Introduction

Weight regain after initially successful weight loss is common (Box 1). This weight regain seems to be due to the activation of compensatory physiological mechanisms that are triggered by the negative energy balance and accompanying weight loss during a period of calorie restriction (and/or other lifestyle changes), and which can remain activated after the energy-restricted diet has stopped. Persistent activation of these mechanisms will contribute to weight regain. Although these compensatory mechanisms are probably a natural response to weight loss in order to maintain sufficient body mass and energy stores, these mechanisms are not functional in people with obesity who want to lose weight.

In a previous review we discussed a number of adipose tissue-related factors that could influence weight regain, including the role of adipocyte stress, adipokines, fatty acid metabolism and microRNAs1. In this Review we address in more detail other potentially causal biological factors: the persistence of obesity-associated immune cell phenotypes in adipose tissue after weight loss, weight loss-induced adaptations of the adipocyte extracellular matrix (ECM), metabolic adaptations such as reduced energy expenditure and decreased lipid oxidation and lipolysis, changes in gut hormone secretion patterns, and neuronal changes that increase appetite and reward responses to food intake. We focus on lifestyle-induced weight loss, as it is probable that the response to surgery-induced weight loss differs from that to lifestyle-induced weight loss, especially in relation to the signalling from the gastrointestinal tract to the central nervous system to alter food intake2. Furthermore, we touch upon some strategies to prevent weight regain after weight loss based on these factors.

Obesity-associated immune cell phenotypes in adipose tissue

Obesity is accompanied by changes in the innate and adaptive immune systems of adipose tissue in humans3,4 and in mice5,6,7. A massive invasion of macrophages is characteristic, attracted by adipocytes stressed by hypoxia, hypertrophy or cell death. Crown-like structures can be seen under the microscope when the adipocytes are surrounded by macrophages. At the same time, tissue resident macrophages differentiate into a continuum of multiple macrophage populations. These populations include the M2 and M1 states that utilize oxidative phosphorylation and glycolysis, respectively, and macrophages with both M2 and M1 metabolic characteristics, which secrete pro-inflammatory cytokines including IL-1, IL-6 and tumour necrosis factor (TNF)8,9. In addition to macrophages, the proportion and function of many other classes of immune cells are changed in adipose tissue during the development of obesity10.

Studies investigating the change in expression of inflammatory genes in adipose tissue with weight loss have given inconsistent results. The findings could indicate that weight loss by calorie restriction lowers the expression of those genes compared with the pre-weight-loss state11. Alternatively, the results could suggest that the expression of inflammatory genes is upregulated by calorie restriction and downregulated with maintenance of the reduced weight, even to levels lower than those in the pre-weight-loss state12,13. In 2022, it was demonstrated that lipopolysaccharide-induced cytokine production by adipose tissue macrophages is increased during obesity development and is not subsequently reduced by weight loss14. These observations support the concept that obesity followed by weight loss establishes an ‘obesogenic memory’, also referred to as an ‘obesity memory’15; that is, a combination of factors originating during obesity and subsequent weight loss, which confers a risk of returning to the obese state. Some of the key factors involved in obesity memory are persisting inflammatory cells in adipose tissue. Underlying the persistence of immune cell populations could be epigenetic changes of gene expression in response to environmental cues9,16. A 2023 article17 hypothesized that long-term imprinting of obesity and weight loss on immune cells of adipose tissue, referred to as ‘trained immunity’ of the innate immune system18,19, underlies chronic inflammation as well as the innate immune memory. The idea is that such ‘training’ (that is, immunometabolic and epigenetic reprogramming of myeloid cells) makes immune cells more responsive to future challenges. Interestingly, trained immunity has also been proposed to cause adjustments in bone marrow progenitor cells, which could explain the sustained or persistent presence of certain immune cell populations in adipose tissue during cycles of weight gain and weight loss18.

Novel methods applied to mice passing through a cycle of weight gain, weight loss and weight regain, including fluorescence-activated cell sorting combined with single-cell sequencing and cellular indexing of transcriptomes and epitopes by sequencing of surface proteins, have enabled researchers to study different obesity-associated immune cell populations in adipose tissue in more detail. A 2022 study10 showed that tissue-resident macrophages decrease in abundance during obesity, weight loss and weight regain, whereas lipid-associated macrophages increase with obesity and persist during weight loss and weight regain. Levels of adipose tissue regulatory T cells decline in proportion to the number of adipose tissue αβ T cells during development of obesity, and remain at lower abundance during weight loss and weight regain than before the development of obesity. T cells from adipose tissue are enriched for the exhausted phenotype during obesity, weight loss and weight regain. This phenotype is characterized by a gradual loss of function and specific gene expression changes20,21. Antigen-presenting dendritic cells in adipose tissue are largely unchanged in proportion, but activated subsets of dendritic cells are enriched in obesity and these subsets persist during weight loss and weight regain10. These findings of persisting cell populations indicate that, in mice, weight gain followed by weight loss primes macrophages and other immune cells in adipose tissue to act as an innate immune memory for inflammation. So far, these changes of adipose immune cell populations have mainly been studied in the male mouse. Sex differences have been reported for inflammatory responses to adipose tissue lipolysis in mice with diet-induced obesity22. Therefore, extrapolation of the observations to female mice and to humans should not be done without additional experiments. With novel techniques such as single-cell sequencing and proteomics, as mentioned above, such information could soon become available.

Obesity memory and weight regain

In mice, there is evidence that an obesity memory imprinted in immune cells during obesity and weight loss is involved in an increased risk of weight regain. A 2017 study23 found that after a cycle of weight gain and weight loss, mice on a high-fat diet (HFD) gained body weight faster than mice of similar body weight that had not passed through such a previous cycle of weight gain and loss. Even 2 months after weight loss, this phenomenon was still observed. CD4+ T cells were found to be essential for this increased weight gain after weight loss, as systemic depletion of CD4+ T cells led to ablation of the obesity memory. Moreover, naive immunodeficient Rag1−/− mice without a previous history of obesity acquired the obesity memory by transfer of CD4+ T cells from previously obese mice.

