Visceral Adipose Tissue and Metabolic Complications of Obesity Are Reduced in Prader-Willi Syndrome Female Adults: Evidence for Novel Influences on Body Fat Distribution

Visceral obesity is detrimental to health, but the mechanisms controlling body fat distribution are not fully understood. In premenopausal adult females (30 nonobese, 14 obese [body mass index> 30kg/m²]), variance in fasting insulin, glucose, insulin/glucose ratio, C-peptide/insulin ratio, triglycerides, and high-density lipoprotein/low-density lipoprotein-cholesterol ratio, were independently influenced by visceral but not total sc or abdominal sc adipose tissue, as measured by whole-body magnetic resonance imaging. Adult females with Prader-Willi syndrome (n = 13) had significantly reduced visceral adiposity, compared with obese controls (visceral/total sc adipose tissue ratio: 0.067 ± 0.017 vs. 0.108 ± 0.021), independent of their total adiposity (P < 0.001), or use of exogenous sex steroids. This is in contrast to that expected by their physical inactivity, hypogonadism, adult GH deficiency, and psychiatric problems. Females with Prader-Willi syndrome not receiving sex steroids (n = 8) had significantly reduced fasting insulin, insulin/glucose ratio, and triglycerides and increased C-peptide/insulin ratio, compared with obese controls, adjusting for total (P < 0.05) but not visceral adiposity (P = 0.3–0.6), supporting their association. The cause of the reduced visceral adiposity in Prader-Willi syndrome may reflect novel hormonal, hypothalamic, and/or genetic influences on body fat distribution.

INCREASED BODY FAT, and particularly visceral obesity, has adverse effects on health (1–3). The mechanisms controlling body fat distribution are not fully understood but are thought to include environmental, genetic, and hormonal influences, such as sex steroids, GH, and glucocorticoids (2, 3). A general finding has been that conditions associated with increased fat deposition, such as GH deficiency, hypercortisolism, hypogonadism, and hyperandrogenism in females, are associated with preferential deposition of central or visceral fat (2–4). This may primarily be owing to differences in the metabolic activity, receptor expression, and molecular characteristics of adipocytes in these sites (2–4). Central or visceral obesity is associated with increased insulin resistance, glucose intolerance, hyperlipidemia, cardiovascular disease, and mortality (1, 5).

Prader-Willi syndrome (PWS) is a genetic disorder characterized by hyperphagia and life-threatening obesity from childhood, mental retardation, behavioral problems, and short stature from GH deficiency and hypogonadism (6). Many of the features are thought to arise from developmental abnormalities in the hypothalamus (7). The relationship with the underlying chromosome 15q11–13 imprinted gene defect(s) is currently unclear (8).

A peculiar body composition has been identified in PWS, with a marked increase in fat relative to lean tissue (9, 10). In addition to the increased calorific intake, there may be a contribution from GH deficiency, hypogonadism, and reduced physical activity (11). Previous studies have suggested that some obesity-related metabolic complications, such as insulin resistance and reduced hepatic insulin extraction, may be less than expected for the degree of obesity in PWS (12–14). Body fat distribution in PWS has not previously been examined in detail. We therefore examined adipose tissue (AT) distribution, using whole-body magnetic resonance imaging (MRI) (15), in nonobese and obese healthy control and PWS female adults (16). We assessed the relationships between different regional AT depots and markers of glucose and lipid metabolism and examined differences in these obesity-related metabolic parameters in PWS, adjusting for such relationships. We demonstrate for the first time a selective reduction in visceral adiposity in PWS, which appears to protect against metabolic complications of obesity, despite their overall increased adiposity. Possible explanations for these findings are discussed in view of the other phenotypes, neuroendocrine and genetic abnormalities seen in PWS.

Materials and Methods

Subject characteristics

Ethical approval for the study was obtained from the Research Ethics Committee, Hammersmith Hospital (No. 92/3995 and 96/4807). Control female subjects (n = 44) were recruited from hospital staff, dietetic clinics, and public advertisement. PWS female adults (n = 13) were recruited through the Section of Developmental Psychiatry, University of Cambridge (A.H.), and the United Kingdom Prader-Willi Syndrome Association. Consent was obtained from both the PWS subject and caregiver or next of kin. All subjects were Caucasian, aged over 18 yr, and nondiabetic and had normal renal and hepatic function. All controls were premenopausal and had no known endocrine pathology. One PWS female, but none of the control subjects, was a smoker. There was no history of diabetes mellitus in any first-degree relative. All subjects reported stable body weight in the previous 2 months. All PWS subjects met diagnostic criteria for PWS (6) and suffered from childhood-onset obesity, requiring vigorous behavioral modification, such as locking food and dietary supervision, for extreme hyperphagia and obsession with food. All PWS subjects had experienced primary amenorrhea or severe oligomenorrhea. Four PWS subjects were receiving the oral contraceptive pill (OCP) (two subjects: ethinyloestradiol 20 μg/norethisterone 1 mg; one subject each: cyclical ethinyloestradiol 30 μg/levonorgesterol 0.15 mg, norethisterone 0.35 mg); and one PWS female was on hormone replacement therapy (HRT) (as cyclical estradiol valerate 2 mg/levonorgestrel 75μ g). No PWS subject was currently receiving GH therapy, but one PWS subject on the OCP, aged 22 yr, had received GH during childhood only. No control or PWS subjects were receiving any other pituitary hormone replacement, except for one PWS subject who was biochemically euthyroid on thyroxine replacement. Four PWS females were receiving selective serotonin reuptake inhibitors, and two of these were also on phenothiazines for management of behavioral and psychiatric problems.

