The ancient prescription of Hippocrates (400 BC) that the obese should eat less and exercise more( Reference Procope 1 ) continues nowadays to be a widespread approach for weight management despite its well-documented failures( Reference Bacon and Aphramor 2 ). Weight regain is generally the rule, with long-term follow-up studies indicating that one-third to two-thirds of the weight lost is regained within 1 year and almost all is regained within 5 years( 3 ). Repeated cycles of weight loss and weight regain, referred to as weight cycling or yo–yo dieting are a frequent occurrence, with the potential for increased cardiovascular risks( Reference Montani, Viecelli and Prévot 4 ). Despite these failures, national surveys in Europe and North America indicate that 30–50% women and 10–30% men are currently or have recently attempted dieting to lose weight( Reference Kruger, Galuska and Serdula 5 – Reference Lahti-Koski, Mannisto and Pietinen 7 ). With studies of the long-term outcomes of energy-restricting diets showing that at least one-third of dieters regain more weight than they lost( Reference Mann, Tomiyama and Westling 8 ), together with population groups as diverse as obese and lean, young and old, sedentary and athletes attempting to lose weight on some form of diet therapy( Reference Montani, Viecelli and Prévot 4 , Reference Kruger, Galuska and Serdula 5 ), there is concern as to whether dieting may paradoxically be promoting exactly the opposite of what it is intended to achieve. To put it bluntly, does dieting make people fatter as proposed nearly three decades ago by Cannon and Einzig( Reference Cannon and Einzig 9 ) in a book entitled Dieting makes you fat.
Prospective studies linking dieting to future weight gain
In an article published in 1994 in which a panel of experts for the National Task Force on the Prevention and Treatment of Obesity( 10 ) addressed concerns about the effects of weight cycling and provided guidance on the risk-to-benefit ratio of attempts at weight loss, a main conclusion was that ‘the (then) available evidence is not sufficiently compelling to override the potential benefits of moderate weight loss in significantly obese patients’. In the decade that followed, however, more than a dozen prospective studies( Reference French, Jeffery and Forster 11 – Reference Pietiläinen, Saarni and Kaprio 24 ), conducted over periods ranging from 1 to 15 years, have suggested that dieting to lose weight is associated with future weight gain and obesity, with many of them showing this association even after adjustment for potential confounders such as baseline BMI, age and several lifestyle and behavioural characteristics. Particularly informative are the 3-year follow-up studies of Stice et al. ( Reference Stice, Cameron and Killen 15 ) showing that adolescents with baseline dieting had three times the risk of onset of obesity than the non-dieters, and the 6–15-year follow-up study of Korkelia et al. ( Reference Korkeila, Rissanen and Kaprio 16 ) reporting that initially normal-weight subjects who were attempting to lose weight had two times the risk of major weight gain (>10 kg) than non-dieters. In contrast, the history of weight-loss attempts in initially overweight men and women subjects in the latter cohort( Reference Korkeila, Rissanen and Kaprio 16 ) was not consistently associated with increased risk of major weight gain, thereby raising the possibility that the long-term impact of dieting on proneness to fatness may be greater in the lean than in the obese, albeit in young and middle-aged population groups. Furthermore, the recent analysis of a large population-based cohort with a follow-up from adolescence to young adulthood, suggest a dose-dependent association between the number of lifetime intentional weight losses, gain in BMI and risk of overweight( Reference Pietiläinen, Saarni and Kaprio 24 ). A single episode of weight loss was found to increase the risk of becoming overweight by the age of 25 years almost 3-fold in women and 2-fold in men. In addition, women who reported two or more weight-loss episodes had an even higher (5-fold increased) risk of becoming overweight at the age of 25 years compared with subjects with no intentional weight loss. Weight cycling has also been shown to be a predictor of subsequent weight gain and the risk of obesity in athletes( Reference Saarni, Rissanen and Sarna 25 ). In a national cohort of 1838 male elite athletes who had represented Finland in international sport competitions between 1920 and 1965, men who performed power sports (boxers, weight lifters and wrestlers), where weight cycling is common, gained 5·2 units of BMI from age 20 to 60 years. In contrast, men without an athletic background gained only 4·2 units of BMI, and athletes without weight cycling gained even less (3·3 units of BMI). These findings are therefore in support of the contention that dieters and weight cyclers may be more prone to future weight gain than non-dieters and non-cyclers.
