0021-972X/01/$03.00/0 Vol. 86, No. 2 The Journal of Clinical Endocrinology & Metabolism Printed in U.S.A. Copyright 2001 by The Endocrine Society Assessment of Growth Hormone Dynamics in Human Immunodeficiency Virus-Related Lipodystrophy* PETRA RIETSCHEL, COLLEEN HADIGAN, COLLEEN CORCORAN, TAKARA STANLEY, GREGORY NEUBAUER, JOSEPH GERTNER, AND STEVEN GRINSPOON Neuroendocrine Unit (P.R., C.H., C.C., T.S., S.G.) and the General Clinical Research Center (G.N.), Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 02114; and Serono Laboratories, Inc. (J.G.), Norwell, Massachusetts 02061 ABSTRACT Human immunodeficiency virus (HIV) lipodystrophy (LIPO) is characterized by increased visceral adiposity, peripheral fat atrophy, dyslipidemia, and insulin resistance. GH concentrations are known to vary inversely with excess weight and body fat but have not been investigated in HIV lipodystrophy. Twenty-one subjects with HIV LIPO, 20 HIV-infected nonlipodystrophy subjects (NONLIPO), and 20 control (C) subjects were prospectively recruited for this study and compared. Subjects in the three groups were all male, age-matched [median, 47 yr old (interquartile range, 37 50) LIPO; 41 (37 44) NONLIPO; and 43 (37 49) C], and body mass index-matched [median, 24.3 kg/m 2 (interquartile range, 22.2 26.6) LIPO; 24.4 (23.3 25.9) NONLIPO; and 24.8 (22.7 26.1) C] (P 0.05 for all comparisons). Visceral abdominal fat [16,124 mm 2 (11,246 19,790) LIPO; 7,559 (5,134 11,201) NONLIPO; and 8,803 (6,165 11,623) C; P 0.01 LIPO vs. NONLIPO and LIPO vs. C] and the ratio of visceral abdominal fat to sc abdominal fat [1.37 (0.71 2.44) LIPO vs. 0.57 (0.47 0.78) NONLIPO vs. 0.55 (0.41 0.71) C, P 0.01 LIPO vs. NONLIPO and LIPO vs. C] were significantly increased in the LIPO subjects but were not significantly different between NONLIPO and C. The mean overnight GH concentration, determined from frequent sampling every 20 min (from 2000 h to 0800 h) was decreased in the LIPO subjects [0.38 g/l (0.13 0.67) LIPO vs. 0.96 (0.53 1.30) NONLIPO vs. 0.81 (0.49 1.03) Received August 21, 2000. Revision received October 10, 2000. Accepted October 24, 2000. Address all correspondence and requests for reprints to: Steven Grinspoon, M.D., Neuroendocrine Unit, Bulfinch 457B, Massachusetts General Hospital, Boston, Massachusetts 02114. E-mail: sgrinspoon@ partners.org. * Supported in part by NIH Grants M01-RR-01066 and P32-DK-07028 and Serono Laboratories, Inc. C, P 0.05 LIPO vs. NONLIPO and LIPO vs. C] and not significantly different between NONLIPO and C. Pulse analysis demonstrated decreased baseline GH [0.08 g/l (0.06 0.21) LIPO vs. 0.19 (0.10 0.32) NONLIPO vs. 0.17 (0.12 0.57) C, P 0.05 LIPO vs. NONLIPO and LIPO vs. C] and GH peak amplitude [1.06 g/l (0.46 1.94) LIPO vs. 2.47 (1.22 3.43) NONLIPO and 2.27 (1.36 4.25) C, P 0.05 LIPO vs. NONLIPO and LIPO vs. C] in the LIPO subjects but no significant difference in pulse frequency. No significant differences were observed between NONLIPO and C for any GH parameter. Insulin-like growth factor-i was not different between the groups. Total body fat (r 0.40, P 0.01) and visceral fat (r 0.58, P 0.0001) correlated inversely with mean overnight GH concentrations in the HIV-infected patients. In a multivariate regression model, controlling for age, body mass index, body fat, and visceral fat, only visceral fat was a significant predictor of mean GH concentrations (P 0.0036, r 2 for model 0.40). These data demonstrate normal GH pulse frequency and insulinlike growth factor-i concentrations but reduced mean GH concentrations, basal GH concentrations, and GH pulse amplitude in patients with HIV lipodystrophy. Increased visceral adiposity is the strongest predictor of reduced GH concentrations in HIV lipodystrophy. Further studies are necessary to determine the clinical significance of reduced GH in patients with HIV lipodystrophy. (J Clin Endocrinol Metab 86: 504 510, 2001) HUMAN immunodeficiency virus (HIV) lipodystrophy syndrome is a recently recognized syndrome that occurs in the majority of HIV-infected patients receiving highly active antiretroviral therapy (HAART) and is characterized by increasing visceral adiposity, loss of sc extremity fat, insulin resistance, and dyslipidemia (1 5). The mechanisms of HIV lipodystrophy are, as yet, unknown and may relate to specific drug effects of protease inhibitors (6), nucleoside reverse transcriptase inhibitors (7), or an interaction between drug and nondrug factors. Furthermore, it remains unknown whether HIV-related lipodystrophy represents a single syndrome or several distinct subsyndromes. Reduced secretion of GH is associated with excess total body fat in obese non-hiv- infected patients (8 10). However, GH dynamics have not previously been characterized in patients with HIV lipodystrophy in whom weight and total body fat are normal, but visceral fat increased in association with reduced sc fat. In this study, we sought to characterize GH dynamics in HIV-infected subjects with lypodystrophy (LIPO), in comparison with age- and body mass index (BMI)- matched HIV-infected nonlipodystrophic subjects (NON- LIPO) and healthy age- and BMI-matched normal control subjects. Our data demonstrate normal GH pulse frequency but significantly reduced mean GH concentrations, basal GH concentrations, and GH pulse amplitude. Increased visceral adiposity was the best predictor of reduced GH concentrations in this population. Subjects Materials and Methods Twenty-one HIV-infected men with LIPO, 20 HIV-infected men without lypodystrophy (NONLIPO), and 20 HIV-negative male control subjects (C) were recruited for the study, between October 1999 and June 2000, and underwent standardized anthropometric testing. HIV status was confirmed by enzyme-linked immunosorbent assay and Western blot testing in all subjects. Subjects were recruited from the multidisci- 504
