Growth Response to Growth Hormone (GH) Treatment Relates to ...

0021-972X/97/$03.00/0 Journal of Clinical Endocrinology and Metabolism Copyright © 1997 by The Endocrine Society

Vol. 82, No. 9 Printed in U.S.A.

Growth Response to Growth Hormone (GH) Treatment Relates to Serum Insulin-Like Growth Factor I (IGF-I) and IGF-Binding Protein-3 in Short Children with Various GH Secretion Capacities* ¨ M, C. JANSSON, S. ROSBERG, AND K. ALBERTSSON-WIKLAND B. KRISTRO ON BEHALF OF THE SWEDISH STUDY GROUP FOR GROWTH HORMONE TREATMENT† Department of Pediatrics, University of Umea (B.K.), Umea; and University of Goteborg (B.K., C.J., S.R., K.A.-W.), Goteborg, Sweden ABSTRACT The purpose of the study was to evaluate the relationship between the 1-yr (n 5 193) and 2-yr (n 5 128) growth response and the individual serum concentrations of insulin-like growth factor I (IGF-I) and IGF-binding protein 3 (IGFBP-3) before and during GH treatment. Our study group of prepubertal short children had from very low to high GH secretory capacity, estimated during an arginineinsulin tolerance test, and the ages ranged from 3–15 yr at the start of treatment. Their serum levels of IGF-I and IGFBP-3 were low before treatment compared to those in an age-related reference group of prepubertal children and increased significantly from the start to 1 month of GH treatment. The mean increase in height SD score was 0.80 SD score after 1 yr of GH treatment and 1.26 SD score after 2 yr, with a wide range. In univariate analyses the highest correlation coefficients to the 2-yr growth response were found to be vs. the following variables from the start of treatment: IGF-I SD score (r 5

20.49), log maximum GH concentration (log GHmax) during the arginine-insulin tolerance test (r 5 20.47), difference between the height SD score of the individual child and the midparental height SD score (diffSD score; r 5 20.45), IGFBP-3 SD score (r 5 20.39), age (r 5 20.30), short term change in IGFBP-3 SD score (r 5 0.37), and IGF-I SD score (r 5 0.34). In multivariate stepwise regression analysis, 41% of the variation in the 2-yr growth response could be explained by IGF-I SD score or log GHmax together with age at the start of treatment, weight SD score at 1 yr of age, and diffSD score. When both IGF-I SD score and GHmax were included and when the short term changes in IGF-I SD score were added, 46% and 58% of the variation, respectively, could be explained. The regression algorithms using different combinations of variables and their corresponding prediction intervals are also presented. (J Clin Endocrinol Metab 82: 2889 –2898, 1997)

S

insulin-like growth factor I (IGF-I) in groups of children defined as GH deficient overlap with the serum concentrations of GH and IGF-I in groups of non-GH-deficient children of short and/or normal stature (1, 2). Moreover, most short children increase their growth rate during GH therapy regardless of the diagnosis of GH deficiency (GHD) or idiopathic short stature (3–8). In fact, only children with GH insensitivity syndrome will not respond to GH treatment (9). With the diagnostic criteria used worldwide today, children with very low pituitary GH secretory capacity can be identified. Various GH provocation tests have been used in an attempt to identify predictors of the growth response to GH treatment (10 –12) or in diagnosing GHD, sometimes together with biochemical parameters, such as basal levels of IGF-I and IGF-binding protein-3 (IGFBP-3) (1, 13) or their short term changes during GH treatment (14). The aim of this study was to evaluate an alternative diagnostic approach to possible GH treatment of a short child without using a cut-off serum level of GH during the stimulation test as the crucial value. The relationship between the 1- and 2-yr growth responses to GH treatment and the basal levels of both IGF-I and IGFBP-3 and their short term changes was evaluated in a group of prepubertal short children with different GH secretory capacities, estimated by a stimulation test, together with auxology variables.

