Greater Than Predicted Decrease in Resting Energy Expenditure and Weight Loss: Results From a Systematic Review

Authors


(edoucet@uottawa.ca)

Abstract

Changes in resting energy expenditure (EE) during weight loss are said to be greater than what can be expected from changes of body mass, i.e., fat mass (FM) and fat-free mass (FFM) but controversy persists. The primary focus of this study was to investigate whether there is a greater than predicted decrease in resting EE during weight loss in a large sample size through a systematic review. The study data were weighted and a partial residual plot followed by a multiple regression analysis was performed to determine whether FM and FFM can predict the changes of resting EE after weight loss. Another subgroup of studies from which all necessary information was available was analyzed and compared against the Harris—Benedict (HB) prediction equation to determine whether the changes in resting EE were greater than what was expected. Subjects lost 9.4 ± 5.5 kg (P < 0.01) with a mean resting EE decline of 126.4 ± 78.1 kcal/day (P < 0.01). Changes in FM and FFM explained 76.5% and 79.3% of the variance seen in absolute resting EE at baseline and post-weight loss, respectively (P < 0.01). Analysis of the 1,450 subject subgroup indicated an ∼29.1% greater than predicted decrease in resting EE when compared to the HB prediction equation (P < 0.01). This analysis does not support the notion of a greater than predicted decrease in resting EE after weight loss.

The physiological changes that take place during weight loss may work in concert in order to re-establish depleted energy stores. One such change may be the sustained depression of resting EE (1). Weight loss may be accompanied by greater than expected decreases in EE during weight loss (2) which are sustained over time (3). Although controversy persists (4), the changes in resting EE during weight loss may be explained by factors beyond what would be expected given the concomitant changes in body composition, i.e., fat mass (FM) and fat-free mass (FFM).

The purpose of this study was to establish whether a greater than predicted change in resting EE exists as a consequence of weight loss. This study draws from the same pool of subjects as our previously published systematic review (5) and uses peer reviewed weight loss literature in a large cohort of adults over the last 20 years. To our knowledge, this is the largest study sample to investigate whether there are greater than predicted changes in resting EE after weight loss. The first objective was to study to what degree the variance in resting EE upon weight loss was associated with changes in FM and FFM and whether FM and FFM alone could predict changes in resting EE. The second objective was to compare actual changes in resting EE to those obtained with the Harris—Benedict (HB) equation (6). Based on previous results, we hypothesize that the depression of resting EE will be greater than what can be predicted from the changes in body composition alone.

Methods

For the first objective, data from 2,977 subjects that were included in the analysis of our previously published systematic review was used (5). The selection of papers in the previous review was carried out systematically through specific collection criteria. In order to be included in the study, the publications had: (i) To include specific information on the weight loss interventions; (ii) To be performed on overweight or obese adults who were otherwise healthy, except in the case of surgical interventions where individuals were only considered candidates for some of the procedures if they had comorbidities such as diabetes and blood pressure, and (iii) To have values of resting EE or resting metabolic rate or basal metabolic rate or sleeping metabolic rate and body weight before and after the intervention. For publications dealing with more than one study group, all those groups that fit the inclusion criteria within that study were included and treated as individual sets of data. Of the 2,977 subjects, a total of 815 were included (714 females and 101 males) from 35 study groups based on information that provided the sex, change in body weight, change in FM, change in FFM, and change in resting EE. The data was then used to establish a relationship between the changes in body composition and changes in resting EE.

For the second objective, data was again selected from the 2,977 subject of the original systematic review and excluded if the information required for the HB equation. In fact, reasons for excluding studies were missing information on the sex of subjects (n = 1,002), on height (n = 183), on age (n = 128), on age and height (n = 26), on age, height, or weight for separate groups or subjects who completed the study (n = 149) and changes in weight were presented without pre- and post-weight loss values (n = 39). As a result, 1,527 subjects from 45 studies were not included in the analysis because they were missing the necessary information required for the HB equation. As such, of the 2,977 subjects, 1,450 (1,313 females and 137 males) from 71 studies had all the necessary information for the HB equation analysis related to the second objective.

