Voluntary weight loss: systematic review of early phase body composition changes


SB Heymsfield, Global Center for Scientific Affairs, Merck Research Laboratories, 126 E. Lincoln Avenue, PO Box 2000, RY34A-A238, Rahway, NJ 07065-0900, USA. E-mail: Steven.heymsfield@pbrc.edu


Weight loss follows when adult humans enter a phase of negative energy balance brought about by reducing energy intake and/or increasing energy expenditure. The weight loss period is usually viewed as a continuous process, ending when energy equilibrium is achieved at a lower weight or with death following depletion of fuel stores. However, growing evidence supports the expanded view that induction of negative energy balance leads to well-defined physiological effects characterized by three discrete phases (I-III). At present there are no comprehensive reviews of the ‘early’ phase of weight loss, a gap highlighted by recent interest in rapidly testing new treatments with short-term protocols. Herein we show from earlier reports and with new data that weight loss during phase I is: mathematically quantifiable with a t1/2 < 1-week and 4- to 6-week duration; includes well-defined rapidly evolving body composition and energy expenditure changes; and is moderated by multiple factors including subject sex and activity level, nutrients ingested at baseline and during the negative energy balance period, and hormone and pharmacologic treatments. Our in depth review collectively characterizes phase I as a distinct weight loss period while revealing important knowledge gaps that can be filled with appropriately designed future studies.


Weight loss is a phenomenon universally recognized and even experienced by most adults. Whether voluntary, as with dieting for excess adiposity, or involuntary with famine or disease, weight loss has captured the attention of scientists, clinicians and public health workers for more than a century.

Voluntary weight loss and associated changes in body composition take on new importance in light of the current obesity epidemic. Can observed weight changes with lifestyle interventions, including diet and exercise, be ascribed to loss of body fat or some other less desirable compartments? Is the composition of weight loss with lifestyle interventions ‘constant’ or are there time-dependent changes corresponding to well-orchestrated physiological and metabolic processes? Does subject baseline adiposity, age or sex influence the composition of weight loss? These are only a few of the many questions posed by investigators studying this topic for the past century and who continue to explore this increasingly important research area.

Studies aimed at evaluating dietary, pharmacologic and surgical treatments for obesity typically set body-weight change, a global measure of energy balance, as a primary efficacy measure. Optimally treatment measures should promote mainly loss of energy-dense body fat and limit the losses of functional compartments such as body protein and bone minerals. Accordingly, well-designed weight loss studies are increasingly including measures of body composition with the dual purpose of establishing efficacy based on changes in total and regional fat mass and safety based on changes in the mass of key lean tissue compartments. Several studies over the past two decades report original research or have assembled the published literature on this topic and developed working estimates reporting the expected composition of weight loss as fat and lean mass with fasting (1–3), very-low-calorie diets (VLCDs, (4,5)), low-calorie diets (5–9), pharmacologic interventions (5), exercise programmes (6,7) and bariatric surgical treatments (10). Several generalizations emerge from these previous reports, although important gaps remain. In particular, most of these earlier studies focus on long term treatment effects, usually 12 weeks or more, and very little collective information is available on temporal changes in body composition immediately following induction of negative energy balance. These short-term weight loss effects are relevant as a strong desire exists to rapidly screen and evaluate new therapies.

The aim of the current report is to fill the knowledge gap regarding early changes in body weight and composition following induction of negative energy balance through voluntary measures such as with lifestyle or pharmacologic treatments. Our reference to ‘early’ changes empirically focuses on time spans of up to 12 weeks after initiation of weight loss treatments. Technical features related to body composition evaluation are described throughout our review and we provide an overview of this topic in Supplementary Material, I.

Overview of weight loss phases

Weight loss ensues in three potential phases (I–III) when subjects are placed in a state of negative energy balance. The first phase, lasting several days or weeks, is associated with rapid or ‘fast’ weight loss (2,3). This is the weight loss phase on which we focus the present review.

The second phase that follows is associated with slower weight loss as shown in the upper panel of Fig. 1 by data collected from Subject L (Levanzin, (11)), the voluntary ‘professional’ male faster. The second phase of weight loss experienced by Subject L follows phase I, represented in the figure by a several-day period of what appears to be a rapid curvilinear loss in weight.

Figure 1.

(Upper panel) Daily weight of Subject L (11) over the course of his voluntary total fast with the exception of non-caloric fluids. (Lower panel) Semi-log plot of body weight vs. day of fast. The data are described by a two-compartment model developed using WinNonlin software (Pharsight, St. Louis, MO, USA) with an early rapid phase (open circles) and later slow phase (triangles). The slow-phase line along with half-life and λ2 of the equation 1 are shown in the figure.

A third phase, brief in duration and not well characterized in humans for ethical reasons, occurs once fat stores are depleted and available fuels during periods of negative energy balance are drawn almost entirely from body protein (12). Dulloo and Jacquet (13), however, debate the existence of an acute ‘pre-mortal rise in N excretion’ (i.e. phase III) and instead posit that fat and protein reserves are exhausted simultaneously at which point the subject succumbs from starvation. For practical and ethical reasons most of the available experimental research on humans focuses on phases I and II of weight loss.

