|Home | About | Journals | Submit | Contact Us | Français|
Reduced growth hormone (GH) secretion is observed in obesity and may contribute to increases in cardiovascular disease (CVD) risk. Lipoprotein characteristics including increased small dense LDL particles are known independent risk factors for CVD. We hypothesized that reduced GH secretion in obesity would be associated with a more atherogenic lipid profile including increased small dense LDL particles.
To evaluate this hypothesis, we studied 102 normal weight and obese men and women using standard GH stimulation testing to assess GH secretory capacity and performed comprehensive lipoprotein analyses including determination of lipoprotein particle size and sub-class concentrations using proton NMR spectroscopy.
Obese subjects were stratified into reduced or sufficient GH secretion based on the median peak stimulated GH (≤6.25 μg/l). Obese subjects with reduced GH secretion (n=35) demonstrated a smaller mean LDL and HDL particle size in comparison to normal weight subjects (n=33) or obese subjects with sufficient GH (n=34) by ANOVA (P<0.0001). Univariate analyses demonstrated peak stimulated GH was positively associated with LDL (r=0.50; P<0.0001) and HDL (r=0.57; P<0.0001) but not VLDL (P=0.06) particle size. Multivariate regression analysis controlling for age, gender, race, ethnicity, tobacco, use of lipid lowering medication, BMI and HOMA demonstrated peak stimulated GH remained significantly associated with LDL particle size (β=0.01; P=0.01; R2=0.42; P<0.0001 for overall model) and HDL particle size (β=0.008; P=0.001; R2=0.44; P<0.0001 for overall model).
These results suggest reduced peak stimulated GH in obesity is independently associated with a more atherogenic lipoprotein profile defined in terms of particle size.
Reduced GH secretion observed in obesity is associated with increased cardio-metabolic complications including dyslipidemia, insulin resistance, inflammation and increased carotid intima-media thickness 1, 2. In addition to traditional lipid profiles, specific lipoprotein characteristics such as increased concentrations of small dense low-density lipoprotein (LDL) particles have been associated with increased risk of cardiovascular disease (CVD), including incident CVD events 3. In subjects with GH deficiency due to hypothalamic/pituitary disease, despite similar total LDL concentrations, an increase in the concentration of small dense LDL particles was noted compared to age and BMI matched controls 4. However, a detailed analysis of lipoprotein particle profile, including size and concentration of the lipoprotein sub-species, has not been performed to our knowledge in obese subjects with reduced GH secretion. We hypothesized that obese subjects with reduced GH secretion would have higher concentrations of small LDL particles which may contribute to the increased cardio-metabolic risk associated with reduced GH secretion in obesity. We therefore performed standard GH stimulation tests and determined the particle size and concentrations of lipoprotein sub-species by proton nuclear magnetic resonance (NMR) spectroscopy in a group of well phenotyped normal weight and obese subjects.
Obese (n=69, BMI >30 kg/ m2) and normal weight (n=33, BMI <25 kg/m2) men and women from the Boston area were enrolled. Data from these subjects were previously published in a study evaluating the relationship between reduced GH secretion and increased carotid intima-media thickness in obesity 2, however, detailed lipoprotein profiles of these subjects have yet to be reported. These subjects were relatively healthy adults, age 18 to 55 years, without known pituitary disease. Exclusion criteria included a history of diabetes mellitus, thyroid disorders, chronic medical condition such as HIV infection, or any medical condition known to affect the GH axis. Subjects taking medications known to affect GH secretion were excluded. Subjects with serum creatinine >1.5 mg/dl, hemoglobin <11g/dl, or aspartate aminotransferase >2.5 times the upper limit of the normal range were also excluded. Subjects underwent written informed consent in compliance with the guidelines of the Subcommittee on Human Studies at the Massachusetts General Hospital prior to the administration of any study procedures.
