J. Endocrinol. Invest. 13: 777-786, 1990

Pulsatile secretion of luteinizing hormone in agonadal men before and during testosterone replacement therapy Ä.D. Genazzan!", G. Forti**, M. Maggi**, M. Milloni**, ~ Cianfanelli**, V. Guardabasso****, V. Toscano***, M. Serio**, and D. Rodbard* * Laboratory of Theoretical and Physical Biology, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD; ** Endocrinology Unit, Universitä di Firenze, Firenze, Italy; *** Clinica Medica V, Universitä La Sapienza, Roma, Italy, and **** Istituto di Ricerche Farmacologiche Mario Negri, Consorzio Mario Negri Sud, S. Maria Imbaro (Chieti), Italy

ABSTRACT. We have reevaluated the question regarding the pulsatile pattern of LH secretion in agonadal men before and following testosterone replacementtherapy. Five normal males were used as a reference group and four agonadal men were studied before and during replacement therapy with testosterone enanthate. All the subjects were sampled every 5 min for 12 h (08:00 to 20:00). Data were analyzed using the statistically based and validated pulse detection program DETECT. The normal subjects showed an LH pulse frequency of 10.2 ± 1.7 peaks/12 h (mean ± SEM) and a mean duration of 48.8 ± 14 min, while in agonadal patients without testosterone replacement the frequency of LH peaks (27.5 ± 2 peaks/12 h) was significantly higher than for normal subjects (p < 0.05), and the mean duration of peaks was lower than in controls (17.2 ± 1.2 min; p < 0.01). .Following chronic testosterone enanthate replacement therapy (200 mg im every two weeks) these patients showed an increase in the duration and a significant reduction in the frequency of LH peaks (from 27.5 ± 2 to 18.2 ± 2.1 peaks/12 h; p < 0.01) but pulse frequency remained significantly higher than for normal subjects (p < 0.01). This finding is independent of the choice of p values for false positive detection rate (p = 0.01 or p =

0.005), but it does depend on sampling trequency and is influenced by large (four-fold) chanqes in the thresholds for peak detection. Using a "discrete deconvolution" technique we estimated the instantaneous secretory rate (ISR) for the two groups of patients. The results using ISR corroborated the findings obtained using analysis of observed plasma LH measurements. ISR computation also showed that the duration of the secretory events of the gonadotropes is significantly shorter (p < 0.01) than the one estimated on plasma concentration, both in normal subjects and in agonadal patlents before and during testosterone administration. In conclusion: LH pulse frequency observed in basal conditions in agonadal men was much higher than previously reported in primary testicular failure; during conventional testosterone replacement therapy LH pulse frequency of agonadal men was significantly reduced but still higher (p < 0.01) than in normal men. This finding is probably related to the subnormal plasma levels of testosterone found in agonadal men during the replacement therapy; the analysis of data using a sampling interval of 10 min gave results similar to previous reports, confirming that the choice of sampling interval can markedly affect the evaluation of frequent LH pulsatile secretion. INTRODUCTION It is weil known that the pulsatile secretion of pituitary gonadotropins is dependent on the intermittent discharge of GnRH into the hypophyseal portal circulation (1). In animal studies it is possible to measure episodic secretion of GnRH directly (2, 3). However, due to the obvious difficulties of sampling the portal venous blood, virtually all clinical studies

Present address: Clinica Ostetrica Giilecologica, Universita di Modena, Moderia, 41100 Italy.

1

Key-words: LH, pulsatile secretion, agonadal man, testosterone replacement therapy, sampling frequency, variance models. Correspondence: Gianni Forti, M.D., Endocrinology Unit, University of Flo.rence, Viale Pieraccini 6, 50134 Florence, Italy. Received March 10, 1990; accepted July 7, 1990.

777

A.D. Genazzani, G. Forti, M. Maggi, et al.

