Visceral adipose tissue (VAT) is associated with greater obesity-related metabolic disturbance. Many studies have reported preferential loss of VAT with weight loss.
This systematic review looks for factors associated with preferential loss of VAT relative to subcutaneous abdominal fat (SAT) during weight loss.
Medline and Embase were searched for imaging-based measurements of VAT and subcutaneous abdominal adipose tissue (SAT) before and after weight loss interventions. We examine for factors that influences the percentage change in VAT versus SAT (%δV/%δS) with weight loss. Linear regression analyses were performed on the complete data set and on subgroups of studies. Factors examined included percentage weight loss, degree of caloric restriction, exercise, initial body mass index (BMI), gender, time of follow-up and baseline VAT/SAT.
There were 61 studies with a total of 98 cohort time points extracted. Percentage weight loss was the only variable that influenced %δV/%δS (r=−0.29, P=0.005). Modest weight loss generated preferential loss of VAT, but with greater weight loss this effect was attenuated. The method of weight loss was not an influence with one exception. Very-low-calorie diets (VLCDs) provided exceptional short-term (<4 weeks) preferential VAT loss. But this effect was lost by 12–14 weeks.
Visceral adipose tissue is lost preferentially with modest weight loss, but the effect is attenuated with greater weight loss. Acute caloric restriction, using VLCD, produces early preferential loss of VAT. These observations may help to explain the metabolic benefits of modest weight loss.
Obesity is becoming a major global health hazard owing to the increasing availability of calorie-dense food and the prevalence of sedentary lifestyles. For susceptible humans, the resulting excess adiposity is highly pathogenic, resulting in diabetes, dyslipidemia, hypertension, cardiovascular disease, arthritis and a variety of cancers. A number of dietary, behavioral, pharmaceutical and surgical interventions are available to reduce adipose tissue and ameliorate related disease. However, these vary in their effects on body weight and obesity-related disease.
Evidence that visceral adipose tissue (VAT) is more pathogenic than subcutaneous abdominal adipose tissue (SAT) in humans is rapidly emerging. In obese subjects, basal free fatty acid (FFA) flux,1 lipolysis rates and secretory protein expression are markedly higher in visceral compared with subcutaneous adipose cells.2 Additionally, FFA excretion from excess VAT is insensitive to insulin and highly sensitive to catecholamine stress signals in obese humans.3 Owing to the proximity and portal access of VAT to the liver, these FFAs are thought to promote hepatic insulin resistance in obesity.4, 5 These and other data provide strong evidence that excess VAT is a key factor in the dyslipidemia, insulin resistance and inflammation associated with obesity.6 Hence, a prominent biological role for visceral fat in the regulation of energy metabolism is becoming apparent.
From a clinical perspective, preferential loss of VAT during initial weight loss may be metabolically advantageous. However, despite the variety of weight loss methods, and competing claims of intervention-induced preferential loss of visceral fat, it remains unclear whether relative rates of change in VAT and SAT can be manipulated at all. Smith and Zachwieja7 systematically reviewed the literature on the subject in 1999 and suggested that future therapies may specifically target visceral fat loss. Since then the pool of published imaging-based measurements of changes in subcutaneous and visceral abdominal fat during weight loss has grown considerably, warranting a more detailed review of this area.
The primary aim of this review was to compare weight loss interventions in terms of the resulting ratio of percentage change in VAT versus SAT (%δV/%δS). Broadly, it was hypothesized that the degree of caloric restriction and exercise may influence %δV/%δS in obese and overweight subjects. Additionally, we have tested the hypotheses that gender, initial body mass index (BMI), percentage weight loss (%WL), time of follow-up and the baseline ratio of VAT and SAT (initial VAT/SAT) are predictive of %δV/%δS. It is suggested that the greater transience of visceral fat is indicative of its roles as a short-term energy store and a regulator of energy metabolism.
Medline and Embase were searched up to May 2006 for studies that report significant changes in subcutaneous and visceral abdominal fat after weight loss interventions. The details of this formal search are shown in Table 1.
All abstracts from the final search set were screened and studies that fulfilled the following criteria were retrieved.
Changes in subcutaneous and visceral abdominal fat were estimated from single or multislice magnetic resonance imaging or computed tomography measurements.
Subjects were overweight or obese at baseline.
All changes in VAT and SAT were significant.
Weight loss was reported.
Time to follow-up was reported.
Published in English language journal.
Only studies of currently available weight loss pharmaceuticals were included.
The reference lists of eligible studies were checked for additional publications not indexed electronically and the International Journal of Obesity and Obesity Research were hand-searched covering the period January 1990–May 2006.
