| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
From the Clinical Research Centers |
Department of Internal Medicine, University of New Mexico School of Medicine (D.L.W., R.N.B.); General Clinical Research Center, University of New Mexico (C.R.Q.); and Department of Internal Medicine, New Mexico Veterans Affairs Health Care System (R.D.), Albuquerque, New Mexico 87131; and Center for Biomathematical Technology, General Clinical Research Center, and Department of Internal Medicine, University of Virginia (J.D.V.), Charlottesville, Virginia 22908
Address all correspondence and requests for reprints to: Debra L. Waters, Ph.D., University of New Mexico School of Medicine, 2701 Frontier Plaza NE, Surge Building, Room 215, Albuquerque, New Mexico 87131. E-mail: dwaters{at}salud.unm.edu
 |
Abstract |
|---|
 Top Abstract IntroductionExperimental SubjectsMaterials and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionExperimental SubjectsMaterials and MethodsResultsDiscussionReferences |
|---|
The secretion of GH and other pituitary hormones is subject to complex feedback- and feedforward-dependent neuroregulation (7). There are also evident gender distinctions in GH neuroregulation, inferentially due to differences in the balance between respective stimulatory and inhibitory input by GHRH and somatostatin (8). In anorexia nervosa, increased GH pulse frequency and elevated integrated serum GH concentrations have been recognized, putatively reflecting a reduction of hypothalamic somatostatin tone and/or increased GHRH discharges (9, 10, 11). Whether these characteristics underlie the presumptively disrupted neuroregulation of GH secretion in women with nonanorectic amenorrhea is unknown (6, 12, 13).
The purpose of this study was to evaluate nocturnal GH secretion and the GH response to an exercise stimulus along with insulin-like growth factor (IGF)-binding protein (IGFBP) concentrations in amenorrheic and eumenorrheic athletes (EA) without eating disorders. To this end, we compared two cohorts of matched, strenuously training women who differed principally in menstrual cyclicity.
|
Experimental Subjects |
|---|
|
TopAbstractIntroductionExperimental SubjectsMaterials and MethodsResultsDiscussionReferences |
|---|
|
Materials and Methods |
|---|
|
TopAbstractIntroductionExperimental SubjectsMaterials and MethodsResultsDiscussionReferences |
|---|
Dietary records were collected for 30 days before the in-patient admission. Nutritional intake was determined by 24-h recall from data returned weekly to the dietician. Dietary intakes were analyzed by Nutritionist III software for kilocalories and grams of protein, fat, and carbohydrate consumed (N2 Computing, Salem, OR). Percent body fat and lean body mass were determined by dual energy x-ray absorptiometry (QDR-1000/W, Hologic, Inc., Waltham, MA) using Lunar Corp. (Madison, WI) Scanning Analysis (version 3.6, 1992).
Postexercise/nocturnal blood sampling
Subjects were admitted to the General Clinical Research Center (GCRC) for 2 nights and were provided meals. EA volunteers were admitted during the early follicular phase of the menstrual cycle. AA were admitted on alternate days during 1 month. The first night was used for acclimation to the GCRC. The second night was the nocturnal GH study. After the first night of admission and to mimic typical energy expenditure during a 24-h period, all athletes performed a 50-min submaximal exercise bout (70% VO2max) using their customary exercise modality (i.e. cycling or running). Blood samples were drawn every 10 min during the exercise test and every 5 min during 20 min of recovery. Immediately after the submaximal exercise, the subjects were provided breakfast, given a take-home lunch, and discharged until 1800 h. At 1800 h, the subjects returned to the GCRC and were fed a caffeine-free, balanced meal. At 2100 h, the subjects were allowed only water, and the iv heparin lock was converted to long tubing, which was run into an adjoining room. Lights were put out at 2200 h, and 2-mL blood samples were drawn every 20 min from 2300–0700 h. A one-way mirror in the room was used to monitor sleep.