GH DYNAMICS IN HIV LIPODYSTROPHY 505 plinary HIV practice at the Massachusetts General Hospital and were referred for evaluation of observed changes in fat distribution. Subjects were also recruited from advertisements seeking HIV-infected patients with evidence of fat redistribution. Lipodystrophic subjects were selected based on a waist-to-hip ratio more than 0.95 and a history of significant change in fat distribution in the trunk, extremities, neck, or face. In all LIPO cases, the presence of changes in fat distribution was confirmed by physical examination and scored by a single investigator as severe ( 1.5 on a scale of 0 2) in 1 or more areas. Severe LIPO was scored for changes obvious to the casual observer and mild-to-moderate LIPO for changes noticeable to the patient and confirmed by the single investigator. Objective criteria used in the determination of severe LIPO included, but were not limited to, prominent peripheral venomegaly and a palpable dorsocervical fat pad. In contrast, HIV-positive, NONLIPO subjects were recruited from advertisements seeking HIV-infected men without changes in fat distribution. NONLIPO subjects were selected based on a waist-to-hip ratio less than 0.95 and did not demonstrate significant fat redistribution in any area on physical examination. Lipodystrophy patients were classified as having significant peripheral lipoatrophy if they demonstrated moderate or severe fat loss in the arms or legs. To prevent enrollment of subjects with primary HIV-related wasting, subjects with a BMI of less than 20 kg/m 2 were excluded from all groups. The non-hiv-infected control subjects were in good health, on no medications known to affect GH, with a waist-to-hip ratio less than 0.95. Subjects receiving testosterone, GH, anabolic hormones, glucocorticoid, antidiabetic agents, megestrol acetate, or any other hormone or drug known to affect GH were excluded. Subjects with known diabetes mellitus, hemoglobin level less than 9.0 g/dl, and age more than 60 and less than 18 yr were also excluded. All HIV-infected subjects were on a stable antiretroviral regimen for more than 6 weeks. One subject in the LIPO group was receiving stable thyroid hormone replacement. No other subjects were known to have thyroid disease. Written informed consent was obtained from each subject before testing, in accordance with the Committee on the use of Humans as Experimental Subjects of the Massachusetts Institute of Technology and the Subcommittee on Human Studies at the Massachusetts General Hospital. Study design After completing a screening visit to determine eligibility, subjects returned for an inpatient admission to determine hormonal and body composition parameters. Subjects subsequently returned for a GHRHstimulation test within 3 10 days of the initial visit. Clinical end points Hormonal assessment. The GH insulin-like growth factor (IGF)-I axis was assessed by frequent GH sampling performed every 20 min from 2000 h to 0800 h. Mean overnight GH concentration, basal GH concentration, GH pulse frequency, and GH pulse amplitude (peak height minus the calculated baseline) were determined using the Pulsar computer program. Subjects were not allowed to eat after 1800 h on the day of sampling. The assay coefficients were calculated as a quadratic function using the intraassay sd for this assay: 4.2%, 2.9%, and 2.8% for GH concentrations of 1.4, 6.0, and 12.2 g/l, respectively. The cut-off parameters for accepting peaks 1-, 2-, 3-, 4-, and 5-points wide were 3.63, 2.12, 1.43, 1.00, and 0.70 times the intraassay sd (11). The sensitivity of the assay, defined as the concentration 2 sd above the mean count of the zero standard, was determined to be 0.01 g/l, based on multiple dilutions with a standard sample, and linearity of the assay was confirmed to a GH concentration of 0.05 g/l. Fasting serum IGF-I, insulin, and GH levels were determined in all subjects at 0800 h, immediately before standard arginine stimulation testing [iv administration of arginine hydrochloride, 0.5 g/kg (maximum dose 30 g)], with GH sampling at 30, 60, 90, and 120 min after arginine administration. Subjects also underwent testing with semorelin acetate (GHRH 1 29) (Geref, Serono Laboratories, Inc.), 1 /kg iv, administered at 0800 h, after an overnight fast (at a minimum of no less than 3 days and no more than 10 days after testing with arginine). GH levels were subsequently drawn at 15, 30, 45, 60, 90, and 120 min after semorelin administration. Subjects underwent a standard 75-g oral glucose tolerance test at 0800 h, after a 12-h overnight fast, with insulin and glucose levels drawn at baseline, 30, 60, 90, and 120 min, 1 day before provocative testing with arginine. Hemoglobin A1C (HgbA1C), glucose, cholesterol, low-density lipoprotein (LDL), high-density lipoprotein (HDL), triglyceride, free testosterone, CD4, and viral load were also determined at 0800 h, after a 12-h overnight fast. A 24-h urine specimen was collected for the determination of urine free cortisol. Nutritional assessment and body composition analysis. Weight was determined on the first day of each visit, after an overnight fast. Percent IBW was calculated based on standard height and weight tables (12). Fat and fat-free mass were determined by dual-energy x-ray absorptiometry (DEXA) using a Hologic, Inc.