- Email: [email protected]
Regular fluctuations in growth hormone (GH) release determine normal human growth
Growth Hormone & IGF Research 1999, 9, 114–122 Article No. ghir.1999.0095, available online at http://www.idealibrary.com on
Regular fluctuations in growth hormone (GH) release determine normal human growth 1,2 M. S. , N. K. S. Thalange2, P. J. Foster3, V. Tillmann2, D. A. Price2, P. J. Diggle4 and P. E. Clayton1,2 1
Endocrine Sciences Research Group, Department of Medicine, Stopford Building, University of Manchester, Oxford Road, Manchester M13 9PT, UK; University Department of Child Health, Royal Manchester Childrens Hospital, Hospital Road, Pendlebury, Manchester M27 4HA, UK; 3Department of Mathematics, University of Manchester, Oxford Road, Manchester M13 9PT, UK; 4Medical Statistics Unit, Department of Mathematics and Statistics, University of Lancaster, Lancaster LA1 4YF, UK 2
Summary Growth hormone (GH) is the principal hormone associated with growth through childhood, but in a normal child the amount of GH secretion does not appear to be critical in the generation of normal growth rates. We have assessed the relationship between growth and urinary GH (uGH) output in a longitudinal study of 29 healthy prepubertal schoolchildren (13 male, 16 female; age 5.7–7.8 years) over 1 year. Height and uGH were measured three times a week. Individual height velocity curves were derived using non-linear regression. Growth was expressed in terms of the total increment over the year (∆Ht, cm), height velocity standard deviation score (HVSDS) and the average size of individual growth spurts. Urinary GH data (ng) were expressed as a weekly average. Mean uGH did not correlate with stature or growth over the year. However, the coefficient of variation of uGH was correlated with height standard deviation score (HtSDS, r = 0.38, P < 0.05), while the relative constancy of short-term change in uGH (coefficient of incremental change, ∆INC) was inversely correlated with ∆Ht (r = – 0.44) and HVSDS (r= – 0.42, both P < 0.05) but not with HtSDS. ∆INC was also inversely correlated with the average size of individual growth spurts derived from the height velocity curves (r = – 0.45, P < 0.05). Using time series analysis to identify rhythms in uGH excretion, a positive correlation was found between the magnitude of rhythms of a period of 2 to 4 weeks and HtSDS (r= 0.40, P < 0.05). These data demonstrate that variability in GH is a more important determinant of normal childhood growth rate than the amount of GH alone. Stature is correlated to the overall variability in GH release, while increment in height and the magnitude of individual growth spurts are influenced by the constancy of the GH profile. This would imply that once the GH dose has been replaced in GH deficiency, optimal growth could only be achieved by varying the pattern of GH administration. © 1999 Churchill Livingstone
Key words: body height, growth physiology, somatotropin urine, rhythmicity.
Received 23/9/98 Revised 17/2/99 Accepted 8/3/99 Correspondence to Dr Peter E Clayton, Endocrine Sciences Research Group, Department of Medicine, Stopford Building, University of Manchester, Oxford Road, Manchester M13 9PT, UK. Tel: + 44 161 275 5180; Fax: + 44 161 275 5958; Email: [email protected]
1096–6374/99/020114+09 $18.00/0
INTRODUCTION Childhood growth, from infancy to adulthood, appears to be a relatively smooth, continuous process with changes in growth velocity characterizing the transition between developmental stages.1 However, when examined frequently over short time intervals, growth is non-linear, © 1999 Churchill Livingstone
GH and normal human growth 115
with variation in height increments seen at daily, weekly, monthly and annual intervals.2–7. We have previously described the pattern of mid-childhood growth, throughout the year, as a biphasic process comprising periods of rapid growth (growth spurts), occurring over a number of weeks, and periods of little or no growth (growth stasis)7. A number of other models have been proposed to characterize growth in infants and children2,3, but there has been no attempt to relate growth over and within a year to a pattern of hormone output, over the same time period. Postnatal growth is dependent on normal GH secretion leading to the generation of its effector hormone insulinlike growth factor (IGF)-I, and the expansion of the growth plates. The conventional approach to investigating the relationship between growth and GH has been to examine growth over months or years in relation to GH output over 1 day8–10. Such studies have demonstrated that GH release is related to growth rate in an asymptotic manner10: hence, over 24 h a subnormal growth rate is associated with low GH secretion (as in GH deficiency), but a normal growth rate can occur over a wide range of GH output. Thus, the amount of GH does not appear to be the principal factor generating variation in normal growth rates. In addition, 24-h GH profiles repeated at intervals of days, months and years show considerable variation within an individual11–13, yet the significance of these fluctuations with respect to growth has not been examined. Measurement of GH in urine is the only practical and ethical technique that can generate longitudinal data on GH output in humans. In keeping with the studies where serial GH profiles have been repeated11–13, considerable variability in urinary GH excretion has been described over a number of days14,15. A longitudinal study of GH excretion, examined for periods of 90 to 365 days, in normal healthy prepubertal children, has identified rhythms in GH output implying variation in pituitary output16. Thus, changes in GH release over days, weeks and months may influence the tempo of normal growth. To examine this hypothesis, we have conducted a longitudinal analysis of the relationship between growth over 1 year and urinary GH excretion over the same period in healthy prepubertal schoolchildren.
MATERIALS AND METHODS Subjects Twenty-nine healthy prepubertal schoolchildren were followed from September to June (13 male, 16 female; age 5.7 to 7.8 years; height standard deviation scores – 2.9 to + 2.3). Subjects were recruited from local schools, with ethical approval and informed parental consent, and
formed part of a larger study of short-term growth7. All children were free from chronic illness. Height was measured three times a week during the school term, at the same time of day, by two observers using a standard stretch technique. Height measurements were made using a free-standing Magnimetre (Raven Instruments, UK), calibrated on each occasion with a machined metre rod. The standard deviation of the differences between ‘blind’ triplicate measurements of the same 25 children for the two auxologists was 0.13 and 0.15 cm, with a mean inter-observer difference of 0.04 cm, and was comparable to other such studies17. Height measurements were made on 94 occasions, the median for one child being 85 (range 75–90). Urinary growth hormone On each measurement occasion, the children provided a timed overnight urine collection for the determination of urinary GH (uGH). Samples were not collected during holiday periods. Urine was collected into preservative-free plastic bottles and 25 ml aliquots were frozen at – 20°C in the presence of 250 µl 10% bovine serum albumin/10% sodium azide. Urinary GH was measured by immunoradiometric after dialysis, as previously described18. A total of 2067 samples were analysed for uGH, and all samples for an individual child were analysed in the same assay. The median number of samples for one child was 71 (range 54 to 86). All results fell within the detection range of the assay. The intra- and inter-assay coefficients of variation (CV) were 6.6–8.8 and 8.8–10.0%, respectively, over the range 6–40 pg/ml. Results were expressed as total amount excreted overnight (tuGH, ng). Of the total number of tuGH results 4.4% exceeded our normal reference range18 and a further 0.6% fell below the normal range, values which are in accordance with our previous longitudinal study16. To further evaluate the relationship between mean tuGH and growth limited tuGH, data was used from a further 14 normal children (ref. 7, < 60% complete collections) (4 M, age 5.8–7.3 years) and 47 short children. The latter comprized 13 GH deficient children (age 2.5–11.6 years; GH peak < 10 ng/ml) and 34 short, slowly growing children with normal GH (age 3.7–14.3 years; GH peak > 10 ng/ml) who had been part of a different study19. Statistical analysis All statistical analysis was performed using the Statistical Package for Social Sciences (SPSS). A P-value < 0.05 indicated statistical significance. Growth was expressed in terms of the increment in height over the period of the study (∆Ht, cm) and the average size of individual growth spurts. The latter were calculated from individual height
