0021-972X/90/7006-1678$02.00/0 Journal of Clinical Endocrinology and Metabolism Copyright © 1990 by The Endocrine Society
Vol. 70, No. 6 Printed in U.S.A.
Basal Plasma Growth Hormone Levels in Man: New Evidence for Rhythmicity of Growth Hormone Secretion* LORI M. WINER, MELISSA A. SHAW, AND GERHARD BAUMANN Center for Endocrinology, Metabolism, and Nutrition, Department of Medicine, Northwestern University Medical School, Chicago, Illinois 60611
and ranged from 40-19,695 ng/L. Dynamic fluctuations occurred within and above the previously undetectable range, with amplitudes varying over 3 orders of magnitude. Women had significantly higher overall GH levels, higher peak amplitudes, and higher valley levels/nadirs than men. GH pulses occurred with an average frequency of about 13/day in both sexes, with a dominant, but not strictly periodic, 2-h rhythmicity. We conclude that in man pulsatile GH secretion occurs throughout the day, and that it is oscillatory rather than episodic. This neurosecretory pattern has eluded recognition heretofore because of the lack of assay sensitivity. Women of reproductive age have higher pulse amplitudes and a higher baseline but equal pulse frequency compared to men. Previous estimates of integrated GH concentrations and GH production rates were too high by a factor of 2 due to overestimation of GH levels in the undetectable range. (J Clin Endocrinol Metab 70: 1678-1686, 1990)
ABSTRACT. Circulating GH levels in man fluctuate widely due to pulsatile GH secretion by the pituitary gland. During much of the time, plasma GH is undetectable by current assays. This is punctuated by occasional secretory episodes, resulting in plasma GH peaks of varying height. The principal diurnal secretory event for GH is that associated with early slow wave sleep, but little is known about the prevailing level and dynamics of GH during the day. We used a new ultrasensitive immunoradiometric assay for GH (Boots-Celltech IRMA; limit of detection, 20 ng/L) to measure plasma GH in the previously undetectable range and to assess its diurnal pattern. Plasma GH was measured every 20 min over a 24-h period in 12 normal subjects (6 men and 6 women, aged 20-47 yr) under physiological conditions. Time series analysis of plasma GH patterns was performed by the Cluster algorithm, autocorrelation, and spectral analysis. Plasma GH, as measured by IRMA, was detectable at all time points
G
H, LIKE other polypeptide hormones, is secreted in a pulsatile fashion in man and animals. Its plasma concentration, therefore, fluctuates over a wide range (1, 2). In humans, the major secretory event for GH coincides with early nocturnal slow wave sleep (3, 4). Other surges occur at seemingly random, widely spaced times, which alternate with quiescent periods (the so-called basal state) (1-4). This type of secretory pattern has been termed episodic (5). During the better part of the day the quiescent state prevails in man, and plasma GH levels are below or near the detection limit of conventional assays. As a consequence, little is known about the predominant GH level to which tissues are exposed or the dynamics of GH secretion during the majority of time. In previous studies we have used an immunoextraction Received October 23,1989. Address all correspondence and requests for reprints to: G. Baumann, M.D., 303 East Chicago Avenue, Chicago, Illinois 60611. * Presented in part at the Annual Meeting of American Association of Physicians, Washington, D.C., April 1989, and reported in abstract form (Clin Res 1989;37:598A). This work was supported by NIH Grants DK-38128, DK-07169, and RR-48.
technique applied to large plasma samples (10-12 mL) to extend the useful range of GH assays in an effort to assess GH levels below 1 ixg/L (6, 7). This technique permitted accurate measurement in the picomolar range and identified GH levels as low as below 40 ng/L (6). In addition, it provided evidence for dynamic fluctuations of GH levels both within and without the previously undetectable range (7). A major drawback of the immunoextraction technique was the requirement for large blood volumes, with attendant limitations on frequency and duration of sampling protocols. Although the combination of a 24-h study with 2-h sampling (6) and a 3-h study with 20-min sampling (7) allowed construction of a rough composite picture of circadian GH dynamics, the information on minute to minute GH dynamics remained sketchy. Recently, ultrasensitive immunoradiometric assays (IRMA) for GH have been developed, which permit measurement in the same range without the limitations mentioned. In the present report we have applied such an assay to extend and amplify our previous observations on GH levels that are undetectable by conventional assays.