Changes in visceral and subcutaneous abdominal fat area or volume were recorded and the ratio of percentage change in VAT to SAT was calculated for each study. Calculation of percentage changes in VAT and SAT allowed inclusion of data reported as changes in area or volume of adipose tissues. Initial visceral and subcutaneous abdominal fat areas or volumes were recorded and initial VAT/SAT was calculated. Initial BMI, weight loss, duration of intervention, number of patients, gender, and reported details of the intervention and imaging techniques were recorded and tabulated for each time point in each study (Tables 2, 3 and 4). Where data was published in multiple studies the latest or most comprehensive one was included. Data were analyzed using SPSS statistical software in a fixed effect model, without weighting by the number of subjects (n). All multiple regression analysis was carried out on loge (natural log=ln) transformed %δV/%δS to ensure linearity of data.
Assessment of heterogeneity
We noted that the reporting, type and intensity of exercise and the composition of diets varies considerably among studies that were pooled in this analysis. We have broadly classified very-low-energy/calorie diets (VLCDs) as those providing less than 800 kcal day−1, low-calorie diets (LCDs) as any degree of caloric restriction that causes weight loss and exercise was noted only when rigorously supervised and reported. Only computed tomography and magnetic resonance imaging-based measurements of abdominal fat distribution were included, however, the choice of lumbar inter-vertebral slice varied between studies.
The distribution of log %δV/%δS among data points was assessed. Outliers were identified and excluded using a box plot and percentiles of ln %δV/%δS were calculated from the remaining normally distributed data. Linear regression was used to assess the confounding influence of the choice of instrument and position of imaging slice. Interventional subgroups (including outliers) were compared in terms of the frequency of data points in either the 95th or 5th percentiles of log %δV/%δS using the χ2-test for heterogeneity.
Analysis of diet and exercise interventions
After excluding outliers, 82 normally distributed data points from studies of diet and exercise were ln transformed and entered into a multiple linear regression model. Caloric restriction (0 for LCD or 1 for VLCD), exercise (0 or 1), gender, initial BMI, initial VAT/SAT, time of follow-up (weeks) and %WL were assessed for their association with ln %δV/%δS.
Analysis of studies with follow-up between 12 and 14 weeks
This subset of studies was analyzed separately to control for the effect of time of follow-up. All studies with follow-up of 12–14 weeks (n=44) were entered into a multiple linear regression model. Caloric restriction (0 for LCD, 1 for VLCD), exercise (0 or 1), gender, initial BMI, initial VAT/SAT and %WL were assessed for their association with ln %δV/%δS.
Analysis of studies of VLCDs
The ln %δV/%δS was normally distributed among all the studies of VLCDs. The data from all studies of VLCDs (13 cohort time points) were entered into a multiple linear regression model. Gender, exercise, initial BMI, %WL, initial VAT/SAT and time of follow-up (weeks) were assessed for their association with ln %δV/%δS after VLCDs.
Medline identified 1126 studies and Embase identified 1617 studies using the search strategy in Table 1. Of these, 61 studies with a total of 98 treatment cohort time points (Tables 2, 3 and 4; Figure 1) and a total of 2153 measurements of changes in SAT and VAT were eligible according to the selection criteria. These studies provided 40 data points of subjects treated with LCD, 20 of LCD+exercise, 13 of VLCDs, eight of LCD+pharmaceuticals and six cohort time points of subjects treated with the surgical procedure laparoscopic adjustable gastric banding (LAGB). There were no reports following other bariatric surgical procedures.
Assessment of heterogeneity
The distribution of ln %δV/%δS among all data points was not normal owing to three outliers that were identified using a box plot. After excluding these three data points the median (interquartile range) %δV/%δS in the remaining 95 cohort time points was 1.405 (0.51). The 95th percentile was above 2.41 and the 5th percentile was below 0.92 (Figures 1a and b). Using linear regression analysis it was shown that no significant systematic variance was added to ln %δV/%δS by the choice of instrument or the position of the imaging slice (lumbar 2–3, 3–4, 4–5 or umbilicus) among these 95 cohort time points.
In interventional subgroup analysis, the frequency of data points (including outliers) in the 95th or 5th percentiles was not significantly greater than expected (1 in 10) among studies of LCD, LCD+exercise, LCD+sibutramine, LCD+orlistat or exercise alone. In contrast, the frequency of data in the 95th percentile of ln %δV/%δS was significantly greater than expected among studies of VLCDs (χ2=30.644, df=1, P<0.001; Figure 1). While two of six LAGB cohort time points gave %δV/%δS less than 0.92 the frequency of data in the 5th percentile was not significantly greater than expected in studies of LAGB.
