Association between Fat-Free Mass Loss after Diet and Exercise Interventions and Weight Regain in Women with Overweight

Abstract

Purpose:

To determine if % fat-free mass loss (%FFML) following diet alone, diet plus aerobic or diet plus resistance exercise, is a predictor of weight regain, in women with overweight.

Methods:

141 premenopausal women with overweight (BMI: 28±1kg/m²; age: 35±6 years) enrolled in a weight loss program to achieve a BMI <25 kg/m² (diet alone, diet plus resistance or diet plus aerobic exercise) and were followed for 1 year. Body weight and composition (with DXA) were measured at baseline, after weight loss and at 1 year.

Results:

Participants lost 12.1±2.6 kg of body weight, 11.3±2.5 kg of FM and 0.5±1.6 kg of FFM during the weight loss intervention, followed by weight regain at 1 year (6.0±4.4 kg; 51.3±37.8%) (P<0.001 for all). %FFML was −3.6±12.4 and a greater %FFML was associated with more weight regain (r = −0.216, P=0.01, n=141), even after adjusting for intervention group (ß:−0.07; 95% CI: −0.13, −0.01; P=0.017).

Conclusions:

%FFML is a significant predictor of weight regain in premenopausal women with overweight. These results support strategies for conserving FFM during weight loss, such as resistance training. Future research should try to identify the mechanisms, both at the level of appetite and energy expenditure, responsible for this association.

INTRODUCTION

Weight regain remains the main challenge in obesity management. Even though clinically significant weight loss is achievable in the short-term, long-term results are disappointing (1, 2), with some patients relapsing to or above baseline weight. The exact reasons for weight recidivist remain unexplained (3, 4). The search for new metabolic determinants of relapse in obesity management is therefore needed to better understand weight regain and help develop new strategies to improve long-term weight loss outcomes.

Recently there has been increased interest in the functional role of body composition, in particular fat free mass (FFM), in modulating eating behavior and body weight (5, 6). Even though the traditional view has been that body weight homeostatic and appetite control are regulated through feedback signals arising from the adipose tissue, in particular leptin, a growing body of evidence suggests that FFM also plays a role in the drive to eat (7–9). It has been proposed that FFM can modulate energy intake and body weight both directly, through feedback signaling between FFM and brain centers involved in appetite control (10, 11), and indirectly, via its effects on resting energy expenditure (EE) (12, 13) and free living activity related energy expenditure (14–16). Loss of FFM following lifestyle interventions may, therefore, contribute to weight regain, not only due to the body’s attempt to restore FFM by overeating, but also to the resulting lower EE and energy needs.

In a reanalysis of the Minnesota Starvation Experiment by Dulloo and colleagues (17), a greater initial loss of FFM was associated with a stronger hyperphagic response during refeeding, even after adjusting for fat mass (FM) loss. Surprisingly, hyperphagia persisted despite complete recovery of body weight and FM, until FFM was totally restored to baseline values. This landmark paper is, nevertheless, limited by the fact that it included lean men only and the energy deficit was extreme (20–26% loss of initial body weight due to semi-starvation). It is unknown if a similar response is activated in individuals with overweight or obesity following less restricted lifestyle interventions, and if FFM loss modulates weight regain. A greater fractional loss of FFM (%FFML), following diet-induced weight loss in 55 individuals with overweight and obesity, was found to be predictive of weight regain at 9 months, even after adjusting for baseline FM% (18). A recent meta-analysis of more than 2000 individuals with overweight or obesity suggested that reductions in FM and FFM better predict weight regain compared with weight loss alone (19).

In a recent reanalysis of the DiOGenes project, %FFML predicted weight regain at 6 months follow-up in men, but not in women, and a larger %FFML was associated with a larger increase in postprandial hunger and desire to eat in men (20). All the available evidence regarding a potential role for %FFML on weight regain arises from studies where weight loss was induced by low-energy or very-low energy diets (5, 18, 19). It remains to be determined if %FFML is associated with weight regain when weight loss is induced by combined lifestyle interventions, including both diet and exercise, given that exercise, in particular resistance exercise, is known to prevent the decrease in FFM that usually accompanies diet-induced weight loss (15, 21, 22).

