ). Glucose and lipid homeostasis can be improved through modest weight-loss, which due to current accepted dietary advice, is most commonly achieved via a modest continuous (daily) energy restriction (CER)( 3 ). Independently of weight-loss, the metabolic regulation of glucose and lipids can also be influenced by other dietary manipulations, including alterations in meal timing( Reference Potter, Cade and Grant 4 ) as well as abrupt changes in energy status such as fasting, which when protracted, elicits profound shifts in fuel utilisation that persist during the subsequent refeeding period( Reference Antoni, Johnston and Collins 5 ).
In recent years, several intermittent fasting variants, including intermittent energy restriction (IER) and time-restricted feeding (TRF) have received considerable interest as alternative dietary strategies for weight-management and improving metabolic health. With these dietary strategies, the pattern of energy restriction and/or timing of food intake are altered, so that individuals undergo frequently repeated periods of fasting (Table 1). Their basic premise compared to CER is that an individual may not need to energy restrict either every day, or at all, to attain metabolic benefits. With these points in mind, the current review aims to provide a commentary on studies within the intermittent fasting literature, with a specific focus on the effects of IER and TRF on glucose and lipid metabolism. Where possible, the relative contribution of weight-loss/energy restriction and the intermittent fasting eating pattern per se to observed metabolic changes will be discussed. In addition, the review highlights potential avenues for future research.
Overview of weekly fasting schedule for the most commonly studied intermittent fasting protocols. ‘Fast’ is used to denote periods of substantial (total or partial, ⩾70%) energy restriction. Intermittent energy restriction (IER) protocols involving more modest energy restriction, or IER–refeeding cycles not included in the review
Intermittent energy restriction
Overview and effects on body weight
The majority of rodent studies( Reference Pedersen, Hagemann and Bock 6 – Reference Wan, Ahmet and Brown 21 ) and a small number of human studies( Reference Halberg, Henriksen and Söderhamn 22 – Reference Catenacci, Pan and Ostendorf 26 ) have used IER protocols, which completely restrict energy intake (i.e. 100 % energy restriction) every other day, with fasting intervals ranging between 20 and 36 h. However, the long-term sustainability of this alternate day total fasting approach in human subjects is questionable due to the persistent hunger reported( Reference Heilbronn, Smith and Martin 24 ). Subsequently, the IER protocols used by most human studies( Reference Ash, Reeves and Yeo 27 – Reference Zuo, Feng and Tinsley 41 ), and by some rodent studies( Reference Varady, Roohk and Loe 11 , Reference Varady, Allister and Roohk 14 , Reference Varady, Hudak and Hellerstein 42 ), have allowed a small amount of ‘fast’ day intake, so that energy is substantially (⩾70 %) but not completely restricted. This is often referred to as modified fasting, such that, the term fasting in this IER context denotes periods of severe (total or partial) energy restriction. These modified fasting forms of IER have predominantly been utilised as weight-loss strategies, whilst a small number of studies have also used protocols intended to encourage weight maintenance( Reference Harvie, Wright and Pegington 37 , Reference Wegman, Guo and Bennion 39 , Reference Zuo, Feng and Tinsley 41 ). The most well-studied examples include alternate day modified fasting and the 5:2 diet, which entails two modified fast days a week (Table 1), although many variations exist. Intakes on non-restricted (‘feed’) days among these studies have ranged from ad libitum ( Reference Ash, Reeves and Yeo 27 – Reference Varady, Bhutani and Church 29 , Reference Varady, Bhutania and Klempel 31 – Reference Eshghinia and Mohammadzadeh 34 , Reference Varady, Bhutani and Klempel 36 , Reference Hoddy, Kroeger and Trepanowski 38 ), hypoenergetic (approximately 15–30 % of energy requirements)( Reference Eshghinia and Mohammadzadeh 34 , Reference Zuo, Feng and Tinsley 41 ), isoenergetic( Reference Harvie, Pegington and Mattson 30 , Reference Harvie, Wright and Pegington 37 , Reference Harvie, Sims and Pegington 40 ) or hyperenergetic (approximately 125–175 % of energy requirements)( Reference Klempel, Kroeger and Norkeviciute 35 , Reference Wegman, Guo and Bennion 39 ). Compliance is reportedly high( Reference Varady, Bhutani and Church 29 , Reference Harvie, Wright and Pegington 37 ), with both acute( Reference Antoni, Johnston and Collins 5 ) and chronic( Reference Varady, Bhutani and Church 29 , Reference Harvie, Pegington and Mattson 30 , Reference Harvie, Wright and Pegington 37 , Reference Harvie, Sims and Pegington 40 ) studies demonstrating a lack of full compensatory hyperphagia following modified fasting days. In overweight/obese( Reference Catenacci, Pan and Ostendorf 26 – Reference Klempel, Kroeger and Norkeviciute 35 , Reference Harvie, Wright and Pegington 37 , Reference Hoddy, Kroeger and Trepanowski 38 , Reference Harvie, Sims and Pegington 40 , Reference Zuo, Feng and Tinsley 41 ) or combined healthy/overweight( Reference Varady, Bhutani and Klempel 36 ) cohorts, the average reported weight-loss through IER (70–100 % energy restriction) has ranged between approximately 4 and 10 % over dieting periods of 4–24 weeks. There are also some data demonstrating the successful applications of IER as a weight-maintenance strategy following weight-loss for periods of up to 1 year( Reference Harvie, Wright and Pegington 37 , Reference Wegman, Guo and Bennion 39 ).