Predictors of ln %δV/%δS in studies of diet and exercise
Using multiple linear regression analysis of 82 normally distributed data points, we found two independent factors that influenced ln %δV/%δS. First, greater %WL was significantly negatively associated with ln %δV/%δS (r2=0.105, P=0.004; Figure 2), and second, the greater caloric restriction with VLCDs was associated with higher ln %δV/%δS (r2=0.055, P=0.034). Together, %WL and the degree of caloric restriction explained 22.2% of the variance in ln %δV/%δS (r2=0.222, P<0.001). No other variables were predictive of ln %δV/%δS among these studies. The combined effect can be seen in Figure 2 where VLCD studies are represented by filled black circles.
Predictors of ln %δV/%δS in data points taken at 12–14 weeks of follow-up
To correct for potentially confounding influence of time of follow-up, a subgroup of studies with follow up between 12 and 14 weeks were analyzed. Fortunately, this subgroup contained 46 data points that were normally distributed for ln %δV/%δS. In this regression analysis, we found that %WL was exclusively negatively correlated with ln %δV/%δS (r2=0.266, P<0.001; Figure 3), while no other variables provided any additional explanation of variance. The early preferential loss of visceral fat with the VLCDs appears to be lost at 12–14 weeks (Figure 3).
Predictors of ln %δV/%δS in studies of VLCDs
As VLCDs seemed to provide greater early loss of visceral fat, they were analyzed separately. The ln %δV/%δS values were normally distributed among all studies of VLCDs. In 13 data points the time of follow-up was exclusively negatively associated with ln %δV/%δS (r2=0.845, P<0.001) (Figure 4a). Percentage weight loss (r2=0.654, P=0.001) (Figure 4b) was strongly negatively associated with ln %δV/%δS, but added no additional explanation of variance to that of time of follow-up. That is, time of follow-up and degree of weight loss were highly dependent on each other among these studies of VLCD. No associations were found between ln %δV/%δS and initial BMI, gender, exercise or initial VAT/SAT after VLCD.
The distribution of abdominal fat loss in studies of sibutramine, orlistat and LAGB
Only two studies of LCD+orlistat were eligible for review. Both reported %δV/%δS within the expected range; 1.75 and 1.93 with weight loss of 10.1 and 7.0 kg over 26 and 21 weeks, respectively (Table 4). Three studies with a total of six time points in which sibutramine was used with LCD show %δV/%δS within the expected range, and the study by Kim and co-workers shows decreasing %δV/%δS with weight loss (Table 4). Four studies of LAGB with six time points were retrieved. The range of %δV/%δS in these cohort time points was 0.19–1.71. Two studies of LAGB showed %δV/%δS in the 5th percentile.
Visceral fat is increasingly seen as a driver of metabolic disturbance in obesity.66 It has been hypothesized that weight loss interventions that target visceral fat preferentially may ameliorate obesity related comorbidity without the need for substantial weight loss. However, the principal finding of this review is that there is no compelling evidence of a weight loss intervention that targets visceral fat preferentially. Rather, it appears that preferential loss of visceral fat is associated with modest weight loss and the effect reduced with greater weight loss (Figures 2, 3 and 4).
In making comparisons of weight loss interventions, we have identified a number of limitations in the literature. Randomized controlled trials22, 39, 67 are uncommon and very few studies have follow-up earlier than 8 weeks (Tables 2, 3 and 4). No information has been published regarding diversionary bariatric surgery, and there are very few studies of weight loss pharmaceuticals and the surgical procedure LAGB (Table 4).
A number of studies show that exercise reduces total abdominal fat as measured using anthropometry or imaging techniques,68 and exercise alone reportedly reduces visceral obesity without significant weight loss.69 However, we have been unable to identify an effect of exercise on %δV/%δS during dietary weight loss. While variance in exercise type, intensity and reporting between studies may have confounded this effect, no effect of exercise was seen in three randomized trials comparing aerobic, resistance and no exercise during caloric restriction.22, 39, 67 Perhaps the effect of weight loss is of greater influence on %δV/%δS than that of any additional exercise. Notably, this review was restricted to changes in VAT and SAT and makes no comparison of abdominal versus peripheral fat loss.
Studies of VLCDs with follow-up during the first 3 weeks were a significant source of heterogeneity in this review with higher ln %δV/%δS related to rapid weight loss in a short period of time (Table 3; Figure 1). Unfortunately, only one study of LCD and one of LCD+sibutramine provide data at 4 weeks for comparison.16, 59 These studies show only mild %δV/%δS (1.28 and 1.39) at this time point, suggesting that early elevated %δV/%δS is associated with VLCDs. Interestingly, among studies in which subjects were followed up after 12–14 weeks, VLCDs had caused greater weight loss and lower %δV/%δS than other studies at this time point (Figure 3). Thus rapid weight loss demonstrated with VLCD studies shows a very early but unsustained exceptional preferential loss of visceral fat (Figures 4a and b). Longer follow-up, however, demonstrates that %WL was exclusively correlated with ln %δV/%δS, suggesting that the degree of weight loss and not time of follow-up is predictive of percentage change in visceral versus subcutaneous abdominal fat.