Anthropometry and body composition

Height; weight; waist; hip; mean arm circumference; and, to assess SCAT distribution, four-site skinfold thickness (limb = triceps and biceps, trunk = subscapular and suprailiac) were measured in all subjects by one trained observer (A.E.B.). Controls were divided into nonobese (body mass index [BMI] < 30 kg/m²) and obese (BMI > 30kg/m²) groups.

Whole-body MRI

Whole-body MRI was acquired in all subjects on a 1.0T HPQ system (Marconi Medical, Cleveland, OH) using a rapid T1-weighted se sequence, as previously described (10, 16–18). Subjects were scanned from fingertips to toes with 10-mm-thick transverse images every 10–30 mm. Interactive computer analysis was used to quantify total body and regional AT volumes: total sc (SCAT), abdominal sc (ASCAT), visceral (VAT), and nonvisceral internal (INAT) adipose tissue, as previously described (17, 18). INAT consisted primarily of AT around muscle fibers. Central AT was defined as VAT + ASCAT. Peripheral AT was defined as total AT –central AT. MRI total fat mass and total and regional fat as percent of body mass were calculated as previously described (17, 18), using body weight to calculate fat-free mass (FFM).

Blood sampling and assays

Blood sampling was performed after a 12-h overnight fast. Plasma glucose was measured using an automated analyzer (RA-1000, Technicon Instrument Co. Ltd., Basingstoke, UK) and serum assayed for total cholesterol and triglycerides using an enzymatic method (Technicon Dax system, Tarrytown, NY); high-density lipoprotein (HDL) cholesterol using direct measurement with an RIAXT machine (Biostat Diagnostics, Cheshire, United Kingdom) and low-density lipoprotein (LDL) cholesterol was calculated using the Friedwald formula. Serum or plasma was assayed using established assays for insulin (IRMA, guinea pig anti-insulin antibody; SAPU, Glasgow, Scotland), C-peptide (enzyme immunometric assay, Immulite, DPC, Los Angeles, CA), total T, dehydroepiandrosterone sulfate (DHEAS) (St. Thomas’s extraction assay, Chelsea Kits, United Kingdom), and SHBG (enzyme tracer and chemiluminescent automated assay, Immulite, DPC). Serum E2 (enzyme tracer and chemiluminescent automated assay; IUX Abbott, Chicago, IL) was also measured in PWS subjects. Coefficient of variation values for all assays were < 10%. Free T index was calculated (100 × total T/SHBG). Fasting insulin resistance index (FIRI) was calculated (fasting insulin × fasting glucose/25) (19). The basal C-peptide/insulin molar ratio was used as a measurement of hepatic insulin extraction (13).

Statistical analysis

Between-group comparisons were made using one-way ANOVA with posthoc Tukey’s test. Pearson product moment correlation coefficients (r) were used to assess the relationship among variables. Log₁₀ transformation was used to normalize data that were not normally distributed. In control females, forward stepwise regression analysis was used to assess the relative contributions of age and each AT depot to the variance in metabolic and hormonal parameters. Examination of the effect of PWS on fat distribution and metabolic or hormonal parameters, correcting for differences in age and total or visceral adiposity, was performed using multiple linear regression analysis to calculate PWS regression coefficient (β). Significance was taken as P < 0.05. Statistical analysis was performed using SigmaStat 2.0 (Jandel Corp., San Rafael, CA) and Systat 8.0 (SPSS, Inc., Chicago, IL).

Results

Subject characteristics and body composition

Subject characteristics are given in Table 1. There were no significant differences between those PWS females not receiving or receiving the OCP/HRT, and PWS data were therefore combined. PWS subjects were significantly shorter than both nonobese and obese controls. PWS subjects had a similar BMI, total AT volume, and MRI percent body fat to obese controls but had a significant lower mean arm circumference and FFM. In view of the reduced height and FFM of PWS subjects, correction was made for both AT volume and fat as percent of body mass, when comparing metabolic parameters between PWS subjects and controls.

TABLE 1.

Subject characteristics, body composition, and fat distribution determined by whole-body MRI