on GnRH secretion have involved the frequent sampling of peripheral blood to determine the periodicity of the episodic release of LH, even though it has been recently demonstrated that human gonadotropes are able to secrete LH in vitro without GnRH control (4, 5). Normal male and female subjects have a weil defined number of LH pulses (2, 6-14), which is altered in a number of physiological and pathological states, e.g. idiopathic hypogonadotropic hypogonadism (15-17), primary testicular failure (18-20), hyperprolactinemia (20), renal failure (21), hypothalamic amenorrhea (22), polycystic ovarian disease (PCOD) (23). Accordingly, we sought to reexamine the pulsatile secretion of LH in normal subjects and agonadal men during conventional testosterone replacement therapy. We used both a very sensitive and specific algorithm for pulse detection, DETECT program (24, 25) and a fairly intensive sampling paradigm (every 5 min for 12 h) since recently Veldhuis et al., (25), have demonstrated that the estimation of pulse frequency is highly dependent on the sampling frequency. This study aimed also to verify the effect of some experimental and methodological choices (i.e. various sampling paradigms, data deconvolution, different thresholds for pulse detection) on the estimation of LH pulse frequency when performing a clinical protoco!. MATERIALS AND METHODS Subjects Four agonadal men underwent this study; their ages ranged between 21 and 43 yr. Three of them had been orchiectomized bilaterally because of a teratocarcinoma at least 2 yr before; in the fourth patient the diagnosis of anorchia had been made in the peripubertal period on the basis of cryptorchidisrn, high levels of gonadotropins, lack of testosterone response to hCG administration and failure to detect gonadal tissue on surgical exploration. The patients had normal SMI and showed no evidence of hepatic or renal disease and plasma levels of cortisol, T4, TSH, PRL and GH were within the normal ranges. The patients, who had suspended testosterone replacement therapy at least 3 months before this study, were admitted to the Endocrine Unit the day before the sampling. An iv heparin lock was inserted in an antecubital vein one h before commencing venous sampling at 5-min intervals for 12 h, from

778

08:00 to 20:00. The subjects were permitted to eat three meals while sitting in bed, but otherwise were generally reclining. We elected to sam pie onlv for 12 instead of 24 h to avoid the problems of overnight sampling and to permit the use of more frequent sampling. The pulsatility study was repeated 3 months after initiation of testosterone replacement therapy (200 mg testosterone enanthate (TE) im every 2 weeks], with the sampling performed on day 14 after the last testosterone injection. Five healthy men between the ages of 27 and 40 yr were selected as controls on the basis of history of normal sexual development, serum LH, FSH and T concentrations within the normal adult male ranges, and normal semen analysis. Each subject was hospitalized and underwent a frequent sampling as described above. Slood sam pies were immediately centrifuged after sampling and serum sampies were stored frozen at -20 C for subsequent measurement of LH. The study protocol was approved by the University of Florence Hospital Human Studies Committee, and all subjects gave informed consent. Assays Serum LH concentrations were determined with a RIA, using BIODATA (ARES SERONO, Milano, Italy) reagents. All sampies from each subject were analyzed in triplicate in the same assay. To obtain a precise evaluation of measurement error, we also assayed in triplicate at least 30 sam pies from a serum pool from the same individual (obtained by combining 50 JLI aliquots from each of the 144 serum sampies). Based on two quality control sampies, the average within- and between-assay % coefficients of variation (CV) were 5.6% and 10.4%, with 22 and 10 degrees of freedom (df) respectively. When the serum pools for 13 studies in 9 subjects (5 controls and 4 agonadal men) were analyzed, the within-assay % CV ranged from 3.7 % to 8.6 % (df = 29) with a mean of 4.6% ± 0.6 (SEM) (df ~ 390). These results were confirmed by analysis of the variance model for the triplicates of the 144 observations in each of the 13 studies with frequent sampling. The concentration corresponding to 90% B/8 0 was 2.2 mUlml of serum. T, dihydrotestosterone (DHT) and estradiol (E2 ) were also measured in one sampie of each serum pool of normal controls and agonadal men, by RIA methods previously described (27), after chromatography of plasma extracts.

LH pu/satile secretion in agonada/ men

ed to provide reduction of noise without significantly affecting the signal. We used the smoothing rule X'i = (X, _ 1 + 2Xi + Xi + 1)/4 also known as the Hanning procedure (30, 31). The variance tor ISR and smoothed ISR were derived from the overall predicted variance of the plasma data, as S21SR = 2 * S2x and S2ISR. smoothed = 3/4 * S2x respectively (32; Rodbard D, Guardabasso V and Veldhuis JD, unpublished data).

Statistical analysis

Statistical analyses were performed using Student's t test, paired or unpaired, as appropriate, after 10garithmic transformation of data. Estimation of variance model of LH data

In addition to the estimate of the random measurement error obtained using 2 quality control sampies and the serum pools from each subject, we analyzed the variance model using program, PREDETEC WK1, which is implemented as series of "macros" in a spreadsheet for LOTUS 1-2-3 for IBM-PC compatible computers. This program evaluates the best of 5 different variance models in terms of the smallest root mean square (RMS) error for s, and provides the coefficients for the variance model to be used in program DETECT.