In humans, a link between the immune status of adipose tissue and weight regain after weight loss has been suggested by the YoYo study, a dietary intervention study in which individuals with overweight or obesity lost 8% of body weight on a low-calorie diet (1,250 kcal per day) or a very-low-calorie diet (VLCD) (500 kcal per day) (weight loss phase), then remained for 4 weeks on a balanced diet to maintain weight loss (weight-stable phase), after which they were followed up for 9 months during which weight regain was recorded24. The expression of 277 ECM genes in the subcutaneous adipose tissue was investigated. In people who lost weight on a VLCD, the change in expression of a cluster of integrin genes during the weight-stable phase correlated positively with percentage weight regain25. The integrin genes appeared to be leukocyte-specific. The findings indicated that people with the lowest reduction in the expression of leukocyte-specific genes in the weeks after weight loss (weight-stable phase) were more prone to weight regain. This finding is in accordance with the role of an obesity memory composed of persistent cells of the myeloid lineage in humans, but at the same time demonstrates the individual variation with respect to this phenomenon.

The persistence of certain immune cell populations after weight loss, the study of CD4+ T cells in mice, and the association between the expression of leukocyte-specific genes and weight regain argue for a role of an inflammatory obesity memory based on immune cells in the risk of weight regain in humans. Although no clear mechanistic data are yet available, it can be hypothesized that ECM-modifying enzymes secreted by macrophages not only play a role in immune cell infiltration but could also be part of an obesity memory. When adipocytes expand, their ECM could lose its dynamic character as hypoxia prevents the proper functioning of oxygen-dependent collagen-modifying enzymes such as lysyl oxidase and prolyl-4-hydroxylase26. Eventually the loss of ECM dynamism will counteract further growth27 and adipocyte signalling will be adjusted to reduce energy intake and storage within the adipocytes (Fig. 1A). During obesity development, ECM-modifying enzymes secreted by macrophages could weaken the ECM and, in that way, diminish this resistance to growth and thereby the resistance to energy intake (Fig. 1A). After weight loss, some immune cells remain in the adipose tissue (Fig. 1B). Upon a novel challenge by a positive energy balance (for instance, by returning to the pre-weight-loss eating pattern), the ECM-modifying enzymes of these persistent immune cells are readily available. As before, these enzymes remodel the ECM to allow adipocyte growth and prevent inhibition of energy intake, thereby facilitating weight regain (Fig. 1B). The involvement of immune cells in this mechanism also supports the possibility that pro-inflammatory and anti-inflammatory cytokines influence the risk of weight regain28,29. However, this possibility has not yet been studied in much detail. A 2022 study found that during the follow-up period of the YoYo study, the change in IL-6 was positively associated with weight regain30.

Fig. 1: Potential mechanism for the influence of macrophages on obesity development and on weight regain after weight loss.
figure 1

A, Adipose tissue expansion during weight gain. Aa, A positive energy balance leads to expansion of adipocytes, generating extracellular matrix (ECM) stress. Consequently, signals from the stressed adipose tissue reduce energy intake and storage. Chemokines, also secreted by the stressed adipocytes, attract macrophages. Ab, Macrophages secrete ECM-modifying enzymes, weakening the ECM and reducing the stress. Ac, Reduced ECM stress enables increased energy intake and further adipocyte growth. B, Macrophage persistence in adipose tissue facilitates weight regain after weight loss. Ba, After weight loss, adipocyte size is reduced and the ECM is remodelled and tightened around the shrunken adipocytes. However, macrophages persist in the tissue. Bb, If overeating occurs again, this challenge leads to rapid secretion of ECM-modifying enzymes by the persistent macrophages already present and weakening of the ECM. Bc, This ECM weakening enables rapid adipocyte growth and weight regain.

Full size image

Notably, in the proposed mechanism, ECM-modifying enzymes play a crucial role. Cathepsin S is part of the expression cluster of leukocyte-specific integrins that correlated with weight regain in the YoYo study, as described above25. Cathepsin S is a member of the cathepsin protease family, a family of proteases that have both intracellular and extracellular functions in relation to health and disease31,32. Cathepsin S is expressed and secreted by macrophages and might have a role in infiltration of immune cells into adipose tissue. In the YoYo study, people with the lowest reduction in levels of cathepsin S RNA in adipose tissue during the weight-stable phase after weight loss had the highest risk of weight regain25. As such, cathepsin S persistence could be part of the obesity memory potentiating the risk of weight regain by immune cell infiltration and inflammation. Two other members of the cathepsin gene family, cathepsin B and cathepsin D, are part of another co-expression cluster identified in the YoYo study, and changes in the expression of these cathepsins in adipose tissue during the weight-stable period was associated with weight regain25. The function of this gene cluster has been defined as ‘ECM modification’25. Like cathepsin S, both cathepsin B and cathepsin D are expressed in macrophages, whereas cathepsin B is also expressed by adipocytes. Cathepsin B is one of the processing enzymes for the formation of cathepsin D33.

In summary, obesity is accompanied by infiltration of macrophages and other immune cells. After weight loss, part of the immune profile persists (Fig. 1). Limited evidence in mice and humans indicates that this ‘obesogenic memory’ is involved in the risk of weight regain. A possible mechanism involves macrophage-secreted ECM-modifying enzymes, some of which could be part of the cathepsin protease family. More experimental evidence is needed to unravel the persisting immune cell profile, in particular in humans, and its role in weight regain.

Energy expenditure and metabolic adaptation

It is generally accepted that a high energy intake rather than a low energy expenditure is the main contributor to weight gain. However, it is also possible that a low energy expenditure, without a corresponding low energy intake, contributes to weight gain. Furthermore, metabolic adaptation to weight loss (that is, decreases in energy expenditure that are greater than expected on the basis of body composition) could potentially have a role in weight regain.

