Main

The study of anterior pituitary hormones has progressed from the documentation of random or fasting levels to the detailed characterization of baseline levels and pulsatility, which may be more important determinants of receptor interaction and function(1, 2). The study of hormone levels in the newborn period, for practical and ethical reasons, has largely been based on single samples. For example, high levels of GH have been documented in the newborn period, but its secretory pattern or function at this time is still largely unknown. The elucidation of the possible biologic roles of GH in the newborn period requires some knowledge of its pulsatility characteristics and thus prolonged periods of serial sampling. Previous dynamic studies of GH secretion in the newborn period have relied on long sample intervals, short periods of study, and multiple heel puncture(3–5). Repetitive sampling has been achieved through umbilical and peripheral arterial catheters, but again the sample period has been short(6, 7) or infrequent(7) and performed only on parenterally fed babies, some of whom were receiving treatment for respiratory distress. Growth hormone levels have also been assessed at 20-min intervals over 6 h during a dilutional exchange transfusion in symptomatic babies with polycythemia(8, 9).

Cord blood GH levels have been found to be significantly higher in SGA babies compared with AGA babies, whether born at term or preterm(10). In addition, after birth, basal GH levels are higher in SGA babies with peak GH levels after i.v. injection of 1 μg/kg GH-releasing hormone significantly increased compared with levels in AGA babies(11). These results demonstrate functional somtotropism in SGA babies in the first few days of life.

We have recently developed a novel automated microsampling system for use in newborn babies adapted from that originally designed at the National Institute for Medical Research, London, for studies in rodents(12). For the first time, we have been able to investigate spontaneous pulsatile hormone levels in healthy babies, both term and preterm, under physiologic conditions. Our system facilitates the automated withdrawal of tiny samples (20-40 μL) of blood diluted in heparinized saline through an indwelling i.v. cannula at 10-min intervals over a 12-h period, minimizing the amount of blood required to establish hormone profiles. The system complies with safety, clinical, and practical considerations in a neonatal unit. This report describes the validation of the automated microsampling technique and a pilot study of GH pulsatility in newborn babies.

METHODS

Subjects. Volunteer for validation studies. A healthy adult female volunteer aged 46 y was subject for the validation studies.

Babies. The microsampling technique has been used in 10 healthy newborn babies with gestational ages between 32 and 40 wk admitted to the Neonatal Unit at the John Radcliffe Hospital. Postnatal ages at the time of study ranged between 2 and 13 d. The birth weight of the babies ranged from 1.05 to 3.27 kg, and all were receiving between 1 and 3 hourly enteral milk feeds. Six infants were of appropriate weight and length for gestation (AGA)(>10th centile) and four infants were below the 10th centile for weight and length (SGA) (Revised Gairdner-Pearson growth charts 1988). All infants were clinically well and were not requiring supplemental oxygen. Six infants at risk of infection were receiving parenteral antibiotics, but the subsequent culture results proved negative, and antibiotics were discontinued after 48 h in all cases. All infants were normoglycemic and euthyroid on routine screening. The venous hematocrit ranged between 40 and 54%.

Ethical permission was obtained from the Oxford District Ethics Committee and informed parental consent was obtained in all cases.

The microsampling system. The basic microsampling system as applied to experimental rodents has been previously described(12). Essentially it consists of a pump, 3-way solenoid valve, a fraction collector, and microcomputer interfaced to the system (Fig. 1). A peripheral i.v. cannula is connected to one port of a low dead space solenoid 3-way valve (LYFA 1218032H Lee Products, Buckinghamshire, England). The common port of the valve is connected via a peristaltic pump (Gilson Minipuls, M312) to a reservoir of heparinized saline(7.5 U/mL) contained in a collapsible bag placed above the level of the baby. The valve is held vertically with the ports pointing down, approximately 20 cm above the baby. The normally closed valve port is connected to a fraction collector (Gilson model 203, Gilson, France). The interface is a single channel version of the unit developed and built by John Lawin(12). The device controls the pump speed and direction, operates the valve, and controls movement of the fraction collector arm. The procedure is controlled by software running on a BBC Master 128 microcomputer. Essentially, the pump withdraws blood through the valve, a mid-portion of whole blood is diverted toward the fraction collector, and the remaining blood/saline mixture is then returned to the infant with an additional flush of approximately 20 μL of heparinized saline. The sample is then pumped into a collecting tube with a flush of saline to make a final whole blood dilution of 1:4. To complete the cycle, the fraction collector is flushed with approximately 40 μL of heparinized saline.