Effects of intermittent energy restriction on glucose and lipid metabolism
Rodent studies
Although findings from rodent studies cannot be readily extrapolated to human subjects, they do permit a greater level of control over dietary intakes and environmental confounders. Rodent studies of IER have mostly used alternate day total fasting protocols. Over study intervals of 4 weeks to 1 year, this form of IER can reduce levels of fasting glucose( Reference Pedersen, Hagemann and Bock 6 , Reference Anson, Guo and de Cabo 8 , Reference Wan, Camandola and Mattson 9 , Reference Wan, Ahmet and Brown 21 ), insulin( Reference Anson, Guo and de Cabo 8 , Reference Wan, Camandola and Mattson 9 , Reference Higashida, Fujimoto and Higuchi 16 – Reference Gotthardt, Verpeut and Yeomans 18 , Reference Wan, Ahmet and Brown 21 ) and insulin resistance (measured via homeostasis model assessment of insulin resistance (HOMA-IR))( Reference Higashida, Fujimoto and Higuchi 16 , Reference Baumeier, Kaiser and Heeren 17 ), which in turn reflects improved hepatic insulin sensitivity. Additionally, lipid profiles are also favourably altered in wild-type rodents( Reference Krizova and Simek 7 , Reference Tikoo, Tripathi and Kabra 10 , Reference Higashida, Fujimoto and Higuchi 16 ), with similar observations made by one alternate day modified (85 % energy restriction) fasting study( Reference Varady, Hudak and Hellerstein 42 ). Findings from a number studies, which have performed more dynamic assessments of glycaemic indices, through use of gold-standard hyperinsulinaemic–euglycaemic glucose clamp techniques to assess insulin sensitivity or glucose/insulin tolerance testing, have been mixed. Several studies have reported improvements in glycaemic control in rodents fed normal or low-fat chow( Reference Jiang, Koike, Li and Wang 13 , Reference Boutant, Kulkarni and Joffraud 19 ), whilst some( Reference Baumeier, Kaiser and Heeren 17 , Reference Yang, Cao and Mao 20 ) (but not all( Reference Higashida, Fujimoto and Higuchi 16 )) studies have shown that IER is able to protect against high-fat diet-induced insulin resistance. In contrast, one study conducted in obesity-prone Sprague–Dawley rats noted a marked impairment in glucose tolerance after 8 months of IER (normal chow)( Reference Cerqueira, da Cunha and Caldeira da Silva 15 ).