Causality in the link between dieting and propensity to fatness
But whether dieting per se is a causative factor for subsequent weight gain and contributes to the current obesity epidemic is, however, debatable( Reference Hill 26 – Reference Cannon 28 ). Hill( Reference Hill 26 ) has argued that ‘it is not that dieting makes you fat, but that being fat makes you (more likely to) diet’, a contention that is supported by data from the Finnish Twin cohort study( Reference Korkeila, Rissanen and Kaprio 16 ) showing that dieting aggregated in families, and hence suggesting a familial predisposition to gain weight. Indeed, twin studies have demonstrated that, like proneness for obesity, episodes of intentional weight loss have substantial genetic components( Reference Keski-Rahkonen, Neale and Bulik 29 ), such that the possibility arises that subjects who are genetically most prone to obesity end up dieting the most and subsequently gain the most weight. In a recent study that directly addressed this question of whether weight gain associated with dieting is better related to genetic propensity to weight gain than to the weight-loss episodes themselves, Pietiläinen et al. ( Reference Pietiläinen, Saarni and Kaprio 24 ) investigated the association between dieting and weight gain in 4129 individual twins whose weight and height were obtained from longitudinal surveys at 16, 17, 18 and 25 years and examined in relation to the number of lifetime intentional weight-loss episodes of >5 kg at 25 years. In monozygotic twin pairs discordant for intentional weight loss, co-twins with at least one weight-loss episode were found to be 0·4 BMI units heavier at 25 years than their non-dieting co-twins despite no differences in baseline BMI levels. Similarly, in dizygotic twin pairs, co-twins with intentional weight losses gained progressively more weight than non-dieting co-twins (BMI difference 1·7 units at 16 years and 2·2 units at 25 years). Overall, therefore, these findings not only confirm previous studies that dieters may be more prone to future weight gain and that dieters have a genetic propensity for obesity, but they also provide evidence that dieting per se may promote subsequent weight gain, independent of genetic factors, in an essentially normal-weight cohort (i.e. <10% overweight as adolescents). Support for this contention that dieting per se may drive excess weight gain in non-obese adults, can in fact be derived from classic studies of food deprivation and refeeding showing that more weight is recovered than is lost; a phenomenon referred to as post-starvation weight (or fat) overshooting( Reference Dulloo 30 ).
Post-starvation fat overshooting
The phenomenon of post-starvation weight overshoot, first documented by Benedict( Reference Benedict 31 , Reference Benedict, Miles and Roth 32 ) in studies of experimental starvation, was subsequently observed during follow-up studies of famine victims of World War II, in men volunteers subjected to experimental semi-starvation or in young army recruits recovering from an 8-week training that included food and sleep deprivation. As shown in Table 1 and elaborated later, data available on dynamic changes in body weight, food intake and/or body composition are informative about the consequences of large weight losses and subsequent weight regain resulting in weight overshooting.
* Increased energy intake above levels prior to energy deprivation.
† Data calculated from graph published by Martin & Demole( Reference Martin, Demole and Apfelbaum 34 ).
‡ Body fat measured by hydrodensitometry.
§ Body fat measured by dual-energy X-ray absorptiometry.
The Swiss food rationing experiment
In Switzerland, when food rationing was implemented in 1941, the physiologist Fleisch( Reference Fleisch 33 ) was asked by the ‘Federal Commission for Wartime Feeding’ to monitor the health consequences of this intervention. In addition to the assessment of mortality and morbidity related to CVD, digestive and infectious diseases, the nutritional status was also monitored in 700 persons of different ages, social status and living areas, every 1–2 months between autumn 1941 and spring 1946. The data collected on food consumption and body weight, compiled and published about 30 years later by Martin and Demole( Reference Martin, Demole and Apfelbaum 34 ) indicate that food intake and body weight dropped to the lowest level during the summer of 1945. At the end of that year, food rationing was abandoned and products were available again in larger quantities; food consumption and body weight rose rapidly. During the subsequent months when food availability improved further, the phenomena of energy overcompensation (hyperphagia) and weight overshooting became evident (Table 1).