Energy expenditure

A seminal paper from 1986 reported that, in a population of Pima Indians, family membership accounted for part of the variance in resting energy expenditure (REE) (adjusted for the amount of metabolically active tissue, that is fat-free mass (FFM))34, suggesting a genetic component in the variation of REE. However, individuals from families with lower REE did not have any higher levels of obesity than individuals from families with higher REE, which the authors suggested was probably partly explained by the close correlation between FFM and percentage of body adipose tissue mass, which indicates that the REE, adjusted for FFM, was already partly adjusted for obesity. A follow-up study from 1988 found that participants’ adjusted REE at the initial visit predicted their gain in body weight over a 4-year follow-up period. Moreover, 24-h energy expenditure, as measured in a respiration chamber, correlated with the rate of change in body weight over a 2-year follow-up period in another group of Pima Indians35. On the other hand, a 2022 analysis using the international Doubly Labeled Water Database for measurements of 24-h energy expenditure did not find an association between 24-h energy expenditure and weight or adipose tissue mass gain36. However, a non-significant association (P = 0.094) was found in those individuals with a follow-up >4 weeks. Thus, the lack of a clear association could have been related to the short follow-up period.

Although, as described above, there is some evidence that a low REE (adjusted for FFM) is associated with weight gain, to our knowledge there is only one study that showed that a lower baseline REE (adjusted for FFM) was associated with more weight regain after diet-induced weight loss37. Whether a lower baseline REE in this study reflected a familial trait or was the consequence of previous weight loss attempts or a negative energy balance is unclear.

Metabolic adaptation

Metabolic adaptation, also known as adaptative thermogenesis, is not unequivocally defined38. Definitions of metabolic adaptation vary and can have different meanings in different contexts. In the context of weight loss, adaptive thermogenesis refers to the fall in REE and non-REE associated with calorie restriction, which is independent of changes in body weight and body composition38. Many studies have only focused on either REE or energy expenditure during exercise and thus are not able to provide the full picture. The concurrent state of energy balance at the time of the energy expenditure measurements after weight loss can have a profound impact on whether metabolic adaptation is detected, and is not always clearly described in publications. In individuals who have lost weight but are still in negative energy balance when the level of metabolic adaptation is measured, the metabolic adaptation will be more pronounced than in individuals who are in energy balance at the time of the measurement39. Diet-induced energy expenditure is also lower under conditions of energy restriction, which might result in lower REE measurements even after an overnight fast of 12–14 h40. Moreover, precise measurements of body composition taking into account weight-loss-induced changes in both anatomical and molecular composition of the FFM are often lacking and this lack can lead to an overestimation of metabolic adaptation41. Also, the method of calculation of changes in metabolic adaptation is not uniform42. Therefore, the results of studies of metabolic adaptation during weight loss and their role in weight regain are sometimes difficult to interpret and to compare. Nevertheless, most studies have shown that metabolic adaptation to calorie restriction-induced weight loss exists for REE and activity energy expenditure43,44,45. A 2020 review of potential mechanisms of metabolic adaptation to energy restriction46 showed that adaptation includes alterations in the composition of the FFM, reduced activity of the sympathetic nervous system, reduced circulating levels of leptin, insulin and thyroid hormones, and improved mitochondrial efficiency. The degree of weight loss on a calorie-restricted diet appears to be influenced by the level of metabolic adaptation, with people with a strong metabolic adaptation losing less weight47,48, adipose tissue mass49 or both50 than those with less metabolic adaptation.

There is considerable debate about whether this metabolic adaptation persists after energy balance is restored and body mass is stable39,51,52,53 and whether it triggers weight regain. Some studies have suggested that metabolic adaptation is maintained over longer periods of time after the initial diet-induced weight loss (≥1 year)54,55,56,57,58, and other studies have suggested that metabolic adaptation rapidly (within weeks) disappears once energy balance is restored39,45,49. Thus, studies are inconsistent with respect to the persistence of weight-loss-induced metabolic adaptation. The discrepancies could be related to the lack of standardization of the conditions under which the measurements of metabolic adaptation were performed in the different studies, what component(s) of 24-h energy expenditure were measured, and the amount of weight loss and weight regain seen in the study participants. Several studies addressed the question of whether the degree of weight-loss-associated metabolic adaptation influences weight regain, but so far none of them has found evidence for this39,49,56,59,60,61. In participants of the ‘Biggest Loser Competition’, who were re-measured 6 years after the end of the competition, better weight loss maintenance was associated with larger metabolic adaptation at 6 years, suggesting that this larger metabolic adaptation is due to ongoing efforts to reduce weight and might be acting against such efforts56.

In summary, although an individually determined low energy expenditure (REE) might be associated with the risk of weight gain, there is hardly any evidence that low REE in a person with obesity is predictive of weight regain after weight loss. Furthermore, it is still unclear whether the metabolic adaptation that occurs with weight loss on an energy-restricted diet is maintained when a negative energy balance is no longer present (Fig. 2). So far there is no evidence that the degree of metabolic adaptation after weight loss predicts the amount of weight regain.

Fig. 2: The physiology of weight regain.
figure 2

Physiological changes during weight loss under a negative energy balance and after restoring energy balance with possible influence on weight regain. Arrows indicate differences from the pre-weight-loss state: up arrows indicate increase, down arrows indicate decrease and horizontal arrows indicate no significant change. It should be noted that the outcomes of studies were sometimes inconsistent (represented by multiple arrows with different directions for the same factor).

Full size image

Lipid oxidation and lipolysis

For a condition of weight stability, a balance between energy intake and energy expenditure is required. Such a balance can only be obtained if the levels of macronutrient intake and expenditure are also balanced. Carbohydrate and protein stores are closely regulated by adjusting levels of oxidation to levels of intake and thus any day-to-day energy imbalances are mainly resolved by changes in storage and utilization of lipids62. Over the long term, changes in levels of free fatty acids (FFA) and in insulin sensitivity, due to gains or losses of adipose tissue, influence the average rate of lipid oxidation63. Subsequently, body composition will drift towards the degree of adiposity where lipid oxidation matches dietary fat intake63. Lipid oxidation is restrained by high levels of glycogen64, so a greater expansion of adipose tissue mass is needed in individuals who maintain relatively high glycogen reserves.