Therefore, the aims of this exploratory post hoc analysis were to determine the association between %FFML following diet alone, diet plus resistance exercise, or diet plus aerobic exercise, and weight regain at 1 year follow-up in a group of women with overweight. We hypothesized that a greater proportional reduction in FFM (%FFML) would result in greater weight regain at 1 year follow-up.

METHODS

Participants

Participants in this analysis were premenopausal women with overweight (25≤BMI≤30 kg/m²). They were 20–41 years of age, sedentary (no more than one time per week regular exercise), had normal glucose tolerance (2–h glucose ≤140 mg/dL following 75g oral dose), family history of overweight/obesity in at least one first-degree relative, and no use of medications that affect body composition or metabolism. All women were nonsmokers and reported a regular menstrual cycle. The two studies included in this retrospective analysis were both approved by the Institutional Review Board for Human Use at the University of Alabama at Birmingham (UAB). All women provided informed consent before participating in the study.

Study design

Participants included in this retrospective analysis come from 2 different studies performed at the Department of Nutrition Sciences at UAB, with the same sequence of events and same methodology, and both aiming to identify metabolic predictors of weight regain. In the first study, all participants achieved weight loss with diet alone (single arm longitudinal study with repeated measurements). In the second study, participants were randomly assigned to one of three groups: 1) Diet alone (same diet as in the first study); 2) Diet plus aerobic exercise training 3 times/week; and 3) Diet plus resistance exercise training 3 times/week. All participants were provided an 800-kcal diet until reaching a BMI <25 kg/m². Food was provided (20–22% fat, 20–22% protein, and 56–58% carbohydrate) by the General Clinical Research Center (GCRC) Kitchen. During follow-up (1 year), participants were encouraged, but not mandated, to attend regular support group meetings (bimonthly dietary education classes aimed at weight maintenance for the first 6 months, followed by monthly meetings for months 6 to 12) and to continue with their exercise program, if applicable. For detailed information about the two studies, see Weinsier et al, 2000 (23) and Hunter et al, 2008 (21).

Testing was done, after a 4-week weight stabilization period (aiming to maintain body weight within a 2.5 kg range), at baseline, after weight loss, and at 1 year follow up. During the 4-week weight stabilization period, participants were weighed 3 times/week the first 2 weeks while eating own food and weighed 5 times/week with food provided by GCRC the last 2 weeks. Variation in body weight during the last 2 weeks of the stabilization period, after weight loss, was −1.0±1.4 kg. Testing was done 30±2 days after the end of the weight loss phase. All testing was conducted in the follicular phase of the menstrual cycle during a 4-day GCRC in-patient stay (to ensure that physical activity and diet was standardized). Testing was done in a fasted state in the morning after spending the night in the GCRC.

Data Collection

Body weight and composition (fat mass (FM) and FFM) were assessed by dual-energy X-ray absorptiometry (DXA) (DPX-L; Lunar Corp, Madison, WI) with the use of software version 1.5g (Lunar Corp), at baseline, after weight loss and at 1 year follow up, after a 4-week weight stabilisation period (at all-time points).

Statistical analysis

Statistical analysis was performed with SPSS version 22 (SPSS Inc., Chicago, IL), and data presented as mean ± SD, unless otherwise stated. Statistical significance was set at P < 0.05. All variables were assessed for normality by visual inspection of Q-Q plots and histograms. Participants from all groups were combined and included in the analysis if they had anthropometric data at baseline, after weight loss, and at 1 year follow-up (n=141). This was done to ensure enough variance in the response variable (loss of FFM) and to allow for sufficient sample size required to have robust examination of multiple regression models.

Changes in body weight, FM and FFM over time were assessed by a repeated measures ANOVA, with bonferroni correction for post hoc pairwise comparisons.