ELECTING children who will benefit from GH treatment is still a challenge despite nearly 40 yr of experience. Earlier studies of the growth response to GH treatment often used groups of short children defined either as GH-deficient/insufficient, i.e. with a maximum GH level (GHmax) during a pharmacological stimulation test below a certain arbitrary level, or defined as having normal GH secretion, i.e. GHmax during a stimulation test above the cut-off level. Several studies have shown that serum GH concentrations during GH provocation tests and 24-h GH profiles as well as serum concentrations of Received February 10, 1997. Revision received March 24, 1997. Revision received May 20, 1997. Accepted June 2, 1997. Address all correspondence and requests for reprints to: Dr. Berit Kristro¨m, University of Goteborg, Department of Pediatrics, International Pediatric Growth Research Center, Sahlgrenska University Hospital East, S-416 85 Goteburg, Sweden. E-mail: Berit.Kristrom@ pediatri.umu.se. * This work was supported by grants from Swedish Medical Research Council (no. 7509), Barnhusfonden, The Jerring Foundation, The Samariten Foundation, the Medical Faculties at the Universities of Umea and Goteborg, Wilhelm och Martina Lundgrens Foundation, The First of May Flower Annual Campaign, and Pharmacia & Upjohn. † The Swedish Study Group for Growth Hormone Treatment consists of Kerstin Albertsson-Wikland, Jan Alm, Stefan Aronsson, Jan Gustafsson, Lars Hagena¨s, Anders Ha¨ger, Sten Ivarsson, Berit Kristro¨m, Claude Marcus, Christian Moe¨ll, Karl Olof Nilsson, Martin Ritze´n, Torsten Tuvemo, Ulf Westgren, Otto Westphal, and Jan Åman.

2889

¨ M ET AL. KRISTRO

2890 Subjects and Methods

JCE & M • 1997 Vol 82 • No 9

Umea and the Karolinska Institute in Stockholm. Informed consent was obtained from the parents of each child and from the child, if old enough.

Study group A group of 193 short healthy prepubertal Swedish children (30 girls and 163 boys) with a broad range of GHmax in response to an arginineinsulin tolerance test (AITT; 0 –104 mU/L) were treated daily with GH (0.1 IU/kg) and followed during treatment for 1 yr (n 5 193) or 2 yr (n 5 128). The patients were either children with isolated idiopathic GHD, defined as having GHmax below 32 mU/L using the WHO International Reference Preparation (IRP) 80/505 (or ,20 mU/L using the WHO IRP 66/217), reported to the Swedish National Registry for GH treatment (n 5 117) or short children without GHD, defined as having a GHmax above 32 mU/L, included in national clinical trials with GH treatment (n 5 76). All children were well nourished and had normal thyroid, liver, and kidney functions. Children with coeliac disease were excluded, and the children were free from chronic disease and dysmorphic syndromes. All children had a gestational age (GA) at birth of more than 30 weeks, and their birth weights and birth lengths were 22.5 sd score or more for gestational age (15). The characteristics of the patients are given in Table 1.

Auxology Information on GA, birth weight, and birth length was collected from the obstetric report on the mothers kept at the Medical Birth Registry. The growth of the children was recorded at health care units from birth to inclusion in the study, i.e. 1 yr before the start of GH treatment. Thereafter, height was measured using a Harpenden stadiometer at the pediatric units. Height parameters were transformed into sd score, corrected for sex and age using the childhood component of the infancychildhood-puberty (ICP) growth model of Karlberg (16), weight parameters according to Karlberg et al. (17), and weight for height sd score (WHsd score sd score), i.e. weight sd score-b 3 height sd score, according to the method of Karlberg and Albertsson-Wikland (18, 19). Midparental height was expressed as the sd score compared with Swedish reference values (17). The difference between the height sd score of each child at the start of treatment and the midparental height expressed as the sd score (diffsd score) was calculated.

Hormone analysis

Study protocol

GH. Measurements of GH during the AITT were performed in different laboratories (n 5 86) using a polyclonal antibody-based immunoradiometric assay (Pharmacia & Upjohn, Uppsala, Sweden) with WHO IRP 80/505 as the standard. During 1991 and 1992, the laboratories in Sweden switched from WHO IRP 66/217 to WHO IRP 80/505, which meant that a conversion factor of 1.55 had to be used to compare the results of GH levels (n 5 107) (19, 20).

Pretreatment investigations. The endocrine investigation was performed during the pretreatment year. The children underwent an AITT according to the protocol described previously, and the GHmax was used for the analysis (12). Treatment follow-up. All children underwent the same regimen of daily GH treatment (0.1 IU/kg). Blood samples were taken at the start of treatment and after 10 and/or 30 days, 3 months, 1 yr, and 2 yr of treatment for analysis of serum concentrations of IGF-I and IGFBP-3. The samples were mostly taken between 1400 –1800 h, i.e. nearly 24 h after the latest GH injection. Measurements of IGF-I and IGFBP-3 were used only when the child was healthy; any known ongoing infection meant that the blood samples were excluded from the analysis for the prediction models. Serum samples were kept frozen until measurement of GH, IGF-I, and IGFBP-3. The studies were approved by the ethical committees of the Medical Faculties of the Universities of Goteborg, Lund, Uppsala, Linkoping, and