Lipid oxidation

One of the first studies showing a potential connection between lipid oxidation and obesity was a study in Pima Indians that found that a higher 24-h respiratory quotient (a measure of the mixture of substrates oxidized), thus a lower level of lipid oxidation, was associated with weight gain, independent of whether REE was high or low65. A 2020 study66, on the other hand, did not find an association between baseline resting respiratory quotient and weight regain 1 year and 2 years after a weight loss programme in premenopausal women with overweight. This difference might have been due to a difference between resting and 24-h respiratory quotient, a difference between weight gain and weight regain and/or a difference in the populations studied. Weight loss was found to decrease lipid oxidation at rest and during exercise67,68 and 24-h respiratory quotient was higher in adults who had lost a large amount of weight and had maintained this weight loss for at least 2 months than in a weight-matched control group who had not lost weight previously69. Similarly, weight-reduced obese rats showed a lower 24-h lipid oxidation when fed a low-fat diet (LFD) ad libitum after the energy restriction than obese rats on the same ad libitum diet that were also matched for energy intake. This finding suggests that the weight-loss-induced adaptation in lipid oxidation is independent of energy imbalance70. By contrast, a 2019 study showed a reduction in respiratory quotient with weight loss accompanied by increased fasting FFA in adults with obesity71. The reason for this discrepancy is not directly clear, but might have been related to the different measurement conditions between the two studies. However, participants with a less-pronounced weight-loss-induced reduction in respiratory quotient and lower FFA after weight loss had a larger weight regain71. Therefore, weight-loss-induced changes in FFA metabolism might influence weight regain. This result was supported by the finding that adipose tissue-specific gene expression indicated a main role for weight-loss-induced adaptations of fatty acid metabolism in weight regain71.

Lipolysis

Lipid oxidation in the fasted state is mainly fuelled by adipose tissue lipolysis and driven by high circulating levels of FFA72. An interesting question therefore is whether the reduction in lipid oxidation that is often found in people with obesity, and which can be further reduced after weight loss, is accompanied by reductions in lipolysis.

Basal and stimulated lipolysis in obesity

For decades it has been observed that obesity is associated with an impairment in adipose tissue lipolysis73,74. Basal lipolysis in adipose tissue, which is mediated mainly by adipose triglyceride lipase (ATGL)75, does not seem to be affected by obesity based on microdialysis studies of subcutaneous adipose tissue76,77. However, although in a 2008 glycerol tracer study76, total rate of appearance of glycerol at baseline was similar in men with or without obesity, it was significantly lower in men with obesity when corrected for body adipose tissue mass. An explanation for the lower basal lipolysis in obesity could be obesity-associated insulin resistance. Insulin resistance has been found to be associated with reduced expression of ATGL mRNA and protein78.

Stimulation of adipose tissue lipolysis by the β-adrenergic system occurs mainly through the activity of hormone-sensitive lipase (HSL)75. In adults and children with obesity the stimulation of lipolysis through the β-adrenergic system is blunted. This resistance to catecholamine-induced lipolysis is attributed to decreased expression of lipolytic β2-adrenoceptors, increased anti-lipolytic properties of α2-adrenoceptors, and decreased expression of HSL75. Furthermore, single-nucleotide polymorphisms in the β2-adrenergic receptor gene are associated with reduced catecholamine-stimulated lipolysis79 and with obesity80. A microdialysis study showed that the β2-adrenergic stimulation of lipolysis is also blunted in skeletal muscle of people with obesity81. In addition, inflammatory markers such as TNF can influence both basal and stimulated lipolysis82.

Effects of weight loss on basal and stimulated lipolysis

A 2007 study78 found that the expression of ATGL mRNA and protein in subcutaneous adipose tissue is reduced by diet-induced weight loss (in negative energy balance). In agreement with these findings, basal lipolysis of subcutaneous adipose tissue, as derived from interstitial concentrations of glycerol measured by microdialysis, was reduced after diet-induced weight loss in women with obesity (in negative energy balance) and the reduction was maintained during weight regain83.

With respect to stimulated lipolysis, a 1997 study67 found that in vitro adrenaline-stimulated lipolysis in isolated subcutaneous abdominal and gluteal adipocytes was reduced in adipocytes isolated from participants who had undergone diet-induced weight loss compared with pre-weight-loss values. In agreement with this finding is the observation that subcutaneous adipose tissue expression of HSL mRNA and protein was reduced by weight loss (in negative energy balance) in individuals with overweight or obesity79. In a 2012 study83 in premenopausal women with obesity, the anti-lipolytic activity of the α2-adrenoceptor decreased during weight loss, resulting in an increase in adrenaline-induced lipolytic activity in adipose tissue. However, this improvement disappeared during the subsequent weight maintenance phase and thus appears to be associated with the negative energy balance rather than the weight loss itself.

Lipolysis and weight regain

A 2019 study84 compared lipolysis by measuring the ex vivo adipose tissue FFA production in mice going through an intervention of 8 weeks on a HFD followed by 4 weeks on a LFD with lipolysis in mice that were kept on the LFD for 12 weeks. After 12 weeks, noradrenaline-stimulated lipolysis activity did not differ between the two groups. However, when both groups of mice were subsequently put on a HFD, stimulated lipolysis in the intervention-treated group was significantly lower than in the control group. In addition, HSL phosphorylation and the expression of the genes for the β1-adrenergic, β2-adrenergic and β3-adrenergic receptors were lower in the treated mice, although the expression of the anti-lipolytic α2-adrenergic receptor was also reduced. The differences in these parameters between the groups closely resembled the situation after the initial 8 weeks on a HFD or LFD. Therefore, it was proposed that obesity memory not only has an inflammatory component but also a metabolic component, in particular with respect to lipolysis.

The impairments in lipolysis after weight loss observed in humans with overweight or obesity also suggest that reduced lipolysis might have a role in weight regain. It is hypothesized that a reduction in the lipolytic activity of adipocytes after weight loss could shift the net balance of adipose tissue FFA uptake and storage versus FFA secretion and utilization towards uptake and storage, in particular because FFA oxidation is also reduced. So far, however, direct evidence for this hypothesis from studies in humans is limited. A 2013 study found that the change in ATGL protein expression in adipose tissue during weight loss predicted weight regain: the larger the reduction in ATGL, the higher the weight regain85. In the YoYo study, a larger weight loss-induced reduction in plasma levels of FFA, which is likely to reflect reduced lipolysis, predicted more weight regain86.