1678
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
PLASMA GH PATTERN IN THE BASAL STATE
Materials and Methods Study protocol Six normal white men (aged 20-46 yr) and six normal white women (aged 22-47 yr) participated in the study. All were within 89-109% of ideal body weight (8) and within 2 SD for height (9). All women had regular menstrual cycles; four were studied during the early to midfollicular phase, and two during the luteal phase, as assessed by plasma estradiol and progesterone levels. No subject was taking any medication at the time of study. The study had been approved by the Northwestern University Institutional Review Board for human investigation, and all subjects gave informed consent. They were admitted to the Clinical Research Center the evening before the study and were instructed to keep their regular sleep-wake cycle, not to eat after 2000 h, and to stay in bed the next morning until the first hour of the study had been completed. At about 2100 h the evening before the study, a heparin lock was inserted into a forearm vein, which was connected to a slow infusion of normal saline the next morning. Subjects were instructed to carry out their normal daily activities (including ambulation), but to avoid taking naps during the day or to engage in exercise. Activities were monitored and recorded by the nurses and/or investigators. Meals (regular hospital diet) were consumed at 0845,1145, and 1745 h by all subjects. Regular nighttime sleep was allowed to occur naturally. Heparinized blood (3 mL) was drawn starting at 0800 h and at 20-min intervals thereafter for a 24-h period. Blood samples were kept on ice for a maximum of 25 min before centrifugation. Plasma was stored at —20 C for 2-5 months until analysis. GH assay Reagents for assay of human GH (hGH) were purchased in kit form (SUCROSEP HGH IRMA) from Boots-Celltech Diagnostics (Slough, United Kingdom). This assay was selected based on its high sensitivity and excellent performance over a wide range of hGH levels, as assessed in preliminary experiments. Sufficient reagents were obtained to permit assay of all plasma samples with reagents of the same lot numbers, and all measurements were made as quickly as possible (within 1 month) after radiolabeling in order to maximize precision and sensitivity. The Boots-Celltech assay employs a two-site immunoradiometric (IRMA) design with two monoclonal antihGH antibodies, of which one is labeled with 125I and the other is coupled to a solid phase (beads). After incubation of standard or unknown sample with the two antibodies, bound hGH is separated from free hGH by unit gravity sedimentation through a sucrose step gradient. The instructions supplied by the company were followed without modification, except for pooling of corresponding reagents from each kit received. In preliminary studies we had examined the performance of this commercial assay with respect to accuracy, sensitivity, and reproducibility against our polyclonal, double antibody RIA (6). The performance data supplied by the company with the assay kits were essentially confirmed. For the particular lot we used for the present study, the limit of detection (defined as a value 2.5 SD above the zero dose) averaged 22.6 ng/L (range, 15.130.5 ng/L). The intraassay coefficient of variation (CV), deter-
1679
mined on nine replicates of a plasma sample measuring a mean value of 101 ng/L, was 12.5%; the corresponding interassay CV was 16.1%. (It should be noted that at higher GH concentrations, the CV is considerably lower, reaching a nadir of ~3% between 2,000-20,000 ng/L.) GH was measured in duplicate, with all 73 samples from a given subject assayed in the same assay run to eliminate interassay variability. The computerized data reduction program supplied by the company was used to calculate GH levels. (Its performance had been verified against a manual calculation/ plotting method.) Concentrations expressed in milliunits per L in terms of the First International Reference Preparation of hGH (code 66/217) were converted to nanograms per L based on the published potency of 2.0 U/mg for that standard (10, 11). Data analysis Time series analysis of plasma GH patterns was performed by 1) visual comparison of individual profiles, 2) an objective peak detection algorithm [Cluster analysis (12)] of individual profiles, 3) superimposition of all 12 profiles, or those from either men or women separately, after temporal alignment of either the first peaks in the series or the sleep-related peaks, 4) reanalysis of profiles after smoothing by a 3-point moving average algorithm, 5) Box-Jenkins autoregressive integrated moving average (ARIMA) analysis (13), 6) spectral analysis, and 7) Fourier series simulations according to GH(t) = (ao/2) + £ (a; cos wti + bi sin wti), where a and b are the Fourier coefficients, i is the ordinal frequency number, oo is the frequency (2ir/T), and T is the period. In the Cluster program, a power function fit of local variance, a one by one point cluster size, and a t statistic of either 2.32 or 1 for significant increases/decreases was used. [For 20-min GH sampling, a 1 x 1 cluster size and t statistic of 1 have been found to minimize both type I and type II errors in pulse detection (14, 15; see Ref. 16 for review); those criteria were, therefore, used to derive summarized data (Tables 1-3)). Although some of the data were transformed logarithmically for presentation in figures, most mathematical manipulations were performed on nontransformed data. However, ARIMA and spectral analysis were also performed on log-transformed data. Integrated GH concentrations were calculated as the mean of all GH values in a 24-h period, and GH production rates were calculated as (integrated GH) x MCR, using a pooled estimate for MCR derived from published data (17-28). Statistical comparisons of derived parameters between men and women were performed using the t test, Wilcoxon-MannWhitney test, and robust rank order test (29).
