Weight suppression and risk of future increases in body mass: effects of suppressed resting metabolic rate and energy expenditure

Abstract

Background: Weight suppression, which reflects the difference between the highest previous weight and current weight, has predicted future increases in body mass index (BMI) and bulimic pathology; however, the mechanisms underlying these predictive effects are unclear.

Objective: The current study sought to test whether weight suppression predicts future increases in BMI and bulimic symptoms and whether suppressed resting metabolic rate (RMR) and suppressed total energy expenditure (TEE) drive these relations.

Design: A randomly selected subsample of 91 young women in their first year of college with body image concerns completed an RMR assessment—a doubly labeled water assessment of TEE—and provided data on weight suppression and change in BMI and bulimic symptoms over a 6-mo follow-up period.

Results: Weight suppression predicted future increases in BMI and correlated inversely with suppressed RMR and TEE, yet this predictive effect did not decrease when suppressed RMR and TEE were controlled for. Weight suppression, however, did not predict future increases in bulimic symptoms.

Conclusions: The results provide additional evidence that weight suppression predicts future increases in BMI but not in bulimic symptoms. Weight suppression showed moderate relations with suppressed RMR and TEE, but these variables do not appear to drive the predictive effect on future increases in BMI. This trial was registered at clinicaltrials.gov as NCT00433680.

INTRODUCTION

Weight suppression, which is the difference between highest past weight and current weight (1), has predicted increases in body mass index (BMI) scores during college (2) and during eating disorder treatment (3–5). Weight suppression has also predicted the future onset of bulimic syndromes (threshold or subthreshold bulimia nervosa, binge-eating disorder, or purging disorder) and the persistence of these syndromes in initially afflicted individuals (6). Individuals who showed a reduction in body mass of ≥10%, which was directly measured, evidenced a 7-fold increase in the risk of future onset of threshold or subthreshold bulimia nervosa (7), although this effect was not significant. Furthermore, weight suppression predicted a poorer response to cognitive behavioral therapy for bulimia nervosa and treatment dropout (8), although these relations were not replicated (5), potentially because there was less weight suppression in this latter study (6). These data indicate that weight suppression increases the risk of weight gain and bulimic pathology and a poor response to treatment, which suggests that it would be useful to understand the mechanisms by which weight suppression relates to these important outcomes.

Weight suppression may lead to future weight gain because it is associated with increased metabolic efficiency and consequent reductions in habitual energy expenditure (8, 9). RMR accounts for 60% to 70% of total energy expenditure (TEE) in humans (10), and TEE is highly dependent on RMR (11). A meta-analytic review found that formerly obese individuals typically show a lower RMR relative to those without a history of obesity (12). However, Saltzman and Roberts (13) reported that cross-sectional studies have found that weight-stable postobese individuals have TEE and resting energy expenditure (REE) values similar to those of weight-matched nonobese control subjects.

In addition, if weight suppression is related to reduced RMR and TEE, this could result in unhealthy weight control behaviors, such as fasting and compensatory weight control techniques (eg, vomiting), and consequent increased risk of binge eating. These factors may conspire to result in higher dropout from eating disorder treatment and poorer treatment response. However, these questions have not been directly tested.

The first study aim was to test the hypothesis that greater weight suppression predicts future increases in BMI and bulimic symptoms. The second study aim was to test the hypothesis that weight suppression would correlate inversely with depressed RMR and TEE and that the effect of weight suppression on future increases in BMI and bulimic symptoms would be attenuated when RMR and TEE were statistically controlled. We focused on suppressed RMR and TEE, which represents the deviation of observed values from expected values because this reflects the degree of suppression from what one would expect based on sex, height, weight, and age. We used doubly labeled water (DLW) measures of TEE because it is highly accurate and immune to self-presentation bias (14).

SUBJECTS AND METHODS

Participants

Participants were 91 young females (see Table 1 for sample characteristics) with body image concerns, who were recruited from a local university. Participants were 89% white, 2% American Indian or Alaska Native, 2% Asian, and 6% “other” or of mixed racial heritage. Exclusion criteria included pregnancy, diabetes, conditions requiring supplemental oxygen, or DSM-IV ( 4th edition) anorexia nervosa, bulimia nervosa, or binge-eating disorder (15). Participants provided data during 4 visits to the laboratory: baseline (T1), 2–4 wk after baseline (T2), 4–6 wk after baseline (T3), and 6 mo after baseline (T4). Participants were also required to avoid traveling >200 miles (≈320 km) from the study site in the 2 wk between T2 and T3 because of regional differences in concentrations of naturally occurring elements in the drinking water (deuterium and oxygen-18) that can affect the levels in the DLW isotope used to calculate TEE.

TABLE 1.