Human studies: alternate day total fasting
A small number of studies have assessed the metabolic impacts of alternate day total fasting using a range of experimental techniques. Most recently, an 8-week pilot study completed by twenty-six obese male and female participants reported an average weight-loss of 9 %, accompanied by beneficial reductions in LDL-cholesterol and fasting TAG( Reference Catenacci, Pan and Ostendorf 26 ). The earlier 2( Reference Halberg, Henriksen and Söderhamn 22 , Reference Soeters, Lammers and Dubbelhuis 25 )–3( Reference Heilbronn, Civitarese and Bogacka 23 , Reference Heilbronn, Smith and Martin 24 )-week studies conducted in small groups (n ⩽ 16) of healthy and overweight participants were not originally designed as weight-loss studies, although participants on the 3-week study struggled to sustain prescribed energy intakes on feed days, and so incurred a modest weight-loss( Reference Heilbronn, Smith and Martin 24 ). No post-treatment changes in fasting levels of insulin and glucose (when measured after a feed day) were observed by these studies( Reference Halberg, Henriksen and Söderhamn 22 , Reference Heilbronn, Civitarese and Bogacka 23 , Reference Soeters, Lammers and Dubbelhuis 25 ), whereas one study reported improved lipid profiles( Reference Heilbronn, Smith and Martin 24 ). Two studies used hyperinsulinaemic–euglycaemic glucose clamp techniques to assess changes in glycaemic control, yielding conflicting results. The hyperinsulinaemic–euglycaemic glucose clamp method involves concomitant infusions of insulin (to achieve hyperinsulinaemia) and glucose (to achieve a constant basal level of plasma glucose). Under steady-state euglycaemic conditions, the glucose infusion rate equals glucose uptake by all tissues and is therefore a measure of tissue sensitivity to exogenous insulin. Halberg et al.( Reference Halberg, Henriksen and Söderhamn 22 ) reported an increase (improvement) in insulin-mediated whole-body glucose uptake after 2 weeks in healthy men (BMI 25·7 (sd 0·4) kg/m2). In contrast, Soeters et al.( Reference Soeters, Lammers and Dubbelhuis 25 ), the only group among these earlier studies to have used a (cross-over) controlled study design, found no change in insulin sensitivity in lean men following the same duration of time. In a 3-week study by Heilbronn et al.( Reference Heilbronn, Civitarese and Bogacka 23 ), which was conducted in a mixed sex healthy/overweight cohort, significant post-treatment reductions in postprandial insulin responses to a test-meal challenge was noted among male participants, whereas there was a contrasting decline in glucose tolerance among female participants (with no change in the insulinaemic response). On closer examination of the study design, post-treatment metabolic assessments were performed following a fast day (i.e. after 36 h of fasting), whereas baseline assessments were conducted after a shorter 12-h overnight fast. As such, the unstandardised fasting periods makes it difficult to ascribe these findings as a true chronic treatment effect, given that prolonged fasting can also acutely alter postprandial substrate responses during the subsequent refeeding period( Reference Antoni, Johnston and Collins 5 ). Ultimately, the uncontrolled study designs represent a significant limitation of most of these earlier studies.
Human studies: modified fasting
To date, all IER studies allowing fast day intake have solely measured fasting blood markers of cardiometabolic risk. Some( Reference Ash, Reeves and Yeo 27 , Reference Harvie, Pegington and Mattson 30 – Reference Bhutani, Klempel and Kroeger 33 , Reference Klempel, Kroeger and Norkeviciute 35 – Reference Harvie, Wright and Pegington 37 , Reference Zuo, Feng and Tinsley 41 ) but not all of these studies( Reference Johnson, Summer and Cutler 28 , Reference Varady, Bhutani and Church 29 , Reference Eshghinia and Mohammadzadeh 34 , Reference Harvie, Sims and Pegington 40 ) have included CER (standard treatment) or ad libitum (no intervention) control groups, whilst others have compared two or more different IER protocols( Reference Klempel, Kroeger and Norkeviciute 35 , Reference Hoddy, Kroeger and Trepanowski 38 , Reference Wegman, Guo and Bennion 39 ). In individuals with T2DM, IER has been shown to improve glycaemic control and lipid profiles( Reference Ash, Reeves and Yeo 27 ). By contrast, many of the IER studies within heterogeneous non-T2DM populations, which have spanned from 4–24 weeks, have failed to show any significant effect on fasting glucose levels( Reference Johnson, Summer and Cutler 28 , Reference Eshghinia and Mohammadzadeh 34 , Reference Harvie, Wright and Pegington 37 – Reference Harvie, Sims and Pegington 40 ) or glycated Hb( Reference Harvie, Wright and Pegington 37 ); unsurprising findings given that glucose levels are usually kept under tight physiological control. However, improvements in other indices of insulin sensitivity have been observed, including reductions in fasting insulin( Reference Harvie, Pegington and Mattson 30 , Reference Harvie, Wright and Pegington 37 , Reference Wegman, Guo and Bennion 39 , Reference Harvie, Sims and Pegington 40 ) and/or HOMA-IR( Reference Harvie, Pegington and Mattson 30 , Reference Harvie, Wright and Pegington 37 , Reference Harvie, Sims and Pegington 40 ). IER additionally elicits favourable alterations in fasted lipid profiles( Reference Johnson, Summer and Cutler 28 – Reference Bhutani, Klempel and Kroeger 33 , Reference Klempel, Kroeger and Norkeviciute 35 – Reference Hoddy, Kroeger and Trepanowski 38 , Reference Zuo, Feng and Tinsley 41 ) including shifts in LDL sub-fraction distribution and/or size towards larger less atherogenic particles( Reference Varady, Bhutania and Klempel 31 – Reference Bhutani, Klempel and Kroeger 33 , Reference Varady, Bhutani and Klempel 36 , Reference Hoddy, Kroeger and Trepanowski 38 ), even in the absence of overall change in total LDL( Reference Hoddy, Kroeger and Trepanowski 38 ). Glycaemic and lipaemic outcomes do not appear to be influenced by meal timing on modified fast days( Reference Hoddy, Kroeger and Trepanowski 38 ), or by dietary composition on feed days, although in this particular study vascular function worsened slightly when a high-fat (45 % total energy) diet was consumed( Reference Klempel, Kroeger and Norkeviciute 35 ). In healthy/overweight participants, Wegman et al.( Reference Wegman, Guo and Bennion 39 ) compared 3 weeks of IER with antioxidant supplementation or placebo using a randomised cross-over design. The protocol consisted of an energy intake of 25 and 175 % of energy requirements on alternating days, designed to maintain energy balance. The study observed a modest within-treatment decline in fasting insulin following the IER plus placebo intervention only, but did not conduct statistical comparisons between antioxidant and placebo legs.