Hormonal measurements
Blood samples for GH, IGF-I, IGFBP-1, IGFBP-3, and estradiol determinations were centrifuged to separate serum, which was frozen for subsequent analysis. GH and estradiol concentrations were analyzed using a double antibody method (Diagnostic Products, Los Angeles, CA). Intra- and interassay variabilities for GH were 2.8% and 5.3%, respectively, with 0.2 µg/L sensitivity. Intra- and interassay variabilities for estradiol were 7.0% and 8.1%, respectively, with 29 pmol/L sensitivity. IGF-I was analyzed using the RIA acid-ethanol extraction method (Nichols Institute Diagnostics, San Juan Capistrano, CA). Sensitivity was 60 µg/L. Intra- and interassay variabilities for IGF-I were 2.4% and 5.2%, respectively. IGFBP-3 was analyzed using RIA (Nichols Institute Diagnostics). Sensitivity was 0.0625 µg/mL, with intra- and interassay variabilities of 3.8% and 6.3%, respectively. IGFBP-1 was analyzed by immunoradiometric assay (Diagnostics Systems Laboratories, Inc., Webster, TX). Sensitivity was 33 µg/L, with intra- and interassay variabilities of 6.0% and 4.6%, respectively. Solid phase RIA was used to determine plasma progesterone. Interassay variability for progesterone was 6.1%, and sensitivity was 0.28 nmol/L serum. Solid phase RIA was used to determine total T4 (TT4; Nichols Institute Diagnostics), immunoradiometric assay was used to assess TSH (Nichols Institute Diagnostics), and the radioactive T3 resin saturation technique was used to determine T3U (Organon Technica, Durham, NC).
Statistical analyses
Demographics. A two-tailed t test was used to compare differences between the AA and EA groups in age, body composition (percent fat), years of training, VO2max, progesterone values, eating attitude test scores (EAT-26), and thyroid function (T4, T3U, and TSH).
Distributional analysis. A modification of the time occupancy analysis, as described by Matthews et al. (14, 15, 16, 17), was used to analyze nocturnal GH secretory patterns. Time occupancies are the percentage of time spent by data in the series at given concentrations. Modification of the time occupancy analysis allowed us to use a 2-bin histogram instead of a complete time occupancy distribution. We applied Fisher’s exact test directly to the resultant distribution instead of t tests on the percentiles of the distributions. This modification also made logarithmic transformation of GH concentrations unnecessary.
Monte Carlo simulation. In a Monte Carlo simulation, statistical power was estimated based on 1000 simulated samples, each consisting of 5 time series representing EA and 5 representing AA. The type I error was estimated by 1000 samples consisting of 2 groups of 5 time series, both representing EA.
Measures of power: receiver operator characteristics (ROC)
curve. We first optimized the selection of a cut-score for trough
GH concentrations, as this is the only adjustable parameter in time
occupancy analysis. Optimization was based on an ROC-like analysis,
which plots sensitivity (1 - 
in the hypothesis setting)
vs. 1 - specificity (ß) as parameterized by the
cut-score. The cut-score is the threshold below which the values are in
a trough. We estimated sensitivity and specificity in two separate
simulations. Each threshold was applied to the resultant data to obtain
the ROC-like curve, which was not necessarily monotonically increasing.
The ROC-like tables contain rows representing either (AA, EA) or (EA,
EA) and columns representing trough/nadir values.
Deconvolution analysis. A pulsatile model of hormone secretion and clearance was assumed, in which the plasma concentration of GH at any given instant is related to four simultaneous secretory and kinetic features of interest: 1) location of secretory bursts, 2), mass of secretory bursts, 3) duration of secretory bursts, and 4) endogenous hormone half-life. The basal rate of GH secretion was calculated simultaneously to reflect the lowest 5% sample GH concentrations in any given profile. A distinct secretory burst was approximated algebraically as a random (Gaussian) distribution of instantaneous secretory rates. Fitted burst mass values were distinguishable from zero based on 95% statistical confidence intervals. A convolution integral was used to relate the serum GH concentrations to the foregoing specific measure of pulsatile GH secretion and removal. Parameters were quantified by iterative nonlinear least squares parameter estimation. The disappearance function for GH was modeled as a one-component exponential decay function for each subject, assuming that the half-life and distribution volume of GH were approximately constant in each individual throughout the nocturnal period (10).
The above deconvolution analysis estimated 1) pulsatile production rate product of mass per burst and number of bursts, 2) secretory burst half-duration, 3) mass secreted per burst, 4) number of bursts, and 5) t1/2 of GH disappearance. Analyses were carried out blinded to the subject group assignment (10).