IGF-I. Concentrations of IGF-I were measured by an IGFBP-blocked RIA without extraction and in the presence of an approximately 250-fold excess of IGF-II (Mediagnost, Tubingen, Germany) (21). The intraassay coefficients of variation were 8.1%, 4.4%, and 4.5% at concentrations of 55, 219, and 479 mg/L, respectively, and the interassay coefficients of variation were 10.4%, 7.7%, and 5.3% at concentrations of 55, 219, and 479 mg/L, respectively. IGFBP-3. Concentrations of IGFBP-3 were determined using a previously reported RIA method (21). The intraassay coefficients of variation

TABLE 1. Characteristics of the study group of prepubertal children Variable

Auxology at birth Ht SD score Wt SD score Gestational age (weeks) Auxology at start of treatment Age (yr) Ht SD score Wt SD score Wt for htSD score SD score Change in ht SD score during pretreatment year MPH SD score Difference between an individual child’s ht SD score and MPH GH level GHmax at AITT (mU/L, WHO IRP 80/505) IGF-1 (SD score) At start After 10 and/or 30 days After 3 months After 1 yr After 2 yr IGFBP-3 (SD score) At start After 10 and/or 30 days After 3 months After 1 yr After 2 yr MPH, Midparental height.

SD

score

Median

Mean

n

20.89 20.72 40

20.90 20.63 39.2

193 193 193

0.94 0.94 1.84

22.5 to 2.7 22.5 to 2.3 31 to 43

9.1 22.7 21.6 0.7 0.0 1.0 21.9

8.9 22.7 21.6 0.8 0.0 20.9 21.9

193 193 193 193 190 193 193

2.8 0.7 0.6 0.8 0.3 0.8 1.1

2.8 to 14.7 26.0 to 21.1 23.9 to 0.1 21.1 to 4.2 20.8 to 1.1 23.0 to 1.7 26.2 to 0.5

22.6

26.2

193

17.6

0.4 to 103.8

21.26 20.05 0.26 0.48 0.34

21.44 20.14 0.18 0.40 0.27

193 152 142 161 95

1.22 1.10 1.00 1.20 1.24

25.21 to 1.89 23.99 to 3.13 23.32 to 2.77 22.84 to 3.20 23.00 to 2.79

20.88 0.07 0.21 0.23 0.42

21.05 0.07 0.16 0.31 0.42

193 151 142 160 94

1.21 1.06 1.01 1.07 1.14

24.75 to 2.34 23.21 to 2.90 23.04 to 2.72 22.87 to 3.23 23.27 to 4.02

SD

Range

GROWTH RESPONSE AND IGF-I AND IGFBP-3 were 6.2%, 5.6%, and 4.6% at concentrations of 1964, 2927, and 4799 mg/L, respectively, and the interassay coefficients of variation were 6.8%, 9.2%, and 6.9% at concentrations of 1964, 2927, and 4799 mg/L, respectively. As serum concentrations of IGF-I and IGFBP-3 are age dependent (22), the values were converted into sd score using a prepubertal reference, obtained in our laboratory from healthy children of normal (62 sd) stature (23). The following variables were included in the correlation analysis to the growth response: sex; GA; age at onset of the childhood component in the ICP model; weight, length, WHsd score sd score at birth and at 1 and 2 yr of age, and 1 yr before and at the start of GH treatment; together with the change in height, weight, and WHsd score sd score between birth and 1 yr and between birth and 2 yr of age. Also included were the yearly change in height, weight, and WHsd score sd score from 1 yr of age to the start of treatment and from 2 yr of age to the start of treatment together with the same auxological changes during the pretreatment year. The auxological information from birth to 2 yr of age was referred to as early growth. Maternal and paternal height sd scores as well as midparental height sd score were included in the analysis together with the diffsd score and the difference in height sd score between the child at the start of treatment and maternal and paternal height sd scores separately. Included in the statistical analyses were the individual basal levels of IGF-I and IGFBP-3, expressed as the sd score, and levels at 10 and/or 30 days (at either time or the mean) and 3 months together with the individual differences between the level at 10 and/or 30 days or 3 months and the level at the start of treatment. Also included was the ratio of IGF-I sd score/IGFBP-3 sd score for each child and the change from the start of treatment to 10 and/or 30 days and to 3 months. The GH values were log transformed before statistical analysis. Bone age estimations are not included in the analysis due to missing information from a number of patients.

Statistical analyses For testing changes over time, Fisher’s nonparametric permutation test for paired observations was used; for comparison between groups, Fisher’s nonparametric permutation test was used. Correlations were tested using Pitman’s nonparametric permutation test (24). Pearson’s correlation coefficients were used only for descriptive purposes. Exclusively variables correlated (P , 0.10) with the growth response were entered into a multiple stepwise forward linear regression analysis.

