Proportion of Caloric Restriction-Induced Weight loss as Skeletal Muscle

Abstract

Objectives:

To develop models predicting the relative reduction in skeletal muscle (SM) mass during periods of voluntary calorie restriction (CR); and to validate model predictions in longitudinally monitored samples.

Methods:

The model development group included healthy non-exercising adults (n=897) who had whole-body SM mass measured with magnetic resonance imaging. Model predictions of relative SM changes with CR were evaluated in two longitudinal studies, one 12–14 weeks (n=74) and the other 12 months in duration (n=26).

Results:

A series of SM prediction models were developed in a sample of 415 males and 482 females. Model-predicted changes in SM mass relative to changes in body weight (i.e., ΔSM/ΔBW) with a representative model were (mean±SE) 0.26±0.013 in males and 0.14±0.007 in females (sex difference, p<0.001). The actual mean proportion of weight loss as SM in the longitudinal studies were 0.23±0.02/0.20±0.06 in males and 0.10±0.02/0.17±0.03 in females, similar to model-predicted values.

Conclusions:

Non-elderly males and females with overweight and obesity experience respective reductions in SM mass with voluntary CR in the absence of a structured exercise program of about 2–2.5 kg and 1–1.5 kg per 10 kg weight loss. These estimates are predicted to be influenced by interactions between age and body mass index in males, a hypothesis that needs future testing.

INTRODUCTION

People embarking on a low-calorie diet or calorie restriction (CR) experience reductions in not only body weight and fat, also but decrements skeletal muscle mass (SM) and other lean tissues and organs [1–4]. The effects of diet-induced weight loss on non-fat body compartments now comes into focus with two recent developments. First, new highly effective weight loss agents for treating people with obesity are accompanied by relatively large changes in fat-free mass (FFM)[5–8]. Fat-free mass includes the body’s main functional tissues and organs, including SM, and excessive loss is viewed as an adverse pharmacologic effect [9–12]. Effects such as these have not yet shown to be present in humans. Second, a new class of drugs that block the activin type II receptor promote loss of body fat while stimulating gains in SM [13, 14]. Pharmacologic uncoupling of the Δfat-ΔSM relationship present with diet-induced weight loss is thus possible. Despite the growing relevance and interest in this topic, there are no validated prediction models linking changes in SM mass with diet-induced weight loss. Models such as these would provide a foundation for evaluating the muscle-specific effects of pharmacologic and exercise interventions in the context of low-calorie weight loss treatments. Similarly, these kinds of models could be used to explore if and to what extent there are sex, age, and body mass index (BMI) differences in the effects of dieting on reductions in SM mass. The aim of the current study was to fill this gap by developing and evaluating prediction models for estimating changes in whole-body SM mass during periods of negative energy balance induced by voluntary CR. Skeletal muscle mass was evaluated in these cross-sectional and longitudinal protocols with whole-body magnetic resonance imaging (MRI).

METHODS

Experimental Design

The study was conducted in two phases. The first phase involved development of SM prediction equations on a sample of healthy adults evaluated with whole-body MRI at two research centers. These equations were used to model the relationships between SM and body mass and to predict how SM might change with diet-induced weight loss. The development of these models is reported in the Statistical Methods section. The second phase evaluated the SM model predictions in two longitudinal weight loss studies, one 12–14 weeks in duration [15] and the second 12 months in duration [16]. The model development group included participants evaluated at the Institute of Human Nutrition, Kiel University, Germany, and the New York Nutrition and Obesity Research Center (NYNORC), Columbia University, College of Physicians and Surgeons, New York, USA. Each MRI scan was analyzed for whole-body SM and adipose tissue volumes. The measured volumes were then transformed into SM mass and adipose tissue mass (ATM) assuming respective densities of 1.04 kg/l and 0.92 kg/l [15].

The proportions of weight loss as MRI-measured SM and ATM in adult males and females were then evaluated in the two longitudinal weight loss studies, one at Kiel and the other at Pennington Biomedical Research Center (PBRC) in Baton Rouge, Louisiana, USA. The details of the Kiel protocols and measurement methods are reported earlier [15, 17, 18] and relevant information is summarized in Supplementary Table S1. The longitudinal study at PBRC, Comprehensive Assessment of Long-term Effects of Reducing Intake of Energy (CALERIE Phase 2), included two groups, one a control and the other active CR [16]. Data from the CR participants completing the 12-month weight loss phase is included in the current report. Salient features of the CALERIE study are summarized in Table S1.