116 M.S. Gill et al.
velocity curves, constructed using locally-weighted, least-squares, kernel regression, as previously described7. Growth spurts were defined by the identification of maxima and minima within the velocity curve, and the size of each growth spurt was determined by calculating the area under the curve in that region, by the trapezium rule. Stature, at the beginning and end of the study, was expressed in terms of height standard deviation scores (HtSDS) calculated from 1990 UK standards1. Height velocity standard deviation scores (HVSDS) were calculated from annualized height velocities20. TuGH results were expressed as the average amount excreted per week to generate a time series of equally spaced observations. The median number of weekly observations for each child was 31 (range 27–33). The percentage CV for tuGH results (CVG H) was calculated from the weekly tuGH series using the formula 100 × (standard deviation/mean). The coefficient of incremental change (∆INC) was used to assess the regularity of short-term change in tuGH–low values indicate regular or constant changes in magnitude and high values indicate irregular changes in magnitude21. The weekly incremental change in tuGH (∆GH) was calculated from the tuGH series: GH1, GH2, …, GHn as ∆GH = GHith+1 – GHith. The coefficient of variation of the absolute value of ∆GH was then calculated as above. To characterize the variability in the tuGH profiles, time series analysis was used to identify periodicities in GH release. Missing values were replaced by an estimate of the local mean, generating a series of 45 weekly
Table 1 Summary of the auxological and growth characteristics of the 29 subjects. Body mass index (BMI, kg/m2) was calculated at the beginning of the study. Height standard deviation scores at the beginning (HtSDSSTART) and the end of the study (HtSDSEND), and height velocity standard deviation scores (HVSDS) were calculated from UK Standards1,20. The change in height over the period of the study (∆Ht, cm), and the number and the average size of growth spurts were calculated from individual height velocity curves. Mean Age (years)
6.79
BMI (kg/m2)
15.9
HtSDSSTART
Standard deviation 0.40 1.60
Minimum 5.74 13.1
Maximum 7.77
observations. Non-stationarity within the data was removed by subtraction of a simple moving average (width 7) to generate a series of residuals (n = 39) which were used to calculate the periodogram. Approximate point-wise 95% confidence intervals were calculated for the pooled periodogram as the mean ± 2 SEM and compared with the theoretical spectrum for filtered white noise22. The cumulative periodogram test was used to determine whether individual periodograms were significantly different from white noise22. To examine the contribution of low- and high-frequency rhythms to the overall variance, the area under the curve was calculated in the low-frequency region (0 < frequency < 0.25; ≡ period > 4 weeks) and high-frequency region (0.25≤frequency≤0.5; ≡ period < 4 weeks) for individual periodograms. RESULTS Table 1 summarizes the auxological and growth characteristics of the subjects. All children grew well over the period of study, with height velocity SDS within the normal range. There was no relationship between height at the start of the study (HtSDSSTART) and growth rate over the year (HVSDS). Growth within the year was characterized by an average of five individual growth events which on average accounted for 1.28 cm of the total growth over the year (Table 1). Table 2 summarizes the attributes of urinary GH excretion in the 29 subjects. Mean tuGH was not different between the sexes and did not correlate with body mass index. In addition, mean tuGH excreted over the period of study did not correlate with stature (HtSDS) or growth (∆Ht, HVSDS). However, when tuGH data from normal and short children from previous studies7,19 were included, there was a significant correlation between the log10 mean tuGH and HVSDS (r = 0.48, P < 0.01; Fig. 1). Table 2 Summary of the urinary GH data. Mean total urinary GH (ng) was calculated from the weekly series from each child. The coefficient of variation (CVGH, %) and the coefficient of incremental change (∆INC) were calculated as a measure of the variability of GH21. The variance attributable to high and low frequencies was calculated from individual periodograms.
20.2
Mean
Standard deviation
Minimum
Maximum
–0.06
0.99
–2.94
2.32
HtSDSEND
0.15
0.99
–2.73
2.52
∆Ht (cm)
6.40
0.62
5.00
7.40
CVGH (%)
HVSDS
0.56
0.66
–0.87
1.82
∆INC
0.93
0.15
0.62
1.21
Number of growth spurts
5.00
0.82
3.00
6.00
Low frequency variation in GH
0.43
0.32
0.09
1.42
Size of growth spurts (cm)
1.28
0.25
0.87
1.97
High frequency variation in GH
0.71
0.46
0.06
1.80
Weekly TuGH (ng)
3.18 52.7
1.11 16.5
1.80 28.4
6.21 80.7
GH and normal human growth 117
4
Height velocity standard deviation score
2
0
-2
-4
-6
-8
0
1
2
3
4
5
6
Mean tuGH (ng) Fig. 1 Relationship between total urinary GH excretion (tuGH, ng) and height velocity standard deviation score (HVSDS) in 43 normal (circles), 34 short slowly growing (squares) and 13 GH-deficient (triangles) prepubertal children. The solid line represents the best fit line, determined by regression analysis, and is described by the equation HVSDS = 2.43[log10(tuGH)] – 1.32, r 2 = 0.23, P < 0.0001.
Inspection of the tuGH profiles revealed week-to-week variability in the 29 normal children, which was quantified using two standard measures of short-term variation: the coefficient of variation (CVG H) and the coefficient of incremental change (∆INC) (Table 2). CVG H provided a measure of the overall variability, while ∆INC was used as an index of the relative constancy of short term changes in tuGH. There was no relationship between CVG H and growth over the year, but CVG H was positively correlated to both HtSDS at the start [r = 0.38, Fig. 2(a)] and at the end of the study (r = 0.39, both P < 0.05). In contrast, ∆INC was inversely correlated with both ∆Ht (r = – 0.44, P < 0.05) and HVSDS [r = – 0.42, P < 0.05, Fig. 2(b)] but was not related to stature. Thus, increased variability in tuGH was associated with increased stature, while relative constancy of this variability was related to an increased growth rate. There was also a negative relationship between the average size of a growth spurt and ∆INC [r = – 0.45, P < 0.05; Fig. 2(c)]. To characterize the variability in the tuGH profiles, we used time series analysis to identify periodicities in GH release. The pooled periodogram for all children indicated that there were no dominant periodicities in GH excretion common to all individuals (data not shown).
However, individual children had significant rhythms in tuGH excretion. When the total variation in tuGH (total area under the periodogram) was divided into the lowand high-frequency components (Table 2) there was no relationship between the low-frequency variation and any of the measures of growth or stature. However, there was a significant positive relationship between the highfrequency variation and HtSDS, both at the start [r = 0.40, Fig. 2(D)] and at the end (r = 0.38, both P < 0.05) of the study. There was also a positive correlation between the high-frequency variation and CVG H (r = 0.42), P < 0.05), indicating that increased variability in tuGH could be accounted for by an increase in rhythms at 2 to 4 weeks. Figure 3 illustrates the association between the high-frequency component of tuGH variability and stature. Panel (a) shows the periodogram for a female, HtSDS – 1.4, and panel (b) shows a female subject of the same age with HtSDS + 1.3. The periodogram in panel (a) is characterized by low power at all frequencies, with peaks predominantly in the low-frequency range (frequencies 0.00 to 0.25). In contrast, there is greater overall power in the periodogram of the taller subject [panel (b)], with a predominance of high-frequency power (frequencies 0.25 to 0.50). DISCUSSION Longitudinal studies of growth, with measurements at frequent intervals, have begun to delineate the complex pattern by which normal growth occurs2,3,7. The common feature of these studies is that growth in the short term is non-linear, with periods of rapid growth separated by periods of little or no growth. The pattern of childhood growth is likely to be dictated by the interaction between nutritional, environmental and hormonal influences, the latter including GH. Yet the precise nature of these interactions is unknown17. We have previously demonstrated that GH output, measured by urinary GH excretion, varies considerably over days, weeks and months16. In this study, we have now extended these findings to examine the relationship between growth, over and within a year, and urinary GH output in the first longitudinal study of normal children with simultaneous height and uGH measurements. Measurement of GH in urine has been used as a surrogate for pituitary GH release in the clinical investigation of GH deficiency for a number of years23,24. These studies have shown that uGH excretion reflects pituitary GH release, over the period of collection, in provocation tests and in 24-h and overnight serum profiles24. In addition, overnight uGH measurements have the benefit of being relatively free from influences such as physical activity and diet. However, renal handling of GH remains one potential source of variability in uGH measurements. GH
118 M.S. Gill et al.
(a)
(b)
3
3
r= –0.42, P<0.05
2 2 HVSDS
HtSDS START
1 0 -1
1
0 -2 -1
-3 r=0.38, P<0.05 -4 20
30
50
40
70
60
80
-2
90
0.6
0.8
CVGH (%)
1
1.2
∆ INC
(d)
(c) 3
2.5
2 2
HtSDS START
Size of growth spurts (cm)
r= –0.45, P<0.05
1.5
1 0 -1 -2
1 -3 0.5 0.6
r=0.39, P<0.05
-4 0.8
1
1.2
∆ INC
0
0.5
1.5 1 High frequency variation in GH
2
Fig. 2 Relationship between the variability of tuGH over 1 year and growth and stature in 29 normal prepubertal children. (a) Correlation between height standard deviation score at the start of the study (HtSDSSTART) and the coefficient of variation of GH (CVGH, %). (b) Correlation between height velocity standard deviation score (HVSDS) and the coefficient of incremental change (▲ ▲INC). (c) Correlation between the average size of individual growth spurts and ▲INC. (d) Correlation between HtSDS at the start of the study and the high frequency variation in tuGH. In all panels, the best fit line is shown (solid line) with 95% individual prediction intervals (dashed lines).