	Nonobese	Obese	PWS
n	30	14	13
Age (yr)	31 ± 8	36 ± 10	27 ± 7d
Weight (kg)a	65.7 ± 8.3	100.9 ± 14.9c	82.3 ± 26.7b,fb,f
Height (m)	1.66 ± 0.06	1.66 ± 0.05	1.49 ± 0.08c,fc,f
BMI (kg/m2)a	23.8 ± 2.4	36.9 ± 5.6c	36.6 ± 9.9c
Waist (cm)a	75.0 ± 6.1	103.5 ± 9.5c	99.9 ± 17.9c
Hip (cm)a	98.9 ± 6.7	124.6 ± 12.9c	120.7 ± 18.7c
Waist/hip ratio	0.76 ± 0.04	0.83 ± 0.04c	0.83 ± 0.05c
Waist/height ratio	0.45 ± 0.04	0.63 ± 0.06c	0.67 ± 0.11c
Trunk/limb skinfold	1.05 ± 0.27	1.10 ± 0.16	1.37 ± 0.35c,dc,d
Mean arm circumference (cm)a	28.4 ± 2.8	38.0 ± 3.5c	33.3 ± 6.1c,ec,e
Total AT (liters)a	26.8 ± 6.3	59.3 ± 14.0c	55.7 ± 25.8c
Fat mass (kg)a	19.5 ± 4.6	42.7 ± 10.1c	40.1 ± 18.6c
Percent body fata	29.1 ± 4.3	41.9 ± 4.1c	46.9 ± 6.9c
FFM (kg)	46.3 ± 4.9	58.3 ± 6.4c	42.2 ± 9.6f
Fat mass/FFM ratioa	0.42 ± 0.09	0.73 ± 0.13c	0.92 ± 0.25c
VAT (liters)a	1.6 ± 0.6	5.4 ± 1.5c	3.2 ± 1.7c,fc,f
VAT (% total AT)	5.7 ± 1.7	9.1 ± 1.6c	5.7 ± 1.3f
Visceral fat (% body mass)	1.7 ± 0.6	3.8 ± 0.7c	2.7 ± 0.7c,fc,f
SCAT (liters)a	22.9 ± 5.3	49.8 ± 12.1c	48.6 ± 23.0c
SCAT (% total AT)	85.4 ± 2.4	83.9 ± 1.8	86.8 ± 2.5f
ASCAT (liters)a	5.5 ± 1.9	15.3 ± 4.7c	14.5 ± 7.2c
ASCAT (% total AT)	20.3 ± 3.5	25.5 ± 2.5c	25.9 ± 3.2c
INAT (liters)a	2.4 ± 0.7	4.1 ± 1.0c	3.9 ± 1.5c
INAT (% total AT)a	8.9 ± 1.7	7.0 ± 1.3c	7.4 ± 1.9b
Central/peripheral AT	0.357 ± 0.084	0.532 ± 0.085c	0.467 ± 0.088c
VAT/SCAT ratio	0.067 ± 0.021	0.108 ± 0.021c	0.067 ± 0.017f
VAT/ASCAT ratio	0.284 ± 0.073	0.358 ± 0.064c	0.223 ± 0.046b,fb,f

Measurements made in nonobese (BMI < 30 kg/m²) and obese (BMI > 30 kg/m²) control and PWS females. Figures are given as mean ± sd.

^a

ANOVA was performed on log₁₀-transformed data; ^bP < 0.05, ^{cP < 0.005 vs. nonobese control; d}P < 0.05, ^eP < 0.01, ^fP < 0.005 vs. obese control (ANOVA with post hoc Tukey’s test).

Body fat distribution

In controls, obesity was associated with a relative increase in visceral adiposity, as evident by significant increases in the VAT volume as percent of total AT volume, visceral fat as percent of total body mass, VAT/SCAT and VAT/ASCAT ratios (Table 1). The VAT/SCAT ratio was positively correlated with total AT volume (r =+ 0.58, P < 0.001) or percent body fat (r = +0.61, P < 0.001) in controls, but there was no significant additional effect of age in multiple linear regression (both P = 0.7). Correction for total adiposity, but not age, was therefore made when comparing ratios between PWS and controls. Obese controls also had significantly increased surrogate measurements of abdominal obesity, compared with nonobese controls, namely waist, waist/hip ratio (WHR), and waist/height ratio, but the trunk/limb skinfold ratio was not increased (Table 1).

There was a lower level of visceral adiposity in PWS, compared with obese controls, as evident by significantly lower absolute VAT volume, VAT as percent of total AT or visceral fat as percent total body mass (Table 1, Fig. 1). The proportion of VAT in PWS females was in fact similar to nonobese controls. In contrast, there was no significant decrease in the amount or proportion of SCAT, ASCAT, or INAT in PWS subjects, compared with obese controls (Table 1). The VAT/SCAT and VAT/ASCAT ratios were lower in PWS subjects, compared with obese controls, but there was no significant decrease in the central/peripheral AT ratio, waist diameter, WHR, waist/height ratio, or trunk/limb skinfold ratio (Table 1). This reduction in visceral adiposity was significant in both PWS females not receiving or receiving the OCP/HRT (VAT/SCAT ratio: 0.063 ± 0.017, P < 0.001 vs. obese controls, or 0.073± 0.016, P < 0.01; VAT/ASCAT ratio: 0.221 ± 0.056 or 0.226 ± 0.032, both P < 0.001).

Fig. 1.

Reduced visceral adipose tissue in PWS. T1-weighted MRI scan (1.0T HPQ system) at the level of the lower abdomen, in a control adult female (total AT volume 86.4 liters, MRI total body fat 50.5%, VAT (percent total AT) 9.3%, VAT:SCAT ratio 0.111, VAT:ASCAT ratio 0.326), and a PWS adult female (total AT volume 86.9 liters, MRI total body fat 54.4%, VAT (percent total AT) 4.4%, VAT:SCAT ratio 0.048, VAT:ASCAT ratio 0.208).

When adjusting for total adiposity, VAT, but not ASCAT, was significantly lower in PWS, compared with controls. The slope of the regression line for the relationship between VAT and total AT volume was significantly reduced in all PWS subjects, compared with controls (P < 0.001) (Fig. 2A), and this remained significant when correcting for age in multiple linear regression analysis (P < 0.001), although this was not a significant covariate (P = 0.6). At the highest levels of obesity, this resulted in around a 50% lower level of VAT in PWS (Fig. 1). By contrast there was no significant difference in the relationship between ASCAT and total AT volume between all PWS subjects and controls (slope P = 0.1, intercept P = 0.3) (Fig. 2B). Both the VAT/SCAT and VAT/ASCAT ratios were significantly lower in PWS subjects than obese controls, correcting for either total AT volume (PWS β [95% confidence interval]: −0.042 [−0.034, −0.049], −0.136 [−0.114, −0.158], both P < 0.001) or percent body fat (PWS β: −0.041[− 0.032, −0.049], −0.127 [−0.103, −0.151], both P < 0.001), and these findings were independent of OCP/HRT usage, selective serotonin reuptake inhibitor/phenothiazine usage, and height (data not shown).