Kinetic models

Weight loss follows a classic dose-response model so that when subjects are placed into negative energy balance the weight loss that ensues can be modelled by a linear combination of exponential decay functions. Phases I and II of weight loss can be formulated by the sum of two exponential decay terms:


where W(t) represents weight on day t and the coefficients satisfy the property that f1+f2= 1 and λ1 >> λ2(2,3). The kinetics of weight loss observed during Subject L's 31 d fast in 1912 are shown in Fig. 1 as a semi log plot, with a short phase I half-life of 1.9 d and phase II following with a longer half-life of 109.3 d. Nearly all carefully conducted weight loss studies that include a baseline stabilization period show this characteristic two phase weight loss pattern. The weight loss curve during phases I and II appears as one continuous smooth curve even though phases I and II are governed by different rates of decay.

Forbes first formalized these mathematical associations and derived model parameters for equation 1 based on available literature at the time (2,3). We have rearranged and added to Forbes' estimates as shown in Table 1. The half-life of phase I based on the limited available weight data is generally short, typically less than 1 week, and is longer in obese than in lean subjects. If we assume 4–5 half-lives reaches 94–97% completion of phase I, the upper limit based on phase I estimates from weight data in Table 1 is about 1-month with a range of 5–26 d.

Table 1.  Kinetics of body weight and nitrogen loss during periods of negative energy balance
f1λ1 (per day)t1/2f2λ2 (per day)t1/2
  1. Sources: Modified from references 2 and 3.

  2. The table presents parameters of the exponential function: (equation 1).

  3. F, female; f, fraction of body weight; λ, decay constant; M, male; t1/2, half-life.

Sex/weight (n)      
 M/60.6 kg (1)0.0490.3641.90.9510.00634109.3
 M/63.6 kg (6)0.0570.3681.90.9430.00544127.4
 M/128 kg (3)   0.9610.00261265.6
 M/142.6 kg (9)0.0380.2442.80.9620.00364190.4
 F/56.3 kg (1)0.0510.5861.20.9490.00644107.6
 F/99.3 kg (11)0.0120.3771.80.9880.0022315
 F/131 kg (3)   0.9690.00289239.8
 F/131.3 kg (3)0.0430.1345.20.9570.00296234.2
 Non-obese (4)0.010.2842.40.990.00599115.7
 Obese (9)0.060.065510.60.940.0016433.2

The half-life for phase II is considerably longer than that for phase I, ranging in the limited available data from about 100 d in lean subjects to 300 d (i.e. 14–43 weeks) in subjects with very high body weights. The two-phase weight loss model is convenient as it allows us to further explore the underlying components of weight loss and related metabolic effects. We now focus on the biology of phase I, the early weight loss period, in the sections that follow.

Early weight loss phase

Body composition changes

The loss in body weight during phase I reflects losses of both fat and fat-free mass, the latter representing the sum of all major molecular components including protein, glycogen, water, minerals and electrolytes (Supplementary Material, I). We begin our review with an analysis of each fat-free mass component change in response to negative energy balance. We then review the kinetics of fat mass change with voluntary weight loss.

Fat-free mass


The main nutritional concern with voluntary weight loss is depletion of body protein. Excessive loss of body protein is associated with adverse functional effects, for example depletion of cardiac structural and functional proteins with a reduction in myocardial mass, ventricular rupture and cardiac arrhythmias (14). Total body protein is not easily measured in living humans, the established reference method neutron activation analysis (15) of limited availability, high cost, and a need of technical expertise for construction, operation and maintenance. Moreover, in vivo neutron activation analysis (IVNA) is typically associated with radiation exposure, largely limiting evaluated subjects to men and non-reproductive age women (15). Frequently spaced measurements over time with interventions are not possible with IVNA as the cumulative radiation exposure is unacceptable. The alternative approach, applied in most studies, is to evaluate nitrogen (N) balance as a proxy for protein balance; as protein is 16% N, then Δprotein (g) = 6.25 ×ΔN (g) (15). These studies are ideally carried out in patients living on a metabolic ward and include N balance estimates as the difference between N intake and the sum of measured fecal and urinary N losses; miscellaneous N losses from skin are estimated.

A typical N balance study for 11 obese women placed on a 5-week 900 kcal d−1 liquid diet following a 1-week maintenance period is shown in Fig. 2(16). At baseline, with the subjects in N equilibrium, the daily urinary N loss is about 14 g reflecting a protein intake in the range of 80–90 g d−1. Urinary total N and urea losses rapidly decline with the reduction in energy and protein intake; there is a corresponding return in N balance towards baseline equilibrium levels with the metabolic adaptations that follow. Nitrogen balance is maximally negative immediately following the reduction in dietary protein intake, reflecting a daily loss of ∼5–6 g N or 35–40 g protein. Protein in this study represented 5% to 6% of weight loss during phase I and the relative contribution of protein to weight loss then decreased by about one-half during phase II. The half-life of the early weight loss phase is less than 2 d, followed by phase II with ∼135 d required for N balance to reach zero.

Figure 2.

Components of and total N balance (±SEM) observed in 11 obese women switched from a maintenance intake (1-week) to a 900 kcal d−1 formula diet for 5 weeks. Modified from reference 16.

