0021-972X/98/$03.00/0
Journal of Clinical Endocrinology and Metabolism
Copyright © 1998 by The Endocrine Society
Vol. 83, No. 2
Printed in U.S.A.
COMMENTS
Patients with Cushing’s Disease Secrete
Adrenocorticotropin and Cortisol Jointly More
Asynchronously than Healthy Subjects
FERDINAND ROELFSEMA, STEVEN M. PINCUS,
AND
JOHANNES D. VELDHUIS
Department of Endocrinology (F.R), Leiden University Medical Center, Leiden 2333AA, The
Netherlands; Guilford, Connecticut (S.M.P.); Department of Medicine, National Science Foundation
Center for Biological Timing, University of Virginia Health Sciences Center (J.D.V.),
Charlottesville, Virginia 22908
after detrending the series was 0.394 6 0.033 in controls and 0.297 6
0.034 in patients with considerable overlap of the subgroups, giving
a sensitivity for this index of only 5%. In addition to previous findings
of increased individual irregularity of ACTH and cortisol release in
Cushing’s disease, we can now also demonstrate greater joint asynchrony of the circulating concentrations of these hormones. Thus,
Cushing’s disease disrupts ensemble network secretory dynamics
over individual hormone output. We conclude that, like GH-secreting
pituitary and aldosterone-secreting adrenal tumors, ACTH-secreting
pituitary tumors exhibit significant loss of orderly hormone release
patterns. Moreover, Cushing’s disease is marked further by deterioriation of bihormonal synchrony between ACTH and cortisol release,
thus suggesting further erosion of within-axis feedback control.
(J Clin Endocrinol Metab 83: 688 – 692, 1998)
ABSTRACT
We examined serum concentration time-series for ACTH and cortisol in 20 patients with pituitary-dependent ACTH excess (Cushing’s
disease) and in 29 age- and gender-matched controls. For each subject,
blood samples were obtained at 10-min intervals for 24 h. Joint
ACTH-cortisol synchrony was quantified using the recently introduced cross-approximate entropy (cross-ApEn) statistic. In patients,
cross-ApEn was greater than in controls (1.686 6 0.051 vs. 1.077 6
0.039, P 5 3.45 3 10216), giving a sensitivity of 85%. In control
subjects, but not in patients, cross-ApEn was correlated positively
with age (r 5 0.465, P 5 0.011) There was no gender difference in
cross-ApEn, nor a relationship between cross-ApEn and the 24-h
ACTH and cortisol secretion, in patients or controls. In contrast, the
maximal cross- correlation coefficient for the ACTH and cortisol series
C
LASSICALLY, endocrine diseases are thought to be the
result of excessive or decreased secretion of a particular hormone. More recently, abnormalities have often been
characterized by changes in pulsatility or irregularity of secretory dynamics apart from mean level changes. In Cushing’s disease, we recently demonstrated that the secretion of
ACTH, and also that of cortisol, is highly irregular, as quantified by the Approximate Entropy (ApEn) statistic (1– 4).
Disorganized hormone secretion has also been described in
other endocrine pathophysiologies such as GH secretion in
acromegaly and aldosterone secretion in aldosterone-secreting adrenal adenomas, and also for insulin secretion in relatives of patients with noninsulin dependent diabetes mellitus (5–7). Although considerable insight has been obtained
recently in associating irregular hormone secretion with various pathophysiological states by applying ApEn analysis to
individual hormone data series, we wish now to gain insight
into the network aspects operating operating within an axis,
beyond observations taking one hormone at a time.
A recently introduced measure of asynchrony or condi-
tional irregularity, cross-ApEn (8) affords such a statistical
capability applied to a pair of (hormone) signals. The usefulness of cross-ApEn for evaluating coupled endocrine systems was recently demonstrated in an analysis of LH-testosterone (T) synchrony in cohorts of younger vs. older males
(9). In particular, the older males exhibited markedly more
jointly asynchronous LH-T dynamics than did the younger
group, leading us to hypothesize that, analogous to onedimensional signal analysis, regularity and synchronicity
generally correspond more to (normal) physiology, whereas
greater irregularity (apparent process randomness) and increased (two-variable) asynchronicity correspond to pathophysiology, aging, and other variants of network disruption.
Below, we investigate the emerging hypothesis that ACTHcortisol bihormonal secretory dynamics are more asynchronous in Cushing’s disease than in controls. Validating this
theme would both offer further evidence of this broadlystated paradigm and provide some initial insights into explicit network manifestations of secretory tumors beyond the
vivid one-variable increases in irregularity seen for ACTH
and cortisol, individually (4).