Subject characteristics¹

	Mean ± SD	Minimum	Maximum
Age (y)	18.4 ± 0.6	17	20
Baseline weight (kg)	65.1 ± 12.4	45.2	116.9
Baseline BMI (kg/m2)	23.6 ± 4.1	17.6	43.9
Weight suppression (kg)	3.4 ± 4.1	0	25.2
Measured RMR (kcal/d)	1337 ± 218	835	2165
Predicted RMR (kcal/d)	1431 ± 131	1199	1937
Suppressed RMR (kcal/d)	−110 ± 135	−483	252
Measured TEE (kcal/d)	2448 ± 351	1739	3285
Predicted TEE (kcal/d)	2550 ± 183	2161	4031
Suppressed TEE (kcal/d)	−119 ± 245	−647	558

¹n = 91. Data presented are untransformed. RMR, resting metabolic rate; TEE, total energy expenditure.

Measures

Weight suppression

At baseline, participants were weighed and self-reported their previous highest weight ever. In accordance with past studies, weight suppression was calculated as the difference between the participants’ previous highest self-reported weight and measured weight at baseline (2, 6, 8). Weight recalled retrospectively after a 20-y period has shown a strong correlation with measured weight from that time (r = 0.85) (16).

RMR

RMR was measured by indirect calorimetry with a TrueOne 2400 Metabolic Measurement System (ParvoMedics Inc, Sandy, UT). The participants fasted for 5 to 15 h and abstained from exercising for 24 h before testing. For the RMR assessment, the participants rested quietly in a temperature-controlled room for 20 min, after which a transparent plastic hood that was connected to the device was placed over the participant's head. To determine RMR, resting gas exchange was measured by using calculations of oxygen consumption and carbon dioxide production obtained at 10-s intervals for 30 to 35 min. The participants remained motionless and awake, and the last 25–30 min of the measurement were used to calculate RMR. The validity and reliability of this method for assessing RMR have been established (17–19).

Suppressed RMR

Suppressed RMR was calculated as the difference between measured RMR and predicted RMR values. The predicted RMR values were calculated by using the Mifflin-St Jeor equation: 10 × wt (kg) + 6.25 × ht (cm) − 5 × age − 161 (20). Then, suppressed RMR was calculated as measured RMR − predicted RMR.

Energy expenditure

DLW was used to estimate TEE over a 2-wk period. DLW uses isotopic tracers to assess total carbon dioxide production, which can be used to generate accurate estimates of habitual caloric expenditure (21). It is considered to be the gold standard measure of habitual caloric expenditure because participants can be kept blinded to the objective of the study, it provides an estimate of total caloric expenditure over the observational period, and it requires minimal effort on the part of participants (21). Women arrived at the laboratory after having fasted for 5 to 15 h, and DLW was administered immediately after the subject testing negatively for pregnancy. The doses were 1.6–2.0g H₂¹⁸O (10 atom percentage)/kg estimated total body water. Spot urine samples were collected immediately before DLW was administered and then 1, 3, and 4 h after dosing during the second visit to the laboratory. Two weeks later during the third visit to the laboratory, 2 additional spot urine samples were collected at the same time of day as the 3- and 4-h postdosing samples. No samples were the first void of the day.

TEE was calculated by using Equation A6 in the publication by Schoeller et al (21), the dilution space ratios of Racette et al (22), and the modified Weir (23) equation as described by Black et al (24).

Suppressed TEE

Suppressed total energy expenditure was calculated as the difference between measured TEE and predicted TEE. Predicted TEE was calculated based on the 2002 Dietary Reference Intake (DRI) for nonobese adult women (25). DRI values were chosen because they are based on recent scientific knowledge about energy requirements for healthy populations. The DRI prediction requires an estimate of physical activity level (PAL), which was calculated from DLW-measured TEE divided by measured RMR (PAL = measured TEE/measured RMR). The equation used for each participant was as follows:

Suppressed TEE values were calculated as measured TEE − predicted TEE.

Height and weight were measured at each assessment, and BMI (kg/m²) was calculated. Height was measured to the nearest millimeter by using a direct reading stadiometer with the body positioned such that the heels and buttocks are against the vertical support of the stadiometer. Weight was assessed to the nearest 0.1 kg by using digital scales with participants wearing light clothing without shoes or coats. Two measures of height and weight were obtained and averaged.

Eating pathology

The Eating Disorder Diagnostic Interview, a semistructured interview adapted from the Eating Disorder Examination (26), assessed DSM-IV eating disorder symptoms. Items assessing the symptoms in the past month were summed to create an overall bulimic symptom composite for each assessment (observed scale range: 0–92), as done previously (27, 28). The symptom composite showed internal consistency (α = 0.92), 1-wk test-retest reliability (r = 0.90), sensitivity to detecting intervention effects, and predictive validity for future onset of depression in past studies (28–30).

