The increased use of hormone therapy in children requires a detailed understanding of the basic effects of hormones on skeletal growth. Of special interest is testosterone therapy in boys with a constitutional delay in puberty, the purpose being to induce the pubertal growth spurt. Its effect on linear growth has been investigated, but its effect on craniofacial growth is not known. As a number of studies show a relation between the craniofacial growth spurt and the peak height velocity in body length, it can be assumed that an effect will also be reflected in craniofacial growth.

Androgens influence skeletal growth by direct and indirect (i.e., mediated by the growth hormone—IGF-I axis) mechanisms. Independent and additive contributions by gonadal steroid hormones and growth hormone to the adolescent growth spurt are evident (. These findings were demonstrated in vitro, in animals and in clinical studies. Recently, a new concept of the effect of androgens in growth regulation via their conversion to oestrogens has been proved. The stimulatory effect of testosterone on body growth in rats by modulating their hypothalamopituitary functions was shown by Jansson et al. (1983). Neonatally secreted testicular androgens imprint on the high-amplitude pulses, characteristic of growth-hormone secretion in adult male rats, which are more favourable for somatic growth than the low-amplitude pulses in female rats. Further, the much stronger suppression of longitudinal growth by neonatal rather than prepubertal castration indicates the important effect of the amplitude of the growth-hormone pulse on growth.

Very few studies have demonstrated effects of androgens on craniofacial growth, particularly in the pubertal period. Therefore, the present study was designed to clarify further the role of testosterone in pubertal craniofacial growth. To study the effects of endogenous testosterone secretion on the rat craniofacial complex, experiments were designed around neonatal mechanical and prepubertal chemical castration.

For the first study, pregnant inbred Wistar rats (n=18) were obtained from the Animal Laboratories of the University of Leuven. After birth, 97 male offspring were taken and randomly divided into two groups, identified by toe amputation at birth. The experimental group was castrated (n=52) 4 h after birth and the control group was sham-operated (n=45).

For the second study, pregnant inbred Wistar rats (n=14) were similarly obtained. After weaning, male offspring were randomly divided into three groups: two experimental groups, I (n=20) and II (n=20), and a control (n=24).

Animals were pair-fed a diet adequate for their nutritional needs and were kept in an air-conditioned and light controlled room with an ambient temperature of 23°C. The cages each contained five to six rats. The protocol for animal use was reviewed and approved by the Ethics Committee of the Medical School, University of Leuven.

For the second study, testosterone release was pharmacologically suppressed at the pituitary level by the injection of triptorelin (Décapeptyl-Dépôt®; Ipsen-Biotech, France). As puberty in the Wistar rat starts around day 25, the animals were treated at day 25; both experimental groups were injected intramuscularly with 10 ? triptorelin (100–200 mg/kg body wt). Group II received a second injection of the drug on day 45, i.e., in the time period (day 40–50) in which the pubertal testosterone rise was observed in the control rats of the first study. The rats of the control group received no injections. With this chemical castration we intended to induce a delayed pubertal testosterone rise to set up a model for constitutional delay in puberty.

For the first study, every tenth day from day 20 until day 70, seven to nine rats were taken from both groups; for the second study, all rats were taken every tenth day from day 30 to 110. A sagittal cephalometric radiograph of the skull was taken under general anaesthesia with a Röntgendevice Model D9 (Ritter A.G., Karlsruhe-Durlach) in a specially designed craniostat using a standardized technique with an occlusal film (Agfa Dentus M2 Comfort 2.25/3). The dose of anaesthetic (Nembutal) was adapted to the weight of the rat (25–30 mg/kg). General measures of growth, body length (from the nose tip to the anus), and weight (with an electronic balance) were also registered. Before anaesthesia, blood was collected from the tail around the same time of day, between 7 and 10 a.m. Plasma testosterone, growth hormone and IGF-I were measured by radioimmunoassay.

Photographic images were taken from the developed X-ray images and recorded on a Kodak CD–ROM. The craniofacial landmarks were digitized. The linear measurements were extracted from the data-points using a specially made program. The cephalometric technique included linear measurements from the analyses described by Vilmann (1969), Engström et al. (1982), Persson et al. (1989) and Kiliardis et al. (1985). The choice of the craniofacial landmarks was made in relation to the function of different structures represented in the skull. The skull-roof measurements were represented by the distances 1–2, 4–5, 1–3, 1–5; the vertical height by the distance 6–5; the cranial-base bones, including the basi-occipital, the basisphenoid, and the presphenoid, by 12–13, 13–14, 14–15; the mandible by 10–11; and the incisors by 6–7, 8–9. Replicate analyses showed no significant intra- and inter-observer differences.

Plasma testosterone was analysed using kits for testosterone ¹²⁵I radioimmunoassay (Byk-sangtec Diagnostica GmbH and Co. KG, von Hevesy-Strasse, D-63128 Dietzenbach). The kit consisted of testosterone antiserum from rabbits, testosterone ¹²⁵I tracer and testosterone standard dissolved in human serum and goat antirabbit gammaglobulins in polyethylene glycol solution. The detection limit of testosterone was 0.2 ng/ml. All experimental and control groups were assayed at the same time. The intra-assay coefficient of variation was between 5.8 and 6.3%. The interassay coefficient of variation was between 13.5 and 7.4%.

Radioimmunoassay for growth hormone was done in duplicate on all samples, using kits provided by the National Institute of Diabetes and Digestive and Kidney Diseases (Rockville, MD, U.S.A). The kit consisted of rat GH-I-7 for iodination, standard rat GH-RP-2 and monkey antirat GH-S-5. All treatment and control groups were assayed at the same time. The within-assay coefficient of variation was 1%.

Plasma IGF-I was measured by heterologous radioimmunoassay as described by Renaville et al. (1994). Dilution series of rat plasma with radioimmunoassay buffer showed good parallelism with the standard curve. The intra-assay coefficient of variation was 6.5% and the interassay coefficient was 15%.

Cross-sectional data from the first study were tested in the Procedure General Linear Model (2-way ANOVA) in SAS. An ANOVA model with two main effects, age and group differences and their interaction, was fitted to the data collected over the entire period. For variables showing a significant interaction effect, the same analysis was performed over the time period before and after physiological puberty (i.e., from day 20 to 40 and from day 40 to 70). Also the tenth-day time period with the largest difference between the two groups was examined by the two-sample t-test.

Analysis of the longitudinal data in the second study required a statistical technique for repeated measurements so a linear mixed model was used. The inter- as well as the intra-animal and the group variability were taken into account. An initial check of the data was made with the SAS procedure MEANS. The original variables had a normal distribution. The model used was defined as follows. For each time-point and group (experimental I, II, and control), a mean response was estimated. A deviation of growth in the treated rats (groups I and II) from that of the control group was expected from day 25. Owing to the experimental design, groups I and II were combined for the day 25–45 period. As the effect of the drug may disappear some time after the injection, it was reasonable to expect a catch-up towards normal growth for group I. Owing to the second injection at day 45, group II might have continued to deviate further from the control group, and also from group I. Sequential F-tests were used to compare the three groups. For the observed differences, we determined the latency as well as the time period during which the deviation was observed.

Throughout the statistical analysis a significance level of p<0.05 was used.