Received April 11, 1997. Revision received July 30, 1997. Re-revision
received October 7, 1997. Accepted October 22, 1997.
Address correspondence and requests for reprints to: Dr. F. Roelfsema, Department of Endocrinology, Leiden University Medical Center,
Albinusdreef 2, 2333AA, Leiden, The Netherlands. E-mail: roelfsema@
rullf2.MedFac.LeidenUniv.nl
Patients and Methods
Patients
Fourteen female and six male patients (mean age 37 yr, range 17–74
yr) were studied. In all patients the diagnosis of Cushing’s disease was
688
COMMENTS
established by elevated 24-h urinary excretion of free cortisol, subnormal
or absent overnight suppression of plasma cortisol by 1 mg oral dexamethasone, absent or subnormal suppression of urinary cortisol excretion during an oral 2-day dexamethasone test (low-dose Liddle test),
suppression of plasma cortisol concentration by 190 nmol/L or more
during a 7-h iv infusion with dexamethasone at a dose of 1 mg/h (10),
positive immunostaining for ACTH of the pituitary adenoma, and clinical cortisol dependency for several months after selective removal of the
adenoma. Four patients had undergone previous transsphenoidal surgery. Two other patients had been treated many years earlier by unilateral adrenalectomy and pituitary radiation, which was the standard
therapy for Cushing’s disease in our institute before the application of
transsphenoidal surgery in the early 1970’s. All these six subjects had
clinical and biochemical relapse during follow-up studies.
The patients were hospitalized the evening before the sampling studies. On the following morning, an indwelling iv cannula was inserted in
a large vein of the forearm, and blood samples were withdrawn at
10-min intervals starting at 0900 h and for the next 24 h. A slow iv
infusion of 0.9% NaCl and heparin (1 U/mL) was used to keep the line
open. The subjects were free to move around, but not to sleep during
daytime. Meals were served at 0800, 1230, and 1730 h. Lights were turned
off between 2200 –2400 h, depending on the sleeping habits of the patient. As controls, 12 healthy male and 17 healthy female volunteers
underwent the same plasma sampling study. Their mean age was 42 yr,
range 30 – 63 yr (NS vs. patients).
Plasma samples for ACTH were collected on ice in chilled ethylenediamine tetraacetate-containing siliconized glass tubes, and samples
for cortisol were collected in heparinized tubes. The samples were centrifuged at 4 C, and within 30 min the plasma was separated, frozen, and
stored at -20 C until the assay was performed. All samples from any
subject were run in duplicate in the same assay (see below).
Informed consent was obtained from all subjects and patients, and the
study was approved by the ethical committee of the Leiden University
Hospital.
Assays
Plasma ACTH concentrations were measured in duplicate by immunoradiometric assay, using reagents obtained from the Nichols’ Institute
(San Juan Capristrano, CA). In our hands the detection limit of this assay
was 3.0 ng/L. The intra- and interassay precision varied from 2.8 –7.5%.
The cross-reactivity of this assay with aMSH, LH, FSH, TSH, GH, and
PRL was less than 0.1%. Plasma cortisol concentrations were measured
by RIA (Sorin Biomedica, Milan, Italy). The detection limit of this assay
was 25 nmol/L. The intra- and interassay precision varied from 2– 4%.
Cross-approximate entropy (cross-ApEn)
To quantify asynchrony (conditional irregularity), we use crossApEn, as introduced in ref 8, definition 5. As noted there, cross-ApEn
can be employed to compare sequences from two distinct yet intertwined variables in a network, herein applied to the joint ACTH-cortisol
time-series. Larger cross-ApEn values indicate greater joint signal asynchrony. The precise mathematical definition is thematically similar to
that for ApEn:
Let u 5 (u(1), u(2), . . . . u(N)) and v 5 (v(1), v(2), . . . . v(N)) be two
N-length sequences. Fix input parameters m and r. Form vector sequences x(I) 5 (u(I), u(I11),. . . . u(I1m-1)) and y(j) 5 (v(j), v(j11),. . . .