Results Individual circadian GH profiles are shown in Fig. 1 (men) and Fig. 2 (women). All GH levels were well within the detectable range of the assay, with the lowest values exceeding the detection limit by a factor of 2. Thus, all
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
1680
WINER, SHAW, AND BAUMANN
JCE & M •1990
3OOOO 10000k
FIG. 1. Twenty-four-hour pattern of plasma GH in six men. Note the logarithmic ordinate. The shaded lines denote the boundary between the detectable and undetectable range in conventional assays. The arrowheads at the top of each panel denote GH pulses detected by Cluster analysis. Larger arrows correspond to peaks detected using a t value of 2.32 for significant increase/decrease, smaller arrows to those additionally detected using a t value of 1 (see text for details). Black bars on the lower right of each panel denote sleep. Meals were consumed at 0845,1145, and 1745 h.
30000™
? f
i
? i
i
i
i
i
tit
i
t
I
I »
I
O Q.
_l
a. "» ? i
? »? ? ? » I
?
f
d 26y
CLOCK TIME
GH values represented reliable measurements. The hatched line in each panel denotes the boundary between detectable and undetectable values in conventional assays. Inspection of the area above these lines confirms what has been known heretofore about the secretory pattern of GH, namely a major peak during the early hours of sleep, occasionally occurring peaks at other times, and long periods when GH is undetectable, with women tending to have more and higher peaks. The area below the line, however, represents largely unknown territory. Consideration of the two areas combined indicates that GH levels ranged over 3 orders of magnitude in both sexes. Plasma GH fluctuated considerably not only in the previously detectable range, but also in the range below 1 /ug/L. (In conventional assays for GH, only occasional peaks above 0.7-1 Mg/L would be detected.) Women generally exhibited higher peak amplitudes than men. The highest GH levels were achieved during sleep in all but one subject (a 47-yr-old woman who had higher values at other times). Despite the fact that physiological
conditions for the basal state were fulfilled for the first hour (i.e. fasting and supine in bed), several subjects did not exhibit their lowest levels during that time, with two showing a first GH level above 1 ixg/L, and several others showing either decreasing or increasing trends, phenomena that we observed previously (7). The logarithmic transformation of the ordinates in Figs. 1 and 2 was necessary to accommodate the wide range of values and at the same time permit appreciation of local variation. This transformation is potentially misleading, however, in that it magnifies the lower end of the scale and compresses the high end. To verify the degree of visual distortion, we also plotted the profiles on a broken linear scale, with the range below 1 Mg/L expanded. An example is shown in Fig. 3, where the first subject in Fig. 1 is replotted. Figure 3A is reminiscent of conventional profiles, with the low range essentially invisible. Figure 3B shows the same profile, with the low range plotted on an expanded linear scale. The visual impact in the latter is very similar to that in Fig. 1, where a logarithmic scale was used. Similar conclusions
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
PLASMA GH PATTERN IN THE BASAL STATE
1681
3OOOO™
f
I T
i f
i
f
ft
f
10000
30 30000
f
f
T
»
f
»
T
f
f
10000
X
o
1000
FIG. 2. Twenty-four-hour pattern of plasma GH in 6 women. Conditions, presentation, and symbols are described in Fig. 1. The 22- and 47-yr-old women were in the luteal phase; all others were in the follicular phase of the menstrual cycle.
(PG/MI
_ ]
100
i
CL
30Lr 30000™
f
i
n
f
f
f
f
t
f
t
f
10000
1000
100-
30 r
CLOCK TIME were reached for all other subjects. Thus, the logarithmic depiction has a high degree of visual fidelity. Table 1 summarizes the numeric details of the circadian GH profiles. (Special emphasis is given to the range below 1 ng/h, since this represents new information.) The integrated GH concentration was significantly higher in women, although substantial overlap existed between the sexes. Women also had significantly more frequent values above 1 iig/L, higher mean peak values overall and within the previously undetectable range, higher mean peak increments, and higher minimal peak increments. Women also had higher mean nadirs and valley levels. In contrast, the absolute nadirs, zeniths, and maximal increments were similar in both sexes.1 Inspection of the profiles suggested an element of peak 1 It should be noted that the absolute peak (zenith) values in our 12 subjects were relatively low (~20 /zg/L). We believe that this is a fortuitous occurrence which may underemphasize the difference between men and women. Peak values up to 30 or 40 /ug/L are not infrequently seen, particularly in women (43, 44). As our and other studies show, peak amplitudes are highly variable among subjects and, hence, unpredictable.