The Minnesota experiment
Coincidentally, it was during the same time period that Keys et al. ( Reference Keys, Brozek and Henschel 35 ), motivated by the desire to obtain quantitative information upon which to base an efficacious relief programme directed at nutritional rehabilitation of the starvation victims of World War II, were conducting their study of experimental semi-starvation and refeeding in Minnesota. The thirty-two healthy volunteers who completed the study (many of whom were conscientious objectors of war), were in continuous residence at the University of Minnesota during the 12-week control baseline period, 24 weeks of semi-starvation and 12 weeks of restricted rehabilitation; furthermore, twelve of them also remained in the laboratory during the first 8 weeks after removal of dietary control, i.e. ad libitum refeeding. The total loss in body weight of about 25% of the original weight in each individual reproduced the conditions of severe semi-starvation, and is comparable with values found in severe famine. Following removal of dietary control at the end of week 12 of restricted rehabilitation, the food intake increased markedly above the pre-starvation level, and this hyperphagic response persisted for several weeks after body weight had reached the pre-starvation level, and contributed to weight overshooting mostly as fat.
The Army Ranger multi-stressor experiments
In more recent years, similar body weight and fat overshooting have also been reported in young men at the US Army Ranger School recovering from about 12% of weight lost following 8–9 weeks of training in a multi-stressor environment that includes energy deficit and sleep deprivation( Reference Nindl, Friedl and Frykman 36 – Reference Friedl, Moore and Hoyt 38 ). Nindl et al. ( Reference Nindl, Friedl and Frykman 36 ) found that at week 5 in the post-training recovery phase, body weight had overshot by 5 kg, reflected primarily in large gains in fat mass, with all ten subjects showing higher fat mass than before weight lost. Similarly, in young male volunteers participating in another 8-week US Army Ranger course that involved four repeated cycles of restricted energy intake and refeeding, Friedl et al. ( Reference Friedl, Moore and Hoyt 38 ) found that more weight was regained than lost at week 5 of recovery following training cessation, with fat overshooting representing an increase of 40% in body fat above pre-training levels. Data obtained in a parallel group of subjects revealed that hyperphagia peaked at about 4 weeks post-training, thereby suggesting that hyperphagia was likely persisting over the last week of refeeding during which body fat had already exceeded the baseline.
The search for control systems regulating body composition during weight recovery
The earlier observations raise the question of what drives hyperphagia well after body weight or body fat had been fully recovered. Is it an explanation based solely on long-lasting psycho-biological reaction to food deprivation? Preoccupations with food and food obsession have often been described long after episodes of food deprivation or dietary restraint( Reference Polivy and Herman 39 ). However, what is particularly striking from a closer inspection of data on the real-time pattern of hyperphagia relative to changes in body composition in the men recovering weight in the Minnesota Experiment (Fig. 1) is that when their body fat had been completely recovered (i.e. 100% of control period), at which point the fat-free mass (FFM) was not yet fully recovered, the hyperphagia was still very much evident. It only disappeared as FFM recovery approached 100% of the control level. These observations about hyperphagic overcompensation and its relationship with changes in body composition suggest a link between lean tissue recovery relative to fat recovery, sustained hyperphagia and fat overshooting, and raise fundamental questions about how control systems operate to re-establish lean and fat tissues during weight recovery:
(i) What is the relative importance of fat and lean tissue depletion as determinants of post-starvation hyperphagia?
(ii) What drives the faster rate of fat recovery relative to lean tissue recovery even in the absence of hyperphagia as observed in the period of restricted refeeding in the Minnesota Experiment? Could this be explained by the adaptive reduction in energy expenditure (i.e. suppressed thermogenesis) that occurs during weight loss and which persists during weight recovery? What then is the relative importance of fat and lean tissue depletion as determinants of such adaptive suppression of thermogenesis during weight recovery?
(iii) What determines the large inter-individual variability in partitioning of energy into lean and fat tissues during weight recovery?