v(j1m-1)) from u and v, respectively. For each I#N-m11, set Ci m(r)(v u)
5 (number of j# N-m11 such that d[x(I), y(j)]# r)/(N-m11), where d[x(I),
y(j)] 5 maxk 5 1,2. . . . . m( u(I1k-1)-v(j1k-1) ), i.e. the maximum difference
in their respective scalar components. The Ci m(r)’s measure within a
tolerance r the regularity, or frequency, of (v-) patterns similar to a given
(u-) pattern of window length m. Then define Fm(r)(v u) as the average
value of ln Fm(r)(v u), and finally define cross-ApEn (m, r, N)(v u) 5
Fm(r)(v u)-Fm11(r)(v u). For this study, we applied cross-ApEn with
m 5 1, and r 5 0.2 to standardized ACTH ( 5 u) and cortisol ( 5 v)
time-series data, i.e. for each subject we applied cross-ApEn to the
{u*(I),v*(j)}series, where u*(I) 5 (u(I)-mean u)/sd u and v*(j) 5 (v(j)-mean
v)/sd v. This standardization, in conjunction with the choices of m and
r, ensures appropriate replicability properties for cross-ApEn for the
data lengths studied (1–3, 8).
689
Deconvolution analysis
Multiple parameter deconvolution was used to estimate various specific measures of hormone secretion and half-life from all plasma hormone concentrations and their dose-dependent intrasample variances
considered simultaneously (11). In normal subjects basal secretion constituted only a small part of the 24-h cortisol and ACTH production. In
most patients, however, basal release forms a substantial part of total
production. The latter is the sum of basal production and pulsatile
production, which were estimated as described (12, 13). For the present
study only the total production rates are reported. Detailed data on the
deconvolution parameters of ACTH and cortisol secretion in Cushing’s
disease were published recently (14). Here, we relate these data to
cross-ApEn calculations.
Statistical analysis
Data are given as the mean 6 sem, unless otherwise mentioned.
Statistical comparison of data in patients and controls (or male vs. female) was performed with the two-tailed Student’s t-test for unpaired
data. Linear regression analysis and ANOVA were applied where appropriate, and cross-correlation analysis both on raw data and after
prewhitening of the data series. Significant parametric differences were
corroborated by non-parametric tests (Mann-Whitney U test, Wilcoxon
signed rank test). Calculations were made with SPSS Windows version
7.0 and with Systat (SPSS Inc., Chicago, ILL). P , 0.05 was considered
significant.
Results
Cross-ApEn(1, 20%) was 1.686 6 0.051 in patients and
1.077 6 0.039 in controls (P 5 3.45 3 10216). In Fig. 1, the
plasma ACTH and cortisol profiles of a female control subject
and a patient with Cushing’s disease are depicted with the
calculated cross-ApEn for both subjects. We determined a
significant positive correlation between cross-ApEn(1, 20%)
and age in controls (r 5 0.465, P 5 0.011), but not in patients
(r 5 0.067, P 5 0.779). From the 95% confidence interval,
shown in Fig. 2, it is clear that only 3 out of 20 patients had
a cross-ApEn value within the normal age-related range,
giving a sensitivity of 85%. Cross-ApEn did not correlate
with the 24-h ACTH and cortisol production, estimated with
the deconvolution analysis. The classical cross-correlation
function (CCF) was significant for all ACTH-cortisol concentration data series, and the maximal r value obtained in
controls was 0.74 6 0.04, and 0.56 6 0.05 in patients (P 5
0.007). The mean time lag at maximal ACTH and cortisol
correlation was 7.9 min in controls, and 9.5 min in patients
(NS). Because the raw data series exhibited a high degree of
autocorrelation, we removed these by differencing the raw
data series with a lag of 10 min. After this procedure, the
cross correlation decreased to 0.39 6 0.03 in controls and to
0.30 6 0.03 in patients (P 5 0.046). No significant regression
of the CCF on age was present. Notably, the CCF obtained
after detrending the data series resulted in an almost complete overlap of the data of patients and controls (see Table
1). In addition, we also measured copulsatility with a specific
hypergeometric probability-density program (15), after identifying significant peaks in the data series with Cluster analysis (2x1 cluster size, t-statistics 5 2.0 for up- and downstrokes, ref. 16). In a time window of 10 min, patients had 149
coincident peaks, while the expected number by chance
would be 97 6 7.9 (mean 6 sd). In the control group, the
respective numbers were 197 and 114 6 8.9. In terms of
z-score, in controls the coupling was 9.3 sd above expectation, and in patients 6.6 sd.