Settings and Participants

The two model development samples included healthy adults between the ages of 18 and 88 years evaluated with whole-body MRI at Kiel and the NYNORC. The disposition of these study participants and accrual criteria are presented in Supplementary Information, Figure S1. The NYNORC sample was race/ethnically mixed while the Kiel sample was all White. Participants were recruited from the general public who had a normal physical examination, electrocardiogram, and did not have a history of cardiovascular or metabolic diseases; females were premenopausal and were not pregnant or lactating. Neither sample included participants actively engaged in exercise training programs. Body weight was measured ±0.01 kg after an overnight fast with digital scales and height ±0.5 cm with mechanical stadiometers at both centers. Each participant completed the whole-body MRI scan.

The disposition and key features of the two longitudinal samples are presented in Table S1 and Figure S2. Participants were healthy adults between the ages of 19 and 50 years. The studies were all approved by the respective institutional review boards and participants signed informed consents prior to participation. The CALERIE study is listed on Clinicaltrials.gov as NCT02695511.

Magnetic Resonance Imaging

A 1.5-T Magneton Vision or Avanto scanner (Siemens Medical Systems, Erlangen, Germany) was used to quantify total body SM and adipose tissue volumes in the cross-sectional and longitudinal studies at the Kiel site. Details of the Kiel imaging protocol have been reported earlier [17, 19–21]. Cross-sectional MRI scans at the NYNORC site were evaluated using a 1.5-T 6X Horizon scanner (General Electric, Milwaukee, WI, USA). A 3.0-T Signa Excite scanner (General Electric, Milwaukee, WI, USA) was used in the longitudinal CALERIE study at Pennington Biomedical to evaluate total body SM and adipose tissue volumes as previously reported [16]. The acquired MRI images were manually analyzed by trained technicians with SliceOmatic software (version 4.3, Tomovision, Montreal, Canada).

Statistical and Model Development Methods

The SM prediction models were developed similar to the classic fat and FFM prediction models reported by Webster and Garrow in 1984 [11] and validated in many longitudinal studies since then. First, a series of regression models were developed in the male and female samples of the general type shown in Figure 1. In this example, patterned after that introduced by Webster and Garrow [11], BMI (weight/height²) was set as the predictor variable and SM index (SMI, SM/height²) as the dependent variable. Webster and Garrow referred to the β coefficient in their version of this model as the proportion of “excess” weight as fat mass [11]. The inference is that with CR this would be the proportion of weight loss as fat (i.e., Δfat mass/Δweight; height cancels when fat mass index is divided by BMI). Webster and Garrow estimated this value as 0.70–0.78 in females [11]; the predicted relative loss of lean (i.e., FFM) was derived as 1 minus the proportions of fat mass as 0.22–0.30. In addition to this simple model (SMI versus BMI), we also explored covariates including age, center (Kiel or NYNORC), and interaction terms using multiple-variable linear regression. The first series of cross-sectional models included SMI as dependent variable and BMI as the main independent variable, along with other independent variables. A second series of similar models included SM as the dependent variable and weight, rather than BMI, as the main predictor variable, along with height and other independent variables. In addition to our focus on the magnitude of β in these models, we also selected one for predicting expected changes in the proportion of weight loss as SM in the longitudinal models.

Figure 1.

Schematic diagraming the analytical plan for defining the proportion of weight loss as SM in the model development and validation samples. The slope, B, in the upper panel equation represents the estimated increase in skeletal muscle mass index (SMI) per unit increase in body mass index (BMI). Error is represented by ε in the equation. The proportion of weight loss as SM with volitional CR (ΔSMI/ΔBMI) is predicted to follow this relation, as shown in the lower panel, and have a value approximately equal to B. This general strategy was used to estimate the composition of “excess” weight and proportion of weight loss as fat mass by Webster and Garrow in 1984 [11].