The proportion of weight lost as FFM was calculated as the change in FFM during weight loss divided by total weight loss [i.e., %FFML=(∆FFM/∆weight) ∗100]. There was one extreme value for %FFML (40.5%), which was excluded from the analysis. The association between %FFML and weight regain was investigated with Pearson correlation. Univariate linear regressions were also conducted to investigate crude associations between potential predictors and the outcome variable (weight regain). β-Coefficients were reported as unstandardized estimates and 95% CIs, representing the estimate and confidence of a 1-unit change in the predictor variable per 1-kg change in weight regain at 1 year (1 year minus post-intervention). Next, multivariate linear regression models were generated for all individuals. Adjustments were considered for intervention group, the amount of weight lost (because weight loss has been shown to be a strong predictor of weight regain) (19), as well as initial FM, given that baseline body composition has been shown to modulate body composition changes with weight loss (24). Multicollinearity was tested by examining the variance inflation factors (VIF) of the model variables and was deemed acceptable.

RESULTS

One hundred and forty-one women (68 white and 73 black) were included in this analysis. They had an average age of 35±6 years and a BMI of 28±1 kg/m². Anthropometrics at baseline, after weight loss and at 1 year follow-up are presented in Table 1. Significant reductions in BMI, body weight and FM (kg) were seen both after weight loss and at 1 year, compared with baseline (P<0.001 for all). Participants lost on average 4.5±1.0 kg/m², 12.1±2.6 kg of body weight, 11.6±2.5 kg of FM and 0.5±1.6 kg of FFM (range −5.5 to 3.1 kg) during the weight loss intervention. This was followed by weight (6.0±4.4 kg) and FM (5.9±3.9 kg) regain at 1 year (P<0.001 for both), even though values were still below baseline at 1 year follow-up (P<0.001 for both). FFM (kg) was maintained between end of intervention and 1 year (−0.1±1.5 kg, P=0.579). %FFML post-intervention was −3.6±12.4% (range: −32.7 to 28.2%). Average weight regain from end of weight loss to 1-year was 6.0±4.4 kg (51.3±37.8%), with a very large inter-individual variation (−3.6 to 20.3 kg), even though 95% of the participants regained weight (anything above zero) over time. %FFML was significantly larger in the diet group (−9.7±9.9%) compared with both diet + aerobic (−2.2±10.0%) and diet + resistance exercise (3.6±12.8%) (P=0.007 and P<0.001, respectively), and there was also a tendency for %FFML to be larger in the diet + aerobic versus the diet + resistance exercise group (P=0.068).

Table 1.

Anthropometric variables over time (n=141)

	Baseline	Post-intervention	1 year
BMI (kg/m2)	28.3±1.3ab	23.9±1.0ac	26.1±2.2bc
Weight (kg)	77.3±7.1ab	65.1±6.3ac	71.1±8.1bc
FM (kg)	33.3±4.8ab	21.7±4.3ac	27.6±6.1bc
FFM (kg)	44.0±4.0ab	43.5±4.1a	43.5±4.4b
%FFML		−3.6±12.4
Weight regain (kg)			6.0±4.4
Weight regain (%)			51.3±37.8
FM regain (kg)			5.9±3.9
FFM regain (kg)			−0.1±1.5

Data presented as mean±SD. BMI: body mass index; FM: fat mass; FFM: fat-free mass; %FFML: % FFM loss.

Means sharing the same superscript letters denote significant changes over time: ^a,b,c P<0.001.

There was a significant association between %FFML and weight regain at 1 year (r = −0.216, P=0.01, n=141), with a greater %FFML (or lower accretion) being associated with more weight regain (see Figure 1). Table 2 provides univariate regression results predicting weight regain (from end of weight loss to 1 year). Baseline FFM (kg), but not weight or FM (kg or %), significantly predicted weight regain (β: 0.220 kg; 95% CI: 0.041, 0.399 kg; R2 = 0.0334, P=0.016). Moreover, FFM loss (kg) (β: −0.628 kg; 95% CI: −1.083, −0.174 kg; R2 = 0.044, P=0.007) and %FFML (β: −0.076 kg; 95% CI: −0.134, −0.019 kg; R2 = 0.040, P=0.01) were also predictors of weight regain. Age, race, and weight or FM loss were not significant predictors of weight regain. There was a tendency for intervention group to predict weight regain (β: 0.821 kg; 95% CI: −0.090, 1.732 kg; R2 = 0.015, P=0.077).

Figure 1.

Scatterplot for the association between the proportion of weight lost as fat-free mass (%FFML) with the weight loss intervention and weight regain at 1 year.

Table 2.