Outstanding questions
Weight-management is an integral part of reducing cardiometabolic risk; even modest amounts of weight-loss can improve glycaemic control and lipid profiles, which reflect improved glucose and lipid homeostasis. The weight-loss efficacy of IER is not in doubt however, from a metabolic standpoint, a number of questions remain outstanding:
Can intermittent energy restriction influence metabolism in the absence of overall energy restriction or weight-loss?
If IER could be shown to improve metabolic control in the absence of overall energy restriction or weight-loss, it would highlight additional applications of IER in healthy weight populations. There is some supportive evidence in the rodent literature to suggest that IER is able to improve glycaemic control markers, to a similar extent as CER (40 % energy restriction), despite no overall energy deficit( Reference Anson, Guo and de Cabo 8 ). In healthy/overweight human subjects, improvements in insulin sensitivity have been observed after 2–3 weeks of alternate day total and modified fasting in the absence of weight-change; however, these studies lacked appropriate ad libitum control group/interventions( Reference Halberg, Henriksen and Söderhamn 22 , Reference Wegman, Guo and Bennion 39 ). There is clear paucity of data in this area and so more controlled studies in healthy weight populations are required to address this particular question.
Does the mode of energy restriction (intermittent v. continuous) influence metabolic responses during weight-loss?
Ultimately it is the degree of dietary adherence and sustainability, rather than the type of dietary strategy, which will predict weight-loss outcomes. Understanding metabolic differences between dieting approaches is nonetheless important as it may identify potential applications of IER within certain patient subgroups. Rodent studies (examining glycaemic indices) have demonstrated comparable short-term outcomes between IER and CER( Reference Anson, Guo and de Cabo 8 , Reference Cerqueira, da Cunha and Caldeira da Silva 15 , Reference Baumeier, Kaiser and Heeren 17 ), but also worse longer-term outcomes with IER( Reference Cerqueira, da Cunha and Caldeira da Silva 15 ), which is discussed further in the following section. Considering the rodent literature in its entirety, a large proportion of studies have reported a lack of full energy compensation on feed days, irrespective of background diet composition( Reference Wan, Camandola and Mattson 9 , Reference Tikoo, Tripathi and Kabra 10 , Reference Jiang, Koike, Li and Wang 13 , Reference Higashida, Fujimoto and Higuchi 16 – Reference Boutant, Kulkarni and Joffraud 19 ). Inclusion of a separate pair-fed group, i.e. a group fed an identical quantity of food but on a daily basis, is one way in which future studies could ascertain the relative contribution of the IER eating pattern and overall energy restriction to observed metabolic changes.
Of the relatively few human studies to directly compare IER to CER, some studies have demonstrated comparable metabolic outcomes between the two dietary approaches( Reference Catenacci, Pan and Ostendorf 26 , Reference Ash, Reeves and Yeo 27 ). The 5:2 diet studies, which were conducted in overweight/obese women, found reductions in fasting insulin and HOMA-IR following IER that have exceeded those of CER (25 % energy restriction/day) after 3 months( Reference Wegman, Guo and Bennion 39 ) and 6 months( Reference Varady, Bhutani and Klempel 32 ), with both studies reporting comparable average weight-loss between the two dieting groups. In a study by Varady et al.( Reference Varady, Bhutani and Klempel 32 ), this time using an alternate day modified fasting regimen, IER led to a significant greater reduction in fasting TAG than CER after 12 weeks, with both dieting groups attaining a comparable 5 % weight-loss.