FIG. 1. Serum concentrations of IGF-I (top) and IGFBP-3 (bottom) in individual short children (E, girls; f, boys) plotted against reference values (shaded area) from healthy prepubertal children of normal stature (62 SD). Values from the start of GH treatment (left) and after 1 yr (middle) and 2 yr (right) of treatment are shown.

2891 Results

Changes in IGF-I levels

The serum concentrations of IGF-I in individual children at the start of GH treatment and after 1 and 2 yr of treatment compared with the age-related prepubertal reference values are shown in Fig. 1 (top panel). After 1 and 3 months of treatment, the mean IGF-I sd score for the group did not differ from the reference group and did not change significantly during the following 2 yr (Table 1). Compared with the level at the start of treatment, there was a highly significant change in individual serum IGF-I levels. Expressed as the sd score (sd, range) and shown in Fig. 2, the change from the start to 10 and/or 30 days of GH treatment was 1.26 (0.90, 20.82 to 4.69; n 5 152; P , 0.0001), that from the start to 3 months of treatment was 1.47 (0.89, 20.90 to 5.16; n 5 142; P , 0.0001), that from the start to 1 yr was 1.71 (1.13, 20.98 to 6.01; n 5 161; P , 0.0001), and that from the start to 2 yr was 1.66 (1.19, 21.32 to 5.03; n 5 95; P , 0.0001). For those 83 children measured at both 1 and 2 yr of treatment, there was no significant difference in the individual IGF-I sd score between these two points [0.07 (0.92, 23.26 to 2.13; n 5 83; P 5 NS]. Changes in IGFBP-3 levels

The serum concentrations of IGFBP-3 in individual children at the start of GH treatment and after 1 and 2 yr of treatment compared with reference values are shown in Fig. 1 (bottom panel). For IGF-I, the mean IGFBP-3 sd score of the group was in the normal range after 1 month of treatment and did not change significantly during the following 2 yr (Table 1). The mean increase in individual serum concentrations of IGFBP-3, expressed as the sd score (sd, range), from the start to 10 and/or 30 days of GH treatment was 1.10 (0.83, 20.95 to 4.17; n 5 151; P , 0.0001), that from the start to 3 months

2892

¨ M ET AL. KRISTRO

JCE & M • 1997 Vol 82 • No 9

total height gain was 1.26 sd score (0.53, 0.42 to 3.14; P , 0.0001 compared with the pretreatment year). There was considerable interindividual variation in the response, even though none of the children was known to be noncompliant (Fig. 3, top panel). The mean attained height at the start of treatment (sd, range) was 22.74 (0.69, 26.00 to 21.12), 21.94 (0.72, 24.49 to 0.26) after 1 yr, and 21.57 (0.84, 23.58 to 1.25) after 2 yr (Fig. 3, middle panel). The growth response can also be expressed as the diminishing distance to the individual midparental height, expressed as the sd score. This mean difference (sd, range), was 21.95 (1.07, 26.16 to 0.46) at the start of treatment, 21.08 (0.90, 24.65 to 1.21) after 1 yr of treatment, and 20.70 (0.90, 24.22 to 1.05) after 2 yr of treatment. During these 2 yr of treatment, the children were showing catch-up growth and were approaching their appropriate target height (Fig. 3, bottom panel). Correlation of the growth response to GH treatment

FIG. 2. Individual change in serum IGF-I SD score (top) and IGFBP-3 SD score (bottom) from the start of GH treatment to 10 and/or 30 days, 3 months, 1 yr, and 2 yr of treatment. Box and whisker plots indicate median, lower, and upper quartiles, and the whiskers show the 1st and 99th percentiles. The mean values are depicted as solid circles in each plot. ***, P , 0.0001 compared with the level at the start of treatment.

was 1.08 (0.87, 21.40 to 3.95; n 5 142; P , 0.0001), that from the start to 1 yr was 1.26 (1.05, 21.25 to 5.59; n 5 160; P , 0.0001), and that from the start to 2 yr was 1.46 (1.20, 21.19 to 4.44; n 5 94; P , 0.0001). For those 83 children measured at both 1 and 2 yr of treatment, there was a significant increase from the first to the second year in individual IGFBP-3 sd score [mean, 0.22 (0.85, 22.05 to 2.20); P , 0.05; Fig. 2]. Growth response