is freely filtered at the glomerulus and then undergoes receptor-mediated renal tubular reabsorption25. Expressing uGH results as a ratio to creatinine excretion or as a total amount excreted has been used to control for variation in renal function24. However, we have previously shown that the day-to-day variability in uGH, in terms of CV, is unaffected by the mode of expression14. Furthermore, spectral analysis undertaken on both tuGH
and the uGH/creatinine ratio yielded identical results providing evidence that renal factors do not contribute significantly to rhythms in uGH output, providing renal function is not impaired16. We have previously identified infradian rhythms in tuGH excretion, but we were unable to demonstrate any relationship between tuGH and growth16. Similarly a study by Ahmed et al. examined daily uGH output and
GH and normal human growth 119
(a) 4
Power
3 2 1 0 0
0.25
0.5
Frequency
(b)
Power
15
10
5
0 0
0.25
0.5
Frequency Fig. 3 Sample periodograms for two subjects with differing patterns of tuGH excretion. (a) The periodogram from a female subject (age 6.9 years) with a HtSDS at the start of the study of – 1.4. The variance of tuGH excretion was characterized by a predominance of low-frequency periodicities (0.00 to 0.25) with very little contribution from higher frequencies (0.25 to 0.50). In contrast, the periodogram in (b) has greater overall power, predominantly in the frequency range (0.25 to 0.50). This child was female, age 6.9 years, with a HtSDS at the start of the study of + 1.3.
lower leg growth, measured by knemometry, over a 4week period in prepubertal children26. Whilst there was no obvious correlation with growth, tuGH excretion was highly variable over the 4-week period, with CVs similar to those reported here. In the present study, we could not discern any relationship between mean tuGH excretion and growth in the 29 normal subjects. However, on inclusion of tuGH data from subjects with low growth rates, there was a significant logarithmic relationship between tuGH and HVSDS, in which normal growth rates were associated with a wide range of tuGH values, confirming the asymptotic relationship previously demonstrated for serum GH and growth rate10. The lack of relationship between the level of GH and growth has led some to propose that the GH receptor (GHR) complement of an individual may be a factor which influences growth rates27. Thus, a child with a high level of GHR expression would require less GH secretion in order to
elicit the same growth rate as a child with high GH release and low GHR expression. We now suggest that the variability of GH over time is an important determinant of growth in normally growing children. We have used the coefficient of variation to provide a measure of the overall variability of the GH series with respect to its mean. Large values of the CV indicate high variability in GH. In addition, we have used the coefficient of incremental change to describe the relative magnitude of changes throughout the GH series, irrespective of their temporal organization. The value of this parameter has been demonstrated in the analysis of 24-h serum GH profiles from normal subjects and those with acromegaly21. The normal pattern of GH secretion is characterized by large infrequent pulses separated by low trough concentrations, generating a high coefficient of incremental change. This is because the magnitude of changes from one time point to the next is highly variable. In contrast, the coefficient of incremental change is low in acromegaly, as baseline concentrations of GH are elevated and discrete GH pulses are not easily discernible. We found that there was considerable variation in the amount of tuGH excreted (high CV) through the year for individual children, and this variability was positively correlated with their height. This implies that large changes in GH output from week to week are associated with tall stature, while short stature is associated with less variation in GH output. Moreover, the regularity or constancy of the changes in GH output, defined by the coefficient of incremental change, was correlated with growth rate. Thus, large swings in GH of constant magnitude from week to week would be associated not only with tall stature, but also with good growth through the year. Conversely, small changes in GH of variable magnitude would be associated with short stature and poor growth. The pattern of GH output within 24 h has been shown to be an important determinant of growth rates in rodents, where the sexual dimorphism in GH secretion parallels differences in growth rate28. Male rats secrete GH in discrete pulses separated by low trough levels and grow at a faster rate than females who exhibit high basal GH levels and less pulsatility. In addition, pulsatile infusion of GH in GH-deficient rats stimulates growth, while continuous infusion is associated with the metabolic actions of GH29. A sexual dimorphism in GH secretion is also evident in humans. Average daily GH output is greater in women compared with men, a difference which disappears when corrected for oestrogen concentrations30. Recently, it has been demonstrated that there are also differences in the pattern of 24-h GH secretion between the sexes31. In men, GH output was characterized by small pulses in daylight hours with large nocturnal pulses, while in women GH secretion was more
120 M.S. Gill et al.
continuous with more frequent pulses. Furthermore, there may be associations between the peak and trough attributes of 24-h GH secretion and different endpoints of GH action: trough concentrations of GH are correlated with body composition and metabolic parameters, while peak concentrations correlate with IGF-I32. Modulation of pulse amplitude is thought to be the principal mechanism by which GH mediates changes in its biological actions. For instance, progression into puberty in humans is associated with an increase in growth velocity which is mediated by an increase in GH pulse amplitude, with no apparent change in pulse frequency, augmented by the presence of sex steroids33. In red deer, frequency modulation of GH has been implicated in the formation of new antlers and change in weight gain34. The acceleration of weight gain and antler growth in spring/summer is associated with an increase in GH pulse frequency and amplitude in the preceding winter and early spring. Maximal IGF-I concentrations were observed ~1 month after the maximal increase in GH but were coincident with live weight changes and antler growth rate. Changes in GH production over time may also be important in generating rhythmicity in its effector peptides, IGF-I and IGFBP-3. Month-to-month variations in the serum concentration of IGF-I and IGFBP-3 have been demonstrated in normal prepubertal children35. Furthermore, the monthly change in IGFBP-3, not the absolute amount, was correlated with the increase in lower leg length over the period of study. The pulsatile secretion of GH from the pituitary arises principally from the interaction between GH-releasing hormone (GHRH) and somatostatin (SS), but also involves other neurotransmitters and neuropeptides. GHRH release from the hypothalamus is thought to determine the magnitude of the GH pulse, while the prevailing concentration of SS influences the timing36. The pulsatile attributes of GH secretion are superimposed over a baseline concentration of GH, which itself may be subject to variability in the form of a circadian rhythm. Using cosinor analysis of 24 GH profiles, Keret et al.37 have identified a subgroup of short children with an absence of circadian rhythmicity of GH who are significantly shorter than those children with evidence of 24-h periodicity. Our data now suggest an additional level of control of GH output, perhaps involving GHRH and SS, which integrates the well-characterized ultradian and circadian events in GH secretion to generate variability over weeks and months. An analogy can be drawn with the changes in follicle-stimulating hormone (FSH) and luteinizing hormone (LH) which occur during pubertal maturation38. The transition from prepuberty into puberty is marked by increases in FSH and LH, which are attributed not only to an increase in basal levels but also to increased diumal and ultradian rhythmicity.
Longitudinal studies of pulsatile hormone output and integration of their signal over time have been few, due to the limitations of performing repeated 24-h serum profiles. However, this application of urinary assays now offers a new approach to the study of how physiological and developmental changes are mediated by variation in hormone output over days, weeks and months. These findings also raise fundamental questions with regard to the manner in which GH replacement therapy is administered. Current treatment regimens are based on daily injections of GH at a constant dose that is adjusted according to body weight. The relationship between growth rate and GH output (Fig. 1 and ref 10) demonstrate that the initial administration of GH to a GH-deficient child will increase growth rate due to a repositioning on the dose–response curve. However, recent data of final height in GH deficient children at the end of GH treatment suggests that, although there is an improvement over predicted adult height, target height is rarely achieved39–41. Furthermore, the results of GH treatment in non-GH-deficient short children demonstrate even less of an impact on final height42,43. We now propose that in order to optimize growth, GH replacement therapy should be tailored to mimic the underlying variability in normal GH output. This study is the first to describe a relationship between GH output and growth using simultaneous hormone sampling and height measurements in a large cohort of healthy prepubertal children over an extended period of time. We have demonstrated that GH variability is a more important determinant of normal mid-childhood growth than the amount of GH alone. Stature is correlated with overall variability in GH release, while increment in height and the magnitude of individual growth spurts are influenced by the regularity of the GH profile. Tall stature is therefore associated with large variation in GH release, while good growth, over and within the year, is related to constant short-term changes in GH.
ACKNOWLEDGEMENTS We thank the children who gave their time and effort, their parents and teachers, Sister Julie Jones and Margaret Downes. Financial support was provided by Serono Laboratories UK (MSG), Lilly Industries (MSG, NKST) and Pharmacia & Upjohn (VT).
REFERENCES 1. Freeman J V, Cole T J, Chinn S, Jones P R M, White E M, Preece M A. Cross-sectional stature and weight reference curves for the UK, 1990. Arch Dis Child 1995: 73: 17–24. 2. Lampl M, Veldhuis J D, Johnson M L. Saltation and stasis: a model of human growth. Science 1992; 258: 801–803.