From a metabolic perspective, without standardisation for weight-loss between dieting groups, it becomes difficult to interpret to what extent any underlying differences can be explained simply by the extent of weight-loss. Whilst most dietary comparison studies have reported similar average weight-loss between IER and CER groups overall, the degree of weight-loss can vary considerably at an individual level, as exemplified by one such study, which found a proportionally greater number of IER participants tending to attain a ⩾5 % weight-loss( Reference Harvie, Wright and Pegington 37 ). Notwithstanding, these data do highlight the potential for underlying differences between the two modes of energy restriction, which requires focused metabolic studies, which conduct assessments following standardised weight-loss as opposed to diet duration.
By what mechanisms does intermittent energy restriction alter metabolism, and how does this compare with continuous (daily) energy restriction?
Acute disturbances in fuel management
Over an acute timescale, fasting (both total and modified) elicits reciprocal shifts in fuel utilisation in a seemingly dose-responsive manner; free-fatty acid (FFA) mobilisation, fatty-acid oxidation (FAO) and ketogenesis progressively rise, favouring the conservation of glucose as its demand drops( Reference Antoni, Johnston and Collins 5 ). During the subsequent refeeding period after a prolonged (36 h) fast, postprandial fatty-acid partitioning remains shifted towards whole-body FAO and hepatic ketogenesis, translating to a marked reduction in postprandial lipaemia( Reference Antoni, Johnston and Collins 5 ). From a chronic perspective, these acute disturbances in fuel management (experienced repeatedly during IER) may elicit very distinct alterations in glucose and lipid metabolism when compared with CER, and may not necessitate an overall energy restriction.
For instance, the premise of IER is that the repeated periods of FAO may promote or augment improvements in insulin sensitivity via the preferential reduction in the ectopic accumulation of lipid and associated cytosolic intermediaries, which are causally implicated in the development of insulin resistance( Reference Shulman 43 ); by extension, this would also be expected to favourably impact other aspects of metabolism, such as the regulation of hepatic TAG metabolism( Reference Fabbrini, Magkos and Mohammed 44 ). In support of this theory are findings from the previously discussed human studies demonstrating superior reductions in HOMA-IR( Reference Harvie, Pegington and Mattson 30 , Reference Harvie, Wright and Pegington 37 ) and fasting TAG( Reference Varady, Bhutani and Klempel 32 ) with IER v. CER, whilst another study found suggestions of increased mitochondrial fatty-acid transporter expression in skeletal muscle after 3 weeks of IER, which reflects long-term adaptation to the repeated elevations in FFA/FAO( Reference Heilbronn, Civitarese and Bogacka 23 ). In rodents, post-treatment reductions in hepatic steatosis have been observed( Reference Baumeier, Kaiser and Heeren 17 , Reference Boutant, Kulkarni and Joffraud 19 ), but the effects in human subjects remains unknown.
Whilst diminished FAO capacity has been implicated as a key instigator of insulin resistance, an alternative mitochondrial overload model by Koves et al.( Reference Koves, Ussher and Noland 45 ) proposes that increased FFA delivery is conducive to excessive and incomplete FAO and it is this that instead precipitates skeletal muscle insulin resistance. A major contributory factor is thought to be lipid-induced mitochondrial stress, which is the primary site of endogenous reactive oxygen species production. In Koves’ model, the liver is considered relatively resilient to lipid-induced mitochondrial stress due to its capacity to channel FFA towards alternate metabolic pathways, with the ketogenic pathway being of particular relevance to fasting. Whilst Koves’ theory is framed in the context of obesity/high-fat feeding, it draws parallels with IER in that metabolic tissues are repeatedly exposed to elevations in FFA/FAO with each fast–refeed cycle. Indeed the studies (in both rodents and human subjects), which report impaired glucose tolerance, all of which have used alternate day total fasting protocols, have also found increased markers of oxidative stress( Reference Cerqueira, da Cunha and Caldeira da Silva 15 ), oxidative insulin receptor modification( Reference Cerqueira, da Cunha and Caldeira da Silva 15 ), reduced GLUT content( Reference Higashida, Fujimoto and Higuchi 16 ), or indications of reduced mitochondrial function markers( Reference Heilbronn, Civitarese and Bogacka 23 ) in skeletal muscle. Interestingly, insights into the time-course of glycaemic changes were provided by one of the rodent studies, which found that despite an initial improvement at 4 weeks, glucose tolerance then deteriorated over time( Reference Cerqueira, da Cunha and Caldeira da Silva 15 ). These data highlight the potential for harm, as well as the potential for tissue-specific responses to IER given that two of these studies reported no change in fasting (hepatic) glycaemic indices( Reference Higashida, Fujimoto and Higuchi 16 , Reference Heilbronn, Civitarese and Bogacka 23 ).