Statistical analysis

We used the method of weighted least-squares for all of the regressions, where the size of the study was the weight in order to compensate for heteroscedasticity. To investigate the marginal significance for each predictor, we constructed a partial residual plot of resting EE against each of the two predictors (FM and FFM) while adjusting for the effects of the other predictor. We performed the Brown—Forsythe test to verify that the dispersions of the weighted residuals are similar for both groups. We then performed a regression analysis of resting EE against both FM and FFM both at baseline and post-weight loss. To compare the preweight loss parameter estimates to the post-weight loss estimates, we computed the Mahalanobis distance between the vector of the parameter estimates preweight loss compared to post-weight loss. We then computed the prediction error by using the preweight loss parameter estimates from the baseline model.

For the second objective, paired t-tests were performed to determine differences between actual and predicted values in resting EE before and after weight loss. All data was weighted to reflect the source study sample size and statistical analyses were performed using Statistical Product and Service Solutions software, version 17.0 (SPSS, Chicago, IL). Effects were considered significant at P < 0.05 and data are presented as mean ± SD.

Results

Subject group characteristics along with the studies from which the data was taken from are presented in Supplementary Table S1. On average, the 815 subjects used for the first objective looking at the changes of FFM and FM lost 9.4 ± 5.5 kg (P < 0.01) of body weight, 7.1 ± 4.5 kg of which was FM and 2.1 ± 1.8 kg was FFM (P < 0.01). The mean decline in resting EE was 126.4 ± 78.1 kcal/day (P < 0.01). The results of the partial residual plot (Figure 1) indicate that while adjusting for FM, the regression associated to the partial residual plot of resting EE against FFM gave a coefficient of determination of 51.4% (P < 0.01) for the preweight loss data and it is 63.4% (P < 0.01) for the post-weight loss. When adjusting for FFM, the regression associated to the partial residual plot of resting EE against FM gave a coefficient of determination of 54.8% (P < 0.01) for the preweight loss data and it is 57.4% (P < 0.01) in the case of post-weight loss. Therefore, in the case of preweight loss and of post-weight loss, FM and FFM are significant predictors of resting EE.

Figure 1.

Partial residual plot of resting energy expenditure (EE) against the predictors fat mass (FM) and fat-free mass (FFM) while adjusting for the effects of the other predictor. (a) The regression associated to the partial residual plot of resting EE against FFM while adjusting for FM gave a coefficient of determination of 51.4% (P < 0.01) for the preweight loss data and 63.4% (P < 0.01) post-weight loss. (b) When adjusting for FFM, the regression associated to the partial residual plot of resting EE against FM gave a coefficient of determination of 54.8% (P < 0.01) for the preweight loss data and it is 57.4% (P < 0.01) in the case of post-weight loss. The regression indicates that in both pre-weight loss and post-weight loss, FM and FFM are significant predictors of resting EE and marginally explain a large proportion of the variability in the REE (N = 815) (P < 0.01). Note: Solid line is baseline and dashed line is post-weight loss.

When resting EE was regressed against both FM and FFM, the coefficient of determination was 76.5% (P < 0.01) for the preweight loss studies, while it was 79.3% (P < 0.01) in the post-weight loss case. Finally, in the regression analysis the lower and upper limits for a 95% confidence interval of the mean prediction error are −1,820.254 and 1,837.5892. The prediction error on average is zero. Therefore, a preweight loss model of resting EE against FM and FFM could be used to predict post-weight loss resting EE while using the post-weight loss FM and FFM.