Statistical analyses

Regression analyses were used to test the relation of weight suppression to future changes in BMI and bulimic symptoms (aim 1), whether weight suppression correlated with suppressed RMR and TEE, and whether the 2 prospective effects would be attenuated when suppressed RMR and TEE were controlled statistically (aim 2). For the prospective analyses of BMI, we regressed BMI from the 6-mo follow-up on baseline BMI and weight suppression, including each of the physiologic variables (ie, suppressed RMR and TEE) in separate models due to the colinearity among the physiologic variables. We conducted parallel analyses for the measure of bulimic symptoms. Sixteen of the 91 subjects had measured current weights equal to or higher than their reported highest-ever weights. If current weight was higher than the self-reported highest past weight, the latter was recoded to the same value as the current weight, giving a weight suppression value of zero. This approach was used previously (31). Variables with skew coefficients >2 were normalized with either a square root transformation (eg, weight suppression) or log base₁₀ transformation if necessary (eg, eating pathology and BMI). Transformed data are presented unless otherwise noted. All analyses were performed by using SPSS software (PASW version 18.0.2, 2009; SPSS Inc, Chicago, IL). All tests were 2-sided with significance levels set at P < 0.05.

RESULTS

Relation of weight suppression to future change in BMI and bulimic symptoms

Mean (±SD) weight suppression was 3.4 ± 4.1 kg. Mean measured RMR was 1337 ± 218 kcal/d. Mean measured TEE was 2448 ± 351 kcal/d. Mean suppressed RMR and TEE were −110 ± 135 and −119 ± 245 kcal/d, respectively (Table 1). Weight suppression showed a significant inverse correlations with suppressed RMR and TEE (Table 2; Figure 1). The mean change scores for the 6-mo follow-up of BMI showed an increase (log base₁₀ = 0.001 ± 0.02; untransformed = 0.35 ± 1.12), whereas bulimic symptoms decreased (log base₁₀ = −0.24 ± 0.34; untransformed = −4.56 ± 12.76).

TABLE 2.

Pearson correlations between weight suppression, suppressed resting metabolic rate (RMR), and suppressed total energy expenditure (TEE)¹

	Weight suppression	Suppressed RMR	Suppressed TEE
Weight suppression	—	−0.22^	−0.27*
Suppressed RMR		—	0.88**
Suppressed TEE			—

¹n = 83. ^{^}P = 0.05, *P < 0.05, **P < 0.01.

FIGURE 1.

Pearson correlations between weight suppression and suppressed resting metabolic rate (RMR; A) and total energy expenditure (TEE; B). n = 83. Weight suppression showed a significant inverse correlation with suppressed RMR (panel ; r² = 0.05, P = 0.05) and suppressed TEE (panel ; r² = 0.07, P < 0.05).

Baseline weight suppression predicted increases in BMI over the 6-mo follow-up, which indicated that participants with elevated weight suppression showed greater increases in BMI over follow-up (P < 0.05; Table 3, model 1). Baseline weight suppression did not predict future increases in bulimic symptoms over the 6-mo follow-up (P = 0.45; Table 4).

TABLE 3.

Regression analyses between weight suppression and future increases in BMI¹

	Unstandardized β coefficients ± SE	95% CI for β	Semipartial R	P value
Model 1
Baseline BMI	0.97 ± 0.04	(0.89, 1.04)	0.96	0.001**
Weight suppression	0.01 ± 2.09 × 10minus3	(1.24 × 10minus3, 0.01)	0.09	0.011*
Model 2
Baseline BMI	0.97 ± 0.04	(0.89, 1.04)	0.95	0.001**
Weight suppression	5.49 × 10minus3 ± 2.17 × 10minus3	(1.15 × 10minus3, 0.10)	0.09	0.014*
Suppressed RMR	5.44 × 10minus7 ± 1.74 × 10minus5	(−3.43 × 10minus5, 3.54 × 10minus5)	1.16 × 10minus3	0.975
Model 3
Baseline BMI	0.97 ± 0.04	(0.89, 1.05)	0.92	0.001**
Weight suppression	5.74 × 10minus3 ± 2.29 × 10minus3	(1.16 × 10minus3, 0.01)	0.09	0.015*
Suppressed TEE	2.18 × 10minus6 ± 1.05 × 10minus5	(−1.88 × 10minus5, 2.32 × 10minus5)	0.01	0.836

¹n = 65. The dependent variable is BMI at the 6-mo follow-up. RMR, resting metabolic rate; TEE, total energy expenditure. *P < 0.05, **P < 0.01.

TABLE 4.