690
COMMENTS
JCE & M • 1998
Vol 83 • No 2
FIG. 2. Relationship between cross-ApEn(1, 20%) and age in patients
with Cushing’s disease (‚) and control subjects (E). There is a pronounced statistical separation between controls and patients, with
only three patients having a cross-ApEn value within the control
subject limits. Hence, the sensitivity of cross-ApEn is 85%. The regression line and the 95% confidence lines are shown. The correlation
coefficient was 0.465 (P 5 0.011) for controls, but not significant in the
patients, indicating increasing asynchrony with advancing age in the
healthy cohort only.
FIG. 1. Plasma concentrations of ACTH (dotted line) and cortisol
(continuous line) in a female patient with Cushing’s disease (upper
panel) and a control subject (lower panel), each sampled at 10-min
intervals for 24 h.
Because six patients had undergone previous treatment(s),
we specifically compared the results obtained in this subgroup with those of the other (untreated) patients. None of
the above-mentioned parameters was different in these
groups, and P values were between 0.28 and 0.93 (individual
data not given).
We then explored whether ApEn or cross-ApEn exhibits
diurnal variation in patients and controls. To this end the
24-h period was divided into three equal parts of eight hours,
starting at 0900 h in the morning. ApEn(1, 20%) for ACTH in
patients was 1.085 6 0.036, 1.124 6 0.038 and, 1.087 6 0.027
in the respective 8-h blocks (ANOVA P 5 0.264), and the
respective means for the controls were 0.937 6 0.054, 0.948 6
0.041, and 0.826 6 0.031 (ANOVA P 5 0.076). ApEn(1, 20%)
for cortisol in patients was 1.083 6 0.037, 1.173 6 0.035, and
1.168 6 0.032 (ANOVA P 5 0.016), and for controls 0.923 6
0.036, 0.897 6 0.033 and 0.821 6 0.035(ANOVA P 5 0.135).
The ANOVA’s for the mean concentrations of cortisol and
ACTH during the 8-h blocks were statistically insignificant
for the patients (P 5 0.171 and 0.948, respectively) and highly
significant for controls (P , 0.0005, individual data not
given).
There was an overall significant difference of cross-ApEn
between patients and controls (P , 0.005). Gender had no
FIG. 3. Cross-ApEn (1, 20%) in patients (Œ) and controls (F). The 24-h
period was divided into three periods of 8 h each. Both patients and
controls exhibited a diurnal variation in cross-ApEn, although with
a clearly different pattern. Note that within each time period the
degree of ACTH-cortisol asynchrony in patients is greater (higher
cross-ApEn) than in control subjects.
statistical influence on the results. In controls, the ANOVA
was significant (P 5 0.036), manifesting a clear decrease in
cross-ApEn during the third period, i.e. during the night and
early morning, from 1.224 6 0.059 and 1.262 6 0.051 in the
first two 8-h blocks to 1.086 6 0.036. In patients, we also
found a significant diurnal variation (P 5 0.002), showing
lower cross-ApEn values during the first part of the period,
i.e. daytime. The values determined were 1.361 6 0.048,
1.485 6 0.054, and 1.505 6 0.034, respectively. This diversity
of cross-ApEn is clearly demonstrated in Fig 3.
COMMENTS
691
TABLE 1. Clinical characteristics and joint ACTH and cortisol statistics
Age (yr)
BMI (kg/m2)
Cross-ApEn (1.20%)
CCFa
CCF dentrendeda
ACTH productionb
Cortisol productionb
Controls (n 5 29)
Patients (n 5 20)
42.9 6 2.1
23.4 6 0.6
1.077 6 0.039
0.736 6 0.039
0.394 6 0.033
765 6 80
3520 6 230
37.2 6 3.4
28.2 6 1.2
1.686 6 0.051
0.557 6 0.050
0.297 6 0.034
4880 6 860
26100 6 11700
P value
0.143
0.001
3.45 3 10216
0.007
0.046
1.5 3 1028
2.15 3 1025
Data are given as the mean 6 SEM. Statistical comparisons were made with the two-tailed Student’s t test. Statistical comparisons for the
secretion rates were made after log transformation of the data.
a
Cross-correlation function r values.
b
ACTH and cortisol production rates were calculated with a multiparameter deconvolution method and results expressed as ng/L per 24 h
and nmol/L per 24 h, respectively.