Subjects underwent standard GH stimulation testing with GH releasing hormone (GHRH) and arginine as previously reported 2, 5. Briefly, subjects were administered GHRH 1-29 [1 μg/kg] (Sermorelin acetate, Geref, Serono Laboratories, Inc., Norwell, Ma) intravenously followed by arginine hydrochloride [30 g/300 ml (max 30 g)]. GH levels were measured at 0, 30, 45, 60, 90 and 120 minutes. GH was determined by a paramagnetic particle, chemiluminescent immunoassay (Beckman Coulter, Chastka, MN) with an analytical sensitivity of 0.002 μg/l. The intra-assay CV is 1.90-2.78 % and the inter-assay CV is 1.77-2.65 %. NMR spectroscopy was performed as previously described 6 on fasting plasma samples by a commercial vendor (Liposcience Inc., Raleigh, NC). The NMR technique has previously been validated against lipoprotein subspecies distribution determined by gel electrophoresis with good results 6. Oxidized LDL was measured in fasting plasma samples using commercially available ELISA kit (ALPCO Diagnostics, Salem, NH) with an intra-assay CV of 4-7.6 % and inter-assay CV of 6.2-10.7 %. Subjects also completed a 75 gram oral glucose tolerance test (OGTT) and determination of fasting cholesterol profile using standard methodologies.
Height and body weight were obtained after an overnight fast. Total body fat percentage was determined by dual X-ray absorptiometry (DXA) testing using a Hologic, Inc-4500 densitometer (Hologic, Inc., Waltham, MA). In addition, 1 cm cross-sectional abdominal CT scans were performed at the level of L4 to assess the distribution of abdominal subcutaneous adipose tissue (SAT) and abdominal visceral adipose tissue (VAT) as previously described 7.
Lipoprotein particle size and sub-species concentrations were compared between normal weight and obese subjects using Student’s t-test. Normal weight subjects were included for comparison with obese subjects as an internal control. Obese subjects were then stratified into GH sufficient (GHS) or relative GH deficient (GHD) based on the median peak stimulated GH value amongst the obese subjects (≤6.25μg/l).. Lipoprotein particle size and sub-species concentrations were compared between normal weight, obese GHS and obese GHD using ANOVA, and adjustments for age, gender, race and BMI were also made. Subsequently, Tukey-Kramer post-hoc test to determine significant difference between groups was performed for comparisons with significant ANOVAs. Mean LDL and mean HDL particle size was related to various measures of body adiposity and to various metabolic parameters using univariate regression analyses with Pearson correlations. In addition, the association between peak stimulated GH and lipoprotein particle size and sub-species concentration was also determined using univariate regression analyses with Pearson correlations. Adjustments for possible covariates including age, gender, race, ethnicity, tobacco use, use of lipid lowering medication, anthropometric measurements, HOMA and their respective cholesterol concentrations was performed using multivariate regression analyses with standard least squares modeling. Variables were selected for entry into the multivariate models based on known associations with lipoprotein particle size from the literature (such as age, gender, race and ethnicity, and insulin sensitivity 8-13) and based on significant associations with mean LDL or HDL particle size on univariate analyses (measures of adiposity, cholesterol and insulin sensitivity). In addition, sensitivity analyses using the median peak GH responses among normal weight subjects as cutoffs were performed. Statistical analysis was performed using JMP 8.0.2 (SAS Institute, Cary, North Carolina, USA). Statistical significance was defined as P<0.05.
Thirty-three normal weight and 69 obese subjects were studied. The two groups were similar in age (40.4±1.9 vs. 42.1±1.1 years; P=0.42), gender (64 vs. 57 % male; P=0.49), race (64 vs. 51 % Caucasian; P=0.22) and tobacco use (5.2±1.7 vs. 7.4±1.6 pack years; P=0.39). Obese subjects had higher BMI (22.5±0.3 vs. 38.1±0.8 kg/m2; P<0.0001) and waist circumference (80±1 vs. 119±2 cm; P<0.0001) as expected. The two groups had similar levels of total cholesterol (174±6 vs. 182±4 mg/dl; P=0.26) but obese subjects had higher triglycerides (75±7 vs. 127±10 mg/dl; P=0.0009) and LDL cholesterol (101±5 vs. 116±4 mg/dl; P=0.02) and lower HDL cholesterol (58±3 vs. 45±1 mg/dl; P<0.0001). Obese subjects had lower fasting GH (0.8±0.2 vs. 0.1±0.03 μg/l; P<0.0001), peak stimulated GH (44.7±4.9 vs. 8.8±0.9 μg/l; P<0.0001) as well as a lower IGF-1 (93.1±7.6 vs. 73.5±4.1 μg/l; P=0.01) (all comparisons normal weight vs. obese).