In addressing these questions, we revisited the Minnesota Experiment and applied more elaborate statistical and analytical tools in a re-analysis of the data on the dynamic changes in body composition, food intake and BMR in response to the 24 weeks of semi-starvation, 12 weeks of restricted refeeding and subsequent 8 weeks of ad libitum refeeding. The methodological approaches and results of this re-analysis have been reported in detail elsewhere( Reference Dulloo, Jacquet and Girardier 40 – Reference Dulloo 44 ), and are summarised below.
Energy partitioning between lean and fat tissues: an individual characteristic
The notion that the control of partitioning between the body's protein and fat compartments is an individual characteristic was first proposed by Payne and Dugdale( Reference Payne and Dugdale 45 ) and supported by data from studies of prolonged fasting showing that the fraction of energy mobilised from protein (termed the P-ratio) during weight loss is relatively constant in a given individual but varies considerably between individuals( Reference Henry, Rivers and Payne 46 ). Using a quantitative index of energy partitioning defined as body energy mobilised from protein during weight loss (the semi-starvation P-ratio) or as the proportion of energy gained as protein during weight recovery (the refeeding P-ratio) in the first part of our Minnesota Experiment re-analysis( Reference Dulloo, Jacquet and Girardier 40 ), we showed that the highly variable P-ratio during refeeding is strongly correlated with the P-ratio during semi-starvation, particularly after controlling for inter-individual variability in the degree of fat recovery (Fig. 2(a)). This demonstration that the individual's P-ratio during semi-starvation is conserved during refeeding therefore provides direct evidence to the proposal of Payne and Dugdale( Reference Payne and Dugdale 45 ) that the control of energy partitioning is an individual characteristic. Furthermore, in the search for predictors of the large inter-individual variability in energy partitioning( Reference Dulloo, Jacquet and Girardier 40 ), we found that the initial % body fat (i.e. % body fat prior to weight loss) to be the most important determinant of inter-individual variability in P-ratio (Fig. 2(b)), which suggests that the higher the initial adiposity, the lower the proportion of energy mobilised from body protein, and hence the greater the propensity to mobilise fat during weight loss and to subsequently deposit fat during weight recovery. In order to examine in more detail the relationship between initial adiposity and the partitioning characteristic (Pc) of individuals, we have pooled the Minnesota data on P-ratio in normal-weight Caucasians with those calculated for obese Caucasians undergoing prolonged fasting( Reference Henry, Rivers and Payne 46 ) or semi-starvation( Reference Passmore, Strong and Ritchie 47 ). As can be seen in Fig. 2(b), >85% of the variance in the P-ratio could be explained by the % body fat prior to weight loss. The steep part of the exponential curve lies between 8 and 20% body fat, and a shift from the lower to the upper values in this range, generally considered to reflect a ‘normal’ range of adiposity for men living in affluent societies, results in 2·5–3-fold reduction in P-ratio. This extremely high sensitivity of the P-ratio with regard to the initial body composition emphasises the critical importance of even small differences in the initial % body fat in dictating the individual's energy-Pc and, hence, the pattern of lean and fat tissue deposition during weight loss and subsequent weight recovery.
Suppressed thermogenesis: feedback signals from fat tissue
It is well established from longitudinal studies of fasting and energy restriction that the reduction in BMR and total energy expenditure during weight loss is greater than can be accounted for by the loss of active tissues. This reduction in the energy cost of maintenance is adaptive in that it reduces the rate at which body's tissues are being depleted, and has been demonstrated both in normal-weight subjects( Reference Keys, Brozek and Henschel 35 , Reference Luke and Schoeller 48 – Reference Rosenbaum, Hirsch and Gallagher 50 ) as well as in obese individuals( Reference Luke and Schoeller 48 – Reference Rosenbaum, Hirsch and Gallagher 50 ). In order to determine whether this adaptive reduction in thermogenesis during weight loss persists during weight recovery, and underlies the disproportionately greater rate of recovery of fat mass relative to FFM, the dynamic changes in body composition and BMR of the thirty-two Minnesota men who completed the 24 weeks of semi-starvation and first 12 weeks of restricted refeeding were re-analysed. The data on changes of BMR after adjusting for changes in FFM and fat mass (i.e. an index of energy conservation through suppressed thermogenesis) revealed that (i) the suppression of thermogenesis which was evident during the phase of weight loss persisted during weight recovery and (ii) the extent to which thermogenesis was suppressed during the phases of weight loss and weight recovery was determined not only by the food energy deficit per se but also by the extent to which body fat was depleted( Reference Dulloo and Jacquet 42 ). This continuum in the relation between suppressed thermogenesis and fat depletion (and not with FFM depletion) during both phases of weight loss and weight recovery (Fig. 3) therefore reflects the operation of a control system with a negative feedback loop between a component of adaptive thermogenesis and the state of depletion of the fat stores, such that suppressed thermogenesis during weight recovery would accelerate body fat recovery. Quantitatively, this suppressed thermogenesis during weight recovery is calculated to represent an energy economy of 10–15% in BMR when integrated between 0 and 100% fat recovery( Reference Dulloo, Jacquet and Girardier 40 , Reference Dulloo and Jacquet 42 ).