Discussion
Via cross-ApEn, one can analyze the network aspects of an
interconnected hormone system, herein specifically the
ACTH-cortisol axis. The present findings indicate that, beyond the individual nodal changes in ACTH and cortisol
secretory irregularity manifested in Cushing’s disease, as
previously described in ref. 4, there is also a highly significant
network-timing disruption in joint ACTH-cortisol secretory
dynamics compared with matched controls. Extending the
hypothesis indicated in the Introduction, the present findings
suggest that enhanced bihormonal asynchrony may also be
associated with general classes of tumors, in addition to the
enhanced individual hormonal concentration irregularity
seen in several tumoral contexts. In Cushing’s disease, ACTH
is probably secreted without the coordinate control of hypothalamic hormones, such as CRH and vasopressin, while
the sensitivity to cortisol negative feedback is diminished.
The loss of synchrony between ACTH and cortisol levels in
Cushing’s disease may well further reflect the abnormal
waveform of ACTH release, and the concurrent nonpulsatile/high basal rates of ACTH release, which constitute a
relative less effective physiologic stimulus to responsive
adrenals.
An independent technique for investigating the coupling
between ACTH and cortisol pulses was the application of
copulsatility analysis (15). Although we found for both patients and controls a highly significant coincidence of pulses
within the biologically relevant time window, the coupling
tightness was much higher in controls than in patients, adding independent evidence for the asynchrony of the pituitary-adrenal axis in Cushing’s disease. At this point we want
to stress that copulsatility does not bear on subordinate dynamics, in contrast to both ApEn and cross-ApEn, which
algorithmically do so by comparisons over all pieces of time
series, and not only peaks or nadirs. Thus, if primary ACTH
pulses remained “answered” by an adrenal cortisol pulse,
with the latter’s subordinate features altered, cross-ApEn
would detect this, while the copulsatility analysis would not,
even apart from considerations as to the quality of the copulsatility algorithm and its statistical implementation.
The utility of cross-ApEn is also evidenced when the analyses are compared with those of the cross-correlation function (CCF) of the ACTH and cortisol time sequences. As
noted above, there is a large control-patient CCF value subgroup overlap, upon standard removal of (spurious) auto-
correlations prior to the CCF analyses. As discussed in ref. 9,
this observation is not surprising because, as indicated there,
the autocorrelation function and CCF are typically most effective in linear systems, e.g. ARMA models, oftentimes less
so for other classes of models. Nonetheless, the present results reinforce the complementarity and nonredundancy of
cross-ApEn, copulsatility, and CCF, as techniques to potentially detect changes in bivariate data behavior.
In our normal subjects, we found an age-related increase
of cross-ApEn pointing to a decrease in the synchrony of
ACTH and cortisol release with advancing age. A similar
finding has been reported for the LH-testosterone system in
healthy males and was attributed to a decreased feedback
signal strength or to diminished GnRH-LH-testosterone system responsiveness to feedforward and feedback signal control (9). However, there is a primary mechanistic difference
between the results of the present analyses and those of the
LH-T study. Specifically, in that study, each of LH and T
exhibited a significant increase in serial irregularity with
increasing age, in addition to the joint signal change in asynchrony. In contrast, within the control group of the present
study, there was an insignificant correlation between age and
ApEn of cortisol (r 5 0.207, P 5 0.450) and a borderline
correlation between age and ApEn of ACTH (r 5 0.409, P 5
0.046). This suggests the new inference that aging affects
network or ensemble aspects of the ACTH-cortisol system,
i.e. conduits, links, and/or global control mechanisms, more
so than it does individual glandular secretory activity. Further studies will be necessary to pinpoint the precise physiologic changes that elicit the presently seen statistical
contrasts.
Interestingly, we also found a clear, nonsex-dependent
diurnal variation of cross-ApEn, with the lowest values
(highest synchrony) during the night period, i.e. when the
ACTH-cortisol system comes to its maximal activity during
sleep. At this portion of the 24-h cycle, the normal system
shows tight coupling in contrast with the Cushing’s disease
patients. The physiological pathways by which this coupling
is accomplished are not clear at this time, as the nodal components of the regulatory system did not exhibit evident
diurnal properties. The observation as to the diurnal variation in cross-ApEn is only valid if the application of 48 data
points for the calculation of cross-ApEn is justified and when
the circulating hormone concentrations are well above the
detection limit in the 3 time segments. Monte Carlo simu-
692
COMMENTS
lation experiments show only slight degradation of the typical theoretical limits of (cross-)ApEn replicability for 48 data
points (1–3, 8). The application of 48 data points to determine
diurnal changes is therefore justified, and if anything, the
uncertainty in the application of 8-h blocks would tend to
diminish differences. Furthermore, in all 3 time segments,
both for patients and controls, ACTH and cortisol concentrations were detectable well above the detection limits of the
assays, so that assay noise could not have influenced the
calculation of ApEn and cross-ApEn.