Figure 1
figure 1

Diagrammatic representation of the automated microsampling equipment.

Full size image

To adapt the system for use in newborn babies, a number of fail-safe stages were incorporated in both hardware and software. To provide adequate patient protection from electrical currents, the Gilson peristaltic pump has been enclosed within a sealed Perspex container, and the Gilson 203 fraction collector and BBC computer system are connected to the mains supply by a medical grade isolation transformer. Before automatic sampling, all lines are primed with heparinized saline, and the pump tubing is clamped. Although the system is fully automated, continuous medical supervision was maintained while the system was in operation.

The samples are spun and separated, and hormone levels are determined in the diluted plasma. To determine the concentration of hormone in neat plasma, allowance must be made for the venous hematocrit in the whole blood. Thus the hormone concentration in the diluted plasma is multiplied by a dilution factor(DF), calculated according to the venous hematocrit (Hct) expressed as a fraction of 1. Equation where 4 is derived from the 1 in 4 dilution of whole blood with saline.

Validation experiments. To determine the precision of the microsampling system and to determine whether correction using the calculated dilution factor accurately reflects levels in neat plasma, two validation studies were undertaken. In each study, two sampling cannulae were inserted into different peripheral veins of the adult volunteer. A large one was used for intermittent whole blood sampling, which coincided with samples taken from the other cannula that was attached to the microsampling system. The microsampling conditions, including the size of cannula and the volume and dilution of each sample, were the same as those used in the babies.

The substances measured in both diluted and neat plasma were albumin in one study and inulin in the other. These substances were chosen as they can be measured with accuracy over a wide range of concentrations. For the albumin experiment, 18 consecutive diluted blood samples were collected at 10-min intervals using the automated microsampling technique together with simultaneous whole blood samples taken through the other i.v. cannula. The concentration of albumin was determined in neat plasma separated from the whole blood samples, in the supernatant of the automated samples (the results of which were then corrected using the dilution factor calculated from the venous hematocrit) (vide supra), and in samples diluted 1:8 in the laboratory and then corrected for dilution. The mean levels over the 18 samples from the three types of sample were compared, and the CV between samples was deduced.

The inulin study was performed on two occasions. The subject received a constant infusion of inulin (100 mL/kg/h) starting 1 h before the sampling periods. Nineteen consecutive microsamples of blood were collected at 10-min intervals during each run, together with simultaneous blood samples taken through the other i.v. cannula. The mean ratio of neat plasma inulin concentration to the microsample inulin concentration was determined to give a true dilution factor and compared with the predicted dilution factor using the venous hematocrit (vide supra). Similarly the mean ratio of the inulin concentration in the neat plasma to that in plasma from the whole blood diluted 1:8 in the laboratory was compared with the known manual dilution factor. The CV of the dilution factor measurements of both the microsamples and manually diluted samples were determined.

Studies in newborn babies. All babies had 10-min blood sampling over 12 h using the microsampling system. Samples of venous blood (35 μL) were withdrawn through a peripheral i.v. cannula and collected diluted (1:4) in heparinized saline (total volume of blood withdrawn is 2.5 mL). In the subjects where a cannula had been sited in a large peripheral vein for antibiotic treatment (antecubital fossa or long saphenous), this was used for sampling; in those without a cannula, this was inserted at the time of routine blood sampling. Growth hormone levels were measured at 10-min intervals.