Univariate regression analysis predicting weight regain at 1 year

Predictor	ß-coefficient	Adjusted R2	P value
Age	0.067 (−0.047, 0.181)	0.002	0.248
Race	0.793 (−0.654, 2.241)	0.001	0.280
Group	0.821 (−0.090. 1.732)	0.015	0.077
Baseline weight (kg)	0.080 (−0.022, 0.183)	0.01	0.124
Baseline FFM (kg)	0.220 (0.041, 0.399)	0.034	0.016
Baseline FM (kg)	−0.021 (−0.132, 0.173)	0.001	0.789
Baseline FM (%)	−0.132 (−0.336, 0.071)	0.005	0.201
Weight loss (kg)	−0.050 (−0.329, 0.230)	0.001	0.727
FM loss (kg)	0.195 (−0.097, 0.488)	0.005	0.188
FFM loss (kg)	−0.628 (−1.083, −0.174)	0.044	0.007
%FFML	−0.076 (−0.134, −0.019)	0.040	0.010

Univariate linear regression analyses predicting weight regain at 1 year (from week 9). Each unstandardized ß-coefficient represents 1 kg weight regain at 1 year per unit of the predictor. FFM: fat-free mass; FM: fat mass; %FFML: fat-free mass loss.

Table 3 reports a multivariate linear regression model predicting weight regain. %FFML was a significant predictor of weight regain at 1 year (β: −0.07; 95% CI: −0.13, −0.01 kg; P=0.017), even after adjusting for intervention group, with the model explaining 5% of the variation in weight regain. Baseline FM (kg) and FM loss (kg) were not included in the model, as they were not significant predictors and reduced the R² and statistical significance of the model (see Supplemental Table, SDC 1, Multivariate linear regression models predicting weight regain at 1 year).

Table 3.

Multivariate linear regression model predicting weight regain at 1 year

Predictor	ß-coefficient (95% CI)	P value	Adjusted R2
A. Multivariate model		0.012	0.048
Constant	4.77 (2.60, 6.94)	<0.001
Group	0.50 (−0.43, 1.454)	0.292
%FFML	−0.07 (−0.13, −0.01)	0.017

Multivariate linear regression analyses predicting weight regain at 1 year. Each unstandardized ß-coefficient represents 1 kg weight regain at 1 year per unit of the predictor. FM: fat mass; %FFML: fat-free mass loss. Variance inflation factors (VIF) <1.1.

Those with a negative %FFML (n=88), denoting loss of FFM during the intervention, regained significantly more weight compared with those with a positive %FFML (n=53), denoting an increase in FFM during weight loss (6.4±4.9 versus 5.2±3.1 kg, P=0.004).

DISCUSSION

The present secondary analysis aimed to determine the association between %FFML during weight loss induced by diet alone, diet plus aerobic exercise, or diet plus resistance exercise, on weight regain at 1 year follow-up in women with overweight. There was an inverse association between %FFML after weight loss and weight regain at 1 year, with a greater %FFML (or lower accretion) being associated with more weight regain. Moreover, those with a negative %FFML, denoting loss of FFM during the weight loss intervention, regained significantly more weight at 1 year follow-up (on average more than 1.2 kg), compared with those with a positive %FFML (i.e. gained FFM during the intervention. Finally, %FFML was a significant predictor of weight regain, even after adjusting for intervention group.

Previous studies have found an association between %FFML and weight regain in men (20) or in a sample composed mainly (90%) of men (18), and in heavier samples: Vink et al study (average BMI: 31 kg/m²) (18); Turicchi et al (weight: 99.6±16.3 kg) (20). Weight regain is likely to be weaker in fatter individuals, given that the proportion of weight lost as FFM decreases with increasing BMI and FM (25). Moreover, the association between %FFML and weight regain is likely to be stronger in men, given that they have less FM compared with women at baseline and are, therefore, more likely to lose FFM following weight loss interventions (26). This study adds to the literature by showing an association between %FFML and weight regain in premenopausal women with overweight.