Fig. 2.

Reduced visceral but not abdominal sc adipose tissue in PWS. Relationship between total AT volume and (A) visceral AT volume or (B) abdominal sc AT volume, in control (○) females, PWS females not on OCP/HRT (▪) and PWS females receiving OCP/HRT (+); r indicates Pearson correlation coefficient. Linear regression line is solid for control and dashed for PWS females (combined for both PWS groups). Note that PWS subjects have a reduction in visceral AT but not abdominal sc AT, relative to their total AT.

Metabolic parameters and body fat distribution in control females

In control females, fasting insulin, fasting glucose, insulin/glucose ratio, and triglycerides were positively correlated, and C-peptide/insulin, and HDL/LDL-cholesterol ratios were negatively correlated to the total body AT, independent of age (Tables 2² and 3³). In forward stepwise multiple linear regression analysis, similar correlations were found with VAT, but there was no independent additional relationship between any of these metabolic parameters and either SCAT or ASCAT (Table 3). Similar results were found using FIRI instead of insulin/glucose ratio (data not shown). Similar results were obtained when using fat depots as percent of body mass rather than AT volume in stepwise multiple linear regression (data not shown). There was no significant influence of height on any metabolic parameters in forward stepwise multiple regression analysis, independent of age and total or visceral AT volume (P^E 0.45 − 0.99).

TABLE 2.

Metabolic and hormonal parameters in female controls and PWS females

	Nonobese	Obese	PWS (no OCP/HRT)
n	30	14	8
Fasting insulin (pmol/liter)a	29.4 ± 16.4	114.1 ± 77.2d	83.2 ± 64.0d
Fasting glucose (mmol/liter)a	5.0 ± 0.5	5.2 ± 0.4	5.1 ± 0.3
Insulin/glucose ratioa	5.9 ± 3.3	21.8 ± 14.5d	16.6 ± 13.2d
FIRIa	0.80 ± 0.45	3.19 ± 2.20d	2.23 ± 1.67d
C-peptide (pmol/liter)	367 ± 195	509 ± 270	440 ± 243
C-peptide/insulin ratioa	17.8 ± 17.0	6.7 ± 3.5d	11.4 ± 5.8
Fasting triglycerides (mmol/liter)a	0.78 ± 0.29	1.77 ± 0.94d	0.98 ± 0.38e
Total cholesterol (mmol/liter)	4.6 ± 0.7	5.5 ± 1.0b	5.1 ± 1.1
HDL/LDL cholesterol ratio	0.63 ± 0.20	0.38 ± 0.16d	0.38 ± 0.15c
Total T (nmol/liter)	2.0 ± 0.5	1.8 ± 0.5	1.5 ± 0.5b
SHBG (nmol/liter)a	73 ± 44	43 ± 25b	39 ± 30c
Free T index	3.4 ± 1.8	5.2 ± 2.8b	5.0 ± 2.2
DHEAS (μmol/liter)	6.2 ± 2.7	4.1 ± 2.9b	4.2 ± 1.5

Measurements made in nonobese (BMI < 30 kg/m²) and obese (BMI > 30 kg/m²) control females (data as in Table 1) and PWS females not receiving the OCP or HRT (age 30± 7 yr, BMI 36.1 ± 10.5 kg/m², percent body fat 46.7 ± 7.1, total AT 54.6 ± 23.1, VAT 3.0 ± 1.4, VAT/SCAT ratio 0.063 ± 0.017). Figures are given as mean ± sd.

^a

ANOVA was performed on log₁₀-transformed data; ^bP < 0.05, ^cP < 0.01, ^dP < 0.005 vs. nonobese control; ^eP < 0.005 vs. obese control (ANOVA with post hoc Tukey’s test).

TABLE 3.

Effect of age and regional adiposity on glucose homeostasis and lipids in control females