Assays. GH. Plasma GH concentrations in the diluted plasma samples were determined by an immunoradiometric assay (NETRIA, St. Bartholomew's Hospital, London, UK). Fifty microliters of the diluted plasma samples were added to 350 μL of working assay buffer [10 mL of 0.54 M phosphate buffer (pH 7.4), 5 mL of 10% Tween 20, 1 g of BSA up to 100 mL distilled water]. Iodinated (125I) anti-GH was reconstituted in 1 mL of distilled water and diluted in 5 mL of working assay buffer to give 100 000 cpm. Fifty microliters of the GH tracer and 50 μL of solid phase were added to the sample and incubated overnight on a rotator at room temperature. One milliliter of 0.1% Triton X-100 was added to each tube and spun at 3000 rpm for 15 min and decanted, and the latter stage was repeated. Calibrators ranging from 0.125 to 100 μg/L were reconstituted in 2 mL of distilled water and similarly used in a volume of 50 μL in the immunoradiometric assay. The lowest detectable concentration of the assay was 0.05 μg/L, and intrassay CVs were 14.5, 8.9, 4.2, and 5.7% at concentrations of 0.25, 1.25, 2.5, and 125 μg/L, respectively. The working range of the assay reproducible on diluted samples was between 0.25 and 6.25 μg/L. Plasma GH levels between 2.0 and 100 μg/L can be detected in the samples obtained using the microsampling technique.

Albumin. Plasma albumin was measured using an ELISA(13). A human albumin calibrator was obtained from Sigma Chemical Co. (Poole, UK) and polyclonal rabbit anti-human albumin and horseradish peroxidase conjugated anti-human antibodies from DAKO (High Wycombe, UK). Automatically diluted plasma was further diluted 1:100 000 and neat plasma diluted 1:800 000.

Inulin. Plasma Inutest was measured enzymatically on an automated spectrophotometer (Cobas MIRA, Roche Diagnostics, Welwyn, UK) using sorbitol dehydrogenase (Sigma Chemical Co., Poole, UK) after perchloric acid hydrolysis of Inutest to fructose(14).

Data Analysis. To estimate mean, baseline, and peak GH levels, the distribution of concentrations of GH in the 12 hourly GH profiles are expressed in terms of the observed concentrations for 5, 50, and 95%(OC5, OC50, and OC95)(15). The OC5 is the threshold at or below which the hormone concentrations are measured for 5% of the time and gives an estimate of baseline, the OC50 is the value below which 50% of observed concentrations lie giving an estimate of the mean, and the OC95 the level above which 5% of the concentrations lie and gives an indication of peak GH levels.

GH profiles were also analyzed by Fourier transform, which is an unbiased method of examining all the oscillatory signals within an array(2). Fourier analysis allows the power of each oscillatory function to be displayed as a histogram of power versus frequency (power spectrum). Fourier transforms can thus produce composite spectra of data showing all the dominant and subdominant harmonics together with an estimate of power and amplitude. The data can be pooled using parametric statistics to yield spectra that can composite for the groups data.

Pre- and postprandial changes in GH levels were analyzed by serial array averaging(16). Using this technique, the average percentage change in GH at 10-min intervals was determined from a set point(the time of a feed) within the GH profile. By combining data from a number of feeds from the babies studied, other factors affecting GH levels randomly, are averaged out. Postprandial changes in GH were averaged at 10-min intervals over a 90-min period after 30 feeds given to 8 of the 10 infants, 7 of whom were preterm. The two who were excluded from analysis were receiving hourly feeds. Preprandial changes in GH were averaged over a 60-min period before 28 feeds in the 8 babies who were fed 2 to 4 hourly.

Data are expressed as mean ± SD unless otherwise stated. All other statistical analyses were made by standard methods.