The longitudinal evaluations included measured (mean±SD) values for the proportion of weight loss as SM (i.e., ΔSMI/ΔBMI or ΔSM/Δbody weight; Figure 1). The measured values were then qualitatively compared to the corresponding estimates provided by the cross-sectional models. The available data also allowed us to evaluate the proportion of weight loss as adipose tissue. Our exploratory models provided an opportunity to compare the composition of weight loss as SM across males and females and between adults varying in age. Our hypothesis was that the decrease in SM mass with body weight loss (ΔSM/ΔBW) in the longitudinal studies would have the same magnitude as the slope (± error) of our sectional regression lines found in the model development sample.

Since the current report utilizes existing data from completed studies, calculated power primarily focuses on assessing the margin of error for the model development sample regression equation slopes. Specifically, for males, 415 participants were included from the Kiel and NYNORC studies with an anticipated standard deviation of 1.3 kg/m² for SMI and 3.5 kg/m² for BMI. With 80% power, a two-sided confidence interval, and an anticipated regression slope of 0.2 based on earlier studies [22, 23], the estimated margin of error is 0.02. This margin of error corresponds to approximately 10% of the slope magnitude. For females, a total of 482 participants were included with an anticipated standard deviation of 1.0 kg/m² for SMI and 4.7 kg/m² for BMI. With an anticipated regression slope of 0.10, the estimated margin of error is 0.01. All analyses were conducted using SAS (Windows version 9.3; SAS Institute, Cary, NC) and the statistical program R version 4.0.2 (R Core Team, 2020) (https://cran.r-project.org/).

RESULTS

Samples

The characteristics of the model-development sample are summarized in Table 1. As expected, males had a higher percentage of body weight as SM than females, about 38% versus 30%, respectively. Females had a higher percentage of body weight as adipose tissue than males, about 35% and 23%, respectively.

Table 1.

Sample characteristics.

	Model Development				Model Validation
Sample	Kiel		NYNORC		Kiel		CALERIE
	Males	Females	Males	Females	Males	Females	Males	Females
N	218	256	197	226	14	60	8	18
Age (y)	49.9±18.1	49.4±18.2	39.2±13.9	44.5±16.2	38.1±6.1	34.0±6.9	41.8±6.9	39.5±7.4
Weight (kg)	84.8±12.3	68.6±12.4	79.8±12.7	67.0±15.1	111.8±12.8	100.1±17.5	82.4±7.6	68.0±7.2
Height (cm)	178.7±6.4	165.7±6.7	177.1±6.8	162.0±7.2	180.7±5.0	169.1±7.4	178.2±8.1	164.0±5.9
BMI (kg/m2)	26.5±3.3	25.0±4.1	25.4±3.6	25.5±5.4	34.3±3.8	34.8±4.3	25.8±1.4	25.2±1.8
AT (kg)	19.5±6.8	22.2±8.2	18.4±7.8	25.7±12.4	32.1±7.8	42.4±11.2	15.6±3.3	21.8±5.4
(%)	22.5±5.6	31.5±6.8	22.4±7.0	36.6±10.4	28.5±4.4	41.8±4.8	18.7±2.5	31.7±5.8
ATMI (kg/m2)	6.1±2.1	8.1±2.9	5.9±2.4	9.8±4.7	9.9±2.4	14.7±3.3	4.9±0.7	8.1±1.9
SM (kg)	30.7±4.1	19.9±3.1	31.7±5.5	19.9±3.4	39.6±4.1	24.7±3.8	32.4±2.4	20.7±2.5
(%)	36.4±3.8	29.3±6.4	40.0±4.6	30.3±4.9	35.6±3.5	24.9±2.6	39.5±2.5	30.5±3.5
SMI (kg/m2)	9.6±1.0	7.2±0.9	10.1±1.6	7.6±1.1	12.1±1.2	8.6±0.9	10.2±0.8	7.7±0.7

Results are mean±SD. AT. Adipose tissue; ATMI, adipose tissue mass index; BMI, body mass index; NYNORC, New York Nutrition and Obesity Research Center; SM, skeletal muscle; SMI, skeletal muscle mass index.