CONTROL

LH pulse evaluation

The presence of LH pulses was determined using the program DETECT, which detects peaks using a previously described algorithm (24) using two types of logic: a) analysis of first derivatives of data, for the detection of rapid events relative to sampling frequency; b) fitting of linear segments, for the detection of slow events. Slowly rising or falling sections of data, with slope significantly different from 0, were detected. Data from each subject were analyzed using DETECT with two different values of p (p = 0.005 and p = 0.01 ) tor the nominal false positive rate and also with different threshold levels, ie twice and four times the threshold estimated from the predicted measurement error. Program DETECT also allows one to compute the instantaneous secretory rate (ISR) to estimate the spontaneaus LH secretory rate using a discrete deconvolution algorithm (24, 28). Calculation of ISR permits one to reduce or eliminate the peak broadening or smearing which arises due to the slow metabolic clearance of the hormone (29). Since the discrete deconvolution algorithm is applied to hormone concentration measurements which are subject to experimental error, the deconvoluted series tend to convert this noise into rapid oscillations, when sampling frequency is high relative to hormone clearance kinetics. These oscillations might increase the number of false positive errors. To reduce the effect of such measurement error, the ISRvalues can be subjected to linear smoothing. The degree of smoothing of the data can be adjust-

---l

~

I.S.R.

,-- -- - - - - - -

- T- - --,

SMOOTHED I.S.R. c:

E

10

3

C

2

E ::>

E

~

0 -1-=------\l--4-~=--------"__j

8AM

TIME

8PM

Fig. 1 - Represenlative pattern ot plasma LH pulsatile secretion in anormal subjecl (panel A), compared to calculated ISR (panels B) and 10 Ihe smoothed ISR (panel Cl. The arrows indicale Ihe significanl pulses aetecteä by program DETECT.

779

AD. Genazzani, G. Forti. M. Maggi, et al.

panel A) and during (Fig. 2, panel B) replacement therapy . Panels C and 0 show the ISR and panels E and F show the smoothed ISR. The arrows show the locations of peaks detected by pragram OETECT at the nominal p = 0.01 level for false positive errors. In normal men the mean LH values in serum pools ranged fram 3.8 to 7.4 mUlml (5.8 ± 1.1 mU/ml; mean ± SEM). Using 5min sampling the detected pulse frequency was 10.2 ± 1.7 pulses/12 h (range 5 to 14). The absolute height of the peaks (above 0) ranged from 3.3 to 12.4 mUlml (mean : 7.7 ± 0.3) and the mean duration was 48.8 ± 14 min (Table 1). T, OHT, E2 and LH levels in both groups of subjects are summarized in Table 2. lt is evident that during treatment T levels in agonadal men are below the lower limit of the normal range, whereas OHT and E2 are in the normal range. The LH serum pool levels of the agonada l men before TE administration ranged fram 24 to 56.9

To examine the rate of false positive peaks we also used pragram OETECT to analyze the data of the plasma pools of each subject together with the time series. At the nominal p level of 0.01 (1 %) for false positive errors the pragram found only 4 significant peaks in 390 sampies (30 measurements of a pool for each of the 13 time series) . The observed false positive rate of 0.9% was in excellent agreement with the expected value of 1% for cr iteria at the p = 0.01 level. An independent algorithm for pulse detection, CLUSTER method (33), was used to compare and validate results obtained with OETECT. CLUSTER was used with a 1 x 1 cluster size. RESULTS Figure 1 shows a typical LH pattern for anormal subject (panel A: plasma ; panel B: computed ISR; panel C: smoothed ISR) and Figure 2 shows the typical patterns for an agonadal man before (Fig. 2,

«JA

es

AGONADAL

AGONAD AL

BASELINE CO NDITIONS

DURING TE REPLACEMENT THERAPY

,,

:J)B

,,

26

E

22

~

18

E 24

:5

14

20

16 ~-----------1 16

c

10

I.S.R.

C

--l

~_-

16

0

I.S.R.

12

12

E

4

~

0 f---i V--:-:--Hf-flHl--l--'-H-'''-fJ-+-t1l'-'t ----'l-IH

o lhJYJ~~'W:!JJ':»/1l~l1b'l~

-4

-4

E

8

U')

E

:5

.a