Approximate entropy (ApEn). ApEn, as described by Pincus (8), was used to analyze the orderliness/disorderliness of GH profiles. Normalized ApEn is a scale- and model-independent statistic for assessing the regularity of time-series data. ApEn is a distributional method that allows consideration of the joint distribution of adjacent points (m > 1) instead of marginal time occupancy distribution. A single nonnegative number is thereby assigned to a time series to quantify the serial orderliness or regularity of the data. ApEn specifically measures the logarithmic likelihood that patterns of data length (m) that are similar remain similar within a tolerance (r) on the next incremental (m + 1) comparison (8). Smaller ApEn values indicate a greater likelihood of successive comparisons remaining close and imply greater regularity, and vice versa. ApEn is stable to repeated small change in noise characteristics or infrequent large data artifacts (18).
|
Results |
|---|
|
TopAbstractIntroductionExperimental SubjectsMaterials and MethodsResultsDiscussionReferences |
|---|
The characteristics of the subjects are presented in Table 1
. No differences were noted in age,
years of training, current training frequency or duration,
V02max, percent body fat, thyroid status, or age
at menarche between the groups. Significant differences emerged for
progesterone and unstimulated mean GH concentrations and eating
attitude test scores (EAT-26). EAT-26 scores were within the normal
range for both groups (<20), but were significantly higher in the AA.
Values for TT4, T3U, and
TSH were within normal ranges and not significantly different between
amenorrheics and eumennorheics, respectively
(TT4, 75.9 ± 12.87 vs. 79.9
± 25.74 nmol/L; T3U, 0.63 ± 0.03
vs. 0.63 ± 0.04 nmol/L; TSH, 2.8 ± 2
vs. 2.9 ± 0.9 mU/L).
|
Distributional analysis. The modified time occupancy analysis
of the trough GH concentration revealed that AA spent a significantly
greater portion of time at a higher GH trough concentration than EA
(P < 0.001). Time occupancy analysis of nocturnal GH
concentration data via Monte Carlo simulations (see Materials and
Methods) was performed to estimate the type I error and power in
this approach. Figure 1 is presented as
an ROC-like curve parameterized by the threshold definition of troughs.
For a type I error of 0.01, a threshold of 1.8 µg/L (equal
to the 25th percentile of the time occupancy distribution for AA in the
simulation data) had 79% power. Amenorrheic women had a 39% time
occupancy at or below the 1.8 µg/L threshold, whereas eumenorrheic
individuals had 65% time occupancy at 1.8 µg/L. Using a threshold of
1.0 µg/L for a type I error of 0.04, statistical power was 88%.
There was a negative correlation between high percent time occupancy of
low trough values and mean GH concentrations (r = -0.73;
P = 0.02).
|
illustrates pulsatile profiles of GH release in each volunteer. Table 2 demonstrates the differences between
the groups based on deconvolution analysis. GH secretory burst
half-duration and mass of GH secreted per peak were significantly lower
in AA (P = 0.04 and P = 0.04,
respectively). The number of peaks and GH half-life were significantly
greater in AA (P = 0.03 and P = 0.04,
respectively). The dispersion of key measures of pulsatile GH secretion
(mass or frequency) among the 10 volunteers studied is given in Fig. 3, A and B. There was a negative
relationship between mass per burst and half-life (r = -66;
P = 0.04). Basal secretion, total mass (pulsatile
production rate), and mean and integrated GH concentrations were not
significantly different between groups (Table 2).
|
|
|
], indicating more irregular GH
release. A post-hoc correlation analysis between methods
revealed that ApEn was positively related to the number of peaks
(r = 0.81; P = 0.004) and was negatively
correlated to a high percent time occupancy of low trough values
(r = -0.74; P = 0.01). Nocturnal serum IGF-I, IGFBP-3, and IGFBP-1 concentrations. There were no significant differences between the groups in baseline parameters or peak and integrated area under the curve (AUC) for IGF-I, IGFBP-1, and IGFBP3 concentrations. GH peak and AUC values did not differ, and baseline GH concentrations showed a trend toward significance (P = 0.06). For both groups, peak IGFBP-1 was significantly correlated to peak GH regardless of time (r = 0.58; P = 0.03). IGFBP-1 was significantly correlated to GH AUC (r = 0.39; P = 0.05).