periodicity above and beyond the main circadian sleeprelated surge. Cluster analysis also suggested some regularity in peak occurrence, regardless of whether a t value of 2.32 (Figs. 1 and 2, large arrows) or 1 (Figs. 1 and 2, large and small arrows) was used for peak definition. The dominant, although not sole, frequency of peak occurrence was fairly constant (~2-h) among subjects. The possible existence of peak periodicity was investigated in detail within each individual subject as well as between subjects. The frequency distribution of peak intervals, derived from both raw data and moving average smoothed data, centered around a 2 h value in both sexes, but there were large deviations ranging from 40280 min (Table 2). ARIMA approached, but did not reach, significance (P < 0.05) in showing periodicity for some women, but not for men. Spectral analysis suggested the presence of certain dominant frequency bands, corresponding to periods of about 1490, 480,160, 90, and 60 min, but none of these reached statistical significance above noise. Thus, there were elements of peak regularity, but at the same time sufficient deviations from that
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
WINER, SHAW, AND BAUMANN
1682
15000r
1OOOO
5000
0^ O
15000
B.
Q.
1OOOO
1 I/)
5000 200Q 1000 800 600 400 200 0
a/A/
8
12
16
20
24
8
CLOCK TIME FIG. 3. Linear plots of one subject to address the question of whether the logarithmic transformation of the ordinate in Figs. 1 and 2 seriously distorts the visual impression. A, Profile of one subject (top left, Fig. 1) plotted on a conventional scale. The low range is not well appreciated. B, The same profile, with the range below 1 Mg/L linear but expanded in scale. Note the similarity with the log-transformed profile (Fig. 1, top left).
regularity to preclude mathematical proof. Nevertheless, the dominant temporal pattern seemed to be fairly similar in all subjects. To further probe the possible existence of a global intersubject rhythmicity, a composite profile encompassing all subjects was constructed. (The same was also done for men or women separately). The first peak in each profile was temporally aligned, and subsequent timeincremental values pooled, resulting in superimposition and averaging of all profiles. Alternatively, the main sleep-related peak was aligned, based on the notion that sleep may be the master pacesetter/zeitgeber (either method resulted in similar data). Figure 4 depicts the composite profile resulting from pooling of all subjects' data. The fact that the pooling of the data of 12 subjects did not result in destructive interference, but, rather,
JCE & M • 1990 Vol 70 • No 6
yielded the same overall pattern as that obtained for individuals, provides evidence for an underlying dominant rhythm characteristic of the human species (pooling of males only or females only yielded similar temporal patterns). However, application of rigorous tests for periodicity, such as autoregressive ARIMA or spectral analysis, to the composite (or a concatenated) profile again failed to reach statistical significance. When Fourier analysis of GH patterns was performed, both individual and pooled profiles could be closely simulated by as few as nine Fourier coefficients. An example is shown in Fig. 5. Data on temporal patterns are summarized in Table 2. The number of peaks per 24 h was not significantly different in men and women. Peaks occurred, on the average, approximately every 2 h, and no significant difference in this mean interpeak interval was observed between the sexes. However, there was a wide range of interpeak intervals. Peak width was also similar in both sexes, and it varied over a considerable range, partly as a function of peak height. The first (zero frequency) Fourier coefficient (ao/2) averaged 1104 ng/L for men and 1472 ng/L for women, closely approximating the means for actual integrated GH levels (Table 1). Table 3 compares circadian GH characteristics for three ranges: 1) above 1 ixg/L, 2) below 1 iig/L, and 3) the full range of GH levels. The first range permits comparison with previously available data obtained in conventional assays. It is evident that interpretations can differ depending on which range is analyzed. Inspection of Figs. 1 and 2 helps in interpreting Table 3. In the range above 1 /ug/L, averaged pulse amplitude characteristics were similar for men and women because women tended to have more peaks of intermediate height, whereas men showed only very high peaks in that range, with most of their pulses being excluded due to low amplitudes. The sex difference in pulse amplitude became apparent only in the range below 1 /ug/L and the full range. Similarly, men and women had indistinguishable averaged valley levels in the range above 1 fig/L because the notches in high peaks are counted as valleys. However, when "true" valleys are considered (i.e. those in the previously undetectable range), women exhibited significantly higher valley levels than men. The number of peaks per 24 h in the range above 1 /xg/L was higher in women, but the opposite was true in the range below that limit. When the full range is considered, peak frequency was the same in both sexes. These seemingly contradictory findings are due to the fact that in women many peaks were of sufficient amplitude to penetrate the 1 /ug/L detection limit, whereas in men most peaks had amplitudes below that level. For the same reason, peak intervals showed disparate values for the two sexes in the range above 1 /ug/L, but not in the other ranges. These data demonstrate that caution should used in
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
PLASMA GH PATTERN IN THE BASAL STATE
1683
TABLE 1. Characteristics of circadian plasma GH patterns in normal men and women Range of individual values
Range of means Integrated GH level (ng/L) Men
Women % of values >1 /xg/L Men
Women Mean peak ht (ng/L)° Men Women Mean peak ht in range
Vol. 70, No. 6 Printed in U.S.A.