The early N loss with fasting is represented mainly by gastrointestinal tract and liver proteins involved with nutrient processing, while later losses are from skeletal muscle and to a less extent from visceral organs (12). An example is shown in Fig. 3 of a rodent semi-starvation model in which rapid loss of liver mass and non-collagen protein occur early during underfeeding in contrast to slower rates of body weight and skeletal muscle non-collagen protein losses (17). Colles et al. (18) also observed a rapid loss of total liver volume by a combination of magnetic resonance imaging (MRI) and computed tomography (CT) in severely obese subjects over the course of a 12-week VLCD study. At 12 weeks there was a 10.6% loss in weight while the corresponding loss of liver volume was 18.7%, the majority of which (∼80%; P < 0.001) was observed after 2 weeks of diet treatment; no measurements were made of liver lipid or glycogen. The brain is typically spared until the later stages of famine and severe starvation in which atrophy has been reported (19). The rates of organ and tissue depletion with negative energy balance are thus highly variable and information on relevant human effects, particular at the molecular level, remains limited.

Figure 3.

Upper panel: Changes in body weight and composition, expressed as a percent of baseline, following a reduction in food intake of 60% in mature male Long Evans rats. Lower panel: Corresponding percent changes in skeletal muscle (quadriceps) and liver non-collagen protein. Results are X ± SE. Modified from 17.

Forbes combined estimates of baseline total body protein with balance data to arrive at the N-balance half-lives for non-obese and obese subjects as shown in Table 1(2,3). The half-lives for N depletion of non-obese subjects are 2.4 and 115.7 d for phases I and II, respectively. Appreciably longer half-lives are observed for obese subjects, 10.6 and 433.2 d, respectively.

Phase I during a typical diet or fast is thus characterized by relatively large losses of body protein that diminish in magnitude during phase II. Assuming an upper limit half-life of ∼10 d, phase I based on N balance would maximally extend about 6 weeks, somewhat longer than the 4-week estimate from weight loss kinetics. Given the very limited data used to derive these estimates, it appears as if phase I would be complete in most subjects within the first 6 weeks of voluntary weight loss. We continue to explore this question with data presented in later sections.

Proteins exist within the intracellular and extracellular environments in association with water and electrolytes. A water network links secondary structures within the protein molecule and determines the fine detail of the protein's structure. Fenn and Haege carried out a classic study reported in 1940 that defined the water associated with cat liver proteins (20). The authors observed that each gram of deposited protein is accompanied by (X ± SD) 3.35 ± 0.11 g of water. MacKay and Bergman in their earlier study (1934, (21)) of young albino rats fed casein arrived at a lower value for water association with protein, 2 g g−1. McBride et al. (1941, (22)) raised concerns related to the earlier experimental studies of both Fenn and Haege (20) and MacKay and Bergman (21) as the studies relate to water association with glycogen. Hall recently (2008, (23)) examined this question in relation to the energy deficit per unit weight loss in humans and suggested an empirical hydration of 1.6 g H2O per g protein. Proteins, particularly intracellular proteins, thus strongly associate with water and this effect accounts for a larger change in body mass than can be accounted for solely by protein catabolism. Water associated with protein catabolism thus is an important contributor to the rapid weight loss observed in phase I.

Starvation response.  Bloom first reported in 1959 the use of total starvation for treating severe obesity (24). Subsequent balance studies revealed that starvation leads to disproportionate protein loss compared with a low-calorie diet and this weight control option was eventually abandoned. As an example, Ball et al. (25) placed five obese subjects on an 800 kcal d−1 liquid formula for 14 d with evaluation of fat and ‘lean mass’ (sum of water and protein) changes using a combination of tritiated water dilution and N balance. Subjects were then switched to a starvation protocol for the next 16 d following which there was a 7 d refeeding period. The rate of weight loss with low-calorie dieting was similar to that observed with total starvation, 769 g d−1 compared with 651 g d−1. The rate of fat-free mass loss during low-calorie dieting was 224 g d−1, much less than the 576 g d−1 observed during total starvation. The rate of fat loss during the low-calorie diet period was 553 g d−1, substantially greater than with starvation (75 g d−1). Other small scale studies using balance methods similarly conclude that massive protein loss accompanies the early phase of total starvation (26).


While not a major functional concern as with protein, glycogen provides a short-term energy supply and negative glycogen balance is an important contributor to phase I weight loss. Glycogen is present within skeletal muscle and liver cell granules that also include tightly bound glycogen phosphorylase and synthase (12,27). The glycogen molecule is highly branched and the observed rapid glucose residue turnover in vivo occurs mainly at the outer branches with a stable inner core (12). Glycogen within hepatocytes can contribute up to 8% of postprandial liver mass, which amounts to ∼100–120 g in the average adult (12). Glycogen within skeletal muscle cells accounts for ∼1% of wet mass or ∼200–300 g, depending on subject conditioning and diet (12). Acheson et al. (28) evaluated adult volunteers using a dietary restriction-refeeding protocol and found that glycogen stores maximally approached or even exceeded 1 kg; once beyond saturation levels, carbohydrates are disposed of by high oxidation rates and de novo lipid synthesis.