-4500 densitometer (Hologic, Inc., Waltham, MA). The technique has a precision error of 3% for fat and 1.5% for fat-free mass (13). Cross-sectional abdominal computed tomography (CT) scanning was performed to assess the distribution of sc and visceral abdominal fat. A lateral scout image was obtained to identify the level of the L4 pedicle, which served as a landmark for the single-slice image. Scan parameters for each image were standardized (144 cm table height, 80 kv, 70 ma, 2 seconds, 1-cm slice thickness). Fat attenuation coefficients were at 50 HU as described by Borkan et al. (14). Abdominal visceral (VAT) and sc fat (SAT) and the ratio of VAT: total abdominal tissue (TAT) and VAT:SAT were determined and compared between groups. Laboratory methods Serum IGF-I was determined after an acid-alcohol extraction using an RIA kit with an intraassay coefficient of variation of 2.4 3.0% (Corning, Inc. Nichols Institute Diagnostics, San Juan Capistrano, CA). GH was measured by two-site radioimmunometric assay with an intraassay coefficient of variation of 2.8 4.2% (Corning, Inc. Nichols Institute Diagnostics). Insulin was assessed by RIA (Diagnostic Products, Los Angeles, CA), with an intraassay coefficient of variation of 4.7 7.7%. Urine free cortisol was assessed by RIA. Serum free testosterone was measured by RIA kit (Diagnostic Products), with intraassay coefficients of variation of 3.2 4.3%. CD4 cell counts, viral burden, HgbAIc, cholesterol, LDL, HDL, and triglyceride levels were determined using previously published methods (15). Statistical analysis Comparisons were made between the groups [LIPO vs. NONLIPO, LIPO vs. control (C), and NONLIPO vs. C] by the Wilcoxson rank sum test. Chi-square analysis was used to assess group differences for categorical variables. Univariate regression analyses were performed, comparing GH and indices of body fat and composition among all HIVinfected patients. Age, BMI, total body fat, and intraabdominal fat were tested in a multivariate regression model to determine mean overnight GH concentration. Statistical analyses were made using JMP Statistical Database Software (SAS Institute, Inc., Cary, NC). Statistical significance was defined as P 0.05. Results are median plus interquartile range. Results Clinical demographics Age and BMI were similar between the groups (Table 1). CD4 cell count and viral load were not significantly different between the groups. Duration of HIV and total months on antiretroviral therapy were greater in the LIPO, compared with the NONLIPO group (Table 1). The percentage of patients using protease inhibitors (PIs) (90 vs. 45%, P 0.01) and nucleotide reverse transcriptase inhibitors (NRTIs) was different between the LIPO and NONLIPO groups (100 vs. 60%, P 0.01). The overall LIPO score was significantly different between LIPO and NONLIPO groups [5.5 (4.25 6.25) vs. 0.5 (0.5 1.0), respectively, P 0.01]. Body composition Whole-body lean and fat mass, determined by DEXA, were not significantly different between the groups. Regional
506 RIETSCHEL ET AL. JCE&M 2001 Vol. 86 No. 2 TABLE 1. Group comparison by lipodystrophy and HIV status HIV-infected lipodystrophic (n 21) HIV-infected nonlipodystrophic (n 20) Normal controls (n 20) Age and disease status Age (yr) 47 (37-50) 41 (37-44) 43 (37-49) CD4 T-cells (#/mm 3 ) 364 (135-581) 313 (153-434) NA Viral burden (copies/ml) 89 (50-16575) 10475 (73-42925) NA Duration of HIV (yr) 9 (6-10) a 6 (4-8) NA Total months on HAART 36 (31-44) a 4 (0-24) NA Body composition and body habitus BMI (kg/m 2 ) 24.3 (22.2-26.6) 24.4 (23.3-25.9) 24.8 (22.7-26.1) Body fat by DEXA (kg) 14.1 (12.0-18.4) 14.6 (11.0-18.9) 16.7 (13.4-19.4) Lean body mass by DEXA (kg) 57.7 (55.1-63.0) 58.3 (56.8-62.9) 59.0 (54.1-64.3) %LBM by DEXA 77.5 (74.1-80.8) 78.5 (73.3-81.3) 75.9 (72.7-79.5) %Fat by DEXA 19.2 (16.2-22.5) 18.3 (15.1-23.5) 20.8 (17.3-24.2) Trunk Fat:Total Fat by DEXA 0.61 (0.56-0.65) a,b 0.46 (0.41-0.49) 0.49 (0.42-0.52) Extremity Fat:Total Fat by DEXA 0.31 (0.27-0.35) a,b 0.46 (0.44-0.51) 0.44 (0.42-0.51) Trunk Fat:Extremity Fat by DEXA 2.10 (1.63-2.36) a,b 1.01 (0.82-1.11) 1.11 (0.84-1.23) Abdominal area by CT (mm 2 ) 58929 (53423-65378) 53138 (47315-60954) 58344 (50323-64605) Subcutaneous fat area by CT (mm 2 ) 10961 (6791-17668) 11764 (9900-18060) 15936 (12395-23523) Intra-abdominal fat area by CT (mm 2 ) 16124 (11246-19790) a,b 7559 (5134-11201) 8803 (6165-11623) VAT:TAT by CT 0.26 (0.19-0.31) a,b 0.15 (0.11-0.18) 0.14 (0.12-0.20) VAT:SAT by CT 1.37 (0.71-2.44) a,b 0.57 (0.47-0.78) 0.55 (0.41-0.71) Anthropometrics Waist-to-hip ratio 0.98 (0.97-1.02) a,b 0.91 (0.88-0.93) 0.92 (0.86-0.95) Neck circumference (cm) 39.0 (38.0-41.0) 39.5 (37.9-41.0) 38.5 (37.2-40.5) Chest circumference (cm) 102.0 (99.4-106.2) 99.9 (94.8-103.1) 100 (96.5-104.5) Waist circumference (cm) 93.1 (91.0-96.4) a 88.0 (84.3-92.6) 91.5 (84.8-97.1) Hip circumference (cm) 94.0 (90.9-99.1) c 98.6 (93.2-100.9) 98.5 (96.1-104.3) Arm circumference (cm) 31.1 (29.3-34.1) 32.8 (30.9-34.8) 32.7 (31.5-34.5) Leg circumference (cm) 49.1 (46.1-53.0) a,b 53.5 (51.0-56.0) 52.6 (51.0-56.0) Growth hormone dynamics Overnight frequent sampling (q20) Mean GH ( g/l) 0.38 (0.13-0.67) d,c 0.96 (0.53-1.30) 0.81 (0.49-1.03) Baseline GH ( g/l) 0.08 (0.06-0.21) d,c 0.19 (0.10-0.32) 0.17 (0.12-0.57) Mean peak amplitude ( g/l) 1.06 (0.46-1.94) d,c 2.47 (1.22-3.43) 2.27 (1.36-4.25) Frequency (peaks/h) 0.25 (0.17-0.33) 0.33 (0.25-0.42) 0.33 (0.17-0.33) Pulse length (h) 1.42 (1.14-1.75) 1.43 (1.33-2.10) 1.67 (1.10-1.88) Arginine stimulation test AUC [ g/l (120 min)] 131 (45-323) 272 (67-454) 126 (41-354) GHRH stimulation test AUC [ g/l (120 min)] 293 (131-554) a 729 (541-1125) 638 (280-818) IGF-1 ( g/l) 226 (153-283) 225 (206-287) 224 (160-268) Metabolic status [median] HgbA1c (%) 5.4 (5.2-5.7) 5.5 (4.8-5.9) 5.3 (5.1-5.5) 0.054 (0.052-0.057) 0.055 (0.048-0.059) 0.053 (0.051-0.055) Fasting glucose (mg/dl) 90 (85-98) 94 (89-97) 91 (86-95) (mmol/l) 5.00 (4.72-5.44) 5.22 (4.94-5.38) 5.05 (4.77-5.27) OGTT glucose response, AUC [mg/dl (120 min)] 18180 (15195-19553) b 16493 (14760-18518) e 14393 (13140-16549) (mmol/l) 1008.99 (843.32-1085.19) b 915.36 (819.18-1027.75) e 798.81 (729.27-918.47) Fasting insulin ( IU/mL) 11.5 (8.4-21.3) b 8.7 (7.4-12.8) 7.8 (6.6-8.5) (pmol/l) 80 (58-148) b 60 (51-89) 54 (46-59) OGTT insulin response, AUC [ IU/mL (120 min)] 8389 (3784-20539) c 4781 (3856-6338) 4748 (3508-6190) (pmol/l) 58262 (26280-142643) c 33204 (26780-44017) 32975 (24363-42990) HDL (mg/dl) 35 (30-42) d,b 44 (36-50) 47 (39-60) (mmol/l) 0.91 (0.78-1.09) d,b 1.14 (0.93-1.30) 1.22 (1.01-1.55) LDL (mg/dl) 114 (82-143) 92 (76-123) 93 (76-123) (mmol/l) 2.95 (2.12-3.70) 2.38 (1.97-3.19) 2.41 (1.97-3.19) Triglycerides (mg/dl) 181 (124-320) d,b 117 (74-182) e 68 (49-90) (mmol/l) 2.05 (1.40-3.62) d,b 1.32 (0.84-2.06) e 0.77 (0.55-1.02) Cholesterol (mg/dl) 190 (159-220) c 164 (141-198) 165 (136-186) (mmol/l) 4.92 (4.12-5.70) c 4.26 (3.65-5.13) 4.27 (3.52-4.82) Urine free cortisol ( g/24 h) 49 (33-68) 51 (38-71) 52 (41-69) (nmol/24 h) 135 (91-188) 141 (105-196) 143 (113-190) OGTT, Oral glucose tolerance test; LBM, lean body mass; NA, not applicable. For HIV-infected lipodystrophic patients vs. controls, c indicates P 0.05 and b indicates P 0.01. For HIV-infected non-lipodystrophic patients vs. controls, e indicates P 0.01. For HIV-infected lipodystrophic vs. HIV-infected non-lipodystrophic patients, d indicates P 0.05 and a indicates P 0.01. Results are median (interquartile range). trunk fat, determined by DEXA, was increased and extremity fat decreased in the LIPO, compared with the NONLIPO and C subjects. In contrast, no significant differences in either truncal fat or extremity fat were observed between NONLIPO and C subjects. Visceral abdominal fat, the ratio of visceral abdominal fat to total abdominal fat and the ratio of visceral abdominal fat to sc abdominal fat were significantly increased in the LIPO subjects, compared with NONLIPO and C subjects, respectively. No significant differences in visceral fat were observed between NONLIPO and C subjects.
GH DYNAMICS IN HIV LIPODYSTROPHY 507 GH The mean overnight GH concentration, determined from frequent sampling, was decreased in the LIPO subjects, compared with NONLIPO and with C subjects (Fig. 1) but not significantly different in NONLIPO vs. C. Pulse analysis demonstrated decreased baseline GH and peak amplitude but no significant difference in pulse frequency. No significant differences were seen between NONLIPO and C for any GH pulse parameters (P 0.05 for all comparisons). A significant difference in mean GH concentration persisted between LIPO and NONLIPO patients in multivariate models controlling for PI and NRTI use [P 0.02 for group status (LIPO vs. NONLIPO), P 0.35 for PI use, and P 0.83 for NRTI use] or duration of PI and NRTI use [P 0.02 for group status (LIPO vs. NONLIPO), P 0.65 for duration PI, and P 0.42 for duration of NRTI use]. The area under the curve (AUC) of the GH response to GHRH was reduced in the LIPO, compared with NONLIPO, and tended to be lower in the LIPO, compared with C subjects (P 0.08). In contrast, significant differences in AUC GH response to arginine were not seen. GH responses to arginine and GHRH were not different between NONLIPO and C subjects. GH response to GHRH was also compared based on response cutoff parameters for peak GH. Using a cutoff of 3.0 g/l to define a normal GH response to GHRH stimulation, 7 of 21 (33%) LIPO, 0 of 20 (0%) NONLIPO, and 2 of 20 (10%) C subjects demonstrated GH responses below the cutoff (P 0.01, LIPO vs. NONLIPO; and P 0.06, LIPO vs. C). Using a cutoff of 5.0 g/l, the respective fail rates were 48%, 5%, and 20% for LIPO, NONLIPO, and C, respectively (P 0.01, LIPO vs. NONLIPO; and P 0.06, LIPO vs. C). IGF-I levels were not significantly different between the groups and did not correlate with any GH pulse parameters in either the HIV-infected or control subjects. Glucose, lipid, and other hormonal parameters Fasting glucose and HgbA1c were not different between the groups (Table 1). Glucose AUC was