Exercise data: deconvolution and classical parameter analysis of serum GH concentrations after exercise
Figure 4 summarizes contrasts in the
GH response to exercise. The baseline GH concentration at the start of
exercise was significantly greater in AA than EA (7.9 ± 6.7
vs. 0.94 ± 0.7 µg/L; P = 0.04) and
was significantly correlated to nocturnal GH concentrations
(r2 = 0.65; P = 0.04). Adjusting
for differences in baseline GH concentrations, GH AUC and GH peak
values were significantly lower in AA [AUC, 1571 ± 910
vs. 406 ± 230 µg/L (P = 0.01); peak,
33 ± 15 vs. 12 ± 7 µg/L (P =
0.02)]. Deconvolution analysis revealed that GH mass per burst and
total GH production rate were significantly less in AA [mass, 26
± 7 vs. 120 ± 17 µg/L (P = 0.003);
production rate, 30 ± 9 vs. 124 ± 41
µg/L·min (P = 0.04)]. Figure 5 demonstrates a reversal in the normal
relationship between GH AUC and peak IGF-I during exercise in AA
compared with EA (P = 0.004).
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionExperimental SubjectsMaterials and MethodsResultsDiscussionReferences |
|---|
The increased interpeak basal (trough) serum GH concentration in
anorexia nervosa ensued from augmentation of the basal (nonpulsatile)
GH secretory rate rather than from a prolongation of the GH half-life
(10). In contrast, in AA we observed an apparent 60%
increase in GH half-life, whereas calculated basal (nonpulsatile) GH
secretion estimates were unchanged. Although GH half-life is correlated
inversely with BMI or estradiol in some analyses (19), our
cohorts were not distinguishable in their dual energy x-ray
absorptiometry-determined percent body fat or their serum estradiol
concentrations. Other studies show that the elimination properties of
GH are controlled by the rate of hormone entry into the circulation
(20). The prolonged half-life in AA individuals correlated
negatively with GH secretory burst mass (i.e. smaller
peaks were associated with delayed GH removal). This observation may be
relevant, because kinetic analyses of the impact of GH-binding protein
(GHBP) t on GH elimination rates (21) predict that GH
peaks less than the circulating GHBP capacity and will show half-lives
more closely related to the off-rate of GH’s dissociation from its
plasma binding protein (
20 min) than to free GH (approximate
half-life, 3–7 min). Regardless of the kinetic mechanism involved, the
somewhat longer GH half-life in AA could contribute to the relative
hypersomatotropism observed here.
A dual pathophysiological mechanism of augmented pulsatile GH secretion has been postulated in anorexia nervosa, in which weight loss elicits greater GH secretory burst mass, and hypoestrogenemia results in increased GH pulse frequency (6). This inference has not been tested directly. However, in postmenopausal women, estrogen administration stimulates GH pulse mass without altering burst frequency (22). In contrast, GH secretory burst mass was reduced by 300% in our women with AA. Pulse mass is believed to reflect joint somatostatinergic and GHRH inputs, whereas pulse frequency may be driven primarily by variations in somatostatin output (7). Frequent low mass bursts of GH, as observed in AA, may influence physiological responses to GH. For example, GH secretory patterns in the rat control body growth, GH receptor and GHBP levels, and liver and muscle gene expression (23). Inferentially, GH peaks seem to be critical to stimulate linear growth, whereas GH trough values may especially impact body composition and metabolic parameters (15). Indeed, sustained low levels of GH can exert powerful (lipolytic) effects on fuel metabolism, resulting in protein and glucose sparing (24). In addition, pulses of GH are coordinated with GH receptor turnover (25), which may be essential for other tissue responses (14, 16). If sustained (nonpulsatile) trough GH concentrations secondarily affect GH signaling (16, 26, 27), then the high trough GH concentrations in AA patients could account for their inversion of the normal (EA) reciprocal relationships between GH, IGF-I, and its binding proteins. Evidently, a sufficient GH stimulus is still maintained in AA to support normal mean fasting plasma concentrations of IGF-I, IGFBP-1, and IGFBP-3.