GH and normal human growth 121
3. Hermanussen M, Geiger-Benoit K, Burmeister J, Sippell W G. Periodical changes of shortterm growth velocity (“mini growth spurts”) in human growth. Ann Hum Biol 1988; 15: 103–109. 4. Marshall W A, Swan A V. Seasonal variation in growth rates of normal and blind children. Hum Biol 1971; 43: 503–516. 5. Butler G E, McKie M, Ratcliffe S G. The cyclical nature of prepubertal growth. Ann Hum Biol 1990; 17: 177–198. 6. Gelander L, Karlberg J, Albertsson-Wikland K. Seasonality in lower leg length velocity in prepubertal children. Acta Paediatr 1994; 83: 1249–1254. 7. Thalange N K S, Foster P J, Gill M S, Price D A, Clayton P E. A model of normal prepubertal growth. Arch Dis Child 1996; 75: 427–432. 8. Albertsson-Wikland K, Rosberg S. Analysis of 24-hour growth hormone profiles in children: relation to growth. J Clin Endocrinol Metab 1988; 67: 493–500. 9. Hindmarsh P C, Matthews D R, Brook C G D. Growth hormone secretion in children determined by time series analysis. Clin Endocrinol (Oxf) 1988; 29: 35–44. 10. Hindmarsh P C, Smith P J, Brook C G D, Matthews D R. The relationship between height velocity and growth hormone secretion in short prepubertal children. Clin Endocrinol (Oxf) 1987; 27: 581–591. 11. Donaldson D L, Hollowell J G, Pan F, Gifford R A, Moore W V. Growth hormone secretory profiles: variation on consecutive nights. J Pediatr 1989; 115: 51–56. 12. Saini S, Hindmarsh P C, Matthews D R, Pringle P J, Jones J, Preece M A, Brook C G. Reproducibility of 24-hour serum growth hormone profiles in man. Clin Endocrinol (Oxf) 1991; 34: 455–462. 13. Albertsson-Wikland K, Rosberg S. Reproducibility of 24-hour growth hormone profiles in children. Acta Endocrinol (Copenh) 1992; 126: 109–112. 14. Skinner A M, Clayton P E, Price D A, Addison G M, Soo A. Variability in the urinary excretion of growth hormone in children: comparison with other urinary proteins. J Endocrinol 1993; 138: 337–343. 15. Main K M, Jarden M, Angelo L, Dinesen B, Hertel N T, Juul A, Muller J, Skakkebaek N E. The impact of gender and puberty on reference values for urinary growth hormone excretion: a study of 3 morning samples in 517 healthy children and adults. J Clin Endocrinol Metab 1994: 79: 865–871. 16. Thalange N K S, Gill M S, Gill L, Whatmore A J, Addison G M, Price D A, Clayton P E. Infradian rhythms in urinary growth hormone excretion. J Clin Endocrinol Metab 1996; 81: 100–106. 17. Wales J K H, Gibson A T. Short-term growth: rhythms, chaos, or noise? Arch Dis Child 1994; 71: 84–89. 18. Skinner A M, Price D A, Addison G M, Clayton P E, Mackay R I, Soo A, Mui C Y. The influence of age, size, pubertal status and renal factors on urinary growth hormone excretion in normal children and adolescents. Growth Regulation 1992; 2: 156–160. 19. Tillmann V, Buckler J M, Kibinge M S, Price D A, Shalet S M, Wales J K, Addison G M, Gill M S, Whatmore A J, Clayton P E. Biochemical tests in the diagnosis of childhood growth hormone deficiency. J Clin Endocrinol Metab 1997; 82: 531–535. 20. Tanner J M, Whitehouse R H, Takaishi M. Standards from birth to maturity for height, weight, height velocity and weight velocity: British children 1965. Arch Dis Child 1966; 41: 613–635. 21. Hartman M L, Pincus S M, Johnson M L, Matthews D H, Faunt L M, Vance M L, Thorner M O, Veldhuis J D. Enhanced basal and disorderly growth hormone secretion distinguish acromegalic from normal pulsatile growth hormone release. J Clin Invest 1994; 94: 1277–1288.
22. Diggle P J. Time series analysis: a biostatistical introduction. Oxford Science Publications, Oxford. 1990. 23. Skinner A M, Clayton P E, Price D A, Addison G M, Soo A. Urinary growth hormone excretion in the assessment of children with disorders of growth. Clin Endocrinol (Oxf) 1993; 39: 201–206. 24. Hourd P, Edwards R. Current methods for the measurement of growth hormone in urine. Clin Endocrinol (Oxf) 1994; 40: 155–170. 25. Johnson V, Maack T. Renal extraction, filtration, absorption and catabolism of growth hormone. Am J Physiol 1977; 233: F185–F196. 26. Ahmed S F, Barnes S I, Wallace W H B, Kelnar C J H. Short-term changes in urinary growth hormone excretion and lower leg length in healthy children. Horm Res 1997; 48: 72–75. 27. Martha P M, Rogol A D, Carisson L M S, Gesundheit N, Blizzard R M. A longitudinal assessment of hormonal and physical alterations during normal puberty in boys. I. Serum growth hormone binding protein. J Clin Endocrinol Metab 1993; 77: 452–457. 28. Jansson J-O, Eden S, Isaksson O. Sexual dimorphism in the control of growth hormone secretion. Endocr Rev 1985; 6: 128–150. 29. Gevers E F, Wit J M, Robinson I C A F. Growth, growth hormone (GH)-binding protein and GH receptors are differentially regulated by peak and trough components of the GH secretory pattern in the rat. Endocrinology 1996; 137: 1013–1018. 30. Ho K Y, Evans W S, Blizzard R M, et al. Effects of sex and age on the 24-hour profile of growth hormone secretion in man: importance of endogenous estradiol concentrations. J Clin Endocrinol Metab 1987; 64: 51–58. 31. Jaffe C A, Ocampo-Lim B, Guo W, Krueger K, Sugahara I, DeMott F R, Bennann M, Barkan A L. Regulatory mechanisms of growth hormone secretion are sexually dimorphic. J Clin Invest 1998; 102: 153–164. 32. Hindmarsh P C, Fall C H D, Pringle P J, Osmond C, Brook C G D. Peak and trough growth hormone concentrations have different associations with the insulin-like growth factor axis, body composition, and metabolic parameters. J Clin Endocrinol Metab 1997; 82: 2172–2176. 33. Mauras N, Blizzard R M, Link K, Johnson M L, Rogol A D, Veldhuis J D. Augmentation of growth hormone secretion during puberty: evidence for a pulse amplitude-modulated phenomenon. J Clin Endocrinol Metab 1987; 64: 596–601. 34. Suttie J M, Fennessy P F, Corson I D, Laas F J, Crosbie S F, Butler J H, Gluckman P D. Pulsatile growth hormone, insulin-like growth factors and antler development in red deer (Cervus elaphus scoticus) stags. J Endocrinol 1989; 121: 351–360. 35. Gelander L, Petterson M, Rosberg S, Albertsson-Wikland K. Longitudinal inter-individual changes in insulin-like growth factor-I (IGF-I) and its binding protein 3 (IGFBP-3) in relation to growth in healthy children. 1996; Horm Res 46(Suppl. 2): 100 (Abstract). 36. Vance M L, Kaiser D L, Evans W S, Furlanetto R, Vale W, Rivier J, Thorner MO. Pulsatile growth hormone secretion in normal man during a continuous infusion of human growth hormone releasing factor (1–40). J Clin Invest 1985; 75: 1584–1590. 37. Keret R, Ashkenazi I E, Wasserman M, Bauman B, Pertzelan A, Ticher A, Laron Z. Two types of growth hormone rhythm in boys with constitutionally short stature. J Ped Endocrinol Metab 1996; 9: 599–607. 38. Albertsson-Wikland K, Rosberg S, Lannering B, Dunkel L, Selstam G, Norjavaara E. Twenty-four-hour profiles of luteinizing hormone, follicle-stimulating hormone, testosterone, and estradiol levels: a semilongitudinal study throughout
122 M.S. Gill et al.
puberty in healthy boys. J Clin Endocrinol Metab 1997; 82: 541–549. 39. Price D A, Ranke M B. Final height following growth hormone treatment. In Ranke MB, Gunnarsson R, editors. Progress in Growth Hormone Therapy – 5 Years of KIGS. J, J Verlag, Mannheim. 1994: 129–144. 40. Blethen S L, Baptista J, Kuntze J, Foley T, LaFranchi S, Johanson A. Adult height in growth hormone (GH)-deficient children treated with biosynthetic GH. J Clin Endocrinol Metab 1997; 82: 418–420.
41. Coste J, Letrait M, Carel J C, Tresca J P, Chatelain P, Rochiccioli P, Chaussain J L, Job J C. Long term results of growth hormone treatment in France in children of short stature: population, register based study. BMJ 1997; 315: 708–713. 42. Hindmarsh P C, Brook C G D. Final height of short normal children treated with growth hormone. Lancet 1996; 348: 13–16. 43. Bernasconi B, Street M E, Volta C, Mazzardo G, and the Italian Multicentre Study Group. Final height in non-growth hormone deficient children treated with growth hormone. Clin Endocrinol (Oxf) 1997; 47: 261–266.