Obese subjects had smaller mean LDL (21.6±0.1 vs. 20.8±0.1 nm; P<0.0001) and HDL (9.3±0.08 vs. 8.8±0.05 nm; P<0.0001) and tended to have a larger VLDL (49.0±1.5 vs. 52.4±1.2 nm; P=0.09) particle size compared to normal weight subjects. Obese subjects also had greater concentrations of small (376.9±44.4 vs. 808.4±51.6 nmol/l; P<0.0001), medium small (79.1±9.0 vs. 168.1±10.5 nmol/l; P<0.0001) and very small (297.7±35.6 vs. 640.3±41.5 nmol/l; P<0.0001) LDL particles with lower concentrations of large LDL particles (441.9±24.2 vs. 325.4±21.0 nmol/l; P=0.001) compared to normal weight subjects. Similarly, obese subjects had greater concentrations of small HDL (18.2±0.9 vs. 21.0±0.6 nmol/l; P=0.01) but lower concentrations of large HDL (9.2±0.6 vs. 5.5±0.4 nmol/l; P<0.0001) particles. The two groups had similar concentrations of medium HDL particles (3.5±0.5 vs. 3.5±0.4 nmol/l; P=0.98). Obese subjects had higher concentrations of IDL particles compared to normal weight subjects (29.8±7.1 vs. 56.1±6.1 nmol/l; P=0.01). There were no differences in the concentration of oxidized LDL (2394±607 vs. 3001±1019 mU/l; P=0.69) (all comparisons normal weight vs. obese, respectively).
The obese subjects had a mean GH value of 8.8±0.9 μg/l and a median GH value of 6.25 μg/l (25-75% range: 4.4-9.5 μg/l). The median GH value of 6.25 μg/l was utilized as the cut-off criteria to define relative GH deficiency in obesity so as to obtain equal number of subjects in the GH sufficient and the relative GH deficient group.
The obese GHD subjects (peak GH ≤6.25 μg/l; n=35) tended to be older (P=0.07 by ANOVA) and more likely to be Caucasian (P=0.06 by ANOVA) and had higher percentage of male subjects (P=0.03 by ANOVA) and higher BMI (P<0.0001 by ANOVA) compared to normal weight and obese GHS subjects (peak GH >6.25 μg/l; n=34) (Table 1). Therefore, all results comparing normal weight, obese GHS and obese GHD subjects for lipoprotein particle size and sub-species concentration were adjusted for age, gender, race and BMI.
The obese GHS and GHD subjects had similar percent adipose tissue by DXA and similar abdominal SAT. However, the obese GHD subjects had a greater waist circumference and higher VAT compared to the obese GHS subjects (Table 1).
The obese GHD group demonstrated a smaller mean LDL particle size in comparison with normal weight and obese GHS group by overall ANOVA (P<0.0001). Using Tukey-Kramer post-hoc analyses, the obese GHD group had smaller mean LDL particle size compared to normal weight subjects (P<0.05). The obese GHS group also had smaller mean LDL particle size compared to normal weight subjects (P<0.05). The obese GHD group had higher mean concentrations of small, medium small, and very small LDL particles in comparison to the normal weight group. The obese GHD subjects also had lower mean concentrations of large LDL particles in comparison to normal weight subjects but similar levels compared to obese GHS subjects. The obese GHS subjects similarly had greater levels of small, medium small and very small LDL sub-species in comparison with normal weight subjects (Table 1).
The obese GHD group had smaller mean HDL particle size compared to normal weight and the obese GHS group by ANOVA (P<0.0001) and Tukey-Kramer post-hoc test (P<0.05). The obese GHD group had higher mean concentrations of small HDL particles and lower mean concentrations of large HDL particles compared to normal weight subjects. The obese GHS group had lower mean concentrations of large HDL particles compared to normal weight subjects as well but had similar levels of medium and small HDL particles (Table 1).
There was no difference in the mean VLDL particle size. The obese GHD group had higher mean concentrations of medium and large VLDL and chylomicron particles compared to normal weight and obese GHS subjects, however, there were no differences in the small VLDL particle concentrations. The obese GHS subjects had similar level of VLDL and chylomicrons in comparison with the normal weight subjects (Table 1).
There were no differences in oxidized LDL among the three groups (2394±607 vs. 2196±451 vs. 3783±1966; adjusted P=0.93 comparing normal weight vs. obese GHS vs. obese GHD; Table 1).