Based on these data, we speculate that the alternate day total fasting paradigm may present too great or frequent metabolic challenge, which could potentially over time lead to aberrant rather than beneficial changes in peripheral insulin sensitivity. In human subjects, the effects of this form of IER on glycaemic control beyond 8 weeks (where no change was observed) is unknown( Reference Catenacci, Pan and Ostendorf 26 ). Further research over longer timescales is now required to evaluate the potential harms and benefits of IER protocols that in fast frequency and the severity of fast day energy restriction, as well as tissue-specific changes in insulin sensitivity and ectopic lipid content. In addition, dynamic assessment of glucose and lipid metabolism, such as hyperinsulinaemic–euglycaemic glucose clamp techniques and postprandial challenges, may help to detect any underlying differences between IER and CER, which may not be immediately apparent when solely looking at singular blood measurements taken in the fasted state.
Change in adipose tissue morphology, physiology and distribution
Adipose tissue is integral to the maintenance of glucose and lipid homeostasis; not only does it act as important sink and storage organ for FFA, but also as a source of an array of adipokines capable of influencing whole-body metabolism. An important driver of obesity-associated metabolic disorders appears to be the pathological expansion of adipose tissue among predisposed individuals, the manifestations of which include: impaired adipogenesis and excessive adipocyte hypertrophy in subcutaneous depots; macrophage infiltration, resulting in low grade inflammation and altered adipokine secretion; impaired insulin action and dysregulated lipid storage; visceral adipose tissue accumulation. The result is the creation of an unfavourable metabolic and hormonal milieu, which drives the development of systemic insulin resistance and associated metabolic disorders( Reference Bays 46 ).
Several studies have shown that rodents maintained on IER exhibit reduced visceral adipose tissue( Reference Varady, Allister and Roohk 14 , Reference Varady, Hudak and Hellerstein 42 ), as well as favourable reductions in fat cell size and augmented adipocyte differentiation within subcutaneous and visceral fat pads( Reference Varady, Roohk and Loe 11 ). Levels of adiponectin, which has anti-atherogenic and insulin-sensitising properties, are usually reduced in states of obesity but increased through IER( Reference Varady, Allister and Roohk 14 , Reference Wan, Ahmet and Brown 21 ). On the basis of one study, these beneficial changes in adipose tissue physiology and morphology appear to be comparable with CER( Reference Varady, Allister and Roohk 14 ), but what was unique to IER is that that it did not necessitate an overall energy deficit, which is perhaps attributable to the alternating periods of feeding and fasting. In human subjects, one uncontrolled alternate day total fasting found improved insulin-mediated lipolysis inhibition, reflecting improved insulin action, after 2 weeks( Reference Halberg, Henriksen and Söderhamn 22 ). In addition to reductions in total adiposity, IER weight-loss studies have reported reductions in visceral adipose tissue( Reference Hoddy, Kroeger and Trepanowski 38 ), truncal adiposity( Reference Catenacci, Pan and Ostendorf 26 ) as well as beneficial changes in circulating adipokine profiles. For example, pro-atherogenic adipokines, including leptin( Reference Catenacci, Pan and Ostendorf 26 , Reference Harvie, Pegington and Mattson 30 , Reference Varady, Bhutani and Klempel 36 , Reference Harvie, Wright and Pegington 37 ) are reduced, whilst adiponectin levels are increased( Reference Harvie, Pegington and Mattson 30 , Reference Varady, Bhutani and Klempel 36 ). Among studies comparing changes in adiposity measures (total, truncal and waist circumference)( Reference Catenacci, Pan and Ostendorf 26 , Reference Ash, Reeves and Yeo 27 , Reference Harvie, Pegington and Mattson 30 , Reference Harvie, Wright and Pegington 37 ) and/or circulating adipokines( Reference Catenacci, Pan and Ostendorf 26 , Reference Harvie, Pegington and Mattson 30 , Reference Harvie, Wright and Pegington 37 ) between IER and CER, no consistent differences have been observed; however, additional studies employing robust assessments of regional adiposity are needed.