Basal Plasma Growth Hormone Levels in Man: New Evidence for Rhythmicity of Growth Hormone Secretion* LORI M. WINER, MELISSA A. SHAW, AND GERHARD BAUMANN Center for Endocrinology, Metabolism, and Nutrition, Department of Medicine, Northwestern University Medical School, Chicago, Illinois 60611
and ranged from 40-19,695 ng/L. Dynamic fluctuations occurred within and above the previously undetectable range, with amplitudes varying over 3 orders of magnitude. Women had significantly higher overall GH levels, higher peak amplitudes, and higher valley levels/nadirs than men. GH pulses occurred with an average frequency of about 13/day in both sexes, with a dominant, but not strictly periodic, 2-h rhythmicity. We conclude that in man pulsatile GH secretion occurs throughout the day, and that it is oscillatory rather than episodic. This neurosecretory pattern has eluded recognition heretofore because of the lack of assay sensitivity. Women of reproductive age have higher pulse amplitudes and a higher baseline but equal pulse frequency compared to men. Previous estimates of integrated GH concentrations and GH production rates were too high by a factor of 2 due to overestimation of GH levels in the undetectable range. (J Clin Endocrinol Metab 70: 1678-1686, 1990)
ABSTRACT. Circulating GH levels in man fluctuate widely due to pulsatile GH secretion by the pituitary gland. During much of the time, plasma GH is undetectable by current assays. This is punctuated by occasional secretory episodes, resulting in plasma GH peaks of varying height. The principal diurnal secretory event for GH is that associated with early slow wave sleep, but little is known about the prevailing level and dynamics of GH during the day. We used a new ultrasensitive immunoradiometric assay for GH (Boots-Celltech IRMA; limit of detection, 20 ng/L) to measure plasma GH in the previously undetectable range and to assess its diurnal pattern. Plasma GH was measured every 20 min over a 24-h period in 12 normal subjects (6 men and 6 women, aged 20-47 yr) under physiological conditions. Time series analysis of plasma GH patterns was performed by the Cluster algorithm, autocorrelation, and spectral analysis. Plasma GH, as measured by IRMA, was detectable at all time points
G
H, LIKE other polypeptide hormones, is secreted in a pulsatile fashion in man and animals. Its plasma concentration, therefore, fluctuates over a wide range (1, 2). In humans, the major secretory event for GH coincides with early nocturnal slow wave sleep (3, 4). Other surges occur at seemingly random, widely spaced times, which alternate with quiescent periods (the so-called basal state) (1-4). This type of secretory pattern has been termed episodic (5). During the better part of the day the quiescent state prevails in man, and plasma GH levels are below or near the detection limit of conventional assays. As a consequence, little is known about the predominant GH level to which tissues are exposed or the dynamics of GH secretion during the majority of time. In previous studies we have used an immunoextraction Received October 23,1989. Address all correspondence and requests for reprints to: G. Baumann, M.D., 303 East Chicago Avenue, Chicago, Illinois 60611. * Presented in part at the Annual Meeting of American Association of Physicians, Washington, D.C., April 1989, and reported in abstract form (Clin Res 1989;37:598A). This work was supported by NIH Grants DK-38128, DK-07169, and RR-48.
technique applied to large plasma samples (10-12 mL) to extend the useful range of GH assays in an effort to assess GH levels below 1 ixg/L (6, 7). This technique permitted accurate measurement in the picomolar range and identified GH levels as low as below 40 ng/L (6). In addition, it provided evidence for dynamic fluctuations of GH levels both within and without the previously undetectable range (7). A major drawback of the immunoextraction technique was the requirement for large blood volumes, with attendant limitations on frequency and duration of sampling protocols. Although the combination of a 24-h study with 2-h sampling (6) and a 3-h study with 20-min sampling (7) allowed construction of a rough composite picture of circadian GH dynamics, the information on minute to minute GH dynamics remained sketchy. Recently, ultrasensitive immunoradiometric assays (IRMA) for GH have been developed, which permit measurement in the same range without the limitations mentioned. In the present report we have applied such an assay to extend and amplify our previous observations on GH levels that are undetectable by conventional assays.