Mean LDL particle size was inversely associated with measures of adiposity including BMI (r=−0.48; P<0.0001), waist circumference (r=−0.47; P<0.0001), percent body fat by DXA (r=−0.36; P=0.0002), VAT (r=−0.51; P<0.0001) and SAT (r=−0.36; P=0.0004) (Table 2). While LDL particle size was not associated with total cholesterol (P=0.23), it was inversely associated with serum concentrations of LDL cholesterol (r=−0.26; P=0.007) and triglycerides (r=−0.60; P<0.0001) and positively associated with serum concentrations of HDL cholesterol (r=0.68; P<0.0001) (Table 2). In addition, mean LDL particle size was inversely associated with measures of insulin sensitivity including HbA1c (r=−0.31; P=0.001), fasting glucose (r=−0.27; P=0.006), 2-hour glucose (r=−0.43; P<0.0001), fasting insulin (r=−0.37; P=0.0001) and HOMA (r=−0.37; P=0.0002) (Table 2).
Mean HDL particle size was also inversely associated with measures of adiposity including BMI (r=−0.44; P<0.0001), waist circumference (r=−0.51; P<0.0001), percent body fat (r=−0.33; P=0.0007), VAT (r=−0.58; P<0.0001) and SAT (r=−0.39; P<0.0001) (Table 3). Mean HDL particle size was not associated with serum total cholesterol concentrations (P=0.11) but it was associated with serum HDL concentrations (r=0.81; P<0.0001) and was inversely associated with serum LDL concentrations (r=−0.41; P<0.0001) and triglyceride concentrations (r=−0.49; P<0.0001) (Table 3). HDL particle size was inversely associated with HbA1c (r=−0.31; P=0.002), fasting glucose (r=−0.24; P=0.02), 2-hour glucose (r=−0.38; P<0.0001), fasting insulin (r=−0.38; P<0.0001) and HOMA (r=−0.35; P=0.0003) (Table 3).
Among all subjects, fasting GH (r=0.27; P=0.006), peak stimulated GH (r=0.50; P<0.0001) (Figure 1A) and IGF-1 (r=0.25; P=0.02) were all associated with mean LDL particle size (Table 2). Similarly, fasting GH (r=0.23; P=0.02), peak stimulated GH (r=0.57; P<0.0001) (Figure 1B) and IGF-1 (r=0.24; P=0.02) were all associated with mean HDL particle size (Table 3). While mean VLDL particle size trended to an association with peak stimulated GH (r=−0.19; P=0.06) (Figure 1C), it was not significantly associated with fasting GH (P=0.15) or IGF-1 levels (P=0.10).
In regards to lipoprotein sub-species, peak stimulated GH was inversely associated with mean concentrations of small (r=−0.48; P<0.0001), medium small (r=−0.50; P<0.0001) and very small LDL (r=−0.47; P<0.0001) particles while it was positively associated with large LDL particle concentrations (r=0.27; P=0.007). Similarly, peak stimulated GH was inversely associated with small HDL (r=−0.31; P=0.002) and positively associated with large HDL (r=0.49; P<0.0001) particle concentrations. Peak GH tended to an inverse association with medium HDL particle concentration (r=−0.18; P=0.07). Peak stimulated GH was not associated with oxidized LDL (P=0.86) (Table 4).
In multivariate regression modeling controlling for the effects of age, gender, race, ethnicity, tobacco use, lipid lowering medication use, BMI, and HOMA, peak stimulated GH remained significantly associated with LDL particle size (β=0.01; P=0.01; R2=0.42; P<0.0001 for overall model) and HDL particle size (β=0.008; P=0.001; R2=0.44; P<0.0001 for overall model) in separate models. Peak stimulated GH remained significantly associated with LDL or HDL particle size when the model was further controlled for LDL or HDL concentrations respectively in separate models (Model for LDL: Peak GH: β=0.01; P=0.02; R2=0.43; P<0.0001 for overall model; Model for HDL: Peak GH: β=0.006; P=0.0009; R2=0.73; P<0.0001 for overall model) (Table 5A and B). The addition of waist-circumference to the overall model also did not change the significant relationship between GH and LDL or HDL particle size (Model for LDL: Peak GH: β=0.01; P=0.01; R2=0.43; P<0.0001 for overall model; Model for HDL: Peak GH: β=0.006; P=0.003; R2=0.73; P<0.0001 for overall model). Furthermore, the use of direct measurements of adiposity such as body fat percentage by DXA or measures of central adiposity such as VAT by CT scan did not change the significant relationship between peak stimulated GH and mean LDL or HDL particle size. The use of other measures of insulin sensitivity such as HbA1c, fasting glucose, 2-hour glucose, or fasting insulin also did not change the significant relationship between peak GH and mean LDL or HDL particle size.