An unexpected finding in our study was the positive relationship between elevated trough GH concentrations and disordered pulsatility. This may indicate that elevated trough concentrations reflect mechanisms of GH production that are less effectual at orderly pulse generation. A plausible basis for both alterations is reduced somatostatin tone in AA, as high trough GH concentrations are believed to mirror somatostatin withdrawal. Sustained somatostatin withdrawal also impedes the generation of high amplitude GH pulses, which require recurring somatostatin exposure (28, 29). Interestingly, some exercise studies indicate that exercise stimulates GH secretion in part via somatostatin withdrawal (7, 30).
Exercise is a well known stimulus of GH secretion, and several mechanisms have been proposed to mediate the exercise-induced release of GH (2, 6, 18, 31). Regardless of the particular neurotransmitters involved, the final common pathway probably entails either an increase in GHRH and/or a decrease in somatostatin release or action (2). We observed that AA subjects had significantly lower (4- to 5-fold) peak and integrated GH responses to exercise. These deficits would be consistent with reduced GHRH secretion or excessive somatostatin inhibition. We favor the former possibility, as elevated nocturnal trough concentrations of GH are more consistent with reduced somatostatin release. Although daytime exercise may reduce nocturnal GH secretion (2), alternate day exercise does not (32). Thus, we suggest that elevated nocturnal trough GH concentrations and a reduced mass of GH release per burst both before and after exercise in our AA volunteers result from a combined reduction in somatostinergic and GHRH inputs.
IGFBP-1 is acutely regulated by exercise and is elevated during prolonged exercise. This binding protein is elevated in AA and anorexia nervosa and may serve as a marker of metabolic stress or insulinopenia (1, 33, 34, 35). In our AA patients, IGFBP-1 rose throughout exercise and peaked 20 min after recovery. Baseline IGFBP-1 concentrations were in the high normal range, but not significantly so. This trend could reflect insulin actions, as insulin (negatively) regulates IGFBP-1 production (36). Although the clinical role of this binding protein is not well understood, IGFBP-1 may modulate sex steroid hormone action and fuel homeostasis (6). For example, together with IGF-I, insulin, ovarian sex steroids, cytokines, and other factors, IGFBP-1 is involved in a complex system that impacts menstrual cyclicity, ovulation, decidualization, blastocyte implantation, and fetal growth (37).
In conclusion, we employed trough time occupancy, deconvolution analysis, and ApEn to evaluate the dynamic properties of GH secretion in amenorrheic and eumenorrheic athletes with similar percent body fat. Time occupancy analysis disclosed an increased time spent in elevated troughs. Deconvolution analysis revealed an increased number of GH peaks, a prolonged half-life, and brief GH secretory bursts of reduced mass in the AA. ApEn analysis disclosed more disorderly patterns of GH secretion, which paralleled elevated trough concentrations. Both disorderly and elevated trough GH release were correlated to a blunting of the GH secretory response to acute exercise. Concentrations of plasma IGF-I and its associated binding proteins remained normal. We conclude that in the absence of eating disorders and when controlled for their reduction in percent body fat, volunteers with AA, compared with EA subjects, display multiple alterations in GH secretory control. The precise neurochemical basis for disrupted neuroregulation in AA, the role of estrogen depletion in this context, and the degree and tempo of its reversibility are not yet known.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received August 23, 2000.
Revised December 1, 2000.
Accepted December 4, 2000.
|
References |
|---|
|
TopAbstractIntroductionExperimental SubjectsMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  A. B. Loucks The Response of Luteinizing Hormone Pulsatility to 5 Days of Low Energy Availability Disappears by 14 Years of Gynecological Age J. Clin. Endocrinol. Metab., August 1, 2006; 91(8): 3158 - 3164. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. J. De Souza, H. J. Leidy, E. O'Donnell, B. Lasley, and N. I. Williams Fasting Ghrelin Levels in Physically Active Women: Relationship with Menstrual Disturbances and Metabolic Hormones J. Clin. Endocrinol. Metab., July 1, 2004; 89(7): 3536 - 3542. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. J. de Souza, J. Van Heest, L. M. Demers, and B. L. Lasley Luteal Phase Deficiency in Recreational Runners: Evidence for a Hypometabolic State J. Clin. Endocrinol. Metab., January 1, 2003; 88(1): 337 - 346. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