As with protein, glycogen is hydrated in vivo and to some extent the amount of bound water depends on the molecular weights and associated structure of various analysed glycogen fractions. Fenn and Haege in their study (20) estimated that each 1 g of glycogen deposited in liver will be accompanied by (X ± SD) 1.46 ± 0.21 g of water. This level of hydration (59%) is very similar to that observed by Mas et al. (55%; (29)) in E. coli K12 during periods of glycogen accumulation. Nilsson (30) evaluated post-absorptive human liver tissue and estimated water binding as 2.4 g per g liver glycogen. McBride et al. (22) in their study of male rats arrived at a higher value, 2.7 g H2O per g liver glycogen.

With starvation or semi-starvation, as is typical with the use of a low-calorie diet, there is rapid depletion of the glycogen pool over several days. Although the baseline amount present and the rate of glycogen loss is highly variable between subjects, we can assume that with about 250–300 g catabolized on a traditional composition low-calorie diet that the overweight or obese subject will lose another 350 to 450 g of water and associated electrolytes from the intracellular compartment. This prediction is supported by Benedict's observations on Subject L (11) whose rapid weight loss (Fig. 1) is paralleled by rapid depletion of body ‘carbohydrate’ over about a 10 d period (Supplemental Material, III).


The regulation of electrolyte and water excretion is highly complex with many interactions among neurological and hormonal stimuli. Most of the currently available information on this topic relates to fluid-electrolyte effects of fasting or total starvation. A partial reduction in food intake is more complex with respect to fluid balance than total starvation as the amounts and proportions of diet-derived macronutrients, minerals and electrolytes are highly variable.

The typical early changes in fluid balance associated with low-calorie diet treatment are demonstrated by the study of Wynn et al. (31). The authors examined Na, K and N balances during 68 d while subjects ingested a 655–789 kcal d−1 liquid formula diet. Subjects were variable numbers of obese men (n= 6–9) and women (n= 7–16), depending on phase of study. Sodium balance was markedly negative during week 1, particularly on d 1 (∼−30 to 60 mmol d−1; men > women), and was closely associated with fluid balance (−100 to −500 g d−1). Potassium balance paralleled N balance, reflecting intracellular protein and fluid losses; both K and N balance approached stable negative levels after the first 28 d of treatment towards the end of phase I. The main source of fluid loss after the first week of treatment derived mainly from the intracellular compartment.

Starvation response.  During the early phase of a total fast in obese subjects, notably within the first 7–10 d, there is a kaliuresis with a total potassium loss of about 300 mmol (32). As noted earlier, intracellular glycogenolysis and proteolysis release bound water and electrolytes that likely account for most of the urinary potassium losses. In addition to potassium and nitrogen, ‘protoplasmic loss’ during a fast also releases phosphate, calcium, magnesium and small amounts of sodium that are then excreted through renal mechanisms (32).

Benedict in 1915 first reported that the early phase of fasting is also associated with an increase in urinary sodium excretion, the so-called ‘natriuresis of fasting’(11). The level of observed sodium loss is greater than can be accounted for by the lowering of or absence of dietary sodium intake. Extensive subsequent studies following Benedict ascribe this period of negative sodium and extracellular fluid (ECF) balance to four main factors related to the cessation of food intake and the metabolic events accompanying starvation that follow: a decrease in dietary sodium with a rapid period of related negative sodium balance (decline of ∼50% d−1) that is independent of weight loss (33); generation of ketone bodies (e.g. β-hydroxybutyrate) secondary to free-fatty acid catabolism with metabolic acidosis, ketosis and ketonuria (34,35) with cationic sodium matching the organic acids to maintain electrical neutrality; and hypoinsulinemia leading to natriuresis through insulin's direct effects on renal tubular sodium transport (35,36) and to a less extent effects on the sympathetic nervous system with activation of the renin-angiotensin system (37).

The combined effects of glycogen, protein and fluid loss largely account for the rapidity of phase I weight loss compared with the slower rates observed in phase II. The rapid phase I weight loss phenomenon is the basis for countless ‘quick fix’ diets promoted over the last century and that are still pervasive in our diet-conscious culture.


The remaining fat-free mass component is bone or fat-free skeleton which for the 70 kg Reference Man weighs about 8.1 kg and includes 1.9 kg protein, mainly in the form of collagen, with a total mineral content of ∼2.8 kg (38). Skeletal weight is increased in obese subjects and bone remodelling and formation is strongly influenced by mechanical loading effects (39). The larger bone mass observed in obese post-menopausal women may also relate to aromatization of androstenedione to the bone-preserving hormone estrone that takes place in adipose tissue (40). The topic of bone loss during weight reduction treatment with low-calorie diets has been of great interest, particularly as the majority of treated patients are women in whom osteoporosis is a concern.

The main technique for monitoring changes in bone mass with dieting is dual-energy X-ray absorptiometry (DXA) that provides a measure of bone mineral content, primarily in the form of calcium hydroxyappetite (15,41). Bone mineral density, of the whole body or regions, can also be evaluated by DXA but is not relevant to the current discussion. CT, MRI and ultrasound methods are also available for the study bone changes with voluntary weight loss (42). Other biomarkers include those found in blood and urine that estimate the rate of bone turnover (43).