Previous studies on the association between %FFML and weight regain have used diet alone interventions to induce weight loss (18, 20). The present analysis represents the first study where a combination of interventions was used, including diet alone, diet plus aerobic exercise and diet plus resistance exercise. This helps to explain why average %FFML in the present study was only −3.6±12.4%, while in Turicchi and colleagues study it was −30% (20), and in Vink et al study −8.8% and −1.3% in the VLED and LED-groups, respectively (18). Exercise, in particular resistance exercise, has been shown to minimize the decrease in FFM that follows diet-induced weight loss, or even to increase FFM (15, 21, 22). The great variation in %FFML in the present study, from loss (−33%) to accretion (28%), probably helped in finding an association between %FFML and weight regain.

In the present study weight regain at 1 year follow-up was 6.0±4.4 kg (51.3±37.8%), ranging from −3.6 to 20.3 kg. This is in line with Vink et al study: 4.2 kg (59%) in the low-energy diet group and 4.5 kg (55%) in the very-low energy diet (VLED) group at 1 year follow-up (18), but much larger than in Turicchi et al study (1.6 kg (14%) overall; men 3.0 kg (23%)), likely due to the shorter follow-up (6 months) (20). Body composition in the present study was measured under conditions of weight stability, which is in line with the methodology followed by Vink et al (18). This is opposite to Turicchi et al study, where body composition was measured immediately after weight loss (in negative energy balance (EB)). Moreover, Turicchi et al used an 800 kcal/day diet. Negative EB is accompanied by glycogen depletion and, with it, water loss, while refeeding is followed by glycogen replenishment and with it increased water content (27, 28). Turicchi et al used DXA (20), a method that fails to account for changes in the hydration of lean tissue, and, as such, losses of FFM might be exaggerated when measurements are done under negative EB, particularly when in glycogen-depleted conditions.

The exact mechanisms through which %FFML modulates eating behavior and body weight homeostasis remain to be fully elucidated. However, changes in the size and functional integrity of FFM may be involved (17, 29). It has been suggested that proteinstats, signals released from the muscle (both organs and skeletal muscle) during FFM loss may act at the level of the brain to modulate appetite control (30). Additionally, FFM may also modulate the hedonic system. In a 12-week aerobic exercise intervention study, in young adults with overweight/obesity, Flack and colleagues (31) showed a greater increase in reward-driven feeding (using a behavioral choice task) in those who lost the greatest amount of FFM with the exercise intervention (31). Another potential mechanism for increased weight regain following FFM loss may be because of reduced muscle function. It has been shown that ease of walking is related to participation in free living physical activity (14, 15). It has also been established that increased muscle and strength results in increased ease of walking (16). It is thus possible that those individuals who maintained muscle mass with weight loss were able to walk with less effort and were thus inclined to be more physically active. Indeed, we have previously shown that resistance training during weight loss may decrease weight regain (32). Regardless of the mechanisms, these results suggest that strategies to maintain, or increase, muscle mass during weight loss should be followed in order to limit weight regain in the long-term.

This study has both strengths and limitations. A strength of this study is that measurements of body composition were done under conditions of weight stability at all time points. This is important because it ensures that body composition results were not affected by negative EB and/or shifts in body water compartments that accompany active weight loss. Moreover, all testing was conducted in the follicular phase of the menstrual cycle. Menstrual cycle is known to modulate body composition, even when performed by DXA, with FM (kg and %) being lower in the late follicular phase compared to the mid-luteal phase in women in a natural menstrual cycle (33). Finally, the present analysis displayed a large range of %FFML from loss to accretion and a large range of weight regain, important to establish an association between these two variables. However, this study also suffers from some limitations. First, it includes a very homogenous sample of premenopausal (20–41 years) women with overweight. This prevents the generalisation of our results to men, other BMI groups and older subjects. Second, this is a secondary analysis of data, and as such, it is likely underpowered to examine the association between %FFML and weight regain. More studies, with larger sample sizes, a balanced sex distribution that allows for the investigation of potential important sex interactions, and a large range of BMI and FM are needed. Additionally, further research should employ advanced methods of body composition and multicompartment models in order to try to establish the underlying mechanisms involved.

CONCLUSIONS

In conclusion, %FFML is a significant predictor of weight regain at 1-year in premenopausal women with overweight, suggesting that strategies such as resistance training during weight loss may be beneficial for future weight loss maintenance. Future research should try to identify the mechanisms, both at the level of appetite and energy expenditure, responsible for this association.