In summary, people with obesity are on average characterized by low levels of lipolysis and lipid oxidation. Both lipolysis and lipid oxidation are further impaired upon weight loss, and there is some evidence that this impairment persists after weight loss has ceased and that a stronger impairment of lipolysis after weight loss is associated with more weight regain (Fig. 2).

Appetite-related factors

It is probable that environmental and behavioural pressures and their interaction with genetics determine the level at which body weight is homeostatically regulated87. Appetite has both behavioural and biological aspects, with signals coming from outside the body (behavioural, cognitive and environmental) and inside the body (hormonal, neuronal and metabolic) being integrated to determine energy intake. Measuring appetite in all its forms is a complex task and should involve evaluations of the strength of the motivation to eat, how food choices affect eating behaviour and the hedonic processes that modulate the homeostatic system in the context of the environment and daily energy expenditure88. Moreover, the roles of circulating levels of appetite-related hormones such as leptin, ghrelin, cholecystokinin, glucagon-like peptide 1 (GLP1), peptide YY, amylin, pancreatic polypeptide and gastrointestinal peptide (GIP) and of afferent neuronal activity both before and after a meal as biomarkers of appetite are not yet fully elucidated.

Obesity can arise from a strong drive to eat, inappropriate food choices and/or weak meal-induced satiation, but large interindividual differences exist for each of these factors, and different phenotypes for appetite control are likely to be present within any population88. For example, individuals who are unable to gain weight even if they want to (constitutionally obesity-resistant) appear to sense positive energy balance following short-term overfeeding appropriately as evidenced by changes in hunger and satiation along with reductions in subsequent energy intake89. On the other hand, individuals at high risk of weight gain (individuals with obesity who have lost weight) do not seem to sense the excess calories associated with overfeeding appropriately. These findings were accompanied by clear differences in brain activation patterns in response to pictures of foods with high or low hedonic value between the groups as measured by functional MRI89.

In another study, intragastric glucose and lipid infusions induced orosensory-independent and preference-independent, nutrient-specific cerebral neuronal activity and striatal dopamine release in lean participants. By contrast, participants with obesity had severely impaired brain responses to post-ingestive nutrients, which could cause overeating90. A 2022 systematic review and meta-analysis91 reviewed the findings of studies investigating the differences in levels of gastrointestinal hormones involved in appetite regulation in individuals with or without obesity. The meta-analysis showed that basal and postprandial total ghrelin concentrations as well as postprandial peptide YY concentrations were lower in individuals with obesity than in those without, whereas no differences were found in the concentrations of fasting and postprandial GLP1 and cholecystokinin. However, there was a lot of heterogeneity among the studies assessed91. Appetite-related signals from the gut directly or indirectly (via afferent vagal nerve activation) influence brain centres involved in homeostatic as well as reward-driven food intake. Gastrointestinal hormones and neural signals  released during a meal appear to signal via the brainstem, whereas homeostatic signals act directly through the hypothalamus2. The development of obesity appears to be associated with both leptin and ghrelin resistance92.

With weight loss, adaptations in the appetite system occur, such as increased appetite and food reward, food cravings and orosensory sensations93, that tend to counteract the weight loss, comparable with the adaptations of metabolism. In addition, the impaired neuronal responses to intragastric administration of nutrients in individuals with obesity were not restored after diet-induced weight loss90. In parallel, a systematic review concluded that, besides increased appetite, levels of the orexigenic gut hormone ghrelin are increased and levels of the anorexigenic gut hormones GLP1, peptide YY, cholecystokinin, pancreatic polypeptide and amylin are decreased after energy restriction in most studies94. Another review showed that, although circulating levels of ghrelin increase with diet-induced weight loss, this is not the case for weight loss induced by VLCDs and ketogenic diets. Circulating levels of peptide YY and GLP1 are either reduced or unchanged after weight loss, while levels of amylin and cholecystokinin are reduced and the level of GIP is increased95. However, weight-loss-induced effects on levels of satiety hormones are far from consistent between studies, which could be related to differences in diet, the amount of weight loss, the conditions under which these hormones were measured (in or out of energy balance), and how they were measured (such as whether fasting, postprandial, active or total concentrations were assessed). Circulating leptin concentrations are reduced by diet-induced weight loss96. The reduction is due to a combination of the decrease of adipose tissue mass and the presence of a negative energy balance96. Both factors reduce leptin secretion from adipose tissue.

Persistence of weight-loss-induced appetite changes

Several studies have investigated the persistence of changes in appetite-related factors after weight loss. In the most cited study in this area by Sumithran and colleagues97, 1 year after the initial weight loss, when approximately 40% of the lost weight had been regained, the reduction in fasting plasma concentrations of leptin (adjusted for changes in adipose tissue mass) was partly maintained. Fasting and postprandial plasma concentrations of ghrelin and pancreatic polypeptide increased after weight loss and the increase was partly maintained at 1 year. The fasting and postprandial concentrations of the anorexigenic hormones peptide YY, cholecystokinin and amylin were reduced by weight loss and these changes were still partly present after 1 year. Self-reported fasting and postprandial hunger increased after weight loss and these increases were maintained at 1 year. However, there is considerable inconsistency in the literature about the persistence of weight-loss-induced changes in appetite-related factors (Table 1). Some studies found, after initial weight-loss-induced changes, a return to pre-weight-loss values of some appetite-related factor(s) despite maintenance of (some of the) the lost weight, whereas other studies found, in agreement with the study by Sumithran and colleagues97, persistence of some weight-loss-induced changes in such factors (Table 1). There are likely to be many explanations for this lack of consistency, among them what appetite-related factor(s) were measured and under what conditions (for example, energy balance, fasting or postprandial, level of obesity or weight loss or weight regain, diet, physical activity and sample size). Similar conflicting results are found with respect to persistence of weight-loss-induced appetite sensations (Table 1).