However, when multivariate modeling was performed for VLDL particle size, HOMA, and not peak stimulated GH, was the significant variable associated with VLDL particle size (HOMA: β=1.2; P=0.01; Peak GH: β=−0.01; P=0.82; overall model: R2=0.18; P=0.03). The use of percent body fat by DEXA or VAT by CT scan did not affect the significant relationship between HOMA and VLDL particle size.
We performed additional analyses limited to normal weight subjects. Normal weight subjects had a mean peak stimulated GH of 44.7±4.9 μg/l with a median peak stimulated GH of 43.8 μg/l (25-75% range: 18.9 to 66.7 μg/l). There were no differences in mean LDL, HDL or VLDL particle size comparing normal weight subjects with peak stimulated GH above or below the median value of 43.8 μg/l. Among normal weight subjects only (n=33), peak stimulated GH was significantly related to mean HDL particle size (r=0.34; P=0.05) but was not associated with LDL particle size (r=0.27; P=0.13) or VLDL particle size (r=−0.03; P=0.86).
In this study, we provide the first comprehensive report of lipoprotein profile, including particle size and sub-species concentrations, among obese patients in whom GH secretory capacity was characterized. Our data demonstrate a significant association between peak stimulated GH on standard stimulation testing and LDL and HDL particle size. The results suggest that reduced peak stimulated GH is associated with a more atherogenic lipoprotein profile with decreased LDL and HDL particle size, which may contribute to the increased CVD risk associated with reduced GH secretion in obesity. In our data, peak stimulated GH was not associated with oxidized LDL.
For the purposes of this study, we used the median peak stimulated GH value among obese subjects to define relative GH sufficiency or deficiency of obesity. Using this cut-off value of 6.25 μg/l, we obtained equal numbers of obese subjects with relative GH sufficiency or deficiency and demonstrated a strong relationship of GH status on LDL and HDL particle size. More importantly, we demonstrate a continuum of abnormality in lipoprotein particle size and subspecies concentrations associated with reduced GH secretory capacity as demonstrated in the univariate analyses and confirmed in the multivariate analyses.
Previous studies have demonstrated an association between small dense LDL cholesterol particles and increased CVD 3 and small HDL particles have also been associated with the presence of coronary artery disease 14. Our results demonstrate a significant association between reduced GH secretion in obesity and reduced LDL and HDL particle size with a shift to a greater concentration of smaller LDL and HDL particles. The shift in the specific lipoprotein sub-species and size may contribute to the recently demonstrated association between reduced GH secretion and increased CVD risk in obesity 1, 2.
As age, gender, race, ethnicity and insulin resistance have been shown to play a role in lipoprotein distribution 8-13, we controlled for these covariates, as well as the use of lipid lowering medications and their respective serum cholesterol levels, in our multivariate regression modeling. Our results suggest the association of peak stimulated GH with LDL and HDL particle size may be independent from the effects of these covariates.
The current study was not designed to specifically address the mechanism through which reduced GH secretion in obesity may be associated with circulating lipoprotein concentrations or particle size. Previous studies have demonstrated GH reduces total cholesterol in hypercholesterolemic and normocholesterolemic men 15 either through induction of hepatic LDL receptor to potentiate clearing of LDL 16 or through activation of cholesterol 7α-hydroxylase to metabolize cholesterol to bile acids 17 or even through decreased de novo lipogenesis as seen in subjects with HIV infection 18. In regards to particle size, LDL particle size is inversely related to hepatic lipase activity 19. As hepatic lipase activity is low in conditions of GH excess such as acromegaly 20, 21, reduced GH secretion associated with obesity may be associated with elevated hepatic lipase activity contributing to reduced LDL particle size. The size of the lipoprotein particles, HDL and VLDL/chylomicrons in particular, may also be affected by the activity of cholesterol ester transfer protein (CETP) which is expressed in the GH target organs of adipose tissue and liver 22. While GH excess, as seen in acromegaly, was associated with higher plasma CETP activity in one study 20, another study demonstrated reduced CETP activity in acromegaly and an inverse association between plasma IGF-1 levels and CETP activity 23. GH treatment has also been shown to decrease CETP 24, 25, further supporting this negative association. Increased CETP activity in the setting of reduced GH secretion in obesity may result in increased transfer of cholesterol esters from HDL to VLDL/chylomicrons and IDL resulting in smaller, more dense HDL particles and to larger VLDL/chylomicrons and IDL particles, as seen in our study. The combination of increased CETP activity and hepatic lipase activity also may contribute to the formation of more dense, lipid poor, LDL particles.