Time-restricted feeding
Overview
Over the past few years, there has been an emergence of interest in the concept of chrononutrition, i.e. the interaction between meal timing and our circadian system, which comprise self-sustained approximately 24 h oscillations in physiology, metabolism and behaviour( Reference Garaulet, Ordovas and Madrid 47 ). Examples include the daily rhythms in glucose homeostasis and insulin sensitivity, which declines over the course of the day( Reference Johnston 48 ). These rhythms are driven by a series of molecular ‘clocks’( Reference Garaulet, Ordovas and Madrid 47 ). The ‘master’ clock is located in a small brain region within the anterior hypothalamus, the suprachiasmatic nuclei. Light is the dominant entrainer (Zeitgeber) of the suprachiasmatic nuclei, which controls many essential physiological processes, including the sleep–wake cycle and endocrine rhythms( Reference Garaulet, Ordovas and Madrid 47 ). Peripheral clocks are located in many tissues integral to glucose and lipid metabolism such as the liver, pancreas, skeletal muscle and adipose tissue. Feeding time acts as the major Zeitgeber of peripheral clocks, with the suprachiasmatic nuclei acting as the ‘central conductor’, ensuring correct synchronisation between peripheral clocks( Reference Johnston 48 ). Possession of such rhythms permits effective co-ordination of endogenous processes to changes in the environment such as the daily light–dark cycle, and consequent cyclical food availability. TRF is one example of a timed dietary approach, which also falls under the intermittent fasting umbrella (Table 1). TRF involves limiting food intake to within a shortened window of time, which thereby extends the length of the daily fasting interval. In contrast to IER, TRF is usually performed on a daily basis and does not necessarily entail a prescribed energy restriction; however, the two approaches are not mutually exclusive, indeed, alternate day total fasting can be seen as a prolonged form of TRF. An overview of findings from the rodent and human literature is discussed next.
Effects of time-restricted feeding on glucose and lipid metabolism
Rodents
Being nocturnal animals, mice consume the majority of their energy during the dark/active phase. TRF (⩽12 h) where food intake is consolidated to the dark phase, has been shown to both protect against and reverse the harmful metabolic consequences of diverse nutritional challenges, including high-fat and high-sugar obesogenic diets( Reference Hatori, Vollmers and Zarrinpar 49 , Reference Chaix, Zarrinpar and Miu 50 ). TRF mice display reduced adiposity and liver steatosis, as well as improved glucose tolerance and reduced cholesterol levels when compared with mice fed ad libitum with the same high-fat diet; these changes may reflect improved homeostasis in multiple tissues( Reference Hatori, Vollmers and Zarrinpar 49 , Reference Chaix, Zarrinpar and Miu 50 ). Importantly, these improvements occur in the absence of changes in energy intake.
Human studies
In human individuals, TRF has been achieved through avoidance of night-time intake( Reference LeCheminanta, Christenson and Bailey 51 ), by limiting food intake to one evening meal daily (referred to herein as evening TRF)( Reference Carlson, Martin and Stote 52 , Reference Stote, Baer and Spears 53 ), or by allowing participants to self-select their own shortened (10–11 h) eating window( Reference Gill and Panda 54 ). Islamic religious fasting restricts food access to nocturnal hours, between sunrise and sunset. Recent meta-analyses of these Ramadan studies, which by their nature have been observational by design, demonstrate improvements in some biochemical risk markers despite the nocturnal eating pattern( Reference Kul, Savaş and Öztürk 55 ). With regard to interventional TRF studies, only two publications derived from the same cross-over controlled study conducted within a healthy-weight middle-aged study cohort, have reported metabolic outcomes. Stote et al.( Reference Stote, Baer and Spears 53 ) found both pro-atherogenic (increasing LDL-cholesterol) and anti-atherogenic (increasing HDL-cholesterol and decreasing TAG) changes following 8 weeks of evening TRF. Findings may have been confounded by circadian variations in these parameters, as baseline and post-intervention blood measurements were taken in the morning and evening, respectively. In the second publication, the group also reported significant elevations in fasting and postprandial glycaemia( Reference Carlson, Martin and Stote 52 ). In this instance, oral glucose tolerance testings were conducted in the morning, following a period of time in which participants may have become metabolically adapted to this evening eating pattern. In addition, as there was no standardisation of dietary intakes on the day preceding their two metabolic assessments, participants would have consumed a greater quantity of food the night before their post-intervention visit, which might explain the rise in fasting glucose relative to baseline.
Outstanding questions
Whilst pervasive rodent data exist highlighting the substantive health benefits of TRF, human data are sparse and their interpretation hindered by methodological issues. As such, a number of fundamental questions remain outstanding:
By what mechanisms might time-restricted feeding reduce disease risk, and to what extent can any metabolic benefits be attributable to fasting?