Table 1 Overview of effects of diet-induced weight loss and follow-up on body weight and subjective appetite ratings and gastrointestinal hormones
Full size table

Association with weight regain

Based on a systematic review of the literature on clinical weight loss trials that assessed pre-weight-loss–post-weight-loss changes in fasting leptin and ghrelin, a 2014 article98 concluded that decreases in circulating levels of leptin, increases in levels of ghrelin or baseline or post-intervention concentrations of these hormones consistently predicted subsequent weight regain. Several studies have also investigated the predictive value of appetite-related factors, with inconsistent results. Some studies found that future weight regain could be predicted from baseline, post-weight loss or the change from baseline to post-weight loss values60,61,99, although in one of these studies60 the associations disappeared after adjustment for age, sex and the amount of weight loss, and some of the associations were in a counterintuitive direction60,61, others did not find such associations100,101. In a cross-sectional study comparing weight loss maintainers with regainers, a satiety-induced attenuation of brain activation (as assessed by functional MRI) during receipt of a food-related reward was found in weight maintainers only. This attenuation was associated with lower levels of ghrelin in participants who maintained their weight loss102.

In summary, in people with obesity, circulating basal and postprandial ghrelin concentrations are reduced and concentrations of some of the anorexigenic hormones are also reduced compared with those in people without obesity. With weight loss on a moderately energy-restricted diet, but not on a VLCD or ketogenic diet, plasma concentrations of ghrelin increase. The effect of weight loss on the circulating concentrations of anorexigenic hormones is very variable, with increased, unchanged or reduced values being reported. The results on the long-term maintenance of these changes in the absence of a negative energy balance are also equivocal (Fig. 2), as is also the case for the association with weight regain.

The role of diet, exercise, pharmacotherapy and biomedical strategies in the prevention of weight regain

More insight into strategies to prevent or limit weight regain after weight loss is clearly needed in order to attain better long-term weight reduction and health in people with overweight and obesity (Table 2). Here, we describe several strategies to reduce or prevent weight regain based on the physiological mechanisms discussed throughout this Review.

Table 2 Potential strategies to tackle weight regain
Full size table

Diet

Diet composition and dietary regimens can influence several weight-regain-associated mechanisms, such as appetite, energy expenditure and body composition. We have previously reviewed dietary strategies for prevention of weight regain1,103,104. A more recent review from 2022 specifically addressed the role nutrition in inflammation105, although not in direct relation to the prevention of weight regain.

There is some evidence that a higher dietary protein content in combination with a lower glycaemic index and/or glycaemic load is associated with less weight regain104, but more data are needed about the mechanisms involved. No difference in weight regain after weight loss has been found between diets that contain isocaloric amounts of more satiating foods and diets that contain less satiating foods106. Anti-inflammatory diets (that are rich in olive oil, tomatoes, green leafy vegetables, nuts, fatty fish and fruits, and low in refined carbohydrates, fried foods, sugar-sweetened beverages, red meat and margarine, such as the Mediterranean diet) might reduce weight regain107, but this possibility needs to be studied in randomized controlled trials specifically designed to test this hypothesis. Also, more data are needed on the role of dietary regimens such as time-restricted eating and alternate day fasting on weight regain after weight loss.

Exercise

Regular exercise (or a high level of physical activity) has many beneficial effects on health, and some of the mechanisms for weight regain discussed before are attenuated by regular exercise. Exercise can help to maintain weight loss through mechanisms such as improvement in leptin sensitivity, increased sympathetic nervous system tone, decreased hunger, increased satiety feelings, increased dietary fat oxidation and preservation of muscle mass108. Exercise can also suppress chronic low-grade inflammation, and this beneficial effect is also seen in people with obesity109.

The evidence for a potential positive effect of exercise on weight regain mediated by increased leptin sensitivity is very limited and comes from animal studies110. According to a 2023 study111, obesity-associated leptin resistance is irreversible by diet-induced weight loss alone, because adipose tissue inflammation and the associated hypothalamic inflammation are sustained after diet-induced weight loss and leptin resistance is maintained. Exercise has been shown to reduce hypothalamic inflammation in obese animals and, based on surrogate peripheral markers, has a similar effect in humans with obesity112. It would be interesting to see whether exercise is also able to reduce this sustained hypothalamic inflammation after weight loss and thus increase leptin sensitivity and thereby potentially reduce weight regain.

A series of randomized controlled trials has investigated the role of exercise in different amounts added to a dietary intervention on long-term weight loss (≥1 year) in humans. In many of these studies no differences in long-term weight loss were found between the groups108,113,114, but some studies showed better long-term weight loss in the exercise group115 and that more exercise was associated with more weight loss116. However, randomized controlled trials in which participants were randomized to an exercise or no exercise maintenance programme are few117,118,119,120. Three of these studies found no significant difference in weight maintenance between the exercise and control groups117,118,119, but one trial found that weight regain over 1 year was less in the exercise group than in the control group120. In a study in which participants were randomized after the weight loss intervention to three different levels of physical activity, no differences in weight regain after weight loss were found among the groups121.

Post-hoc analyses of the negative trials often showed that individuals who exercised more had better long-term weight loss118,122,123,124,125,126. A 1997 study demonstrated that a higher physical activity-related energy expenditure after a weight-loss intervention, measured with doubly labelled water, predicted less future weight regain in women who had been weight stable for 1 month after the intervention but started to regain weight thereafter127. Furthermore, a higher self-reported level of physical activity during a follow-up period after a weight loss intervention has been found to be associated with less weight regain24,128,129. Individuals who successfully maintained their weight loss reported high levels of physical activity130, and had a higher physical activity-related energy expenditure and total daily energy expenditure, measured with doubly labelled water, than weight-matched controls131. Subsequent decreases in physical activity over time in these individuals were associated with weight regain132.