When considering the bone loss accompanying voluntary weight loss, moderating factors include diet (i.e. magnitude of energy deficit, macronutrient profile and calcium/vitamin D intake), duration (e.g. bone remodelling cycle ∼6 months in older individuals), subject characteristics (e.g. pre- vs. post-menopausal women), magnitude of adiposity and bone evaluation method (e.g. regional vs. whole-body DXA for bone mineral density or bone mineral content). Carefully controlled studies lasting up to several months show that with voluntary weight loss no or very small changes in total body bone mineral content as measured by DXA are observed when overweight or obese subjects are provided adequate amounts of calcium and vitamin D (Supplementary Material, II). A reasonable assumption based on this available information is that skeletal mass remains minimally changed during the early phase of voluntary weight loss. Confirmation of this observation, with appropriate measurement methods, would be useful when planning future weight loss studies.

Fat mass

Very few studies have closely examined the kinetics of changes in fat mass during the early period of weight loss, except for the limited publications using energy-nitrogen and metabolic balance methods (11,44). The other available techniques for assessing body composition are not very precise, have associated radiation exposure, apply assumptions that are not valid or validated in the short term, or are impractical to carry out at closely spaced time intervals (Supplementary Material, I). Nevertheless, we can construct a general model of how total body fat mass changes during the early period of underfeeding.

Sources of metabolic fuel during periods of negative energy balance include glycogen, protein and fat. Glycogen stores, as noted earlier, are largely consumed within the first week when subjects substantially reduce their energy intake. A two phase loss of protein is observed with half-life examples presented in Table 1. The ‘early phase’ protein loss is thus also relatively short, extending maximally several weeks depending on the magnitude of energy restriction and baseline subject characteristics. The remaining energy balance deficit must therefore come from fat. We can then surmise that the fraction of weight loss as fat is minimal during the first days or weeks of energy restriction and then increases as early losses of glycogen and protein slow or cease as the subject reaches the end of phase I. On the other hand, the magnitude of energy imbalance is also slowing as subjects lose body mass with lowering of energy expenditure and adaptive mechanisms additionally decrease the resting heat production rate (45). This model predicts that the fraction of weight loss as fat increases during phase I and that over time relative loss of body fat can be fit by curvilinear models as described for weight and nitrogen, as shown in Table 1.

While ‘fat’, defined mainly by triglyceride, is the appropriate molecular entity related to energy stores, we now recognize that fat is distributed in ‘adipose tissue’ with specific regional properties (46). With induction of negative energy balance and weight loss, corresponding relative rates of adipose tissue loss are highly variable and range from rapidly mobilized visceral adipose tissue (47) to minimally influenced bone marrow adipose tissue (48). Thus, loss of total body fat with voluntary dieting reflects the integrated changes within several regional adipose tissue compartments. How these differences in adipose tissue fat loss rates distribute across the three weight loss phases has not been firmly established.

Influencing factors

Temporal and sex effects

Phase I up to this point in our review includes variable maximal rates of protein, glycogen and fluid loss that greatly diminish with entry into phase II. We can then surmise that the corresponding rate of collective fat-free mass loss is also relatively large during phase I, particularly during the initial period following induction of negative energy balance.

The temporal changes in relative weight loss composition during the first month of treatment is provided by the VLCD study of Krotkiewski (49) that evaluated diet supplementation with medium chain triglycerides (MCT) in obese women. In addition to subjects placed on the MCT formula, Krotkiewski also included groups treated with VLCDs containing either long chain triglycerides or that were low in fat. The DXA-derived body composition results are shown in Fig. 4. As predicted, the fraction of weight loss as fat-free mass was high at 1 week of VLCD treatment, about 0.4–0.6 across all three diet groups. The fraction of weight loss as fat-free mass subsequently decreased at weeks 2 and 4 in the three groups to 0.3–0.4.

Figure 4.

Upper panel: Changes in body weight in three groups of obese women receiving a 579 kcal d−1 VLCDs differing in lipid source as reported by Krotkiewski (49). Lower panel: The corresponding changes in the ratio ΔFFM/Δweight. FFM, fat-free mass; VLCD, very-low-calorie diet.

The relative contributions of fat and fat-free mass to early weight loss beyond 4 weeks are provided by results of the Calerie (50–52) and Kiel studies. The primary data were available to us from both of these studies that we use in the analyses that follow. Additional details of these studies are provided in Supplemental Material, IV and V and in the original respective publications (50–52).

The main evaluated database for which serial DXA measurements were available was from the 24-week Calerie study (50–52) that had four overweight subject groups including a non-diet control, calorie restriction (CR) at 25% diet-imposed energy deficit, CR plus structured exercise (Ex) as walking, running and cycling (25% energy deficit, 12.5% diet and 12.5% exercise), and a VLCD (890 kcal d−1) until 15% weight loss followed by a weight maintenance diet (Supplemental Material, IV). The VLCD maintenance diet phase was reached at approximately 8 weeks in women and 11 weeks in men and we therefore focus the main analyses that follow on this 3-month period of collective negative energy balance. Subjects were ethnically mixed men (age < 50 years) and women (age < 45 years) randomly assigned to the four groups.

The pooled weight loss and body composition results for Calerie study participants in the CR and VLCD groups up to week 12 are shown in the upper panel of Fig. 5. Fat-free mass loss was rapid during the early weeks of the protocol and then stabilized around weeks 10–12. Fat mass and body-weight loss continued up to week 12.

Figure 5.