Regular fluctuations in growth hormone (GH) release determine normal human growth 1,2 M. S. , N. K. S. Thalange2, P. J. Foster3, V. Tillmann2, D. A. Price2, P. J. Diggle4 and P. E. Clayton1,2 1
Endocrine Sciences Research Group, Department of Medicine, Stopford Building, University of Manchester, Oxford Road, Manchester M13 9PT, UK; University Department of Child Health, Royal Manchester Childrens Hospital, Hospital Road, Pendlebury, Manchester M27 4HA, UK; 3Department of Mathematics, University of Manchester, Oxford Road, Manchester M13 9PT, UK; 4Medical Statistics Unit, Department of Mathematics and Statistics, University of Lancaster, Lancaster LA1 4YF, UK 2
Summary Growth hormone (GH) is the principal hormone associated with growth through childhood, but in a normal child the amount of GH secretion does not appear to be critical in the generation of normal growth rates. We have assessed the relationship between growth and urinary GH (uGH) output in a longitudinal study of 29 healthy prepubertal schoolchildren (13 male, 16 female; age 5.7–7.8 years) over 1 year. Height and uGH were measured three times a week. Individual height velocity curves were derived using non-linear regression. Growth was expressed in terms of the total increment over the year (∆Ht, cm), height velocity standard deviation score (HVSDS) and the average size of individual growth spurts. Urinary GH data (ng) were expressed as a weekly average. Mean uGH did not correlate with stature or growth over the year. However, the coefficient of variation of uGH was correlated with height standard deviation score (HtSDS, r = 0.38, P < 0.05), while the relative constancy of short-term change in uGH (coefficient of incremental change, ∆INC) was inversely correlated with ∆Ht (r = – 0.44) and HVSDS (r= – 0.42, both P < 0.05) but not with HtSDS. ∆INC was also inversely correlated with the average size of individual growth spurts derived from the height velocity curves (r = – 0.45, P < 0.05). Using time series analysis to identify rhythms in uGH excretion, a positive correlation was found between the magnitude of rhythms of a period of 2 to 4 weeks and HtSDS (r= 0.40, P < 0.05). These data demonstrate that variability in GH is a more important determinant of normal childhood growth rate than the amount of GH alone. Stature is correlated to the overall variability in GH release, while increment in height and the magnitude of individual growth spurts are influenced by the constancy of the GH profile. This would imply that once the GH dose has been replaced in GH deficiency, optimal growth could only be achieved by varying the pattern of GH administration. © 1999 Churchill Livingstone
Key words: body height, growth physiology, somatotropin urine, rhythmicity.
Received 23/9/98 Revised 17/2/99 Accepted 8/3/99 Correspondence to Dr Peter E Clayton, Endocrine Sciences Research Group, Department of Medicine, Stopford Building, University of Manchester, Oxford Road, Manchester M13 9PT, UK. Tel: + 44 161 275 5180; Fax: + 44 161 275 5958; Email: [email protected]
1096–6374/99/020114+09 $18.00/0
INTRODUCTION Childhood growth, from infancy to adulthood, appears to be a relatively smooth, continuous process with changes in growth velocity characterizing the transition between developmental stages.1 However, when examined frequently over short time intervals, growth is non-linear, © 1999 Churchill Livingstone
GH and normal human growth 115
with variation in height increments seen at daily, weekly, monthly and annual intervals.2–7. We have previously described the pattern of mid-childhood growth, throughout the year, as a biphasic process comprising periods of rapid growth (growth spurts), occurring over a number of weeks, and periods of little or no growth (growth stasis)7. A number of other models have been proposed to characterize growth in infants and children2,3, but there has been no attempt to relate growth over and within a year to a pattern of hormone output, over the same time period. Postnatal growth is dependent on normal GH secretion leading to the generation of its effector hormone insulinlike growth factor (IGF)-I, and the expansion of the growth plates. The conventional approach to investigating the relationship between growth and GH has been to examine growth over months or years in relation to GH output over 1 day8–10. Such studies have demonstrated that GH release is related to growth rate in an asymptotic manner10: hence, over 24 h a subnormal growth rate is associated with low GH secretion (as in GH deficiency), but a normal growth rate can occur over a wide range of GH output. Thus, the amount of GH does not appear to be the principal factor generating variation in normal growth rates. In addition, 24-h GH profiles repeated at intervals of days, months and years show considerable variation within an individual11–13, yet the significance of these fluctuations with respect to growth has not been examined. Measurement of GH in urine is the only practical and ethical technique that can generate longitudinal data on GH output in humans. In keeping with the studies where serial GH profiles have been repeated11–13, considerable variability in urinary GH excretion has been described over a number of days14,15. A longitudinal study of GH excretion, examined for periods of 90 to 365 days, in normal healthy prepubertal children, has identified rhythms in GH output implying variation in pituitary output16. Thus, changes in GH release over days, weeks and months may influence the tempo of normal growth. To examine this hypothesis, we have conducted a longitudinal analysis of the relationship between growth over 1 year and urinary GH excretion over the same period in healthy prepubertal schoolchildren.
MATERIALS AND METHODS Subjects Twenty-nine healthy prepubertal schoolchildren were followed from September to June (13 male, 16 female; age 5.7 to 7.8 years; height standard deviation scores – 2.9 to + 2.3). Subjects were recruited from local schools, with ethical approval and informed parental consent, and
formed part of a larger study of short-term growth7. All children were free from chronic illness. Height was measured three times a week during the school term, at the same time of day, by two observers using a standard stretch technique. Height measurements were made using a free-standing Magnimetre (Raven Instruments, UK), calibrated on each occasion with a machined metre rod. The standard deviation of the differences between ‘blind’ triplicate measurements of the same 25 children for the two auxologists was 0.13 and 0.15 cm, with a mean inter-observer difference of 0.04 cm, and was comparable to other such studies17. Height measurements were made on 94 occasions, the median for one child being 85 (range 75–90). Urinary growth hormone On each measurement occasion, the children provided a timed overnight urine collection for the determination of urinary GH (uGH). Samples were not collected during holiday periods. Urine was collected into preservative-free plastic bottles and 25 ml aliquots were frozen at – 20°C in the presence of 250 µl 10% bovine serum albumin/10% sodium azide. Urinary GH was measured by immunoradiometric after dialysis, as previously described18. A total of 2067 samples were analysed for uGH, and all samples for an individual child were analysed in the same assay. The median number of samples for one child was 71 (range 54 to 86). All results fell within the detection range of the assay. The intra- and inter-assay coefficients of variation (CV) were 6.6–8.8 and 8.8–10.0%, respectively, over the range 6–40 pg/ml. Results were expressed as total amount excreted overnight (tuGH, ng). Of the total number of tuGH results 4.4% exceeded our normal reference range18 and a further 0.6% fell below the normal range, values which are in accordance with our previous longitudinal study16. To further evaluate the relationship between mean tuGH and growth limited tuGH, data was used from a further 14 normal children (ref. 7, < 60% complete collections) (4 M, age 5.8–7.3 years) and 47 short children. The latter comprized 13 GH deficient children (age 2.5–11.6 years; GH peak < 10 ng/ml) and 34 short, slowly growing children with normal GH (age 3.7–14.3 years; GH peak > 10 ng/ml) who had been part of a different study19. Statistical analysis All statistical analysis was performed using the Statistical Package for Social Sciences (SPSS). A P-value < 0.05 indicated statistical significance. Growth was expressed in terms of the increment in height over the period of the study (∆Ht, cm) and the average size of individual growth spurts. The latter were calculated from individual height