increased in the LIPO and NONLIPO, compared with control, subjects. Insulin AUC was increased in the LIPO, compared with NONLIPO and C, groups. No significant difference in insulin AUC was FIG. 1. Median and interquartile range for mean overnight GH concentrations derived from approximately 20-min frequent sampling in HIV LIPO (n 21), HIV NONLIPO (n 20), and 20 age- and BMImatched control subjects. *, P 0.05 vs. NONLIPO;, P 0.05 vs. CONTROL. seen between the NONLIPO and C groups. HDL was decreased and triglyceride and cholesterol increased in the LIPO vs. C subjects. In contrast, only triglyceride was increased in the NONLIPO vs. C subjects. Urine free cortisol levels were not significantly different between the groups. Free testosterone levels were normal in all LIPO subjects and not significantly different from control subjects [16.4 (22.6 4.7) pg/ml, 57 (78 51) pmol/l vs. 17.2 (20.3 14.2) pg/ml, 60 (70 49) pmol/l ; P 0.05]. Relationship of GH to body composition and LIPO score HIV-infected patients. Mean overnight GH concentrations, GH amplitude, baseline GH concentration, and GH responsivity to GHRH were highly inversely related to visceral adiposity determined by CT scan and to total body fat determined by DEXA. For example, total body fat and intraabdominal visceral fat correlated inversely with mean overnight GH (r 0.40, P 0.01 for total body fat; and r 0.58, P 0.0001 for intraabdominal visceral fat, Fig. 2) and, as well, with GH AUC to GHRH (r 0.35, P 0.02 for total body fat; and r 0.62, P 0.0001 for intraabdominal visceral fat, Fig. 2). In contrast, GH pulse frequency was not significantly associated with visceral adiposity, BMI, or total body fat (Table 2). Among the subjects in the LIPO group, GH concentrations were not related to the percent extremity fat by DEXA (r 0.26, P 0.25), arm (r 0.09, P 0.69), or leg circumference (r 0.09, P 0.70). Mean GH concentrations were not significantly different between those with significant peripheral lipoatrophy [n 15, GH 0.44 (0.13 0.60) g/l] and those with mild lipoatrophy [n 6, GH 0.22 (0.10 1.08) g/l] (P 0.53 for the comparison of GH). Among all the HIV-infected subjects, GH pulse amplitude (r 0.35, P 0.02), baseline GH (r 0.32, P 0.04), and mean GH concentrations (r 0.42, P 0.007) were inversely correlated with overall LIPO score summing changes in the neck, face, extremities, and abdomen. In a multivariate regression model predicting mean overnight GH among the HIV-infected patients (P 0.0008 for whole-model test, R 2 0.40), controlling for age, BMI, body fat, and visceral fat, only visceral fat was a significant predictor of mean GH levels (P 0.0036, Table 3). The parameter estimate for the effect of visceral fat on GH demonstrated that a 1500-mm 2 increase in intraabdominal fat predicts a 0.0825- g/l decrease in mean overnight GH. Control subjects. In a multivariate regression model, controlling for age, BMI, total body fat, and visceral fat, only BMI (P 0.03) was a significant predictor of mean GH (Table 3). Discussion HIV lipodystrophy is characterized by significant excess visceral adiposity and reduced sc fat mass in association with insulin resistance and dyslipidemia (1, 4, 16 19). The mechanism of HIV lipodystrophy is not known, but it may relate to specific effects of antiretroviral medications, nondrug factors, or an interaction between the two. Furthermore, it is unknown whether the syndrome represents a single pathophysiological entity with varying phenotypic expression vs. separate overlapping syndromes. In non-hiv-infected pa-
508 RIETSCHEL ET AL. JCE&M 2001 Vol. 86 No. 2 tients, obesity is associated with reduced GH secretion (8 10). In such patients, the degree of body fatness is associated inversely with indices of GH secretion. HIV lipodystrophy is unique, in that patients do not have severe generalized obesity but instead demonstrate fat redistribution, such that total fat is not increased. GH dynamics have not previously been assessed in HIV lipodystrophy, and the relationship between indices of fat redistribution and GH secretion remain unknown in this syndrome. Our data demonstrate reduced mean overnight and basal GH concentrations, with preserved GH pulse frequency. Increased visceral adiposity is a highly significant predictor of GH in this population. In prior studies of HIV-infected patients with the wasting syndrome, IGF-I concentrations were shown to be reduced in association with increased GH levels in men with AIDS wasting, suggesting GH resistance in association with severe undernutrition (20). In 1996, Heijligenberg et al. demonstrated normal GH pulse amplitude, peak interval, and mean GH concentrations in asymptomatic HIV-infected patients without significant weight loss, compared with age and BMImatched control subjects (21). The patients reported by Heijligenberg et al. were receiving only zidovudine and were studied in 1996, before the recognized occurrence of lipodystrophy. The current study was performed among less severely immunocompromised patients reporting fat redistribution, FIG. 2. Relationship of intraabdominal fat to mean overnight GH concentration, and AUC for GH response to GHRH testing, in HIV-infected subjects. the majority of whom were receiving highly active antiviral therapy. Fat redistribution was verified by a single investigator