Loss in weight, fat and fat-free mass (FFM) in the Calerie Study (50–52) participants assigned to the caloric restriction and VLCD groups, expressed as absolute mass (upper panel) and as a percent of baseline mass (lower panel). Results are presented as X ± SE. BL, baseline; BW, body weight; VLCD, very-low-calorie diet.

On a relative basis, only about 5% of fat-free mass was lost by week 12, in sharp contrast to a much larger percentage of fat mass for which depletion of total body fat approached 25% of the baseline amount (Fig. 5, lower panel). The combined losses of fat-free mass and fat mass led to a body-weight loss of ∼10% by week 12 of the protocol.

The differing rates of fat-free mass and fat mass loss led to changes in the fractional contributions of each compartment to body-weight loss over time (Fig. 6). As in Krotkiewski's 4-week study (49), the fraction of weight loss as fat-free mass was high early during the Calerie protocol (mean ± SD, week 4; 0.54 ± 0.23) with a curvilinear decrease up to week 12 (0.35 ± 0.17). The fraction of weight loss as fat, the reciprocal of the fraction of weight loss as fat-free mass, began low at week 4 and gradually increased up to week 12. At week 24 (data not shown) the fraction of weight loss as fat-free mass decreased further from week 12 to 0.29 ± 0.14 in the combined CR and VLCD groups.

Figure 6.

Fraction of weight change (Δ) as fat and fat-free mass (FFM) vs. study week for Calerie Study (50–52) participants assigned to the caloric restriction and VLCD groups. Results are presented as X ± SE. VLCD, very-low-calorie diet.

These observations are thus consistent with the previously reported kinetics of individual component losses as protein, glycogen, fluid and fat during the early phase of voluntary weight loss and the phase I half-lives of weight and protein loss as summarized in Table 1. The relative contributions of fat and fat-free mass with voluntary weight loss are thus not ‘constant’, but change rapidly during phase I and appear to change at a much slower rate in phase II. Again, we can reaffirm that the energy density of weight change is not constant during phase I, but increases rapidly and then slows with entry into phase II.

More experimental data at closely spaced time points in various populations would ideally provide higher resolution phase I estimates for component half-lives. From Krotkiewski's study (Fig. 4; (49)) and the Calerie results (Fig. 6) we can again estimate that the early rapid phase of weight loss with associated body composition effects extends from diet inception up to about 4 to 6 weeks with sustained negative energy balance.

The larger Kiel VLCD study (Supplemental Material, V) provided DXA body composition estimates at baseline and at approximately 10 weeks of weight loss. The fractions of weight loss as fat and fat-free mass are shown separately for men and women participating in the Kiel study along with the corresponding representative 10-week Calerie data in Fig. 7. Men had a larger fraction of weight loss as fat-free mass than women (Calerie, P= 0.078; Kiel, P= 0.059; and pooled men vs. women from the two studies, P= 0.010). Similarly, the fraction of weight loss as fat at the 10-week time point was less in the men than in the women. Chaston et al. in their systematic review (5) found that for dietary and behavioural weight loss interventions the fraction of weight loss as fat-free mass was also larger in men (X ± SD, 27 ± 7%) than in women (20 ± 8%). These observations strongly support the need to further evaluate sex differences in the kinetic parameters of phase I with appropriately powered samples.

Figure 7.

Fraction of weight loss as fat-free mass observed in men and women participating in the Calerie study caloric restriction group (50–52) and Kiel-VLCD studies. Results are X ± SE. P-values are for men vs. women in each respective study group and for all subjects combined. VLCD, very-low-calorie diet.

Magnitude of energy deficit

Some early VLCDs in the range of 300–400 kcal d−1 that facilitated rapid and large amounts of weight loss were associated with depletion of body cell mass leading to serious adverse events (53), although modern formula diets tend to be higher in energy content, have high-quality protein sources, and include adequate amounts of electrolytes, minerals, vitamins and trace elements. Chaston et al. (5) in their systematic review reported a greater percentage of weight loss as fat-free mass with VLCDs compared with low-calorie diets. Their sample of appropriate studies was, however, relatively small (n= 4) and body composition methods were varied across studies, including underwater weighing (n= 3) and DXA (n= 1). The Calerie sample CR and VLCD groups are also small (n= 12 and 11 subjects) and include both men (n= 10) and women (n= 13) so that the subject number within each cell has limited power to test the hypothesis that FFM is preferentially lost with VLCD treatment. For the combined men and women in CR and VLCD groups at 10 weeks, the ΔFFM was (X ± SD) −1.47 ± 0.90 kg and −4.10 ± 2.0 kg, respectively; the corresponding ΔFFM/ΔW were 0.32 ± 0.20 and 0.36 ± 0.11 (P= NS). The ΔFFM/ΔW over the full 24-week course of the Calerie study for CR and VLCD groups is shown in Fig. S4. Despite large weight loss differences between the groups, there were no discernable between-group relative differences in FFM loss.

A key question emerging from the previous reports and the current data review is if and how the magnitude of relative FFM loss is influenced by the degree of negative energy balance during the early phase of weight loss. As for other areas discussed in our review, more information is needed on early body composition effects in appropriately powered samples based on subjects selected to create sex, age, body mass index (BMI) and physical activity balance across groups.