The longitudinal Kiel and CALERIE samples included 74 and 26 adults, respectively. The Kiel sample had mean baseline BMIs in the obese range of about 35 kg/m² while the CALERIE participants were on average overweight with BMIs of about 25 kg/m².

Model Development

In the first series of analyses, SMI was significantly correlated with BMI in the males (R², 0.35; p<0.001) and females (R², 0.39; p<0.001) (Figure 2) with respective BMI slopes (X±SE) of 0.22±0.015 and 0.13±0.008 (Model 1, Table 2; p<0.001 for male-female slope difference). These initial results suggest that about 22% and 13% of “excess” weight is SM in the males and females, respectively. Age added significantly to the SMI prediction equations (Model 2), increasing R² to 0.53 and 0.46 in males and females, respectively. The BMI slopes in these models minimally increased in both the males (0.22 to 0.26±0.013) and females (0.13 to 0.14±0.007). Center (Kiel or NYNORC) added significantly (p<0.01–0.001) to the BMI and age SMI prediction equations (Model 3), although changes in the BMI slopes from the previous two models were minimal. However, explorations of the various SMI scatter plots suggested the possibility of a BMI × age interaction, and this turned out to be the case in males. The interaction term, age × BMI, added significantly to the models with BMI and age (Model 4) and center (Model 5) in the SMI prediction model for males. The respective BMI slopes in these models were 0.42±0.037 and 0.41±0.036, substantially larger than those of the simpler models ranging from 0.22 to 0.26 in the males. We next solved the interaction term equation in males (Model 4) for ages 30, 40, 50, and 70 years and then plotted the respective predicted SMIs against BMI as shown in Figure 3. As demonstrated in the figure, the BMI slopes decreased from 0.30 at the youngest age of 30 years to 0.14 in the oldest group of 70 years. These cross-sectional observations suggest that the proportion of voluntary weight loss as SM is age-dependent in men.

Figure 2.

Skeletal muscle mass index (SMI) versus body mass index (BMI) in the model development male sample (upper panel; n=415) and female sample (lower panel; n=482). Additional details are provided in Table 3, model 1.

Table 2.

Skeletal muscle mass index (SMI) prediction models for males and females.

SMI Model	BMI		BMI+Age		BMI+Age+Center		BMI+Age+ Age × BMI		BMI+Age+Center+Age × BMI
Model	1	1	2	2	3	3	4	4	5	5
	Males	Females	Males	Females	Males	Females	Males	Females	Males	Females
Intercept	4.10***	4.04***	4.61***	4.59***	4.03***	4.50***	0.65	4.37***	0.44	4.01***
BMI (kg/m2)	0.22***	0.13***	0.26***	0.14***	0.27***	0.14***	0.42***	0.15***	0.41***	0.16***
Age (yrs)			−0.03***	−0.02***	−0.03***	−0.01***	0.06**	−0.01	0.05**	−0.004
Center ‡					0.49***	0.19**			0.46***	0.20**
Age × BMI (yrs × kg/m2)							−0.004***	−0.0002	−0.003***	−0.0004
R2 †	0.35***	0.39***	0.53***	0.46***	0.56***	0.47***	0.55***	0.47***	0.58***	0.48***

^‡,NYNORC, 0; Kiel, 1.

^†,10-fold validation.

^*p<0.05;

^**0.01;

^***0.001.

BMI, body mass index; SMI, skeletal muscle mass index.

Figure 3.

Skeletal muscle mass index (SMI) versus body mass index (BMI) in males at four different hypothetical ages based on SMI values predicted using Model 4 in Table 3. The values for B are for ages 30, 40, 50, and 70 years 0.30, 0.26, 0.22, and 0.14, respectively.