It is clear from these regression analyses, and in studies with multiple time points59, 64, 65 that preferential loss of visceral fat is greatest during initial modest weight loss (Figures 1b, 3 and 4). Furthermore, extrapolation of the regression lines in Figures 3 and 4 leads to speculation that measurements of abdominal fat loss made after about 20% weight loss may start to show preferential loss of SAT or retention of VAT. While the point at which percentage reductions in VAT and SAT reach equilibrium remains undefined, these data suggest that it is almost entirely related to %WL and is not related to method of weight loss, initial BMI or VAT/SAT ratio. Major sustained weight loss following surgically procedures may provide some information regarding the distribution of abdominal fat loss after 25–40% weight loss and indicate if, with extreme weight loss, there is any preferential retention of visceral fat. The limited data we currently have from surgical weight loss studies are inconsistent and serial measures following long term sustained weight loss are recommended.62, 63, 64, 65
The metabolic disturbances associated with visceral obesity are reflected in numerous clinical observations. In the ‘Portal hypothesis’ it is suggested that the proximity of VAT lipolysis, hormone and cytokine production upstream from the liver exacerbates hepatic insulin resistance and increases glucose output.70 VAT appears to further undermine the efficiency of energy metabolism by inducing systemic insulin resistance and influencing a variety of inflammatory pathways.6, 71
Increasing evidence of the biological role of VAT adds greatly to the strength of the conclusion that %δV/%δS is greatest during rapid weight loss with VLCDs or modest weight loss. Many studies have shown preferential improvement in hormonal and metabolic function occurs soon after modest weight loss and/or exercise,60, 72 indicating a tight association of obesity comorbidity on neutral or positive energy balance. Perhaps early preferential loss of visceral fat is physiological, driven by an immediate need for energy, in the face of acute negative energy intake, from a metabolically active fat store uniquely situated up stream from the liver. The early restoration of efficient energy metabolism upon negative energy balance may also be important.
Conversely, under circumstances of over nutrition the hormonal, cytokine and lipid output of VAT may be protective against excessive peripheral adiposity and weight gain. Interestingly, the pharmaceutical administration of thiozoladinedione insulin sensitizers to obese patients with type-2 diabetes is associated with weight gain.73 Thus, the correction of metabolic abnormalities may be associated with weight gain. In support of this is a prospective study showing that insulin resistant Pima Indians gain less weight, over several years, than those without insulin resistance.74
Several other lines of evidence support a physiological role for visceral fat. VAT may provide FFAs for energy during bouts of stress, acute caloric restriction or periods of intense physical activity. Exercise counteracts the insulin insensitivity associated with VAT,75 exerts controls over VAT lipolysis through adrenergic signaling3 and stimulates systemic FFA uptake and oxidation.75 Moreover, VAT lipolysis has been shown to be sensitive to glucocorticoid levels76 and the parasympathetic nervous system,77 indicating the degree to which VAT is integrated in systemic energy utilization and stress signaling. These metabolic nuances of VAT may be particularly advantageous for hunting carnivores and may have evolved in such species. Indeed, the hormonal and metabolic differences between VAT and SAT are not seen in grazing mammals such as rodents, but they are in carnivorous humans, dogs and primates,78, 79 and are more marked in men than women.80
Preferential loss of VAT compared with SAT is greatest with modest weight loss and the effect is attenuated, and possibly lost completely with increasing weight loss. VLCDs inducing rapid weight loss in a short period of time produce the greatest preferential loss of visceral fat, but the effect is transient. Preferential loss of VAT is reduced and perhaps negated altogether with greater weight loss and is not related to the method of weight loss. Exercise does not appear to have an effect beyond that related to the weight loss achieved.
The rapid VAT loss with the acute caloric restriction of VLCDs suggests that VAT has a physiological role to provide energy at times of acute negative energy balance, and the sustained preferential loss of VAT with modest weight loss may add to our understanding of how modest weight loss appears to provide significant metabolic and clinical benefits.
Timothy B Chaston—Contribution included design of the review, extraction of data, analysis of the data and writing the manuscript.
John B Dixon—Contribution included design of the review, assistance with the extraction of data, analysis of data and writing of the manuscript.
This study was funded by Monash University and there are no conflicts of interest to declare.
About this article
Nutrition & Metabolism (2017)