Regression analyses between weight suppression and future increases in bulimic symptoms¹

	Unstandardized β coefficients ± SE	95% CI for β	Semipartial R	P value
Baseline bulimic symptoms	0.49 ± 0.09	(0.30, 0.67)	0.50	0.001**
Weight suppression	0.02 ± 0.03	(−0.04, 0.08)	0.07	0.447

¹n = 78. The dependent variable is bulimic symptoms at the 6-mo follow-up. **Significant at P < 0.01.

Effect of suppressed RMR and TEE on the relation of weight suppression to future change in BMI

The predictive effect of weight suppression on increases in BMI remained significant after suppressed RMR and TEE were added to the models (all P < 0.05; Table 3; models 2 and 3). Suppressed RMR and TEE were not significantly related to future increases in BMI over the 6-mo follow-up (Table 3; models 2 and 3). The medium predictive effect of weight suppression on future increases in BMI (semipartial R = 0.09) was not reduced by adding each of the physiologic measures to the models (semipartial R for all = 0.09; Table 3). Weight suppression did not predict increases in bulimic symptoms at 6 mo of follow-up; consequently, further analyses including suppressed RMR and TEE were not performed.

DISCUSSION

As hypothesized, the degree of weight suppression predicted future increases in BMI, which supports previous work in clinical samples (3–5) and in one nonclinical sample (2). Thus, the current finding extends the previous literature in that a community sample was studied. The replication of the predictive effect suggests that this is a robust relation. Weight suppression did not predict future increases in bulimic symptoms, which is in line with findings from one past study (7), but not another (6). The fact that the average weight suppression (3.4 kg) was smaller in the current study than in the latter study (5.3 kg) may explain why we did not replicate the earlier finding.

Weight suppression showed moderate significant inverse correlations with suppressed RMR and TEE. These results provide support for the hypothesis that weight suppression would be related to reduced RMR and TEE (8. 9). However, the predictive relation of weight suppression to future increases in BMI was not attenuated after control for suppression in RMR or TEE. Thus, the current findings suggest that reduced RMR does not play a substantively important role in driving the predictive effect that weight suppression has on future increases in BMI. Likewise, depressed energy expenditure does not appear to drive this predictive effect. This was the first study, to our knowledge, to test whether weight suppression correlated with objective and accurate measures of RMR and TEE. The use of an objective measure of caloric expenditure circumvents the potential of biases associated with self-reported measures. In this context it should be noted that with a sample size of 91, we had power of 0.83 to detect a medium effect (32), which suggested that we had adequate sensitivity to detect meaningful effects.

The 6-mo weight change of a hypothetical subject with the average level of weight suppression was 1.9 kg greater than that with no weight suppression. Assuming that this was adipose tissue with an energy density of 7800 kcal/kg, the average daily energy imbalance for that weight gain would have been 81 kcal/d. This is not very different from the degree of TEE suppression that was observed for this cohort (−119 ± 245 kcal/d). Suppression of TEE, however, did not explain any of the variance in the 6-mo change in BMI (r² = 0.003, NS). We also measured weight change during the 2-wk DLW period and again, assuming that the weight change was adipose tissue, energy intake could be estimated by adding change in energy stores to TEE. In doing this, we calculated and estimated total energy intake (TEI) at baseline, but the predictive effect of weight suppression to increases in BMI remained significant when TEI was added to the model (P < 0.05).

It is important to consider the limitations of the current study. First, the sample only included young women in their first year of college, so results should be generalized with caution to other demographic groups. Second, the range of weight suppression in this relatively young sample was somewhat smaller than observed in other studies (2, 5, 8, 33), which might have attenuated effects. The average weight suppression score in the current study was 3.4 kg, which was similar to values reported in the studies that involved community samples (2.7–5.3 kg) but lower than those observed in studies that involved clinical samples (7.1–12.0 kg) (2, 4–6, 8). Third, the use of self-reported retrospective highest past weight may have introduced error. Using this method, 16 subjects in the current sample reported that their highest previous weights were less than or equal to their objectively measured current weight. However, previous studies of weight suppression have likewise used self-report measures for highest previous weight, and the percentage of subjects who had actual weights equal to or higher than their reported highest-ever weights has been reported to be as high as 48% (33). This may be unavoidable given the infeasibility of objectively weighing individuals at the point when individuals were at their highest weight. Last, we acknowledge that in the current study that most of the variance in increases in BMI at follow-up was mainly a function of baseline BMI. However, this was expected given that BMI shows high rank stability over time and baseline BMI is a strong predictor of future BMI scores over relatively short follow-up periods.

These results add to a growing literature concerning the predictive significance of weight suppression with relation to future weight gain. In addition, the negative relation of weight suppression to suppressed RMR and TEE are novel contributions to the literature. However, because the current findings provided little support for the hypothesis that suppressed RMR and TEE are the mechanisms underlying the predictive effect of weight suppression on subsequent increases in BMI, future research should investigate other objective measures that may be driving this predictive relation.