Dependent variable	Model	Independent variable	Δrb (%)	rb (%)	β	se	Pβ	PE
Insulina	A	Total AT	51.8	51.8	0.018	0.002	<0.001	<0.001
		Age	12.3	71.9	−0.016	0.004	<0.001	<0.001
	B	VAT	53.0	53.0	0.154	0.018	<0.001	<0.001
		Age	12.0	65.0	−0.016	0.004	<0.001	<0.001
		SCAT	1.4	66.4	Not included			0.20
	C	VAT	53.0	53.0	0.154	0.018	<0.001	<0.001
		Age	12.0	65.0	−0.016	0.004	<0.001	<0.001
		ASCAT	0.1	65.1	Not included			0.79
Glucose	A	Age	10.8	10.8	0.014	0.008	0.07	0.03
		Total AT	6.8	16.8	0.007	0.004	0.07	0.07
	B	VAT	12.1	12.1	0.065	0.032	0.05	0.02
		Age	6.8	18.9	0.014	0.008	0.07	0.07
		SCAT	0.2	19.1	Not included			0.86
	C	VAT	12.1	12.1	0.065	0.032	0.05	0.02
		Age	6.8	18.9	0.014	0.008	0.07	0.07
		ASCAT	0.4	19.3	Not included			0.47
Insulin/glucose ratioa	A	Total AT	49.2	49.2	0.017	0.002	<0.001	<0.001
		Age	14.9	64.1	−0.017	0.004	<0.001	<0.001
	B	VAT	50.0	50.0	0.148	0.018	<0.001	<0.001
		Age	14.6	64.6	−0.017	0.004	<0.001	<0.001
		SCAT	1.2	66.2	Not included			0.17
	C	VAT	50.0	50.0	0.148	0.018	<0.001	<0.001
		Age	14.6	64.6	−0.017	0.004	<0.001	<0.001
		ASCAT	0.1	64.7	Not included			0.73
C-peptide/insulin ratioa	A	Total AT	17.0	17.0	−0.008	0.003	0.007	0.007
		Age	4.0	21.0	Not included			0.17
	B	VAT	21.7	21.7	−0.104	0.027	<0.001	<0.001
		Age	8.3	28.0	0.011	0.006	0.07	0.07
		SCAT	1.1	29.1	Not included			0.64
	C	VAT	21.7	21.7	−0.104	0.027	<0.001	<0.001
		Age	8.3	28.0	0.011	0.006	0.07	0.07
		ASCAT	3.3	31.3	Not included			0.51
Triglyceridesa	A	Total AT	40.8	40.8	0.008	0.001	<0.001	<0.001
		Age	1.6	42.6	Not included			0.27
	B	VAT	51.9	51.9	0.077	0.012	<0.001	<0.001
		Age	2.1	54.0	Not included			0.18
		SCAT	0.9	54.9	Not included			0.33
	C	VAT	51.9	51.9	0.077	0.011	<0.001	<0.001
		Age	2.1	54.0	Not included			0.18
		ASCAT	1.3	55.3	Not included			0.42
HDL/LDL-cholesterol ratioa	A	Total AT	44.7	44.7	−0.007	0.001	<0.001	<0.001
		Age	0.7	45.4	Not included			0.45
	B	VAT	47.5	47.5	−0.059	0.010	<0.001	<0.001
		Age	0.8	48.3	Not included			0.43
		SCAT	0.3	48.6	Not included			0.63
	C	VAT	47.5	47.5	−0.059	0.010	<0.001	<0.001
		Age	0.8	48.3	Not included			0.43
		ASCAT	0.1	48.4	Not included			0.99

These results demonstrate the greater importance of VAT than sc AT depots in explaining the variance in fasting insulin, fasting glucose, insulin/glucose and C-peptide/insulin ratios, triglycerides, and HDL/LDL-cholesterol ratio in adult female control subjects (n= 44). Forward stepwise multiple linear regression was performed with markers of glucose homeostasis or lipids as dependent variable, and age (yr) and total or regional AT volumes (liters), as independent variables. Independent variables in model A are: age, total AT; in model B: age, VAT, total sc AT; in model C: age, VAT, abdominal sc AT.Δ r^b indicates the change in variability (%) in the dependent variable explained by each independent variable in the final model; r^b indicates the cumulative variability (%) explained by the model after each step; β represents the regression coefficient for each independent variable included in the final equation ± se; P^β represents the significance of β in the final model; P^E represents the significance of the F-to-enter for that independent variable in stepwise multiple linear regression, in addition to those variables already included in the equation (if P^E > 0.10 the independent variable was not included in the final model).

^a

The dependent variable was log₁₀-transformed.

Metabolic and hormonal parameters in PWS females

In view of the complex effects of exogenous sex hormones on metabolism (20), analysis of metabolic parameters in PWS was restricted to those subjects not receiving OCP/HRT (n = 8) (Table 2). In these PWS subjects, as with controls, VAT was positively related to fasting insulin (r² = 0.31, P = 0.15), insulin/glucose ratio (r² = 0.31, P = 0.15), and triglycerides (r² = 0.62, P= 0.02), and negatively related to C-peptide/insulin ratio (r² = 0.27, P = 0.19) and HDL/LDL-cholesterol ratio (r² = 0.58, P = 0.03), although owing to the smaller sample size, this did not always reach statistical significance (all dependent variables log₁₀-transformed).

In view of the strong dependency of metabolic parameters on obesity seen above, comparisons between PWS subjects and obese controls used multiple regression analysis, adjusting for differences in total and regional adiposity (Table 4). Adjusting for total adiposity (total AT volume or percent body fat), PWS subjects had significantly lower fasting insulin, insulin/glucose ratio, FIRI, and triglycerides and significantly higher C-peptide/insulin ratio than obese controls (Table 4). However, there was no significant difference in fasting insulin, insulin/glucose ratio, FIRI, triglycerides, or C-peptide/insulin ratio between PWS subjects and obese controls, adjusting for visceral (in contrast to total) adiposity (as AT volume or percent body mass, P = 0.3 − 0.6) (Table 4).

TABLE 4.