Finally, although we have shown increased cross-ApEn of
ACTH-cortisol secretion in Cushing’s disease, it is not
proven at this point whether the increased cross-ApEn is
caused by Cushing’s disease or by the increased ACTHcortisol secretion. Additional studies in other states of hypercortisolism, such as untreated congenital adrenal hyperplasia, and the cortisol resistance syndrome will ultimately
be required to provide an answer to this question. Nonetheless, the present study further shows that cross-ApEn values
can be dissociated from secretion levels per se, as increased
cross-ApEn is found with advancing age without increased
secretion and with no relation existed between cortisol and
ACTH secretion rates and cross-ApEn values in Cushing’s
disease.
References
1. Pincus SM. 1991 Approximate entropy as a measure of system complexity.
Proc Natl Acad Sciences USA. 88:2297–2301.
2. Pincus SM, Huang WM. 1992 Approximate entropy: statistical properties and
applications. Commun Statis-Theory Meth. 21:3061–3077.
JCE & M • 1998
Vol 83 • No 2
3. Pincus SM, Goldberger AL. 1994 Physiological time-series analysis: what does
regularity quantify? Am J Physiol. 266: H1643–H1656.
4. Van den Berg G, Pincus SM, Veldhuis JD, Frölich M, Roelfsema F. 1997
Greater disorderliness of ACTH and cortisol release accompanies pituitarydependent Cushing’s disease. Eur J Endocrinol. 136:394 – 400.
5. Hartman ML, Pincus SM, Johnson ML, et al. 1994 Enhanced basal and disorderly growth hormone secretion distinguish acromegalic from normal pulsatile growth hormone release. J Clin Invest. 94:1277–1288.
6. Siragy HM, Vieweg WVR, Pincus SM, Veldhuis JD. 1995 Increased disorderliness and amplified basal and pulsatile aldosterone secretion in patients
with primary aldosteronism. J Clin Endocrinol Metab. 80:28 –33.
7. Schmitz O, Porksen N, Nyholm B, et al. 1997 Disorderly and nonstationary
insulin secretion in relatives of patients with non-insulin-dependent diabetes
mellitus. Am J Physiol. 272:E218 –E224.
8. Pincus SM, Singer BH. 1996 Randomness and degrees of irregularity. Proc
Natl Acad Sciences USA. 93:2083–2088.
9. Pincus SM, Mulligan T, Iranmanesh A, Gheorghiu S, Godschalk M,
Veldhuis JD. 1996 Older males secrete luteinizing hormone (LH) and testosterone more irregularly, and jointly more asynchronously, than younger males.
Proc Natl Acad Sci USA. 93:14100 –14105.
10. Biemond P, de Jong FH, Lamberts SWJ. 1990 Continuous dexamethasone
infusion for seven hours in patients with the Cushing syndrome. Ann Int Med.
112:738 –742.
11. Veldhuis JD, Johnson ML. 1992 Deconvolution analysis of hormone data.
Methods in Enzymology. 210:539 –575.
12. Roelfsema F, van den Berg G, Frölich M, et al. 1993 Sex-dependent alteration
in cortisol response to endogenous adrenocorticotropin. J Clin Endocrinol
Metab. 77:234 –240.
13. Veldhuis JD, Moorman J, Johnson ML. 1994 Deconvolution analysis of neuroendocrine data:waveform-independent methods and applications. Methods
in Neurosciences. 20:279 –325.
14. Van den Berg G, Frölich M, Veldhuis JD, Roelfsema F. 1995 Combined
amplification of the pulsatile and basal modes of adrenocorticotropin and
cortisol secretion in patients with Cushing’s disease: evidence for decreased
responsiveness of the adrenal gland. J Clin Endocrinol Metab. 80:3750 –3757.
15. Veldhuis JD, Johnson ML, Seneta E. 1991 Analysis of the copulsatility of
anterior pituitary hormones. J Clin Endocrinol Metab. 73:569 –576.
16. Veldhuis JD, Johnson ML. 1986 Cluster analysis: a simple, versatile and robust
algorithm for endocrine pulse detection. Am J Physiol. 250:E486 –E493.