The SM prediction models including separate covariates for weight and height are presented in Table 3 for representative equations. The SM prediction model including weight, height, and age as covariates (Model 6) had weight slopes of 0.26±0.013 and 0.14±0.007 for males and females, respectively (p<0.001 for male-female slope difference). These models had higher R² values for both the males (0.66) and females (0.62) compared to the corresponding models including BMI in place of weight and height (R²s, 0.53 and 0.46). When added to these models, the age × weight interaction term was significant in males (p<0.001) and non-significant in females (Model 7); Model 7 is thus a useful prediction equation for SM in males and Model 6 serves that function in females:

$$\text{Males\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}SM}=-13+0.38\times \text{W}+0.18\times \text{H}+0.11\times \text{A}-0.003\times \text{A}\times \text{W}$$ (1)

$$\text{Females\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}SM}=-7.14+0.14\times \text{W}+0.12\times \text{H}-0.04\times \text{A}$$ (2)

with SM in kg; W, weight, in kg; H, height in cm; and A, age, in years. The weight slope in Model 7 again was larger in this SM prediction model for males, 0.38±0.032 versus 0.26±0.013 in the model without an interaction term. The weight slope was minimally changed with addition of the interaction term in the females. Thus, the same overall findings related to “excess” weight as SM emerge when models are developed using BMI or separate weight and height.

Table 3.

Skeletal muscle mass prediction models for males and females.

SM Model	Weight+Height+Age		Weight+Height+Age+Age × Weight
Model	6	6	7	7
	Males	Females	Males	Females
Intercept	−3.67	−7.14**	−13.00**	−8.83***
Weight (kg)	0.26***	0.14***	0.38***	0.16***
Height (cm)	0.10***	0.12***	0.10***	0.12***
Age (yrs)	−0.11***	−0.04***	0.11*	−0.004
Age × Weight (yrs × kg/m2)			−0.003***	−0.001
R2 †	0.66***	0.62***	0.67***	0.62***

^†,10-fold validation.

^*p<0.05;

^**0.01;

^***0.001.

BMI, body mass index; SM, skeletal muscle.

Model Validation

The longitudinal changes in body composition for the two studies are presented in Table 4. Males, on average, lost more weight than females in both studies (~10–13 kg vs. ~8–9 kg. BMI, SMI, and ATM index decreased by about 3–4, 0.5–1, and 2–3 kg/m², respectively. The proportions of weight loss as adipose tissue were larger in the females (e.g., 0.67±0.32–0.79±0.11) than in the males (0.56±0.10–0.63±0.09).

Table 4.

Results of longitudinal body composition evaluations.

	Kiel		CALERIE
	Males	Females	Males	Females
N	14	60	8	18
ΔWeight (kg)	−13.00±4.40	−8.34±3.84	−9.90±1.61	−9.16±3.11
ΔBMI (kg/m2)	−4.00±1.43	−2.93±1.30	−3.12±0.52	−3.41±1.12
ΔAT (kg)	−7.38±2.76	−5.85±3.77	−6.20±1.29	−7.29±2.66
ΔATMI (kg/m2)	−2.27±0.90	−2.05±1.28	−1.95±0.36	−2.72±0.98
ΔSM (kg)	−2.94±1.44	−0.79±0.92	−2.00±0.77	−1.49±0.83
ΔSMI (kg/m2)	−0.91±0.46	−0.28±0.31	−0.62±0.21	−0.55±0.29

Results are X±SD. AT, adipose tissue; ATMI, adipose tissue mass index; BMI, body mass index; SM, skeletal muscle; SMI, skeletal mass index. Kiel, 12–14 week protocol; CALERIE, 12-month protocol. All changes from baseline are significant at p<0.001.

The mean proportion of weight loss as SM was (mean±SE) 0.23±0.02 in the Kiel males and 0.20±0.02 in the CALERIE males. Corresponding predicted values for ΔSM/ΔBW (Model 7) were 0.27±0.01 and 0.25±0.01. The mean proportion of weight loss as SM was 0.10±0.02 in the Kiel females and 0.17±0.03 in the CALERIE females. Corresponding predicted values for ΔSM/ΔBW were 0.16±0.003 and 0.15±0.001. The measured proportion of weight loss as SM was significantly larger in combined values from Kiel and CALERIE for males (n=22, 0.22±0.02) compared to females (n=78, 0.11±0.02; sex difference, p=0.001). Corresponding predicted values were 0.26±0.01 for all 22 males and 0.16±0.01 for all females (sex difference, p<0.001).