1678
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
PLASMA GH PATTERN IN THE BASAL STATE
Materials and Methods Study protocol Six normal white men (aged 20-46 yr) and six normal white women (aged 22-47 yr) participated in the study. All were within 89-109% of ideal body weight (8) and within 2 SD for height (9). All women had regular menstrual cycles; four were studied during the early to midfollicular phase, and two during the luteal phase, as assessed by plasma estradiol and progesterone levels. No subject was taking any medication at the time of study. The study had been approved by the Northwestern University Institutional Review Board for human investigation, and all subjects gave informed consent. They were admitted to the Clinical Research Center the evening before the study and were instructed to keep their regular sleep-wake cycle, not to eat after 2000 h, and to stay in bed the next morning until the first hour of the study had been completed. At about 2100 h the evening before the study, a heparin lock was inserted into a forearm vein, which was connected to a slow infusion of normal saline the next morning. Subjects were instructed to carry out their normal daily activities (including ambulation), but to avoid taking naps during the day or to engage in exercise. Activities were monitored and recorded by the nurses and/or investigators. Meals (regular hospital diet) were consumed at 0845,1145, and 1745 h by all subjects. Regular nighttime sleep was allowed to occur naturally. Heparinized blood (3 mL) was drawn starting at 0800 h and at 20-min intervals thereafter for a 24-h period. Blood samples were kept on ice for a maximum of 25 min before centrifugation. Plasma was stored at —20 C for 2-5 months until analysis. GH assay Reagents for assay of human GH (hGH) were purchased in kit form (SUCROSEP HGH IRMA) from Boots-Celltech Diagnostics (Slough, United Kingdom). This assay was selected based on its high sensitivity and excellent performance over a wide range of hGH levels, as assessed in preliminary experiments. Sufficient reagents were obtained to permit assay of all plasma samples with reagents of the same lot numbers, and all measurements were made as quickly as possible (within 1 month) after radiolabeling in order to maximize precision and sensitivity. The Boots-Celltech assay employs a two-site immunoradiometric (IRMA) design with two monoclonal antihGH antibodies, of which one is labeled with 125I and the other is coupled to a solid phase (beads). After incubation of standard or unknown sample with the two antibodies, bound hGH is separated from free hGH by unit gravity sedimentation through a sucrose step gradient. The instructions supplied by the company were followed without modification, except for pooling of corresponding reagents from each kit received. In preliminary studies we had examined the performance of this commercial assay with respect to accuracy, sensitivity, and reproducibility against our polyclonal, double antibody RIA (6). The performance data supplied by the company with the assay kits were essentially confirmed. For the particular lot we used for the present study, the limit of detection (defined as a value 2.5 SD above the zero dose) averaged 22.6 ng/L (range, 15.130.5 ng/L). The intraassay coefficient of variation (CV), deter-
1679
mined on nine replicates of a plasma sample measuring a mean value of 101 ng/L, was 12.5%; the corresponding interassay CV was 16.1%. (It should be noted that at higher GH concentrations, the CV is considerably lower, reaching a nadir of ~3% between 2,000-20,000 ng/L.) GH was measured in duplicate, with all 73 samples from a given subject assayed in the same assay run to eliminate interassay variability. The computerized data reduction program supplied by the company was used to calculate GH levels. (Its performance had been verified against a manual calculation/ plotting method.) Concentrations expressed in milliunits per L in terms of the First International Reference Preparation of hGH (code 66/217) were converted to nanograms per L based on the published potency of 2.0 U/mg for that standard (10, 11). Data analysis Time series analysis of plasma GH patterns was performed by 1) visual comparison of individual profiles, 2) an objective peak detection algorithm [Cluster analysis (12)] of individual profiles, 3) superimposition of all 12 profiles, or those from either men or women separately, after temporal alignment of either the first peaks in the series or the sleep-related peaks, 4) reanalysis of profiles after smoothing by a 3-point moving average algorithm, 5) Box-Jenkins autoregressive integrated moving average (ARIMA) analysis (13), 6) spectral analysis, and 7) Fourier series simulations according to GH(t) = (ao/2) + £ (a; cos wti + bi sin wti), where a and b are the Fourier coefficients, i is the ordinal frequency number, oo is the frequency (2ir/T), and T is the period. In the Cluster program, a power function fit of local variance, a one by one point cluster size, and a t statistic of either 2.32 or 1 for significant increases/decreases was used. [For 20-min GH sampling, a 1 x 1 cluster size and t statistic of 1 have been found to minimize both type I and type II errors in pulse detection (14, 15; see Ref. 16 for review); those criteria were, therefore, used to derive summarized data (Tables 1-3)). Although some of the data were transformed logarithmically for presentation in figures, most mathematical manipulations were performed on nontransformed data. However, ARIMA and spectral analysis were also performed on log-transformed data. Integrated GH concentrations were calculated as the mean of all GH values in a 24-h period, and GH production rates were calculated as (integrated GH) x MCR, using a pooled estimate for MCR derived from published data (17-28). Statistical comparisons of derived parameters between men and women were performed using the t test, Wilcoxon-MannWhitney test, and robust rank order test (29).