Integrating the control of energy-partitioning and adaptive thermogenesis
An integration of these control systems in the regulation of body composition during a cycle of weight loss and weight recovery is discussed with the help of a schematic diagram presented in Fig. 4. This diagram embodies the findings that the control of body energy-partitioning between protein and fat is an individual characteristic, i.e. individuals vary in their Pc during weight loss and weight recovery, and takes into account the two distinct control systems for adaptive thermogenesis which can operate independently of each other: (i) one that is a direct function of food intake and energy balance (referred to as the non-specific control of thermogenesis, which is under the control of the sympathetic nervous system) and (ii) the other that is a direct function of the state of depletion of the fat stores (referred to as the adipose-specific control of thermogenesis)( Reference Dulloo and Jacquet 51 ). During starvation, the control of partitioning determines the relative proportion of protein and fat to be mobilised from the body as fuel (i.e. the individual's Pc), and the energy conserved due to suppressed thermogenesis is directed at reducing the energy imbalance, with the net result that there is a slowing down in the rate of protein and fat mobilisation in the same proportion as defined by the Pc of the individual. During starvation, therefore, the functional role of both control systems underlying suppressed thermogenesis is to reduce the overall rate of fuel utilisation. During refeeding, the control of partitioning operates in such a way that protein and fat are deposited in the same relative proportion as determined by the Pc of the individual during starvation, and the increased availability of food leads to the rapid removal of suppression upon the non-specific (sympathetic nervous system-mediated) control of thermogenesis. In contrast, the suppression of the thermogenesis under adipose-specific control is only slowly relieved as a function of fat recovery, such that the energy that continues to be spared is directed specifically at the replenishment of the fat stores. The net effect, as demonstrated( Reference Dulloo, Jacquet and Girardier 40 ) using both statistical and numerical approaches in our re-analysis of data from the Minnesota Experiment, is that fat is deposited in excess of that determined by the Pc of the individual, thereby contributing to the disproportionate rate of fat relative to lean tissue recovery. Direct evidence in support for the existence of mechanisms that suppress thermogenesis and accelerate specifically fat mass (and not FFM) can be obtained from animal studies of refeeding after energetic restriction( Reference Dulloo and Jacquet 51 ). In human subjects, a role for suppressed thermogenesis driving catch-up fat can also be derived from patients recovering from malnutrition resulting from non-neoplastic gastrointestinal disease( Reference Carbonnel, Messing and Rimbert 52 ), as well as from men and women recovering weight (essentially fat) over 6 months after 2 years of sustained energy restriction in the biosphere 2 experiment( Reference Weyer, Walford and Harper 53 ).
Compensatory hyperphagia: feedback signals from fat and lean tissues
To gain insights into the determinants of post-starvation hyperphagia in the Minnesota Experiment, the individual data on food intake, body fat and FFM of the twelve subjects who remained in the laboratory during the 8 weeks period of ad libitum refeeding (Fig. 1) were used to calculate the following variables( Reference Dulloo, Jacquet and Girardier 41 ):
(i) A quantitative index of hyperphagia, i.e. the total hyperphagic response during the 8-week ad libitum refeeding period, calculated as the energy intake in excess of that during the pre-starvation (control) period;
(ii) A quantitative index of the degree of fat and FFM depletion just before ad libitum refeeding, calculated as the deviation in fat and FFM from their respective pre-starvation values; and
(iii) A quantitative index of the deficit in energy intake just before ad libitum refeeding, calculated as the difference between the energy intake during the period of restricted refeeding and that during the pre-starvation (control) period.