in all subjects. Patients were not chosen for this study based on the evidence of lipoatrophy alone but rather demonstrated combined lipoatrophy and visceral adiposity, as demonstrated by the significantly increased abdominal visceral fat, visceral-to-sc fat ratio, and reduced leg and arm circumference measurements. Because we did not recruit patients with lipoatrophy alone, our results cannot be generalized to this subpopulation of patients with HIV lipodystrophy. However, we did not observe any relationship between the loss of sc fat and GH, in contrast to the robust relationship observed between increased visceral fat and GH (see below). Age and BMI-matched HIV-negative subjects were used as a comparison group to control for the effects of weight and overall adiposity on GH secretion. Furthermore, we included age- and BMI-matched NONLIPO HIV-infected subjects to control for the potential effects of HIV infection on the GH axis. Although there was no difference in overall adiposity between the groups, visceral fat mass was significantly increased, and sc fat mass reduced, in the comparisons between HIV-infected LIPO and NONLIPO subjects and between HIV LIPO and normal control subjects. Similarly, lipid levels, insulin, and other metabolic parameters were significantly worse in the LIPO group. In contrast, visceral fat mass, in- TABLE 2. Univariate correlations with BMI, total body fat, intraabdominal visceral fat, abdominal sc fat, and percent extremity fat in HIV-infected subjects (n 41) BMI Total body fat Intraabdominal sc fat Intraabdominal visceral fat Percent Extremity Fat r-value P-value r-value P-value r-value P-value r-value P-value r-value P-value GH frequency 0.27 0.08 0.21 0.19 0.13 0.42 0.29 0.07 0.15 0.36 GH amplitude 0.22 0.17 0.34 0.03 0.23 0.15 0.52 0.001 0.26 0.10 GH interpulse length 0.14 0.39 0.04 0.81 0.05 0.76 0.17 0.30 0.19 0.24 GH pulse length 0.10 0.53 0.19 0.23 0.17 0.28 0.10 0.55 0.03 0.84 GH baseline 0.28 0.07 0.37 0.02 0.24 0.13 0.48 0.001 0.18 0.25 Mean overnight GH 0.22 0.16 0.40 0.01 0.25 0.12 0.58 0.0001 0.28 0.07 Peak GH to arginine 0.36 0.02 0.35 0.03 0.16 0.32 0.44 0.004 0.28 0.08 Response to arginine (AUC) 0.34 0.03 0.31 0.05 0.13 0.41 0.40 0.009 0.25 0.11 Peak GH to GHRH 0.28 0.07 0.35 0.03 0.14 0.39 0.62 0.0001 0.57 0.0001 Response to GHRH (AUC) 0.28 0.08 0.35 0.02 0.14 0.38 0.62 0.0001 0.55 0.0002 BMI 0.65 0.000 0.60 0.001 0.37 0.02 0.03 0.84 Body fat by DEXA 0.6491 0.0001 0.91 0.001 0.36 0.02 0.13 0.42
GH DYNAMICS IN HIV LIPODYSTROPHY 509 TABLE 3. Multivariate correlation with mean overnight GH dices of total and regional fat, and other metabolic parameters were not different in the HIV-infected NONLIPO subjects, compared with the control subjects. Mean overnight GH concentrations, basal GH concentrations, and GH pulse amplitude were reduced, whereas GH pulse frequency was not different in the LIPO subjects, in comparison to either HIV NONLIPO or C subjects. Similarly GH responses to GHRH were reduced in the LIPO subjects, in comparison with C subjects. Using peak GH stimulatory response cutoffs of 3.0 g/l and 5.0 g/l to GHRH (22), 33% and 48% of LIPO patients did not achieve adequate GH responses to GHRH, respectively. Thirty-three percent of LIPO patients demonstrated peak GH responses of less than 3.0 g/l on both arginine and GHRH testing. Although our data do not establish whether such patients are GH deficient, they suggest significantly abnormal GH stimulatory responses in a substantial proportion of patients with LIPO. Our data stand in partial contrast to the data obtained in studies of non-hiv-infected obese patients, in which mean GH concentrations, GH pulse frequency, and GH responsivity to GHRH are reduced and highly correlated with obesity indices (8 10). In contrast, the lipodystrophic subjects in this study were not significantly overweight and were of similar weight to the control groups. Although mean and basal GH concentrations were reduced, pulse frequency was preserved in patients with HIV LIPO. We used a highly sensitive GH assay (sensitivity, 0.01 g/l), to detect discrete low amplitude pulses. Visceral adiposity was the most highly significant predictor of reduced GH concentrations in the HIV-infected patients and was independent of age, BMI, and total body fat. BMI and total body fat correlated less well with GH indices and were not independent predictors of GH secretion controlling for visceral adiposity. In contrast, the mean overnight GH level was most strongly predicted by BMI, and not by visceral fat, in the HIV-negative control subjects who had normal visceral fat. A major unresolved question is whether the strong relationship of GH to visceral fat, observed in the LIPO subjects, is unique to HIV LIPO or similar to a pattern that might be seen in viscerally obese, non HIV-infected patients. To our knowledge, prior studies have not specifically examined the relationship between visceral adiposity and GH in non-hiv-infected patients, and this is an important area of future research. IGF-I levels were not different in the three age- and BMImatched groups. Marin et al. (23) previously demonstrated that IGF-I was reduced in association with visceral adiposity Univariate r Univariate P Multivariate P HIV-infected subjects (N 41, r 2 0.40) Age (yr) 0.26 0.10 0.81 BMI (kg/m 2 ) 0.22 0.16 0.31 Body fat by DEXA (kg) 0.40 0.01 0.06 Intra-abdominal fat by CT (mm 2 ) 0.58 0.0001 0.004 Normal control subjects (N 20, r 2 0.43) Age (yr) 0.15 0.52 0.32 BMI (kg/m 2 ) 0.53 0.02 0.03 Body fat by DEXA (kg) 0.38 0.10 0.11 Intra-abdominal fat by CT (mm 2 ) 0.41 0.07 0.13 in obese overweight men. In contrast, the BMI of patients and controls in this study was normal, and we show no significant relationship of IGF-I to visceral fat mass or overall adiposity in either the HIV-infected or control subjects. IGF-I did not correlate with GH pulse parameters. Low GH levels are seen in association with normal or high IGF-I levels in more generalized obesity, suggesting a relatively greater sensitivity to circulating GH (24). Normal IGF-I levels suggest that viscerally obese subjects with HIV LIPO do not have classically defined GH deficiency. However, IGF-I is a relatively poor marker for GH deficiency, because normal IGF-I levels often occur in patients with adult-onset GH deficiency (11). Further studies are necessary to determine whether there is an enhanced sensitivity of IGF-I to GH in HIV lipodystrophy. The pattern of reduced basal GH and reduced GH pulse amplitude in association with normal GH pulse frequency suggests increased somatostatin tone or decreased GHRH secretion. Increased central somatostatin tone has been suggested in prior studies of obese non-hiv-infected patients in which pyridostigmine (25 27) and/or arginine (10) potentiated responses to GHRH, via a postulated inhibition of somatostatin tone. Alternative hypotheses for abnormal GH secretion in obesity include reduction in the pituitary pool of GH available for release (28) and also to reduced hypothalamic GHRH (29). Patients in this study were sampled overnight, and it is possible that a reduced pulse frequency might have been seen during daytime sampling, when endogenous somatostatin tone is normally highest. In this study, we demonstrate that GH dynamics are normal in NONLIPO HIV-infected men but reduced in association with increased visceral adiposity in HIV lipodystrophic patients. Taken together, our data suggest a primary abnormality in fat redistribution, resulting in secondary changes in GH secretion in HIV lipodystrophy. Less likely, but also possible, is a primary abnormality in GH secretion, with increased visceral obesity as a secondary phenomenon. However, this mechanism would not explain the other features of HIV lipodystrophy, such as peripheral fat loss. In contrast to the highly significant relationships between GH and visceral fat, no relationships were seen between GH pulse parameters and peripheral fat loss, or in a subclassification based on peripheral lipoatrophy. Furthermore, a direct effect of HIV medications is unlikely, because the differences in GH levels remained significant between LIPO and NONLIPO groups, controlling for PI and NRTI use. Reduced GH concentrations could further contribute to a
510 RIETSCHEL ET AL. JCE&M 2001 Vol. 86 No. 2 number of the metabolic abnormalities, including insulin resistance and dyslipidemia, seen in HIV lipodystrophy. Prior studies in non-hiv-infected patients with abdominal obesity suggest a reduction in visceral fat and overall improvement in glucose uptake, on euglycemic hyperinsulinemic clamp testing, in response to low-dose GH (30). Further studies are necessary to determine whether restoration of physiologic GH concentrations may improve related metabolic parameters in HIV lipodystrophy. These data are the first to suggest a significant reduction in GH secretion in men with HIV lipodystrophy. Visceral adiposity seems to be the most significant predictor of reduced GH concentrations in this population. Further studies of the mechanisms of reduced GH concentration and potential gender differences in GH secretion in HIV lipodystrophy are needed. Future studies are necessary to determine the effects of restoration of normal GH concentrations on fat redistribution and metabolic parameters in HIV lipodystrophy. Acknowledgments We acknowledge the nursing and bionutrition staffs of the General Clinical Research Center of the Massachusetts General Hospital for their dedicated patient care; Dr. Anne Klibanski for her helpful suggestions in the preparation of the manuscript; and Drs. Nesli Basgoz, Paul Sax, and Benjamin Davis for help with patient recruitment. References 1. Carr A, Samaras K, Burton S, et al. 1998 A syndrome of peripheral lipodystrophy, hyperlipidemia and insulin resistance in patients receiving protease inhibitor therapy. AIDS. 12:F51 F58. 2. Carr A, Cooper DA. 1998 Lipodystrophy associated with an HIV-protease inhibitor. N Engl J Med. 339:1296. 3. Carr A, Samaras K, Thorisdottir A, Kaufman GR, Chisolm DJ, Cooper DA. 1999 Diagnosis, prediction, and natural course of HIV-1 protease-inhibitor associated lipodystrophy, hyperlipidemia, and diabetes mellitus: a cohort study. Lancet. 353:2093 2099. 4. Miller KD, Jones E, Yanovski JA, Shankar R, Feuerstein I, Fallon J. 1998 Visceral abdominal-fat accumulation associated with use of indinavir. Lancet. 351:871 875. 5. Vigouroux C, Gharakhanian S, Salhi Y, et al. 1999 Diabetes, insulin resistance and dyslipidemia in lipodystrophic HIV-infected patient on highly active antiretroviral therapy (HAART). Diabete Metab. 25:225 232. 