RESULTS

Validation studies. The corrected plasma albumin levels assessed by the microsampling technique were very similar to those measured in neat plasma [36.0 ± 2.5 (±SD) cf. 35.8 ± 2.2 g/L] and in those estimated after manual dilution of plasma from the whole blood in the laboratory (34.5 ± 1.2 g/L) (p > 0.05), indicating that dilution using the microsampling system is precise, and correction using the dilution factor calculated from the venous hematocrit accurately reflects levels in neat plasma (Table 1). The CV of the albumin concentration assessed by the manual dilutional method was slightly less than the CV of the albumin concentration assessed by the microsampling method (3.6 cf. 7.1%), probably reflecting the unvoidable dilutional variation of the microsamples inherent in drop collection rather than pipetting.

Table 1 Results of validation studies
Full size table

The mean measured dilution factor (i.e. the ratio of diluted to neat plasma levels) of the microsamples in both of the inulin studies slightly exceeded the dilution factor calculated from the hematocrit (7.0 cf. 6.5 and 6.8 cf. 6.5), although the mean measured dilution factor of the manually diluted samples was slightly less than the known dilution factor(7.7 cf. 8.0 and 7.4 cf. 8.0) (Table 1). The differences were small, but may suggest that the microsamples were slightly more dilute than predicted because of a failure to collect all the blood sample within the 1:4 dilute volume from the fraction collector tubing. The CVs of the dilution factor measurements of the microsamples exceeded those of the manually diluted samples by 1.8 and 3.2%, implying a similar dilutional variation in the microsamples as assessed by the albumin method. Nevertheless, considering the samples are collected without a volumetric pipetting step, the precision of the sampling procedure was more than adequate to permit valid interpretations of the differences observed in measurements made in the patient samples.

GH pulsatility. GH pulsatility was demonstrated in all of the babies studied. A 12-h GH profile in a SGA (33-wk gestation, aged 3 d) and in a AGA (35-wk gestation, aged 2 d) infant is shown in Figure 2. In the SGA infant, a faster GH pulse frequency and higher baseline and peak GH levels are seen. A preliminary comparison has been made between SGA and AGA babies. Distribution analysis indicated that the observed concentrations (OCs) of 5, 50, and 95 below which 5, 50, and 95% of GH concentrations fell, respectively, were different in the two groups(Table 2). The mean OC5, OC50, and OC95 were not significantly different between the SGA and AGA groups. Fourier transform analysis indicated the dominant pulse periodicity in the AGA group was 180 min, but in the SGA group there were co-dominant pulse periodicities of 90-100 and 140 min (Fig. 3).

Figure 2
figure 2

GH profiles from (A) 33-wk SGA infant age 3 d and (B) 35-wk AGA infant aged 2 d.

Full size image
Table 2 Clinical details and distribution of concentrations of GH (μg/L) in SGA and AGA infants
Full size table
Figure 3
figure 3

Spectral power of Fourier transformation of GH data in(A) SGA and (B) AGA infants. Ordinate shows spectral power as quantitative function of amplitude. The co-dominant pulse periodicities(maximum spectral power) of GH in group A were 90 + 140 min, and the dominant pulse periodicity in B was 180 min.

Full size image

Serial array averaging indicated a relationship between GH and feeds (Fig. 4). A clear postprandial elevation in GH (173± 18% of baseline) was observed reaching a maximum 60 min after a feed. However, no significant preprandial changes were observed.

Figure 4
figure 4

Serial array averaging showing serial changes in GH levels after feeds.