Misalignment between our circadian biology and behavioural patterns, for example through consumption of food outside of the appropriate phase of the endogenous circadian cycle, is thought to be a major contributor to obesity and metabolic disease risk( Reference Johnston 48 ). Defined daily feeding and fasting rhythms support robust oscillation in components of the circadian clock and their output genes, but become dampened in rodent models of diet induced obesity as they start to consume a greater proportion of their food intake in the light (inactive) phase( Reference Hatori, Vollmers and Zarrinpar 49 , Reference Chaix, Zarrinpar and Miu 50 ). Rhythmicity can be restored in these rodent models through TRF, which is made possible by the extensive cross-talk and interaction between molecular clock components and the feed/fasting responsive signals of cellular energy status. For example, fasting induces the activities of the AMP-activated protein kinase and cAMP-response element-binding protein, which promote ATP production during low-energy availability. In contrast, feeding stimulates the mechanistic target of rapamycin pathway, which promotes anabolic processes during increased energy availability. These feed/fast responsive elements can influence the expression of peripheral clock components and vice versa( Reference Peek, Ramsey and Marcheva 56 ). In support of this, TRF has been shown to improve cAMP-response element-binding protein, mechanistic target of rapamycin, and AMP-activated protein kinase rhythms, and hence promote robust oscillations of circadian clock gene components( Reference Hatori, Vollmers and Zarrinpar 49 ). The metabolic benefits of TRF may additionally be attributed to the accompanying elongation of daily fasting intervals, which as previously discussed, may favourably impact on regional and ectopic adiposity. In the comprehensive rodent study conducted by Chaix et al.( Reference Chaix, Zarrinpar and Miu 50 ), the magnitude of the metabolic benefits of TRF were proportional to the length of the fasting interval. Whilst highly suggestive, there are no comparable human data. In addition, whilst one purported benefit of TRF is that it constrains energy intake to the most appropriate phase of the endogenous circadian cycle, in human subjects, the impact of earlier TRF meal timings (when metabolic control is more optimal( Reference Johnston 48 )), has yet to be assessed.
Can time-restricted feeding alter metabolism in the absence of energy restriction in human subjects?
One major difference between human and rodent studies is that human subjects tend to behaviourally reduce food intake during TRF when permitted ad libitum/self-selected intake( Reference LeCheminanta, Christenson and Bailey 51 , Reference Gill and Panda 54 ). Similarly, many Ramadan studies report a mild energy deficit, although this is not a universal finding( Reference Kul, Savaş and Öztürk 55 , Reference Sadeghirad, Motaghipisheh and Kolahdooz 57 ). Where energy restriction is absent, or minimal, current evidence suggests that TRF can influence metabolism (both positively and negatively), necessitating further study into the optimal fasting window( Reference Carlson, Martin and Stote 52 , Reference Stote, Baer and Spears 53 ). Although this often unintentional energy restriction makes metabolic interpretations difficult, the significance cannot be disregarded as it highlights the potential utility of TRF as a ‘small changes’ strategy for weight management. However, from a metabolic standpoint, further studies enforcing tighter control and monitoring over both energy intake and expenditure would be needed to fully explore this.
Conclusion
Put together, there is a growing evidence base demonstrating the benefits of IER over the short-to-medium term on glucose and lipid homeostasis, meriting further research in population groups with (or are at high risk of) T2DM and CVD, including healthy-weight individuals. Whilst the metabolic benefits of TRF appear quite profound in rodents, studies in human subjects are sparse and subject to some methodological issues, which hinder their interpretation. There is some suggestion that the metabolic changes elicited by the two intermittent fasting approaches can occur in the absence of energy restriction, and in the context of IER, may be distinct to those observed following similar weight-loss achieved via modest CER. Specifically, some studies have found greater improvements in indices of hepatic insulin sensitivity and TAG metabolism with IER. The metabolic benefits of intermittent fasting are likely to stem in part from the frequently repeated prolonged fasting intervals, which may favour the preferential reduction of ectopic fat, beneficially modulate aspects of adipose tissue physiology/morphology and distribution (in addition to any changes in overall mass), and may also impinge on circadian clock regulation. However mechanistic evidence is largely limited to findings from rodent studies, thus necessitating focused, controlled, human studies, which also incorporate more dynamic assessments of glucose and lipid metabolism in addition to steady-state measurements. Long-term studies are also needed, which may yet uncover the potential for adverse health consequences, as evidenced by a small number of studies. Ultimately, much remains to be learned about intermittent fasting (in its various forms); however, the positive findings to date serve to highlight promising avenues for future research.
Financial Support
R. A. was funded jointly by the University of Surrey Prize studentship and by Lighterlife
Conflicts of Interest
K. L. J. is the Head of Nutrition and Research at Lighterlife.
Authorship
R. A. wrote the manuscript. K. L. J., A. L. C. and M. D. R. assisted with manuscript editing.