The results indicate that the hyperphagic response over the 8-week ad libitum period was inversely and independently correlated to both the degree of fat depletion and FFM depletion (Fig. 5). Furthermore, the results of stepwise-regression analysis indicated that of the three determinants of hyperphagia, it is the degree of fat depletion that is the strongest, with a significant contribution from lean tissue depletion as well as from the energy intake deficit prior to ad libitum refeeding; the three variables together explained nearly 80% of variability in post-starvation hyperphagic response. A considerable component of the hyperphagic response to energy deprivation can therefore be regarded as the outcome of an autoregulatory control system that contributes to the restoration of body weight and body composition, with feedback loops between the state of depletion of both the fat and FFM compartments and hunger–appetite centres in the central nervous system. These findings therefore suggest the existence not only of adipostatic mechanisms, e.g. via leptin system( Reference Kissileff, Thornton and Torres 54 ), but also of proteinostatic mechanisms in the control of food intake.
The existence of an appetite mechanism driven by the demands for protein generated by lean tissue growth is consistent with the animal and human literature that the nutrient requirements during growth or catch-up growth and the control of food intake are dominated by the impetus for lean tissue growth( Reference Webster 55 – Reference Millward 57 ). More recently, the notion that a signal(s) associated with lean mass exerts a determining effect over self-selected food consumption has also been proposed by Blundell et al. ( Reference Blundell, Caudwell and Gibbons 58 ), who in a reassessment of data on food intake over a 12-week intervention period in adult human subjects found that it was FFM, but not fat mass or BMI, which predicted (as well as correlated with) meal size and daily energy intake. They postulated that this signal may interact with a separate class of signals generated by fat mass. In his proposal for ‘protein–stat’ mechanisms that regulate lean body mass, Millward( Reference Millward 57 ) emphasised that its mode of operation would require an aminostatic component of appetite regulation in which food intake is adjusted to provide the amino acids and protein needs for lean tissue growth or maintenance. However, the evidence in support of the aminostatic theory is fragmentary, and the extent to which the protein requirements for lean tissue (re)growth and appetite control are mediated by plasma changes in amino acids is unknown. An alternative explanation may reside in the increasingly recognised role of the skeletal muscle as an endocrine organ. Recent applications of proteomic approaches to investigate factors secreted by skeletal muscle( Reference Trayhurn, Drevon and Eckel 59 ) have revealed that myocytes are capable of producing several hundred secreted proteins (i.e. myokines), the identity and function of most of which remain to be elucidated. The recognition that a multiplicity of adipokines and myokines are secreted by adipocytes and myocytes, respectively, opens new avenues for research towards their potential role in the control of food intake and adaptive thermogenesis.
Body composition autoregulation: a conceptual model
On the basis of the findings mentioned earlier, the control systems that operate through the control of energy-partitioning, adaptive thermogenesis and hunger–appetite can be incorporated into a conceptual model of autoregulation of body weight and body composition depicted in Fig. 6. In addition to the control of partitioning per se, other control systems operating via the control of food intake and thermogenesis with feedback loops from the lean and/or fat tissue compartments are conceptualised to be dictated by ‘memories’ of the FFM and/or fat compartment; these are, however, viewed as attenuators of energy imbalance and/or accelerators of tissue recovery that are superimposed over a more ‘basal’ control of energy partitioning.