Full size image

DISCUSSION

We describe a unique technique for the automated collection of samples of venous blood diluted in saline from small babies. It enables the withdrawal of frequent consecutive blood samples through an i.v. cannula sited in a peripheral vein and hence far less traumatic(3–5) or invasive(6) than other methods used previously for multiple sampling. In addition, in the technique where blood is taken manually from an umbilical or peripheral arterial catheter, sampling is limited to the few days during which the catheter remains in situ, whereas the automated technique of sampling from a peripheral vein may be applied at any time from birth onward. Furthermore, the microsampling technique permits nonstressful sampling under physiologic conditions in preterm and term babies who may be clinically well and receiving enteral feeds. The value of the automated collection system is that it avoids waste, and very small samples (20-40 μL) can be collected. Sampling is uniform, and the sampling frequency is dependent on the cycling time, which is determined by the speed with which the blood is pumped. Blood is pumped very slowly, as the connecting tubing is of small internal diameter to limit dead space, and by using very low pump pressures the venous wall does not collapse around the tip of the cannula. Nevertheless, it is possible to collect between 6 and 10 samples each hour, which is adequate for the definition of hormone pulsatility. A number of adaptations to the equipment originally used for animal studies have been made to ensure the safety of the equipment, which in addition, is supervised at all times. No complications have been encountered.

With microcomputer control, the sampling pattern, timing, volume dilution, and frequency are easy to vary to accommodate different experimental requirements. The possibility of saline contamination of the blood sample is minimized by withdrawing a relatively large sample of blood and subsampling from this. All of the remaining sample, including the leading edge, which is mixed with the saline, is returned to the baby, and tubing is regularly flushed with heparinized saline. Although whole blood is obtained, the red cells sediment rapidly, leaving a cell-free plasma/saline supernatant. No hemolysis is observed, and if the analyte concentration in the supernatant can be accurately measured, the concentration in plasma can be deduced. A multichannel version of this system has been used extensively in rodents(12, 17). The new single channel version has now been validated using adult volunteers with identical cannulae and sampling conditions. The equipment has been shown to provide a consistent dilution, and correction using the dilution factor calculated from the hematocrit accurately reflects levels of hormone or other substances in neat plasma. The observed dilutional error of <4% is acceptable in situations in which endogenous hormone levels are high and pulsatile. GH levels in the newborn period are elevated, and even diluted are within the working range of the assay. If the sensitivity of the assays permits, several hormones can be measured in the same microsamples, and their different pulsatility characteristics can be determined. The technique does, however, rely on the development of very specific assays, able to detect very low hormone concentrations.

As in other studies(6, 7), we have demonstrated that GH is pulsatile in the newborn period, but this is the first report of GH secretion from healthy babies over a 12-h period under physiologic conditions. Peak, mean, and baseline GH levels were higher than those observed in older children(18). Although the numbers were small, there was a trend toward higher mean, baseline, and peak levels of GH in the SGA compared with the AGA group, although the differences were not statistically significant. In addition, Fourier transformation indicated a shorter pulse periodicity in SGA infants. The higher GH levels in the SGA infants and the faster frequency pulsatility are unlikely to be accounted for by variation in either the timing of feeds, as the mean time between feeds was slightly greater in the SGA group, or in postnatal or gestational age, as the mean age at the time of sampling was higher in the SGA than in the AGA groups (7cf. 4.5 d, respectively), and mean gestational ages were similar in the SGA and AGA groups (33.5 cf. 35 wk, respectively). The pulse periodicity we observed in the AGA babies is similar to that detected by Miller et al.(6).

In conclusion we have demonstrated that this unique microsampling technique can be applied to the study of pulsatile hormones in the newborn period. We have been able to study normal physiologic GH secretion confirming previous observations of high GH levels, and for the first time, we have reliably shown a postprandial elevation in GH levels in preterm infants. Previous cross-sectional data of single GH measurements taken at variable times after a feed failed to show a significant postprandial rise in GH in the first 12 d of life in preterm infants(18). The observed characteristics of GH secretion in the newborn, and particularly in SGA babies, with high pulse amplitude and baseline concentrations and more rapid pulsatility, show many similarities to those observed in strvation and suggest an enhanced lipolytic metabolic role for GH in these subjects. The method that we have developed will permit elucidation of the relationship between GH and other growth factors including insulin, nonesterified fatty acids, and IGF-I to explore this hypothesis further.