Effect of PWS on metabolic parameters, correcting for body fat content and distribution

Obesity-related phenotype (dependent variable)	PWS as % of obese controls (β), adjusting for
	Total AT volume	Total fat (% body mass)	Visceral AT volume	Visceral fat (% body mass)
Fasting insulina	47.4 (26.3, 85.5)b	35.1 (17.7, 69.7)c	80.9 (41.4, 157.8)	71.6 (34.3, 149.6)
Insulin/glucose ratioa	46.9 (25.8, 85.3)b	35.5 (17.7, 71.1)c	77.4 (38.8, 154.5)	68.7 (32.4, 145.5)
FIRIa	48.1 (26.7, 86.7)b	34.6 (17.3, 69.0)c	84.5 (43.8, 163.3)	74.6 (35.9, 155.2)
Triglyceridesa	61.7 (41.3, 92.0)c	55.3 (37.6, 81.5)c	83.8 (53.9, 130.3)	82.0 (55.0, 122.5)

	Absolute difference between PWS and obese control, adjusting for
	Total AT volume	Total fat (% body mass)	Visceral AT volume	Visceral fat (% body mass)
C-peptide/insulin ratio	4.4 (0.4, 8.3)b	5.5 (1.1, 10.0)b	1.8 (−3.8, 7.4)	2.1 (−3.6, 7.8)

Metabolic parameters in PWS females, compared with obese control females (mean and 95% confidence interval), adjusting for age, total or visceral adiposity. Note that fasting insulin, insulin/glucose ratio, FIRI, and triglycerides are significantly lower and C-peptide/insulin ratio is significantly greater in PWS subjects when adjusting for total adiposity (either total AT volume or total body fat[ % body mass]) but not when adjusting for visceral adiposity (AT volume or % body mass). PWS regression coefficients (β) were calculated using multiple regression analysis of combined obese control (n = 14) and PWS females not receiving OCP/HRT (n = 8). The metabolic parameter was the dependent variable, and PWS diagnosis (control = 0, PWS = 1), total or visceral AT volume, or total or visceral fat (% body mass) were independent variables. Age was also included as an independent variable, when suggested by a significant influence (P of F-to-enter of <0.1) of age on the dependent variable in stepwise multiple linear regression analysis in controls (see Table 3). [β represents the value by which the metabolic parameter is altered in PWS compared to obese controls, correcting for total or visceral adiposity. For log₁₀-transformed variables this was converted into a figure for PWS as % of control, equal to 100/10^−β. The interaction coefficients (PWS × dependent AT or fat variable) were all nonsignificant and were therefore excluded from analysis.]

^a

Dependent variable was log₁₀-transformed.

^b

P < 0.05, ^cP < 0.01.

There was no significant difference in HDL/LDL-cholesterol ratio, total T, SHBG, free T index, or DHEAS between PWS subjects not receiving OCP/HRT and obese controls by ANOVA (Table 2), or multiple regression analysis, adjusting for total or visceral adiposity (data not shown). In the eight PWS subjects not on OCP/HRT, serum E2 levels (mean ± sd, range) were 201 ± 107 (57–350) pmol/liter. Two subjects had levels in the postmenopausal range (<100 pmol/liter), and these were the subjects with the lowest percent body fat.

Serum total T (mean ± sd: 1.3 ± 0.4 nmol/liter), SHBG (59 ± 37 nmol/liter), free T index (3.8 ± 3.9), and DHEAS (6.8 ± 3.9 μmol/liter) in the five PWS subjects receiving exogenous sex steroids were not significantly different from obese controls by ANOVA (all P > 0.05) or multiple regression analysis, adjusting for total or visceral adiposity (data not shown).

Discussion

An understanding of the factors controlling body fat distribution is important given the adverse consequences of obesity, particularly visceral fat accumulation, to health. PWS adults are an example of human obesity related to hypothalamic dysfunction, with several interacting endocrine and metabolic abnormalities that have previously been thought to influence fat distribution (2, 3). We have found a selective reduction in visceral adiposity in PWS adult females, which appears to explain a reduction in insulin resistance and triglyceride levels, and an increase in hepatic insulin extraction, correcting for their overall obesity.

Body fat distribution in PWS

PWS adult females had a relative reduction in VAT, but not SCAT or ASCAT, leading to lower VAT/SCAT and VAT/ASCAT ratios, compared with obese controls. This change in visceral adiposity was not reflected by the waist diameter, WHR, waist/height ratio or trunk/limb skinfold measurements. Furthermore, it was not apparent by measurement of the central/peripheral AT ratio, which may explain why no change in truncal/limb fat ratio was seen in a previous dual-energy x-ray photon absorptiometry study in PWS (9), and questions the predominantly central fat distribution mentioned in the PWS diagnostic criteria (6). Normal amounts of ASCAT, as found in the present study, would have masked any reductions in VAT when using dual-energy x-ray photon absorptiometry. This also implies that the reduced VAT in PWS is not caused by any disproportionate reduction in the length or size of the abdomen, compared with the rest of the body.

Body fat distribution, insulin resistance, and hyperlipidemia

We found that, in control premenopausal women, measurements of insulin resistance (fasting glucose, insulin, insulin/glucose ratio, and FIRI) (21, 22) and hypertriglyceridemia were positively correlated and a measurement of hepatic insulin extraction (fasting C-peptide/insulin ratio) (13) and HDL/LDL cholesterol ratio were negatively correlated with total and, in particular, visceral adiposity, as measured by whole-body MRI, in agreement with previous studies (23–25). These associations are thought to result, at least in part, from the direct hepatic effects of visceral AT-derived free-fatty acids via the portal vein (26). Increased free fatty acid efflux reduces hepatic insulin extraction, leading to hyperinsulinemia; enhances hepatic glucose production through increased gluconeogenesis; and increases hepatic very low-density lipoprotein, and hence triglyceride, secretion. There was no significant influence of SCAT or ASCAT, independent of VAT, on measurements of insulin resistance, in agreement with other studies in women (24, 25, 27) but in contrast to studies in men that have found an independent relationship with truncal sc fat (28).