Results Individual circadian GH profiles are shown in Fig. 1 (men) and Fig. 2 (women). All GH levels were well within the detectable range of the assay, with the lowest values exceeding the detection limit by a factor of 2. Thus, all
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
1680
WINER, SHAW, AND BAUMANN
JCE & M •1990
3OOOO 10000k
FIG. 1. Twenty-four-hour pattern of plasma GH in six men. Note the logarithmic ordinate. The shaded lines denote the boundary between the detectable and undetectable range in conventional assays. The arrowheads at the top of each panel denote GH pulses detected by Cluster analysis. Larger arrows correspond to peaks detected using a t value of 2.32 for significant increase/decrease, smaller arrows to those additionally detected using a t value of 1 (see text for details). Black bars on the lower right of each panel denote sleep. Meals were consumed at 0845,1145, and 1745 h.
30000™
? f
i
? i
i
i
i
i
tit
i
t
I
I »
I
O Q.
_l
a. "» ? i
? »? ? ? » I
?
f
d 26y
CLOCK TIME
GH values represented reliable measurements. The hatched line in each panel denotes the boundary between detectable and undetectable values in conventional assays. Inspection of the area above these lines confirms what has been known heretofore about the secretory pattern of GH, namely a major peak during the early hours of sleep, occasionally occurring peaks at other times, and long periods when GH is undetectable, with women tending to have more and higher peaks. The area below the line, however, represents largely unknown territory. Consideration of the two areas combined indicates that GH levels ranged over 3 orders of magnitude in both sexes. Plasma GH fluctuated considerably not only in the previously detectable range, but also in the range below 1 /ug/L. (In conventional assays for GH, only occasional peaks above 0.7-1 Mg/L would be detected.) Women generally exhibited higher peak amplitudes than men. The highest GH levels were achieved during sleep in all but one subject (a 47-yr-old woman who had higher values at other times). Despite the fact that physiological
conditions for the basal state were fulfilled for the first hour (i.e. fasting and supine in bed), several subjects did not exhibit their lowest levels during that time, with two showing a first GH level above 1 ixg/L, and several others showing either decreasing or increasing trends, phenomena that we observed previously (7). The logarithmic transformation of the ordinates in Figs. 1 and 2 was necessary to accommodate the wide range of values and at the same time permit appreciation of local variation. This transformation is potentially misleading, however, in that it magnifies the lower end of the scale and compresses the high end. To verify the degree of visual distortion, we also plotted the profiles on a broken linear scale, with the range below 1 Mg/L expanded. An example is shown in Fig. 3, where the first subject in Fig. 1 is replotted. Figure 3A is reminiscent of conventional profiles, with the low range essentially invisible. Figure 3B shows the same profile, with the low range plotted on an expanded linear scale. The visual impact in the latter is very similar to that in Fig. 1, where a logarithmic scale was used. Similar conclusions
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
PLASMA GH PATTERN IN THE BASAL STATE
1681
3OOOO™
f
I T
i f
i
f
ft
f
10000
30 30000
f
f
T
»
f
»
T
f
f
10000
X
o
1000
FIG. 2. Twenty-four-hour pattern of plasma GH in 6 women. Conditions, presentation, and symbols are described in Fig. 1. The 22- and 47-yr-old women were in the luteal phase; all others were in the follicular phase of the menstrual cycle.
(PG/MI
_ ]
100
i
CL
30Lr 30000™
f
i
n
f
f
f
f
t
f
t
f
10000
1000
100-
30 r
CLOCK TIME were reached for all other subjects. Thus, the logarithmic depiction has a high degree of visual fidelity. Table 1 summarizes the numeric details of the circadian GH profiles. (Special emphasis is given to the range below 1 ng/h, since this represents new information.) The integrated GH concentration was significantly higher in women, although substantial overlap existed between the sexes. Women also had significantly more frequent values above 1 iig/L, higher mean peak values overall and within the previously undetectable range, higher mean peak increments, and higher minimal peak increments. Women also had higher mean nadirs and valley levels. In contrast, the absolute nadirs, zeniths, and maximal increments were similar in both sexes.1 Inspection of the profiles suggested an element of peak 1 It should be noted that the absolute peak (zenith) values in our 12 subjects were relatively low (~20 /zg/L). We believe that this is a fortuitous occurrence which may underemphasize the difference between men and women. Peak values up to 30 or 40 /ug/L are not infrequently seen, particularly in women (43, 44). As our and other studies show, peak amplitudes are highly variable among subjects and, hence, unpredictable.
periodicity above and beyond the main circadian sleeprelated surge. Cluster analysis also suggested some regularity in peak occurrence, regardless of whether a t value of 2.32 (Figs. 1 and 2, large arrows) or 1 (Figs. 1 and 2, large and small arrows) was used for peak definition. The dominant, although not sole, frequency of peak occurrence was fairly constant (~2-h) among subjects. The possible existence of peak periodicity was investigated in detail within each individual subject as well as between subjects. The frequency distribution of peak intervals, derived from both raw data and moving average smoothed data, centered around a 2 h value in both sexes, but there were large deviations ranging from 40280 min (Table 2). ARIMA approached, but did not reach, significance (P < 0.05) in showing periodicity for some women, but not for men. Spectral analysis suggested the presence of certain dominant frequency bands, corresponding to periods of about 1490, 480,160, 90, and 60 min, but none of these reached statistical significance above noise. Thus, there were elements of peak regularity, but at the same time sufficient deviations from that
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
WINER, SHAW, AND BAUMANN
1682
15000r
1OOOO
5000
0^ O
15000
B.