Although the evidence on causality is limited, these data suggest that exercise and physical activity could contribute to better weight maintenance after weight loss. However, long-term adherence is problematic in randomized trials as some participants in control groups might voluntarily choose to become more physically active, while some participants in the exercise intervention group might not be able to adhere to the prescribed exercise, reducing the between-group difference.

Whether differences in effectiveness exist among different types of exercise (for example, aerobic exercise, high-intensity interval exercise, resistance exercise or combinations) has not been studied.

Pharmacotherapy

Currently available pharmacotherapy for the management of obesity mainly targets the appetite system. and these drugs, particularly the newer generation, have been shown to be very effective for long-term weight loss133,134,135. A number of studies specifically addressed the effect of GLP1 receptor agonist treatment on weight regain after weight loss120,136,137 or of subsequent withdrawal of the drug on weight regain after weight loss138. In a 2013 the study, participants were randomized to the GLP1 receptor agonist liraglutide or placebo treatment for 56 weeks after a 12-week weight loss period during which participants in both groups had lost on average 6% of body weight on an energy-restricted diet. At the end of the 56 weeks, participants in the liraglutide group had significantly lower body weight than those in the placebo group. However, this fact was mainly due to continued weight loss in the liraglutide group, and body weight did not change much in the placebo group136. Therefore, no conclusion about the effect of liraglutide on weight regain as such can be drawn from this study. A 2015 study had a similar design but a larger degree of initial weight loss (12%)137. At the end of the 56-week weight maintenance period, body weight was significantly lower in the group treated with liraglutide than in the placebo-treated group, but this was again mainly because weight loss continued in the early phase of the weight maintenance period in the liraglutide group. After reaching the nadir of weight loss, the rate of weight regain did not seem to be less in the liraglutide group than in the placebo group137.

By contrast, in a 2021 study120 there was also some further weight loss in the liraglutide group during the weight maintenance phase, but in this study the weight regain from the nadir of weight loss appeared to be less in the liraglutide group than in the placebo group. Adding an exercise programme during the weight maintenance phase improved weight maintenance in both the liraglutide group and the placebo group120. A 2021 study found that switching some of the participants to placebo after a 20-week run-in treatment with semaglutide resulting in a 10% weight loss, caused weight regain in the placebo group over the 48-week follow-up period, whereas the semaglutide group lost an additional 8% of body weight138.

Rosenbaum and colleagues found that low-dose leptin replacement could reverse the adaptations resulting from a 10% weight loss (reduction in energy expenditure, satiation and perception of amount of food eaten) in humans with obesity. In some of their studies leptin replacement to pre-weight-loss levels resulted in extra weight loss54,139, but in other studies this change in body weight was not statistically significant140,141,142. Whether leptin administration reduces weight regain after weight loss has not been directly tested in humans. However, leptin administration after weight loss in rats with diet-induced obesity did not prevent subsequent weight regain143.

New anti-obesity drugs that are currently being investigated for human use mostly target the appetite system and include GLP1 dual and triple agonists, leptin sensitizers, Y2-receptor agonists, amylin analogues, growth differentiation factor 15 and agents that interfere with the ghrelin pathway (ref. 38). Pharmacological options to target FFM sparing during weight loss and to target the metabolic adaptations induced by weight loss (reductions in energy expenditure, FFA oxidation and lipolysis) include mitochondrial proton leak and uncouplers, mitochondrial dynamics and biogenesis, calcium and substrate cycles, activity of the sympathetic nervous system, leptin, browning of white adipose tissue, G protein-coupled receptor 75, growth hormone, activin type II receptor inhibition and urocortin 2 and 3 (refs. 38,144,145).

Studies in animals have revealed several ways to influence body weight regulation that can be explored to find novel targets for the prevention of weight regain. A 2022 study146 identified obesity-induced genes in mouse adipocytes, the expression of which did not change during weight loss. For 19 of these persisting genes, knockout of the homologous gene in the worm Caenorhabditis elegans led to a decrease in food intake. Knocking out of one of these genes, Atp6v0a1, the gene coding for a component of the vacuolar-ATPase complex, which serves to pump protons across membranes in several cellular organelles by hydrolysing ATP, decreased both food intake and body weight in HFD-fed mice. Gene and protein expression of adipocyte ATP6v0a1 is also increased in adipocytes from people with obesity compared with those from lean controls, and persists after weight loss in human obese adipose tissue. Pharmacological inhibition of vacuolar-ATPase by bafilomycin in HFD-fed mice that had passed through a cycle of weight gain and loss lowered food intake, reduced fat mass and blunted weight regain.

With respect to drugs targeting the immune system to lower adipose tissue inflammation, which often accompanies obesity and weight loss, NSAIDs and glucocorticoids could be used. However, systemic administration of such drugs is currently accompanied by unwanted side effects such as impaired glucose tolerance, storage of lipids in the liver and muscle atrophy147,148. A 2021 study149 targeted dexamethasone to mouse macrophages by linking the drug to dextran, which binds to the mannose receptor and scavenger receptors on the surface of macrophages. This led to a drug that was 20 times more potent than the free drug, as revealed by downregulation of inflammatory genes in the gonadal adipose tissue. Mice receiving peritoneal injection of dextran-linked dexamethasone showed a statistically significant weight loss after 4 weeks, accompanied by increased circulating levels of FFAs due to the induction of lipolysis by dexamethasone150. This finding shows that targeting macrophages with anti-inflammatory drugs could help with weight management and potentially reduce weight regain.

A 2022 study151 showed that lowering mitochondrial levels of iron in mouse macrophages is associated with an increase in the levels of M2 phenotype macrophages and with secretion of fewer pro-inflammatory and more anti-inflammatory cytokines by these macrophages. Lowering macrophage mitochondrial iron levels protected mice from systemic HFD-induced metabolic deterioration.

Specific macrophages in adipose tissue are located close to neurons and import and degrade noradrenaline152,153. Obesity in the mouse leads to increased levels of these macrophages, which acquire a pro-inflammatory phenotype upon stimulation of the sympathetic nervous system. Blocking noradrenaline uptake by these macrophages in mice induces adipose tissue browning, increased thermogenesis and sustained weight loss. Noradrenaline degradation by adipose tissue macrophages in mice is upregulated during ageing and could be responsible for age-related weight gain147. This particular class of macrophages has also been identified in humans146, and could be an interesting target for weight management and prevention of weight regain.