Diet composition

Diet mineral, electrolyte, protein, fat and carbohydrate content all influence components of phase I weight loss. Earlier sections reviewed contributions of diet sodium and calcium content to phase I changes in respective fluid and bone mineral balances. In addition to total energy content, the proportions of protein, fat and carbohydrate have an important influence of the composition of early voluntary weight loss.

The main features of these macronutrient effects are well characterized by the 50 d energy-nitrogen balance study of Yang and Van Itallie (54). Men with severe obesity (n= 6) were treated with three randomly sequenced 10-d diet protocols, including a total fast, a balanced nutrient 800 kcal d−1 VLCD and a low-carbohydrate-high-fat ketogenic 800 kcal d−1 diet. Subjects ingested a balanced 1200 kcal d−1 formula diet for 5 d before and following each of the three experimental periods.

As expected, with the total fast there were large losses of the three evaluated components, protein, fat and water; weight loss was correspondingly large (X ± SD, 750.7 ± 50.9 g d−1; Fig. 8). Weight loss on the ketogenic diet (466.6 ± 51.3 g d−1) was much larger than on the isocaloric balanced nutrient VLCD (277.9 ± 32.1 g d−1), although this difference was primarily accounted for by differences in water balance; fat and protein losses were similar during treatment with the two VLCDs. Ketosis with ketonuria was present during the ketogenic VLCD period.

Figure 8.

Composition of weight loss observed with total fasting and two 800 kcal d−1 VLCDs, one with a balanced (B) nutrient composition and the other a high-fat low-carbohydrate ketogenic (K) diet. Based on 54. VLCD, very-low-calorie diet.

Although this is an extreme example of phase I diet composition effects, diets varying widely in composition are currently used in both clinical research and practice settings. These metabolic and body composition effects should be considered when planning studies that examine diets within the phase I time frame.


Physical activity contributes to total energy expenditure and is thus an important component of energy balance. Physical activities are characterized by type, duration, intensity and other important features that differ between studies. An exercise programme is often prescribed in association with other lifestyle measures as part of a comprehensive overweight and obesity treatment programme.

In addition to contributing to the magnitude of imposed negative energy balance, added exercise purportedly prevents or reduces loss of fat-free mass through trophic and hormonal effects on skeletal muscle and other organs and tissues (55–58). Forbes critically reviewed this hypothesis (56,57) and there are also several meta analyses that explore this topic (6,7). Most authors agree that exercise, depending on type, intensity and duration, has some fat-free mass sparing effects (6,7,55–58), although the composition of this ‘spared weight’ is not well defined. Exercise can influence fluid balance (59), glycogen formation (60) and protein turnover (61).

The important question from the perspective of our review is how exercise influences the early phase of weight loss. Most of the earlier studies did not specifically address this question and we thus provide an initial overview from Calerie study results. Calerie included two groups with a similar prescribed energy deficit of 25% below baseline (Fig. S3). The energy deficit was created by food restriction in the CR group and by food restriction and added exercise in the CR+Ex group (12.5% CR and 12.5% Ex). The loss of fat-free mass in the CR+Ex group up to the 12-week evaluation time point was about one-half that of the CR group (Fig. 9) with variable statistical significance for between-group differences. Three of the 12 subjects in the CR+Ex group had no loss or gain in fat-free mass up to week 12 compared with none in the CR group. These observations are highly suggestive that added exercise influences the kinetics of fat-free mass loss during the early phase of weight loss treatment.

Figure 9.

Loss of fat and fat-free mass in the calorie restricted (CR) and CR+exercise (EX) Calerie study (50–52) groups. Results are X ± SE.

Baseline adiposity

Does baseline subject adiposity moderate the magnitude of phase I fat-free mass loss? Assume that an extreme athlete has a body mass consisting of solely fat-free mass without any body fat. When placed into negative energy balance this subject's metabolic processes would by necessity consume the only two available fuel sources, glycogen and protein.

The data needed to answer this question is extremely limited and the main insight available is from the Minnesota Semistarvation Experiment (8), although the first study time point after baseline is at 12 weeks. This classic study included young men who were placed on a ∼1800 kcal d−1 balanced caloric restriction diet with body composition measured by underwater weighing. Body weight was adjusted through variation in food intake prior to the restriction period in order to ‘normalize’ the men's weights, so as a group they were not in a long-term energy equilibrium state a baseline.

We explored the relation between baseline adiposity and composition of weight change by stratifying the men into quintiles based on baseline %fat. We then plotted the group mean fraction of weight loss as fat-free mass against baseline quintile for the 12-week time point. The men in the lowest quintile of baseline %fat had a substantially larger fractional loss of fat-free mass than did the men in the middle (2–4) or highest baseline %fat quintiles (Fig. S5). Note that baseline %fat in the lowest quintile was extremely low (∼7%); all three groups had a similar BMI of about 22 kg m−2 ((8); Supplemental Material, VI). This observation supports a related earlier analysis reported by Dulloo et al. (13,62).

We also examined Calerie and Kiel study data to test the hypothesis that baseline adiposity influences the fraction of weight loss as fat-free mass. No significant associations between baseline %fat and the fraction of weight loss as fat-free mass were observed at any time point in these exploratory analyses. As with other topics noted in our review, this important question needs resolution in future carefully controlled trials.