116 M.S. Gill et al.
velocity curves, constructed using locally-weighted, least-squares, kernel regression, as previously described7. Growth spurts were defined by the identification of maxima and minima within the velocity curve, and the size of each growth spurt was determined by calculating the area under the curve in that region, by the trapezium rule. Stature, at the beginning and end of the study, was expressed in terms of height standard deviation scores (HtSDS) calculated from 1990 UK standards1. Height velocity standard deviation scores (HVSDS) were calculated from annualized height velocities20. TuGH results were expressed as the average amount excreted per week to generate a time series of equally spaced observations. The median number of weekly observations for each child was 31 (range 27–33). The percentage CV for tuGH results (CVG H) was calculated from the weekly tuGH series using the formula 100 × (standard deviation/mean). The coefficient of incremental change (∆INC) was used to assess the regularity of short-term change in tuGH–low values indicate regular or constant changes in magnitude and high values indicate irregular changes in magnitude21. The weekly incremental change in tuGH (∆GH) was calculated from the tuGH series: GH1, GH2, …, GHn as ∆GH = GHith+1 – GHith. The coefficient of variation of the absolute value of ∆GH was then calculated as above. To characterize the variability in the tuGH profiles, time series analysis was used to identify periodicities in GH release. Missing values were replaced by an estimate of the local mean, generating a series of 45 weekly
Table 1 Summary of the auxological and growth characteristics of the 29 subjects. Body mass index (BMI, kg/m2) was calculated at the beginning of the study. Height standard deviation scores at the beginning (HtSDSSTART) and the end of the study (HtSDSEND), and height velocity standard deviation scores (HVSDS) were calculated from UK Standards1,20. The change in height over the period of the study (∆Ht, cm), and the number and the average size of growth spurts were calculated from individual height velocity curves. Mean Age (years)
6.79
BMI (kg/m2)
15.9
HtSDSSTART
Standard deviation 0.40 1.60
Minimum 5.74 13.1
Maximum 7.77
observations. Non-stationarity within the data was removed by subtraction of a simple moving average (width 7) to generate a series of residuals (n = 39) which were used to calculate the periodogram. Approximate point-wise 95% confidence intervals were calculated for the pooled periodogram as the mean ± 2 SEM and compared with the theoretical spectrum for filtered white noise22. The cumulative periodogram test was used to determine whether individual periodograms were significantly different from white noise22. To examine the contribution of low- and high-frequency rhythms to the overall variance, the area under the curve was calculated in the low-frequency region (0 < frequency < 0.25; ≡ period > 4 weeks) and high-frequency region (0.25≤frequency≤0.5; ≡ period < 4 weeks) for individual periodograms. RESULTS Table 1 summarizes the auxological and growth characteristics of the subjects. All children grew well over the period of study, with height velocity SDS within the normal range. There was no relationship between height at the start of the study (HtSDSSTART) and growth rate over the year (HVSDS). Growth within the year was characterized by an average of five individual growth events which on average accounted for 1.28 cm of the total growth over the year (Table 1). Table 2 summarizes the attributes of urinary GH excretion in the 29 subjects. Mean tuGH was not different between the sexes and did not correlate with body mass index. In addition, mean tuGH excreted over the period of study did not correlate with stature (HtSDS) or growth (∆Ht, HVSDS). However, when tuGH data from normal and short children from previous studies7,19 were included, there was a significant correlation between the log10 mean tuGH and HVSDS (r = 0.48, P < 0.01; Fig. 1). Table 2 Summary of the urinary GH data. Mean total urinary GH (ng) was calculated from the weekly series from each child. The coefficient of variation (CVGH, %) and the coefficient of incremental change (∆INC) were calculated as a measure of the variability of GH21. The variance attributable to high and low frequencies was calculated from individual periodograms.
20.2
Mean
Standard deviation
Minimum
Maximum
–0.06
0.99
–2.94
2.32
HtSDSEND
0.15
0.99
–2.73
2.52
∆Ht (cm)
6.40
0.62
5.00
7.40
CVGH (%)
HVSDS
0.56
0.66
–0.87
1.82
∆INC
0.93
0.15
0.62
1.21
Number of growth spurts
5.00
0.82
3.00
6.00
Low frequency variation in GH
0.43
0.32
0.09
1.42
Size of growth spurts (cm)
1.28
0.25
0.87
1.97
High frequency variation in GH
0.71
0.46
0.06
1.80
Weekly TuGH (ng)
3.18 52.7
1.11 16.5
1.80 28.4
6.21 80.7
GH and normal human growth 117
4
Height velocity standard deviation score
2
0
-2
-4
-6
-8
0
1
2
3
4
5
6
Mean tuGH (ng) Fig. 1 Relationship between total urinary GH excretion (tuGH, ng) and height velocity standard deviation score (HVSDS) in 43 normal (circles), 34 short slowly growing (squares) and 13 GH-deficient (triangles) prepubertal children. The solid line represents the best fit line, determined by regression analysis, and is described by the equation HVSDS = 2.43[log10(tuGH)] – 1.32, r 2 = 0.23, P < 0.0001.
Inspection of the tuGH profiles revealed week-to-week variability in the 29 normal children, which was quantified using two standard measures of short-term variation: the coefficient of variation (CVG H) and the coefficient of incremental change (∆INC) (Table 2). CVG H provided a measure of the overall variability, while ∆INC was used as an index of the relative constancy of short term changes in tuGH. There was no relationship between CVG H and growth over the year, but CVG H was positively correlated to both HtSDS at the start [r = 0.38, Fig. 2(a)] and at the end of the study (r = 0.39, both P < 0.05). In contrast, ∆INC was inversely correlated with both ∆Ht (r = – 0.44, P < 0.05) and HVSDS [r = – 0.42, P < 0.05, Fig. 2(b)] but was not related to stature. Thus, increased variability in tuGH was associated with increased stature, while relative constancy of this variability was related to an increased growth rate. There was also a negative relationship between the average size of a growth spurt and ∆INC [r = – 0.45, P < 0.05; Fig. 2(c)]. To characterize the variability in the tuGH profiles, we used time series analysis to identify periodicities in GH release. The pooled periodogram for all children indicated that there were no dominant periodicities in GH excretion common to all individuals (data not shown).
However, individual children had significant rhythms in tuGH excretion. When the total variation in tuGH (total area under the periodogram) was divided into the lowand high-frequency components (Table 2) there was no relationship between the low-frequency variation and any of the measures of growth or stature. However, there was a significant positive relationship between the highfrequency variation and HtSDS, both at the start [r = 0.40, Fig. 2(D)] and at the end (r = 0.38, both P < 0.05) of the study. There was also a positive correlation between the high-frequency variation and CVG H (r = 0.42), P < 0.05), indicating that increased variability in tuGH could be accounted for by an increase in rhythms at 2 to 4 weeks. Figure 3 illustrates the association between the high-frequency component of tuGH variability and stature. Panel (a) shows the periodogram for a female, HtSDS – 1.4, and panel (b) shows a female subject of the same age with HtSDS + 1.3. The periodogram in panel (a) is characterized by low power at all frequencies, with peaks predominantly in the low-frequency range (frequencies 0.00 to 0.25). In contrast, there is greater overall power in the periodogram of the taller subject [panel (b)], with a predominance of high-frequency power (frequencies 0.25 to 0.50). DISCUSSION Longitudinal studies of growth, with measurements at frequent intervals, have begun to delineate the complex pattern by which normal growth occurs2,3,7. The common feature of these studies is that growth in the short term is non-linear, with periods of rapid growth separated by periods of little or no growth. The pattern of childhood growth is likely to be dictated by the interaction between nutritional, environmental and hormonal influences, the latter including GH. Yet the precise nature of these interactions is unknown17. We have previously demonstrated that GH output, measured by urinary GH excretion, varies considerably over days, weeks and months16. In this study, we have now extended these findings to examine the relationship between growth, over and within a year, and urinary GH output in the first longitudinal study of normal children with simultaneous height and uGH measurements. Measurement of GH in urine has been used as a surrogate for pituitary GH release in the clinical investigation of GH deficiency for a number of years23,24. These studies have shown that uGH excretion reflects pituitary GH release, over the period of collection, in provocation tests and in 24-h and overnight serum profiles24. In addition, overnight uGH measurements have the benefit of being relatively free from influences such as physical activity and diet. However, renal handling of GH remains one potential source of variability in uGH measurements. GH
118 M.S. Gill et al.
(a)
(b)
3
3
r= –0.42, P<0.05
2 2 HVSDS
HtSDS START
1 0 -1
1
0 -2 -1
-3 r=0.38, P<0.05 -4 20
30
50
40
70
60
80
-2
90
0.6
0.8
CVGH (%)
1
1.2
∆ INC
(d)
(c) 3
2.5
2 2
HtSDS START
Size of growth spurts (cm)
r= –0.45, P<0.05
1.5
1 0 -1 -2
1 -3 0.5 0.6
r=0.39, P<0.05
-4 0.8
1
1.2
∆ INC
0
0.5
1.5 1 High frequency variation in GH
2
Fig. 2 Relationship between the variability of tuGH over 1 year and growth and stature in 29 normal prepubertal children. (a) Correlation between height standard deviation score at the start of the study (HtSDSSTART) and the coefficient of variation of GH (CVGH, %). (b) Correlation between height velocity standard deviation score (HVSDS) and the coefficient of incremental change (▲ ▲INC). (c) Correlation between the average size of individual growth spurts and ▲INC. (d) Correlation between HtSDS at the start of the study and the high frequency variation in tuGH. In all panels, the best fit line is shown (solid line) with 95% individual prediction intervals (dashed lines).