Biomedical strategies

Instead of aiming at macrophages, weight management could be attempted by specifically reducing adipose tissue mass. In a 2004 study154 that targeted a proapoptotic peptide to endothelial cells of the adipose vasculature, overall body mass of treated mice reduced by 30% in 4 weeks, whereas their epididymal adipose tissue mass reduced by more than 70%. A 2022 study155 used gold particles modified with adipose homing peptide and phosphatidylserine to coat adipocytes. This coating with phosphatidylserine (which is normally only exposed to the extracellular environment during apoptosis) caused the adipocytes to be recognized as apoptotic cells and they were attacked and cleared by macrophages. Treatment of obese mice with these particles led to a 24% reduction in body weight after 15 days as compared with body weight before treatment. The photothermal property of the gold particles was employed to further increase the weight loss. The abdominal region of the mice was irradiated with laser light inducing a photothermal effect of the gold nanoparticles, which led to depletion of the lipid droplets from coated adipocytes followed by cell shrinkage and necrosis. By this treatment the weight loss was increased to 33%. After the treatment period the mice were put on a HFD for another 15 days. The treated mice regained significantly less weight than all control groups, showing the potential of this approach for weight loss and prevention of weight regain.

In summary, weight regain after weight loss can be influenced to some extent by diet, exercise and currently available pharmacotherapy. However, more specific studies are needed to be able to fully judge the efficacy of all of these strategies. New targets for obesity management based on the physiological mechanisms described in this paper are being discovered and not only should be investigated for their weight-loss effects, but also need to be tested specifically for their effect on weight regain after weight loss. Given the multifactorial nature of weight regain, a combination of different strategies is most probably needed for the optimal prevention of weight regain after weight loss.

Conclusions

We review several aspects of the physiology associated with obesity, the effects of lifestyle-induced weight loss, mainly in the form of energy-restricted diets, and the persistence of such obesity-induced and weight-loss-induced changes. Furthermore, we looked for evidence for the role of these physiological changes in weight regain after weight loss and discuss strategies to limit or prevent weight regain on the basis of these changes. Here, we summarize our findings and provide suggestions for future research (Box 2).

Obesity is accompanied by infiltration of macrophages and other immune cells into adipose tissue. After weight loss, part of this immune profile persists. Limited evidence in mice and humans indicates that this obesogenic memory is involved in the risk of weight regain. However, more studies are needed to characterize the exact combination of cells that contribute to the obesogenic subset, in particular in humans, and to determine the specific role of each cell type in the risk of weight regain. So far, no studies have been performed in humans to investigate whether interfering with the development or persistence of this obesogenic memory results in improved long-term weight-loss outcomes.

A low REE can predict weight gain and, in a single study, a lower REE in people with obesity was shown to predict a greater degree of weight regain after weight loss than a higher REE37. Overall, a reduction in (resting) energy expenditure (also known as metabolic adaptation) is found with weight loss on an energy-restricted diet, but there is controversy about the persistence of this metabolic adaptation once energy balance is restored. So far there is no evidence that the degree of weight-loss-induced metabolic adaptation predicts weight regain.

People with obesity tend to have lower levels of lipolysis and lipid oxidation than people without obesity. Weight loss resulting from an energy-restricted diet might further reduce lipolysis and lipid oxidation. There is some evidence that these reductions in lipolysis and lipid oxidation persist after weight loss has ceased, and that the degree of these reductions are positively correlated with the degree of weight regain. A strategy to counteract the reduction in energy expenditure, lipolysis and lipid oxidation is increased exercise or physical activity. However, there are several reasons why studies on the effects of exercise on long-term weight regain are not always convincing. One reason is that long-term adherence to an exercise regimen is often problematic. Another reason could be that the increased energy expenditure with exercise is to some extent compensated by an increased energy intake. These facts argue for more comprehensive behaviour change strategies and environmental changes that make a persistent increase in physical activity easier and that prevent or limit compensation by increased energy intake, such as by adjusting diet composition.

Obesity is accompanied by disturbed patterns of fasting and postprandial blood concentrations of gut hormones, both orexigenic (ghrelin) and anorexigenic. However, there is large heterogeneity among study results which might reflect the different appetite phenotypes present among people with obesity. The findings on the effects of weight loss resulting from energy restriction on gut hormone concentrations and their persistence after weight loss are inconsistent and the evidence that weight-loss-induced changes in concentrations of gut hormones are associated with weight regain is controversial. Nevertheless, pharmacotherapy targeting the effects of gut hormones, such as GLP1 receptor single or dual agonists, is currently the most effective anti-obesity medication being associated with considerable long-term weight loss. However, whether these drugs also reduce weight regain after the nadir of weight loss has been reached is not clear and needs further study. There are other novel targets for pharmacotherapy for weight regain based on the physiological mechanisms discussed above, and hopefully new drugs that also specifically prevent weight regain will be developed in the near future. So far, most of the evidence for the role of obesity-induced or weight-loss-induced changes in weight regain comes from association studies. Studies on the causality of these changes are largely lacking and need to be done to gain better insights into the underlying mechanisms.

Owing to the multiple factors that are involved in the development of obesity, lifestyle-induced weight loss and weight regain, it is probable that multiple targets need to be tackled for optimal weight management. Improvement in lifestyle habits in combination with pharmacological treatments (with one or multiple drugs) that also prevent weight regain could be the future for long-term obesity management and the wellbeing of people with obesity. Moreover, given the multitude of factors that regulate body weight, it is probable that different phenotypes, not only for obesity but also for weight loss and weight regain, exist and personalization of treatment could help further improve the effectiveness of treatment. Presently, studies addressing weight regain after weight loss are a minority of studies on weight management, in both humans and laboratory animals. The effects that weight regain has on long-term weight control and health, both in individuals and in the population in general, warrant considerably more attention from the scientific and medical field.