DISCUSSION

The current study filled a longstanding gap in defining the magnitude of SM mass loss following induction of negative energy balance and weight loss with CR. The persistence of this gap was brought about largely by the lack of SM reference measurement methods for model development that could be critically evaluated in longitudinal validation samples. Our findings, using whole-body MRI as the reference for quantifying SM mass, show that on average non-elderly males and females with overweight and obesity experience respective reductions in SM mass of about 2–2.5 kg and 1–1.5 kg for each 10-kg weight loss. Although there was wide individual variation, these generalizations largely held across both longitudinal samples that were studied. Outside of the findings reported in the current study, the only other publications we could find in the reviewed literature related to diet-induced weight loss using whole-body MRI to measure changes in SM mass were those by Ross et al. [23] and Janssen et al. [22]. We estimated the proportion of weight loss as SM in these two studies from the mean published values at baseline and follow up. For males, the respective fractions of weight loss as SM were 0.22 and 0.23 and for females 0.11 and 0.10, consistent with the current study results.

Our overall estimates suggest that an adult with obesity who weighs 100 kg and experiences a 15% weight loss will lose about 3 kg and 2 kg of SM if male or female, respectively. These reductions in SM mass would translate to the same relative muscularity reductions in males and females of ~10%. New medicines now in development that target the activin II receptor [13] may eliminate or substantially reduce the amount of SM loss during weight loss treatments. Our developed models are thus only applicable in the context of diet-induced weight loss in the absence of a physical activity protocol and the extent to which they apply with pharmacologic therapies requires specific considerations of the drugs that are studied or evaluated. Moreover, our models may not be accurate in predicting relative SM loss when people rapidly lose weight on very low energy diets or when protein supplements are included in traditional weight loss protocols [9]. Exercise, notably resistance training, is associated with FFM, and presumably SM, conservation during dieting [2, 24].

An unexpected observation emerging from our developed models is that older males are predicted to have a smaller proportion of weight loss as SM with diet-induced weight loss than their younger counterparts at the same BMI or weight. This finding appears in the context of widely held but not well substantiated suggestions that older adults with obesity will lose disproportionate amounts of SM when embarking on a diet [25, 26]. Our longitudinal samples were relatively young (mean age ~40 yrs) and we could therefore not explore this prediction further in the current study. However, these findings suggest that our aforementioned estimates of weight loss composition as SM during dieting are most applicable to middle aged adults. Our study also confirmed that with diet-induced weight loss males lose more absolute SM mass compared to females, an observation reported for FFM in many other studies [3, 27, 28].

The current investigation focused specifically on reductions in SM mass with diet-induced weight loss. In other research, loss of SM mass with voluntary weight loss is accompanied by corresponding functional effects. For example, Alba et al. [29] reported a 13% reduction in lean mass and a 9% decrease in handgrip strength at one-year following Roux-en-Y gastric bypass surgery for severe obesity. Participants with type 2 diabetes in the intensive lifestyle intervention of the Look AHEAD study had a 39% increase in frailty fractures and significant regional bone and FFM loss compared to Diabetes Support and Education-treated controls [30]. Loss of SM, decreased strength, and problems with balance are hypothesized mechanisms [30]. These examples emphasize that functional outcomes accrue the reductions in SM mass with CR as modeled in the current study.

Modeling Approach

The applied modeling approach in the current study was patterned after the approach originally suggested by Webster and Garrow for fat mass index and BMI in 1984 [11]. The Webster-Garrow approach is founded on two main assumptions as applied in the current study: that reductions in SM mass with CR follow a linear trajectory over time and that the trajectory taken follows the path defined by a model developed on a healthy adult sample. One implication of these assumptions is that the proportions of body weight as SM and ATM in the reduced-obese state will be the same as those in people at a lower body weight who have never been obese. While there is some evidence in support of this assumption [31], apart from small and variable effects on FFM hydration with voluntary CR [32, 33], larger studies conducted with accurate body composition methods are needed for confirmation.