Q.
1OOOO
1 I/)
5000 200Q 1000 800 600 400 200 0
a/A/
8
12
16
20
24
8
CLOCK TIME FIG. 3. Linear plots of one subject to address the question of whether the logarithmic transformation of the ordinate in Figs. 1 and 2 seriously distorts the visual impression. A, Profile of one subject (top left, Fig. 1) plotted on a conventional scale. The low range is not well appreciated. B, The same profile, with the range below 1 Mg/L linear but expanded in scale. Note the similarity with the log-transformed profile (Fig. 1, top left).
regularity to preclude mathematical proof. Nevertheless, the dominant temporal pattern seemed to be fairly similar in all subjects. To further probe the possible existence of a global intersubject rhythmicity, a composite profile encompassing all subjects was constructed. (The same was also done for men or women separately). The first peak in each profile was temporally aligned, and subsequent timeincremental values pooled, resulting in superimposition and averaging of all profiles. Alternatively, the main sleep-related peak was aligned, based on the notion that sleep may be the master pacesetter/zeitgeber (either method resulted in similar data). Figure 4 depicts the composite profile resulting from pooling of all subjects' data. The fact that the pooling of the data of 12 subjects did not result in destructive interference, but, rather,
JCE & M • 1990 Vol 70 • No 6
yielded the same overall pattern as that obtained for individuals, provides evidence for an underlying dominant rhythm characteristic of the human species (pooling of males only or females only yielded similar temporal patterns). However, application of rigorous tests for periodicity, such as autoregressive ARIMA or spectral analysis, to the composite (or a concatenated) profile again failed to reach statistical significance. When Fourier analysis of GH patterns was performed, both individual and pooled profiles could be closely simulated by as few as nine Fourier coefficients. An example is shown in Fig. 5. Data on temporal patterns are summarized in Table 2. The number of peaks per 24 h was not significantly different in men and women. Peaks occurred, on the average, approximately every 2 h, and no significant difference in this mean interpeak interval was observed between the sexes. However, there was a wide range of interpeak intervals. Peak width was also similar in both sexes, and it varied over a considerable range, partly as a function of peak height. The first (zero frequency) Fourier coefficient (ao/2) averaged 1104 ng/L for men and 1472 ng/L for women, closely approximating the means for actual integrated GH levels (Table 1). Table 3 compares circadian GH characteristics for three ranges: 1) above 1 ixg/L, 2) below 1 iig/L, and 3) the full range of GH levels. The first range permits comparison with previously available data obtained in conventional assays. It is evident that interpretations can differ depending on which range is analyzed. Inspection of Figs. 1 and 2 helps in interpreting Table 3. In the range above 1 /ug/L, averaged pulse amplitude characteristics were similar for men and women because women tended to have more peaks of intermediate height, whereas men showed only very high peaks in that range, with most of their pulses being excluded due to low amplitudes. The sex difference in pulse amplitude became apparent only in the range below 1 /ug/L and the full range. Similarly, men and women had indistinguishable averaged valley levels in the range above 1 fig/L because the notches in high peaks are counted as valleys. However, when "true" valleys are considered (i.e. those in the previously undetectable range), women exhibited significantly higher valley levels than men. The number of peaks per 24 h in the range above 1 /xg/L was higher in women, but the opposite was true in the range below that limit. When the full range is considered, peak frequency was the same in both sexes. These seemingly contradictory findings are due to the fact that in women many peaks were of sufficient amplitude to penetrate the 1 /ug/L detection limit, whereas in men most peaks had amplitudes below that level. For the same reason, peak intervals showed disparate values for the two sexes in the range above 1 /ug/L, but not in the other ranges. These data demonstrate that caution should used in
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 13 November 2015. at 08:15 For personal use only. No other uses without permission. . All rights reserved.
PLASMA GH PATTERN IN THE BASAL STATE
1683
TABLE 1. Characteristics of circadian plasma GH patterns in normal men and women Range of individual values
Range of means Integrated GH level (ng/L) Men
Women % of values >1 /xg/L Men
Women Mean peak ht (ng/L)° Men Women Mean peak ht in range