Interestingly, while peak stimulated GH remained significantly associated with both LDL and HDL particle size independent of measures of insulin resistance, VLDL particle size was related to HOMA IR and not peak GH in multivariate modeling. This suggests that for VLDL particle size, unlike for LDL and HDL, insulin resistance rather than peak GH is a primary determinant of particle size. This is consistent with the known role of insulin in VLDL metabolism and the association between insulin resistance and hypertriglyceridemia 26. Previous studies have demonstrated the increase in VLDL particle size associated with insulin resistance occurs primarily due to an increases in the concentration of circulating large VLDL particles without significant changes in the concentrations of small or medium VLDL particles 27. This pattern of VLDL particle concentration is confirmed in our subset of obese subjects with reduced GH secretion and may reflect the increased insulin resistance seen in this sub-group of obese subjects.
One potential explanation for our data is that obesity results in reduced GH, and reduced GH further contributes to abnormal lipoprotein particle size in obesity. However, this study is limited by its observational nature and we can not conclude whether the changes in lipoprotein sub-species are a consequence of the reduced GH secretion. Therefore an alternative explanation for our data could involve other covariates such as age, BMI and insulin resistance which may affect lipoprotein particle size independent of GH. Although we controlled for these factors statistically using multivariate modeling, further interventional studies would be necessary to demonstrate whether normalization of the reduced GH secretion would yield any beneficial effects on specific lipoprotein profiles in obesity. Nonetheless, this study is the first study to perform a detailed evaluation of the association between GH secretion and various lipoprotein particle size and sub-species using proton NMR spectroscopy and suggests a potential relationship between reduced GH and a more atherogenic lipoprotein particle size in obesity. In addition, we characterized GH secretory capacity using a standardized GHRH-arginine test. We did not characterize endogenous GH secretion including various pulsatility parameters which would have been difficult in a study of this size but may relate differentially to lipoprotein particle size. Lastly, the obese GHD subjects, compared to the obese GHS subjects, were older, with more central obesity and tended to be male. This is consistent with the known effects of these covariates, particularly, central obesity, on GH secretion, and may account for some of the more unfavorable lipoprotein characteristics. However, we control for these covariates and report both an un-adjusted and an adjusted P value in the stratified analyses. Furthermore, the relationship between GH and lipoprotein particle size and sub-species concentrations was also demonstrated on univariate analyses including all subjects and in more complex multivariate modeling. Further studies are needed into the mechanism by which reduced GH secretion may contribute to abnormal lipoprotein particle size in obesity.
In summary, we demonstrate a more adverse LDL and HDL subpopulation distribution in obese subjects with reduced GH secretion. These changes were independent of obesity and insulin resistance and may contribute to the increased CVD risk previously demonstrated in obese subjects with reduced GH secretion. Further interventional studies are now needed to demonstrate whether improvement of reduced GH secretion by exogenous GH or GH releasing factors would provide beneficial effects on lipoprotein profile and cardio-metabolic risk associated with obesity.
We gratefully acknowledge the MGH bionutrition and nursing staffs and the research volunteers for their participation in the study. In addition, we would like to acknowledge the following funding sources for their financial support: National Institutes of Health grant R01HL085268 to SKG, K24DK064545 to SKG, K23DK087857 to HM, K23DK089910 to TLS and M01RR01066 and UL1RR025758, Harvard Clinical and Translational Science Center, from the National Center for Research Resources. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Center for Research Resources or the National Institutes of Health. This study was registered at www.clinicaltrials.gov as NCT00562796.
Disclosure: The authors have nothing to disclose