Reduced metabolic complications of obesity in PWS

PWS females had reduced insulin resistance (assessed by fasting insulin, insulin/glucose ratio, and FIRI), greater hepatic insulin extraction (assessed by fasting C-peptide/insulin ratio), and lower hypertriglyceridemia, compared with obese controls, when adjusting for age and total adiposity. When correcting for the lower levels of VAT, these metabolic differences were no longer apparent in PWS subjects. This is consistent with an association between these reduced metabolic complications of obesity and the lower amounts of VAT in PWS females. Such differences have been found in other studies of children and adults with PWS, even though appropriate adjustment for adiposity, as in our study, has not been made (12–14, 29). Fasting insulin levels are lower in PWS children than BMI-matched children (13, 14, 29), despite the fact that BMI underestimates the degree of adiposity in PWS (9, 30). Mean fasting insulin was normal and C- peptide/insulin ratio was raised in a study of PWS adults, compared with obese controls, despite a tendency for increased percent body fat, measured by bioelectrical impedance (13), which is still likely to have underestimated their obesity owing to GH deficiency in PWS (31).

Although our study used surrogate measurements of insulin resistance, studies have found good correlations between the ratio or product of fasting glucose and insulin with insulin resistance measured by the iv glucose tolerance test (21, 22). Previous studies have shown greater falls in glucose in PWS children following iv insulin administration, compared with obese controls (12). Schuster et al. (13) meanwhile found lower insulin, but normal glucose responses, following oral glucose in PWS children, and iv glucose in PWS adults (with a tendency to greater rate of glucose disposal), compared with obese controls (13). This dissociation between obesity and insulin resistance in PWS again occurred despite the limitations of BMI and impedance measurements to match adiposity discussed above. Our results are therefore consistent with improved insulin sensitivity in PWS, independent of overall adiposity, but appropriate body composition analysis, in combination with an iv insulin tolerance test or hyperinsulinemic euglycemic clamp, will be needed to confirm this. Furthermore, our conclusions have some limitations in view of the small number of PWS subjects studied.

Reduced visceral adiposity in PWS

The finding of reduced visceral adiposity in PWS females is of particular interest. This is an unusual, perhaps unique, situation in which increased obesity is associated with reduced visceral adiposity. This reduction occurs despite the presence of several phenotypes in PWS, such as reduced physical activity, hypogonadism, adult GH deficiency, and psychiatric problems, which by contrast would be expected to increase the proportion of VAT.

Reduced physical activity and visceral adiposity

Physical exercise leads to greater reductions in VAT than SCAT in women (2, 3, 18). Although we did not formally assess physical activity levels in our study, anecdotally our PWS subjects described reduced mobility and exercise, previous reports have found reduced physical activity in PWS (11), and the reduction in VAT was greatest in the more obese PWS subjects (Fig. 2A), the opposite of that expected if increased physical activity was the explanation.

Hypogonadism and visceral adiposity

Visceral adiposity increases during menopause, independent of age and total adiposity (32), and VAT is stable during puberty in obese females (33). Furthermore the majority of PWS females not receiving the OCP/HRT still had E2 levels above postmenopausal levels, despite their oligo- or amenorrhea, probably from increased aromatization of androgens by peripheral fat because the lowest levels were seen in less obese subjects. Although a lack of cyclical changes in estrogens and progesterone might cause changes in fat distribution (34), the reduction in visceral adiposity was also seen in PWS females receiving cyclical exogenous sex hormones. Hypogonadism and incomplete or delayed puberty are therefore unlikely to explain the reduced visceral adiposity seen in PWS adult females.

GH deficiency and visceral adiposity

GH deficiency is a recognized feature of PWS children and adults, independent of obesity (14, 35), and suggested by the short stature of our subjects. GH has important effects on body fat content and distribution (31, 36, 37), with GH deficiency in adults producing an increase in both total and truncal adiposity, and the VAT/SCAT ratio, in contrast to the reduction in visceral adiposity in PWS adults seen in our study. Care must be taken to distinguish childhood-onset from adult-onset GH deficient adults, who may differ in their body composition and hormonal and metabolic perturbations (38), and to adjust for total adiposity (39), but studies using computed tomography (CT) or MRI have shown greater reductions in VAT than SCAT during GH therapy in adult-onset (40), childhood-onset (41), or mixed-onset (39) GH-deficient adults and greater increases in VAT on discontinuation of childhood-onset GH therapy in adults (42). Interestingly, a recent preliminary report found no change in VAT after 6 months of GH treatment to PWS adults, as assessed by single-slice CT (43), that might reflect already reduced visceral adiposity at baseline, as seen in our study.

Although GH deficiency in children leads to increased abdominal sc adiposity (44) and GH therapy decreases abdominal sc and total body fat in GH-deficient (45) and PWS children (46, 47), the effects on visceral adiposity are unknown from published literature. Although one study found normal visceral adiposity in childhood-onset GH deficient adults using single-slice CT (41), these subjects had received GH treatment for short stature in childhood, a therapy that was received by only one PWS subject as a child in our study.