6. Mulligan K, Grunfeld C, Tai VW, et al. 2000 Hyperlipidemia and insulin resistance are induced by protease inhibitors independent of changes in body composition in patients with HIV infection. J Acquir Immune Defic Syndr. 23:35 43. 7. Carr A, Miller J, Law M, Cooper DA. 2000 A syndrome of lipoatrophy, lactic acidaemia and liver dysfunction associated with HIV nucleoside analogue therapy: contribution to protease inhibitor-related lipodystrophy syndrome. AIDS. 14:F25 F32. 8. Veldhuis JD, Liem AY, South S, et al. 1995 Differential impact of age, sex steroid hormones, and obesity on basal versus pulsatile growth hormone secretion in men as assessed in an ultrasensitive chemiluminescence assay. J Clin Endocrinol Metab. 80:3209 3222. 9. Veldhuis JD, Iranmanesh A, Ho KK, Waters MJ, Johnson ML, Lizzaralde G. 1991 Dual defects in pulsatile growth hormone secretion and clearance subserve the hyposomatotropism of obesity in man. J Clin Endocrinol Metab. 72:51 59. 10. Ghigo E, Procopio M, Boffano GM, et al. 1992 Arginine potentiates but does not restore the blunted growth hormone response to growth-hormone releasing hormone in obesity. Metabolism. 41:560 563. 11. Baum HBA, Biller BMK, Katznelson L, et al. 1996 Assessment of growth hormone (GH) secretion in men with adult-onset GH deficiency compared with that in normal men a clinical research center study. J Clin Endocrinol Metab. 81:84 92. 12. Metropolitan Life Insurance Company. 1983 Height and weight tables. Stat Bull Metrop Insur Co. 64:2 9. 13. Mazess RB, Barden HS, Bisek JP, Hanson J. 1990 Dual-energy x-ray absorptiometry for total-body and regional bone-mineral and soft-tissue composition. Am J Clin Nutr. 51:1106 1112. 14. Borkan GA, Gerzof SG, Robbins AH, Silbert CK, Silbert JE. 1982 Assessment of abdominal fat content by computed tomography. Am J Clin Nutr. 36:172 177. 15. Jordan D, F1ood JG, Laposata M, Lewandrowski KB. 1992 Case records of the Massachusetts General Hospital. Weekly clinicopathological exercises. Normal reference laboratory values. N Engl J Med. 327:718 724. 16. Mynarcik DC, McNurlan MA, Quick J, Fuhrer J, Stiegbigel R, Gelato MC, Serum markers of metabolic dysregulation in HIV disease. Proc of the 81st Annual Meeting of The Endocrine Society, San Diego, CA, 1999; p. 96. 17. Periard D, Telenti A, Sudre P, et al. 1999 Atherogenic dyslipidemia in HIVinfected individuals treated with protease inhibitors. Circulation. 100:700 705. 18. Walli RK, Herfort O, Michl GM, et al. 1998 Treatment with protease inhibitors associated with peripheral insulin resistance and impaired oral glucose tolerance in HIV-1 infected patients. AIDS. 12:F167 F173. 19. Saint-Marc T, Partisani M, Poizot-MArtin I, et al. 1999 A syndrome of peripheral fat wasting (lipodystrophy) in patients receiving long-term nucleoside analogue therapy. AIDS. 13:1659 1667. 20. Grinspoon S, Corcoran C, Lee K, et al. 1996 Loss of lean body mass and muscle mass correlates with androgen levels in hypogonadal men with acquired immunodeficiency syndrome and wasting. J Clin Endocrinol Metab. 81:4051 4058. 21. Heijligenberg R, Sauerwein HP, Brabant G, Endert E, Hommes MJ, Romijn JA. 1996 Circadian growth hormone secretion in asymptomatic human immune deficiency virus infection and acquired immunodeficiency syndrome. J Clin Endocrinol Metab. 81:4028 4032. 22. Bing-You RG, Bigos ST, Oppenheim DS. 1993 Serum growth hormone responses to growth-hormone releasing hormone in non-obese and obese adults with hypopituitarism. Metabolism. 42:790 794. 23. Marin P, Kvist H, Lindstedt G, Sjostrom L, Bjorntorp P. 1993 Low concentrations of insulin-like growth factor-i in abdominal obesity. Int J Obes. 17:83 89. 24. Hochberg Z, Hertz P, Colin V, et al. 1992 The distal axis of growth hormone (GH) in nutritional disorders: GH-binding protein, insulin-like growth factor-i (IGF-I), and IGF-I receptors in obesity and anorexia nervosa. Metabolism. 41:106 112. 25. Cordido F, Dieguez C, Casanueva FF. 1990 Effect of central cholinergic neurotransmission enhancement by pyridostigmine on the growth hormone secretion elicited by clonidine, arginine, or hypoglycemia in normal and obese subjects. J Clin Endocrinol Metab. 70:1361 1370. 26. Cordido F, Casanueva F, Dieguez C. 1989 Cholinergic receptor activation by pyridostigmine restores growth hormone (GH) responsiveness to GH-releasing hormone administration in obese subjects: evidence for hypothalamic somatostatinergic participation in the blunted GH release of obesity. J Clin Endocrinol Metab. 68:290 293. 27. Ghigo E, Mazza E, Corias A. 1989 Effect of cholinergic enhancement by pyridostigmine on growth hormone secretion in obese adults and children. Metabolism. 38:631 633. 28. Renier G, Gaudreau P, Deslauriers N. 1989 In vitro and in vivo growth hormone responsiveness to growth hormone-releasing factor in male and female Zucker rats. Neuroendocrinology. 50:454 459. 29. Tannenbaum GS, Lapointe M, Gurd W. 1990 Mechanisms of impaired growth hormone secretion in genetically obese Zucker rats; roles of growth hormonereleasing factor and somatostatin. Endocrinology. 127:3087 3095. 30. Johannson G, Marin P, Lonn L, et al. 1997 Growth hormone treatment of abdominally obese men reduces abdominal fat mass, improves glucose and lipoprotein metabolism and reduces diastolic blood pressure. J Clin Endocrinol Metab. 82:727 734.