For the second objective comparing the HB equation to actual changes, 1,450 subjects were further analyzed in order to provide a comparison of the actual and predicted changes in resting EE before and after various weight loss interventions. Differences were noted between actual and predicted resting EE at both baseline (1,695.7 ± 247.6 kcal/day vs. 1,708.9 ± 191.9 kcal/day, respectively, P = 0.01) and postintervention (1,539.1 ± 197.1 kcal/day vs. 1,598.0 ± 171.9 kcal/day, respectively, P < 0.01) but the differences postintervention were significantly greater than preintervention (pre: 13.2 ± 149.4 kcal/day vs. post: 58.8 ± 134.6 kcal/day, P < 0.01). The changes in resting EE during weight loss were greater than what could be predicted with the HB equation (−156.5 ± 99.4 vs. −110.9 ± 75 kcal/day, P < 0.01), such that every kg of weight loss (relative decrease), was associated with a significantly greater actual decrease in resting EE (−18.3 ± 14.7 kcal/kg weight loss) than was predicted (−10 ± 1.2 kcal/kg weight loss) (P < 0.01).

Discussion

This study does not support a greater than predicted decrease in resting EE when post-weight loss resting EE is predicted with FM and FFM. It is often assumed that the reduction in resting EE is proportionate to changes in body weight, and in particular, the lean and fat tissue compartments. This is presumably because of the strong relationship between FFM and 24 h EE in a state of weight stability (7). In fact, we also show a strong association between FFM and resting EE in our study population. This study has shown that under typical weight loss conditions, changes in body composition may be strong contributors to the depression of resting EE during weight loss, which is in agreement with other previous work (8,9). In this study resting EE was regressed against both FM and FFM and the coefficient of determination were 76.5% (P < 0.01) for the preweight loss, whereas it was 79.3% (P < 0.01) in the post-weight loss. A recent review posits that the attempts to sustain weight loss are met with an adaptive phenotype which contains multiple systems that regulate EE (10). Some of the EE regulatory factors that are implicated in the greater than predicted reduction in resting EE include sympathetic nervous system-mediated changes (11), thyroid function, and catecholamine excretion (12) and changes in leptinemia during weight loss (13). According to our data, this study does not support such a conclusion.

The actual decrease in resting EE (∼−157 kcal/day) was significantly greater than what was predicted using the HB equation (∼−111 kcal/day) or a 30% difference, approximately. As a result, using the HB equation after weight loss may overestimate energy needs. Previous work has indicated that reductions in resting EE persist well beyond weight reduction (3). As such, current formulas such as the HB equation may not provide an accurate estimate of resting EE after weight loss has occurred since the formula fails to show metabolic reduction (14). Accessible formulas which predict the necessary changes to maintain weight loss are in fact available (15) and it may be more useful to utilize models such as those proposed by Hall et al. which have accounted for greater than predicted changes in EE under conditions of both starvation/refeeding (16).

One of the limitations of this article is the fact that the degree of energy imbalance differed at the time when the resting energy expenditure (EE) measurement was performed post-weight loss amongst studies included in the analyses. It would indeed be reasonable to assume that metabolic adaptations would be commensurate to the degree of energy imbalance. With respect to our data, 1,156 (1,108 female and 48 male) subjects from 26 studies were recorded during the dynamic phase of weight loss. In contrast, 294 subjects (205 female and 89 male) from 9 studies in this analysis were relatively weight stable at the time when the post-weight loss resting EE measurement was performed. A comparison of the relative decrease in resting EE during weight loss revealed no significant difference between groups who were still in a dynamic phase of weight loss and those who were relatively weight stable (−18.5 ± 15.4 vs. −17.3 ± 11.9 kcal/kg of weight loss, respectively, not significant). Even if we cannot conclude that the degree of energy imbalance does not influence metabolic adaptations, our data would seem to indicate that it had little impact on our conclusions.

Our results show that body weight reduction is not associated with a greater than predicted decrease in resting EE when post-weight loss values of FM and FFM are used to predict resting EE in a large cohort using different weight loss interventions. Finally, our results tend to indicate that prudence should be used when applying prediction equations established in weight stable humans that rely on body weight as they may incorrectly inflate post-weight loss resting EE.

SUPPLEMENTARY MATERIAL

Supplementary material is linked to the online version of the paper at http:www.nature.comoby

ACKNOWLEDGMENTS

E.D. is a recipient of a CIHR/Merck-Frosst New Investigator Award, CFI/OIT New Opportunities Award and of an Early Researcher Award.

DISCLOSURE

The authors declared no conflict of interest.

Ancillary