Treatment of neonatal rats with GH antibodies over 8 wk reduces visceral fat much more than sc fat (48). Furthermore, both GH-deficient and PWS children have increased adipocyte volume but no increase, and even a decrease, in adipocyte cell number (49, 50), suggesting contrasting effects of GH to increase lipolysis but also to increase adipocyte differentiation and proliferation (2, 49, 50). The speculation that GH deficiency during early childhood might have opposite effects on visceral adiposity to that seen in adults, could account for our observed reduction in visceral adiposity in adult PWS subjects. Both non-PWS (51) and PWS (12–14, 29, 52) GH-deficient children have increased insulin sensitivity, suggesting that similar processes are at play, that may be not only a consequence of the loss of anti-insulin action of GH but also a reduction in visceral adiposity.

Adrenal steroids, androgens, and visceral adiposity

Stressful social circumstances, anxiety, and depression (the hypothalamic arousal mechanism), associated with chronic activation of the hypothalamic-pituitary (HP)-adrenal axis, may also predispose to the development of obesity, and particularly visceral adiposity (53). It is interesting to note, therefore, that we found a reduction in visceral adiposity in PWS in our study, despite high levels of stress being reported in PWS (54), depression being observed in 27% of PWS adults (55), and 31% of our PWS subjects using antidepressant or antipsychotic medication (56), with this reduction being independent of their usage. Alternatively reduced cortisol secretion in PWS could contribute to the reduced visceral adiposity (57). We did not find any reduction in DHEAS levels in our PWS subjects, and all PWS subjects not receiving OCP/HRT had basal cortisol levels greater than 215 nmol/liter (unpublished observations), suggesting that there is no gross impairment in HP-adrenal axis activity. Although clinical cortisol deficiency is not a recognized problem in PWS, more detailed dynamic testing of glucocorticoid secretion than has previously been reported (12, 52) will be needed to investigate whether there could be a subtle impairment of cortisol secretion during obesity in PWS, perhaps related to hypothalamic abnormalities.

Androgens increase the VAT/SCAT ratio in females (34, 58). However, there was no evidence for androgen deficiency (free T index or DHEAS) in our study to explain the reduced visceral adiposity in PWS females.

Hypothalamic or genetic defects and visceral adiposity

Other factors particular to PWS could account for their reduced visceral adiposity. Developmental abnormalities in the hypothalamus might alter control of peripheral fat distribution, perhaps through the autonomic system (59), distinct from its effects on the HP-axis. However, this remains speculative because fat distribution has not yet been reported in other causes of human hypothalamic obesity (60, 61), and there is no evidence of such changes from animal studies of hypothalamic damage (5, 62). Secondly, genes in the PWS chromosomal region (chromosome 15q11–13) could regulate body fat distribution through direct adipocyte effects. Lack of expression of imprinted PWS candidate genes or haploinsufficiency for other genes in the PWS chromosomal region, in adipocytes, and perhaps regional variation in their expression or degree of imprinting among various fat depots might also lead to changes in regional fat physiology. However, genetic linkage studies have shown no association of this region with visceral adiposity (63).

Conclusion

Using whole-body MRI, a selective reduction in VAT was found in PWS female adults, despite their increased overall obesity, that was independent of the use of exogenous sex steroids. This appeared unrelated to and occurred despite the hypogonadism, physical inactivity, adult GH deficiency, and psychiatric problems seen in adults with PWS. PWS females had lower fasting insulin, insulin/glucose ratio, FIRI, or triglycerides and increased C-peptide/insulin ratio that appeared to be associated with the reduction in visceral adiposity. This protection against the insulin resistance and other metabolic complications of obesity would reduce the risk of development of diabetes mellitus and cardiovascular complications in PWS, relative to their degree of obesity. However, there will still be an absolute increased prevalence of type 2 diabetes mellitus and cardiovascular disease, such as hypertension, despite the relatively less severe visceral adiposity (19% and 17%, respectively, in a study of 232 PWS adults) (64). The cause of the reduction in visceral adiposity in PWS is unclear but could be related to childhood-onset GH deficiency, subtle defects in HP-adrenal axis activity, or novel hypothalamic or genetic influences on the control of body fat distribution.

Acknowledgements

We thank PWS and other patients, their caregivers, and families and the United Kingdom PWS Association for their support and keen participation; Department of Biochemical Endocrinology, Hammersmith Hospital for performing hormone assays; Caroline Doré, Department of Medical Statistics and Evaluation, ICSM, Hammersmith Hospital; and Joan Morris, Department of Medical Statistics, Queen Mary and Westfield College, University of London, for statistical advice.

This work was supported by United Kingdom Medical Research Council and Marconi Medical Systems. A.P.G. and J.H. are United Kingdom Medical Research Council Training Fellows.

E.L.T. and A.E.B. contributed equally to this work.

Anthony P. Goldstone is now with the Department of Endocrinology, St. Bartholomew’s Hospital, West Smithfield, London EC1A 7BE, United Kingdom. E-mail:

Abbreviations:

• ASCAT,

  Abdominal sc adipose tissue;

• AT,

  adipose tissue;

• BMI,

  body mass index;

• CT,

  computed tomography;

• DHEAS,

  dehydroepiandrosterone sulfate;

• FFM,

  fat-free mass;

• FIRI,

  fasting insulin resistance index;

• HDL,

  high-density lipoprotein;

• HP,

  hypothalamic-pituitary;

• HRT,

  hormone replacement therapy;

• INAT,

  nonvisceral internal adipose tissue;

• LDL,

  low-density lipoprotein;

• MRI,

  magnetic resonance imaging;

• OCP,

  oral contraceptive pill;

• PWS,

  Prader-Willi syndrome;

• SCAT,

  total sc adipose tissue;

• VAT,

  visceral adipose tissue;

• WHR,

  waist/hip ratio.