is freely filtered at the glomerulus and then undergoes receptor-mediated renal tubular reabsorption25. Expressing uGH results as a ratio to creatinine excretion or as a total amount excreted has been used to control for variation in renal function24. However, we have previously shown that the day-to-day variability in uGH, in terms of CV, is unaffected by the mode of expression14. Furthermore, spectral analysis undertaken on both tuGH
and the uGH/creatinine ratio yielded identical results providing evidence that renal factors do not contribute significantly to rhythms in uGH output, providing renal function is not impaired16. We have previously identified infradian rhythms in tuGH excretion, but we were unable to demonstrate any relationship between tuGH and growth16. Similarly a study by Ahmed et al. examined daily uGH output and
GH and normal human growth 119
(a) 4
Power
3 2 1 0 0
0.25
0.5
Frequency
(b)
Power
15
10
5
0 0
0.25
0.5
Frequency Fig. 3 Sample periodograms for two subjects with differing patterns of tuGH excretion. (a) The periodogram from a female subject (age 6.9 years) with a HtSDS at the start of the study of – 1.4. The variance of tuGH excretion was characterized by a predominance of low-frequency periodicities (0.00 to 0.25) with very little contribution from higher frequencies (0.25 to 0.50). In contrast, the periodogram in (b) has greater overall power, predominantly in the frequency range (0.25 to 0.50). This child was female, age 6.9 years, with a HtSDS at the start of the study of + 1.3.
lower leg growth, measured by knemometry, over a 4week period in prepubertal children26. Whilst there was no obvious correlation with growth, tuGH excretion was highly variable over the 4-week period, with CVs similar to those reported here. In the present study, we could not discern any relationship between mean tuGH excretion and growth in the 29 normal subjects. However, on inclusion of tuGH data from subjects with low growth rates, there was a significant logarithmic relationship between tuGH and HVSDS, in which normal growth rates were associated with a wide range of tuGH values, confirming the asymptotic relationship previously demonstrated for serum GH and growth rate10. The lack of relationship between the level of GH and growth has led some to propose that the GH receptor (GHR) complement of an individual may be a factor which influences growth rates27. Thus, a child with a high level of GHR expression would require less GH secretion in order to
elicit the same growth rate as a child with high GH release and low GHR expression. We now suggest that the variability of GH over time is an important determinant of growth in normally growing children. We have used the coefficient of variation to provide a measure of the overall variability of the GH series with respect to its mean. Large values of the CV indicate high variability in GH. In addition, we have used the coefficient of incremental change to describe the relative magnitude of changes throughout the GH series, irrespective of their temporal organization. The value of this parameter has been demonstrated in the analysis of 24-h serum GH profiles from normal subjects and those with acromegaly21. The normal pattern of GH secretion is characterized by large infrequent pulses separated by low trough concentrations, generating a high coefficient of incremental change. This is because the magnitude of changes from one time point to the next is highly variable. In contrast, the coefficient of incremental change is low in acromegaly, as baseline concentrations of GH are elevated and discrete GH pulses are not easily discernible. We found that there was considerable variation in the amount of tuGH excreted (high CV) through the year for individual children, and this variability was positively correlated with their height. This implies that large changes in GH output from week to week are associated with tall stature, while short stature is associated with less variation in GH output. Moreover, the regularity or constancy of the changes in GH output, defined by the coefficient of incremental change, was correlated with growth rate. Thus, large swings in GH of constant magnitude from week to week would be associated not only with tall stature, but also with good growth through the year. Conversely, small changes in GH of variable magnitude would be associated with short stature and poor growth. The pattern of GH output within 24 h has been shown to be an important determinant of growth rates in rodents, where the sexual dimorphism in GH secretion parallels differences in growth rate28. Male rats secrete GH in discrete pulses separated by low trough levels and grow at a faster rate than females who exhibit high basal GH levels and less pulsatility. In addition, pulsatile infusion of GH in GH-deficient rats stimulates growth, while continuous infusion is associated with the metabolic actions of GH29. A sexual dimorphism in GH secretion is also evident in humans. Average daily GH output is greater in women compared with men, a difference which disappears when corrected for oestrogen concentrations30. Recently, it has been demonstrated that there are also differences in the pattern of 24-h GH secretion between the sexes31. In men, GH output was characterized by small pulses in daylight hours with large nocturnal pulses, while in women GH secretion was more
120 M.S. Gill et al.
continuous with more frequent pulses. Furthermore, there may be associations between the peak and trough attributes of 24-h GH secretion and different endpoints of GH action: trough concentrations of GH are correlated with body composition and metabolic parameters, while peak concentrations correlate with IGF-I32. Modulation of pulse amplitude is thought to be the principal mechanism by which GH mediates changes in its biological actions. For instance, progression into puberty in humans is associated with an increase in growth velocity which is mediated by an increase in GH pulse amplitude, with no apparent change in pulse frequency, augmented by the presence of sex steroids33. In red deer, frequency modulation of GH has been implicated in the formation of new antlers and change in weight gain34. The acceleration of weight gain and antler growth in spring/summer is associated with an increase in GH pulse frequency and amplitude in the preceding winter and early spring. Maximal IGF-I concentrations were observed ~1 month after the maximal increase in GH but were coincident with live weight changes and antler growth rate. Changes in GH production over time may also be important in generating rhythmicity in its effector peptides, IGF-I and IGFBP-3. Month-to-month variations in the serum concentration of IGF-I and IGFBP-3 have been demonstrated in normal prepubertal children35. Furthermore, the monthly change in IGFBP-3, not the absolute amount, was correlated with the increase in lower leg length over the period of study. The pulsatile secretion of GH from the pituitary arises principally from the interaction between GH-releasing hormone (GHRH) and somatostatin (SS), but also involves other neurotransmitters and neuropeptides. GHRH release from the hypothalamus is thought to determine the magnitude of the GH pulse, while the prevailing concentration of SS influences the timing36. The pulsatile attributes of GH secretion are superimposed over a baseline concentration of GH, which itself may be subject to variability in the form of a circadian rhythm. Using cosinor analysis of 24 GH profiles, Keret et al.37 have identified a subgroup of short children with an absence of circadian rhythmicity of GH who are significantly shorter than those children with evidence of 24-h periodicity. Our data now suggest an additional level of control of GH output, perhaps involving GHRH and SS, which integrates the well-characterized ultradian and circadian events in GH secretion to generate variability over weeks and months. An analogy can be drawn with the changes in follicle-stimulating hormone (FSH) and luteinizing hormone (LH) which occur during pubertal maturation38. The transition from prepuberty into puberty is marked by increases in FSH and LH, which are attributed not only to an increase in basal levels but also to increased diumal and ultradian rhythmicity.
Longitudinal studies of pulsatile hormone output and integration of their signal over time have been few, due to the limitations of performing repeated 24-h serum profiles. However, this application of urinary assays now offers a new approach to the study of how physiological and developmental changes are mediated by variation in hormone output over days, weeks and months. These findings also raise fundamental questions with regard to the manner in which GH replacement therapy is administered. Current treatment regimens are based on daily injections of GH at a constant dose that is adjusted according to body weight. The relationship between growth rate and GH output (Fig. 1 and ref 10) demonstrate that the initial administration of GH to a GH-deficient child will increase growth rate due to a repositioning on the dose–response curve. However, recent data of final height in GH deficient children at the end of GH treatment suggests that, although there is an improvement over predicted adult height, target height is rarely achieved39–41. Furthermore, the results of GH treatment in non-GH-deficient short children demonstrate even less of an impact on final height42,43. We now propose that in order to optimize growth, GH replacement therapy should be tailored to mimic the underlying variability in normal GH output. This study is the first to describe a relationship between GH output and growth using simultaneous hormone sampling and height measurements in a large cohort of healthy prepubertal children over an extended period of time. We have demonstrated that GH variability is a more important determinant of normal mid-childhood growth than the amount of GH alone. Stature is correlated with overall variability in GH release, while increment in height and the magnitude of individual growth spurts are influenced by the regularity of the GH profile. Tall stature is therefore associated with large variation in GH release, while good growth, over and within the year, is related to constant short-term changes in GH.
ACKNOWLEDGEMENTS We thank the children who gave their time and effort, their parents and teachers, Sister Julie Jones and Margaret Downes. Financial support was provided by Serono Laboratories UK (MSG), Lilly Industries (MSG, NKST) and Pharmacia & Upjohn (VT).
REFERENCES 1. Freeman J V, Cole T J, Chinn S, Jones P R M, White E M, Preece M A. Cross-sectional stature and weight reference curves for the UK, 1990. Arch Dis Child 1995: 73: 17–24. 2. Lampl M, Veldhuis J D, Johnson M L. Saltation and stasis: a model of human growth. Science 1992; 258: 801–803.