Implications for ‘Dieting makes some fatter’
One important feature of this model (Fig. 6) is the sharp contrast between the determinants of the two accelerators: whereas the control system operating through increased hunger–appetite is dictated by the degree of depletion both in body fat and FFM, that operating through the adipose-specific suppression of thermogenesis is dictated specifically by the degree of depletion of body fat only (and not by FFM depletion). This differential relationship of hyperphagia and suppressed thermogenesis with regard to the two main energy-containing compartments suggests that there is an asymmetry in the way FFM and fat mass are recovered, with fat being recovered at a faster rate than FFM. Thus, the greater the severity of weight loss (and the degree of fat and FFM depletion), the more the suppression of thermogenesis that enhances specifically fat deposition (and not FFM deposition) and hence the greater the disparity in the rate of fat v. FFM recovery. This would provide an explanation for the fact that when fat recovery in the Minnesota men reached 100% of pre-starvation values, the FFM recovery was still far from complete (Fig. 1). Since depleted FFM can also drive hyperphagia, a consequence of disparity between 100% fat recovery and incomplete FFM recovery is that the hyperphagia is prolonged until FFM is also fully recovered. However, since the completion of FFM recovery can only be achieved through the process of energy-partitioning, more body fat is also deposited, which hence underscores the phenomenon of fat overshooting.
From a perspective of body composition autoregulation therefore, the critical event that eventually leads to the prolongation of hyperphagia and fat overshooting resides in the suppression of thermogenesis which drives fat recovery at a rate that is greater than that determined by the Pc of the individual. As this adipose-specific suppression of thermogenesis (that drives fat acceleration) is a function of fat depletion, and the prolongation of hyperphagia (after 100% fat recovery) is a function of depleted FFM still to be recovered, the extent of fat overshooting would therefore depend on the extent to which both fat mass and FFM are depleted; this in turn depends on the Pc of the individual which is dictated primarily by the initial adiposity (Fig. 2(b)). To substantiate this contention, we show here, from the individual body composition data of the twelve men who completed the Minnesota Experiment, that the extent of fat overshooting decreases exponentially with increasing initial adiposity (Fig. 7(a)). Using a mathematical model (J Jacquet and AG Dulloo, unpublished results) that incorporates the relationships shown in Figs. 2 and 3, the differential body composition responses of an average lean dieter and an average obese dieter losing 12% of their respective body weight is illustrated in Fig. 7(b); their changes in fat mass and FFM being superimposed on Forbes curve that relates fat mass and FFM( Reference Forbes 60 ). Should the lean dieter be subjected to multiple weight cycles, it can also be predicted that the amount of fat overshoot will nonetheless decrease with each successive cycle, since each cycle leads to increased adiposity. The cumulative fat overshoot over several cycles will nonetheless amount to substantial excess of body fat. These results predicting little or no fat overshooting in obese dieters in accord with studies in which obese individuals subjected to one cycle( Reference Wadden, Foster and Stunkard 61 ) or three successive cycles( Reference Jebb, Goldberg and Coward 62 ) of dieting failed to show altered body composition. They are also in line with data from lean individuals who during rehabilitation after losing about 12% of their weight due to food deprivation (Table 1) showed fat overshooting of about 4 kg( Reference Nindl, Friedl and Frykman 36 – Reference Friedl, Moore and Hoyt 38 ); these data are now superimposed on the curve shown in Fig. 7(a).
Our analysis of how dieting makes some fatter suggests that the lean dieters are at greater risk for fat overshooting than the obese dieters. This contention is in line with prospective studies indicating more consistent association with increased risks for major weight gain in initially normal-weight subjects than in initially overweight and obese subjects attempting to lose weight( Reference Korkeila, Rissanen and Kaprio 16 , Reference Pietiläinen, Saarni and Kaprio 24 ). With the prevalence of dieting increasing among individuals in the normal-weight range (due to pressure for a slim image, body dissatisfaction or athletic performance) and accumulating evidence suggesting increased cardiometabolic risks associated with weight fluctuations in the non-obese population groups( Reference Montani, Viecelli and Prévot 4 , Reference Dulloo, Jacquet and Seydoux 63 ), the notion that dieting makes some fatter warrants greater experimental scrutiny and deserves greater public health concern than so far acknowledged.
This work was partly supported by a grant from the Swiss National Science Foundation (Grant no. 31–130481). A. G. D. drafted the manuscript. J. J. contributed importantly to data analysis. J. J. and J-P. M. critically revised the manuscript for important intellectual content. All authors have read and approved the final manuscript. The authors have no conflict of interest to declare.