Non-linear models relating changes in body composition to weight gain or loss during respective periods of positive or negative energy balance were reported earlier by Forbes [34, 35] and Thomas et al. [34, 35]. Forbes, in a classic 1987 paper, described a curvilinear model relating changes in “lean body mass” measured by ⁴⁰K counting to changes in body fat [34] in females. Thomas et al. extended Forbes’s model in males and females by developing a fourth order polynomial equation associating FFM to fat mass in a representative sample of the U.S. population [35]. Both Forbes and Thomas et al. included participants with very low body weights in their samples, notably by inclusion of females with anorexia nervosa. To explore their observations further, we added a group of 22 females with anorexia nervosa to the current healthy adult female sample, all of whom had whole-body MRI studies [36]. A plot of SMI against ATM index for this supplemented sample confirmed a curvilinear function in females (Figure S3). These observations and those of Forbes and Thomas et al. suggest that loss of SM is larger relative to CR-induced weight and adipose tissue loss in females who are very lean compared to those who are overweight or obese. Comparable data in males are lacking. A non-linear decrease in nitrogen balance and FFM was also observed in the early phase of fasting and low-calorie dieting by Benedict [37] (Figure S4) and Forbes [34]. Our models and their predictions are thus applicable only to samples including people whose weights are in the normal range or above and for weight loss periods extending for several weeks or more.

Earlier studies examining changes in FFM with CR-induced weight loss often conflate this large compartment with SM mass. While SM comprises about one-half of FFM in adults [38], some organs and tissues are relatively unaffected during periods of negative energy balance (e.g., brain) while others undergo rapid atrophy (e.g., liver) [15, 39]. Nevertheless, our findings with SM largely follow earlier observations with FFM such as a larger fraction of weight loss as FFM loss in males than in females. Ideally, future studies of drugs or exercise protocols targeting SM during periods of negative energy balance will include methods such as MRI that specifically measure the SM compartment. Methods such as D₃-creatine dilution could probe further into changes with dieting in the contractile and non-contractile components of SM that might provide insights into functional effects of specific protocols [40, 41]. Measuring a specific muscle component such as myofibrillar mass might avoid concerns that relative muscle composition is altered in the weight-reduced state, as for example with increased connective tissue and water and deceased lipid [32, 42].

Study Limitations

The current study model development and validation samples were limited in adult age and race/ethnicity, restricting to some extent the scope of generalizations reported herein. Our measurements did not include estimates of SM composition or mechanical function, both important topics for future studies. The weight loss observed in our two longitudinal studies was not very large and there was considerable individual variation in SM loss. Larger weight and SM loss can be anticipated with greater magnitudes of weight loss, as may be observed with newer pharmacologic agents and bariatric surgery. Lastly, our analyses revealed small but significant between-center differences in model predictions of SM. This observation could be anticipated as there is yet no agreed-upon universal protocol for whole-body MRI scan acquisition and analysis.

Conclusions

The current study establishes the relative amount of SM loss associated with negative energy balance induced with volitional CR. The model predictions can serve as a baseline against pharmacologic and exercise interventions compared for their muscle-sparing effects, for developing power calculations when designing future studies, and to test generated hypotheses related to aging effects on SM loss with voluntary weight reduction. Our prediction models describing SM loss with CR in adults could be extended to CR-induced weight-loss effects in children and adolescents and in people treated with bariatric surgery. Future studies can also explore the associations between weight gain and accretion of SM with models such as those reported in the current study.

What is already known about this subject?

• Voluntary weight loss with low-calorie diets is accompanied by reductions in not only body fat, but fat-free mass and its individual organ-tissue components.

• The proportion of caloric restriction-induced weight loss specifically as skeletal muscle is largely unknown.

What are the new findings in your manuscript?

• This is the first study to report development and validation of models relating caloric-restriction-induced weight loss to proportional reductions in skeletal muscle mass in healthy adult males and females.

• Our findings indicate that, on average, non-elderly males and females experience respective reductions in skeletal muscle mass in the absence of a structured exercise protocol of about 2–2.5 kg and 1–1.5 kg per 10 kg weight loss.

How might your results change the direction of research or the focus of clinical practice?

• These observations provide a framework for evaluating and implementing interventions aimed at preserving skeletal muscle mass and function over the course of dieting and subsequent weight maintenance.

• Our models also generate new hypotheses related to the impact of dieting on skeletal muscle mass in older adults.

DATA SHARING

Data described in the manuscript will be made available upon request pending application and approval by the investigators.