GH and normal human growth 121
3. Hermanussen M, Geiger-Benoit K, Burmeister J, Sippell W G. Periodical changes of shortterm growth velocity (“mini growth spurts”) in human growth. Ann Hum Biol 1988; 15: 103–109. 4. Marshall W A, Swan A V. Seasonal variation in growth rates of normal and blind children. Hum Biol 1971; 43: 503–516. 5. Butler G E, McKie M, Ratcliffe S G. The cyclical nature of prepubertal growth. Ann Hum Biol 1990; 17: 177–198. 6. Gelander L, Karlberg J, Albertsson-Wikland K. Seasonality in lower leg length velocity in prepubertal children. Acta Paediatr 1994; 83: 1249–1254. 7. Thalange N K S, Foster P J, Gill M S, Price D A, Clayton P E. A model of normal prepubertal growth. Arch Dis Child 1996; 75: 427–432. 8. Albertsson-Wikland K, Rosberg S. Analysis of 24-hour growth hormone profiles in children: relation to growth. J Clin Endocrinol Metab 1988; 67: 493–500. 9. Hindmarsh P C, Matthews D R, Brook C G D. Growth hormone secretion in children determined by time series analysis. Clin Endocrinol (Oxf) 1988; 29: 35–44. 10. Hindmarsh P C, Smith P J, Brook C G D, Matthews D R. The relationship between height velocity and growth hormone secretion in short prepubertal children. Clin Endocrinol (Oxf) 1987; 27: 581–591. 11. Donaldson D L, Hollowell J G, Pan F, Gifford R A, Moore W V. Growth hormone secretory profiles: variation on consecutive nights. J Pediatr 1989; 115: 51–56. 12. Saini S, Hindmarsh P C, Matthews D R, Pringle P J, Jones J, Preece M A, Brook C G. Reproducibility of 24-hour serum growth hormone profiles in man. Clin Endocrinol (Oxf) 1991; 34: 455–462. 13. Albertsson-Wikland K, Rosberg S. Reproducibility of 24-hour growth hormone profiles in children. Acta Endocrinol (Copenh) 1992; 126: 109–112. 14. Skinner A M, Clayton P E, Price D A, Addison G M, Soo A. Variability in the urinary excretion of growth hormone in children: comparison with other urinary proteins. J Endocrinol 1993; 138: 337–343. 15. Main K M, Jarden M, Angelo L, Dinesen B, Hertel N T, Juul A, Muller J, Skakkebaek N E. The impact of gender and puberty on reference values for urinary growth hormone excretion: a study of 3 morning samples in 517 healthy children and adults. J Clin Endocrinol Metab 1994: 79: 865–871. 16. Thalange N K S, Gill M S, Gill L, Whatmore A J, Addison G M, Price D A, Clayton P E. Infradian rhythms in urinary growth hormone excretion. J Clin Endocrinol Metab 1996; 81: 100–106. 17. Wales J K H, Gibson A T. Short-term growth: rhythms, chaos, or noise? Arch Dis Child 1994; 71: 84–89. 18. Skinner A M, Price D A, Addison G M, Clayton P E, Mackay R I, Soo A, Mui C Y. The influence of age, size, pubertal status and renal factors on urinary growth hormone excretion in normal children and adolescents. Growth Regulation 1992; 2: 156–160. 19. Tillmann V, Buckler J M, Kibinge M S, Price D A, Shalet S M, Wales J K, Addison G M, Gill M S, Whatmore A J, Clayton P E. Biochemical tests in the diagnosis of childhood growth hormone deficiency. J Clin Endocrinol Metab 1997; 82: 531–535. 20. Tanner J M, Whitehouse R H, Takaishi M. Standards from birth to maturity for height, weight, height velocity and weight velocity: British children 1965. Arch Dis Child 1966; 41: 613–635. 21. Hartman M L, Pincus S M, Johnson M L, Matthews D H, Faunt L M, Vance M L, Thorner M O, Veldhuis J D. Enhanced basal and disorderly growth hormone secretion distinguish acromegalic from normal pulsatile growth hormone release. J Clin Invest 1994; 94: 1277–1288.
22. Diggle P J. Time series analysis: a biostatistical introduction. Oxford Science Publications, Oxford. 1990. 23. Skinner A M, Clayton P E, Price D A, Addison G M, Soo A. Urinary growth hormone excretion in the assessment of children with disorders of growth. Clin Endocrinol (Oxf) 1993; 39: 201–206. 24. Hourd P, Edwards R. Current methods for the measurement of growth hormone in urine. Clin Endocrinol (Oxf) 1994; 40: 155–170. 25. Johnson V, Maack T. Renal extraction, filtration, absorption and catabolism of growth hormone. Am J Physiol 1977; 233: F185–F196. 26. Ahmed S F, Barnes S I, Wallace W H B, Kelnar C J H. Short-term changes in urinary growth hormone excretion and lower leg length in healthy children. Horm Res 1997; 48: 72–75. 27. Martha P M, Rogol A D, Carisson L M S, Gesundheit N, Blizzard R M. A longitudinal assessment of hormonal and physical alterations during normal puberty in boys. I. Serum growth hormone binding protein. J Clin Endocrinol Metab 1993; 77: 452–457. 28. Jansson J-O, Eden S, Isaksson O. Sexual dimorphism in the control of growth hormone secretion. Endocr Rev 1985; 6: 128–150. 29. Gevers E F, Wit J M, Robinson I C A F. Growth, growth hormone (GH)-binding protein and GH receptors are differentially regulated by peak and trough components of the GH secretory pattern in the rat. Endocrinology 1996; 137: 1013–1018. 30. Ho K Y, Evans W S, Blizzard R M, et al. Effects of sex and age on the 24-hour profile of growth hormone secretion in man: importance of endogenous estradiol concentrations. J Clin Endocrinol Metab 1987; 64: 51–58. 31. Jaffe C A, Ocampo-Lim B, Guo W, Krueger K, Sugahara I, DeMott F R, Bennann M, Barkan A L. Regulatory mechanisms of growth hormone secretion are sexually dimorphic. J Clin Invest 1998; 102: 153–164. 32. Hindmarsh P C, Fall C H D, Pringle P J, Osmond C, Brook C G D. Peak and trough growth hormone concentrations have different associations with the insulin-like growth factor axis, body composition, and metabolic parameters. J Clin Endocrinol Metab 1997; 82: 2172–2176. 33. Mauras N, Blizzard R M, Link K, Johnson M L, Rogol A D, Veldhuis J D. Augmentation of growth hormone secretion during puberty: evidence for a pulse amplitude-modulated phenomenon. J Clin Endocrinol Metab 1987; 64: 596–601. 34. Suttie J M, Fennessy P F, Corson I D, Laas F J, Crosbie S F, Butler J H, Gluckman P D. Pulsatile growth hormone, insulin-like growth factors and antler development in red deer (Cervus elaphus scoticus) stags. J Endocrinol 1989; 121: 351–360. 35. Gelander L, Petterson M, Rosberg S, Albertsson-Wikland K. Longitudinal inter-individual changes in insulin-like growth factor-I (IGF-I) and its binding protein 3 (IGFBP-3) in relation to growth in healthy children. 1996; Horm Res 46(Suppl. 2): 100 (Abstract). 36. Vance M L, Kaiser D L, Evans W S, Furlanetto R, Vale W, Rivier J, Thorner MO. Pulsatile growth hormone secretion in normal man during a continuous infusion of human growth hormone releasing factor (1–40). J Clin Invest 1985; 75: 1584–1590. 37. Keret R, Ashkenazi I E, Wasserman M, Bauman B, Pertzelan A, Ticher A, Laron Z. Two types of growth hormone rhythm in boys with constitutionally short stature. J Ped Endocrinol Metab 1996; 9: 599–607. 38. Albertsson-Wikland K, Rosberg S, Lannering B, Dunkel L, Selstam G, Norjavaara E. Twenty-four-hour profiles of luteinizing hormone, follicle-stimulating hormone, testosterone, and estradiol levels: a semilongitudinal study throughout
122 M.S. Gill et al.
puberty in healthy boys. J Clin Endocrinol Metab 1997; 82: 541–549. 39. Price D A, Ranke M B. Final height following growth hormone treatment. In Ranke MB, Gunnarsson R, editors. Progress in Growth Hormone Therapy – 5 Years of KIGS. J, J Verlag, Mannheim. 1994: 129–144. 40. Blethen S L, Baptista J, Kuntze J, Foley T, LaFranchi S, Johanson A. Adult height in growth hormone (GH)-deficient children treated with biosynthetic GH. J Clin Endocrinol Metab 1997; 82: 418–420.
41. Coste J, Letrait M, Carel J C, Tresca J P, Chatelain P, Rochiccioli P, Chaussain J L, Job J C. Long term results of growth hormone treatment in France in children of short stature: population, register based study. BMJ 1997; 315: 708–713. 42. Hindmarsh P C, Brook C G D. Final height of short normal children treated with growth hormone. Lancet 1996; 348: 13–16. 43. Bernasconi B, Street M E, Volta C, Mazzardo G, and the Italian Multicentre Study Group. Final height in non-growth hormone deficient children treated with growth hormone. Clin Endocrinol (Oxf) 1997; 47: 261–266.