Hormones and drugs

Pharmacologic and hormonal treatments are reported to alter early phase protein, fat-free mass and fat mass losses. For example, classical energy-nitrogen and metabolic balance studies have been used to characterize hormone and drug effects (63–65). A key feature of some of these studies is the use of experimental designs that recognize the distinguishing features of phase I and II weight loss. For example, pharmacologic doses of T3 were frequently used in the past in association with caloric restriction as a means of treating obese subjects. T3 as used in this context was hypothesized to have catabolic effects mainly on skeletal muscle. As the protein losses during phase I are likely from mixed sources, including liver, gastrointestinal tract and skeletal muscle, Abraham et al. (63) conducted their T3 study in obese subjects following 30 d of caloric restriction when phase I was largely completed and nitrogen losses reflected primarily skeletal muscle catabolism. In support of the hypothesis, high doses of T3 during this presumed initial phase II period caused striking increases in N loss compared with placebo treatment.

Observations such as those of Abraham et al. (63) emphasize the importance of designing hormonal and pharmacologic weight loss studies with careful consideration of phase I effects.

Early weight loss as a distinct phase

Forbes was the first to fully articulate the idea that weight loss vs. time functions have mathematically quantifiable features allowing separation of these curves into at least two different kinetic components (i.e. phases I and II; (2,3)). Our review extends Forbes' concept by providing a synthesis of the vast amounts of human data in support of the view that voluntary weight loss proceeds in three potentially discrete phases.

We have expanded on Forbes' innovative concept by showing that phase I is characterized by distinct body composition and metabolic effects that differentiate it from the later phases of weight loss. First, as subjects switch from their ‘usual diet’ to a new lower food intake, there are multiple variable changes in macronutrient, mineral and electrolyte balances that stabilize over time on the new diet. Phase I thus reflects these ‘washout’ effects and that are in most cases not present in phase II.

Second, switching from one plane of nutrition to a new lower level of food intake triggers hierarchal changes in fuel stores and related body composition effects with depletion of glycogen stores and early phase protein losses that come into equilibrium with the onset of phase II.

Third, well-established neuro-hormonal mechanisms are invoked early in the course of weight loss such as in the leptin and thyroid hormone axes (66), only briefly touched upon in our review, that slow energy expenditure and protein turnover with activation of a host of other metabolic and neurobiological processes that help to sustain life in the face of negative energy balance. The ‘adaptive’ phase I period can thus for heuristic purposes be considered a distinct period of rapid weight loss appearing when subjects change from a maintenance plane of nutrition to a new lower level of nutrient intake.

From available literature and our own pooled data it tentatively appears as if at least five factors moderate the kinetic features of phase I, including (i) the magnitudes of energy, nitrogen and sodium balances as derived from the differences between subject baseline intakes and losses; (ii) sex, with more relative fat-free mass loss in men; (iii) diet macronutrient composition, (iv) physical activity level, with greater activity reducing relative fat-free mass losses and (v) selected hormonal and pharmacologic treatments. Our evaluation samples, Calerie and Kiel, tended to be small and we likely did not have adequate power to detect age effects; our subjects were mainly Caucasians so that race effects could not be adequately explored. The extent to which these and other factors (e.g. baseline adiposity) impact specifically on the composition of phase II weight loss remains unclear as previous critical reviews often combined short and long term weight loss studies in their pooled analyses.

An important observation emerging from our findings is that the phase I – phase II composition of weight loss difference impacts on classically assumed stable ratios such as the energy density of body-weight change (23) and the Pratio, defined as the fraction of energy mobilized as protein with food restriction (13,62). Further exploration of these time-related effects on classic ratios is justified based on observations reported in the current review.

Our comprehensive evaluation of phase I weight loss also provides critical new ideas for developing human energy balance models, a burgeoning area of biological research (Supplementary Material, VII).


Our findings support the view that early weight loss following reduction in or cessation of food intake reflects the combined kinetics of multiple compartmental changes with varying half-lives. Many of these processes are not well characterized in humans as some of the required in vivo studies are difficult or impractical to carry out and the available methods are not sufficiently accurate to make quantitative assessments. With these limitations, our findings suggest that the early phase of weight loss reflects the combined losses of water/electrolytes, glycogen, protein and fat and to a less extent bone minerals. Although we empirically set 12 weeks as the upper boundary for investigating the early period of weight loss, an exact length of phase I cannot be assigned and likely varies in duration depending on the nature of dietary intervention and baseline subject characteristics. A reasonable assumption emerging from our findings is that for most subjects phase I likely lasts 4 weeks or less, perhaps in some cases up to 6 weeks, based on the limited published half-life data for body weight and protein balance along with the supportive new experimental data provided in our review.

An important conclusion emerging from our analyses is that planned weight loss studies exploring various interventions should be designed to accommodate the differing early and late phase body composition effects. For example, an intervention focusing on body fat loss would not ideally be designed with body weight as the primary outcome measure unless the duration extended beyond several weeks or months as the fraction of weight loss as fat increases rapidly and is highly variable during phase I. The use of fat-free mass as a safety measure with pharmacologic treatments would need to consider the differing early and later phase effects on fluid, glycogen and protein losses. Many other relevant examples abound and the present study highlights the important need to further elaborate on the underlying biology of early weight loss and to similarly explore the main features of phase II weight loss as approached in the current report.

Conflict of Interest Statement

No conflict of interest was declared.