The mean increase in height sd score (sd, range) during the first year of GH treatment was 0.80 (0.34, 0.08 to 2.27; P , 0.0001 compared with the pretreatment year change in height sd score) and during the second year was 0.46 (0.25, 20.11 to 1.51; P , 0.0001 compared with the pretreatment year change in height sd score). After 2 yrs of treatment, the mean

Univariate analysis. The variables that correlated (P , 0.10, by Pitman’s permutation test) to the growth response after 1 or 2 yr of GH treatment are presented together with the correlation coefficients in Table 2. The correlation to IGF-I and IGFBP-3 variables are shown in Fig. 4 and to other variables selected in the multivariate analysis in Fig. 5. Note the broad range in GHmax during the AITT as well as the continuum in the growth response. The correlation coefficient between the IGF-I sd score at the start of treatment and the IGFBP-3 sd score was 0.76 (P , 0.0001) and that between IGF-I sd score at the start of treatment and log GHmax at AITT was 0.34 (P , 0.0001). When the IGFBP-3 sd score was compared with log GHmax at AITT, the correlation coefficient was 0.43 (P , 0.0001). There was a week correlation between the GHmax at AITT and the short term (10 and/or 30 days) change in IGF-I sd score or IGFBP-3 sd score (r 5 20.20; P , 0.05 for both). Multivariate analysis. In the initial linear regression analysis, the commonly used variables when diagnosing GHD were included: parental heights, pretreatment growth pattern, age, and GHmax at AITT. The GHmax at AITT was the first selected variable, followed, in order of selection, by diffsd score, attained weight sd score at 1 yr of age, and age at the start of treatment (analysis 1, Table 3); these variables accounted for 41% of the variance in the growth response. In the next analysis, the IGF-I sd score and IGFBP-3 sd score were included, whereas GH values were excluded (analysis 2a). The same level of variance (41%) in the growth response was explained, with the IGF-I sd score being the most informative variable. Thus, the IGF-I sd score was as informative as the GHmax at AITT at the start of treatment. In the third analysis, both the GHmax from the AITT together with the IGF-I sd score and the IGFBP-3 sd score were included (analysis 3a). This resulted in a small increase in the level of variance that could be explained (46%). The highest level of explanation of the variance (58%) was obtained when the GHmax from the AITT and the IGF-I sd score at the start of treatment together with short term changes in the IGF-I sd

GROWTH RESPONSE AND IGF-I AND IGFBP-3

2893

score after the start of GH treatment were combined (analysis 3b). When only the variables most often available in a clinical setting at the start of treatment were included, i.e. auxology during the pretreatment year, parental heights, IGF-I sd score, and IGFBP-3 sd score, but not GH values or the auxological information from before 3 yr of age (here described as early growth), the following order of selection was obtained, predicting the 2 yr growth response: IGF-I sd score at the start, age at the start, midparental height sd score, and change in WHsd score sd score during the pretreatment year; these variables explained 39% of the variance (analysis 4a). Adding the short term changes (both 10 and/or 30 days and 0 –3 months) in the IGF-I sd score and the IGFBP-3 sd score, the 0 –3 months change in IGF-I sd score was selected, and the variance that now could be explained was 43% (analysis 4b). The estimated regression algorithms for the yearly 2-yr growth response, based on the different variables used in each model, are shown in Table 3. The regression algorithms for the 1-yr growth response were also calculated. Using almost the same variables, 0% (analysis 4a) to 12% (analysis 3b) less of the variance could be explained. Using the different regression algorithms presented in Table 3, a 95% prediction interval was calculated, using the mean values, 61sd, and 62sd for the included variables (reminding the reader that the amplitude of the prediction interval is wider than the amplitude of the 95% confidence interval of the algorithm). The values obtained are presented in Table 3, right column. An illustration of the pretreatment growth and the prediction interval of the 2-yr growth response for one boy, with growth response about the mean of the group, is shown in Fig. 6. Discussion

FIG. 3. Top, Growth response, expressed as change in the height SD score, of the study group of short children with different GH status during the pretreatment year and during the first, second, and 2-yr periods of GH treatment (0.1 IU/kgzday). Middle, Attained height SD score of the group at 1 yr before, at the start, and after 1 and 2 yr of GH treatment. The group mean midparental height is indicated (dashed line) 6 1 SD (dotted lines). Bottom, Growth response to GH treatment, expressed as the distance to the individual midparental height SD score at the start of treatment and as the diminishing

Our study can be seen as an attempt to mathematically/ statistically evaluate the clinical information we already are using. When using the classic diagnostic criteria for GH deficiency, we have previously shown that it is possible to explain only 33% of the variation in the growth response (12). In the present study it was possible to explain 41% of the variance in the 2-yr growth response when the infancy growth data from the child were included together with the growth during the pretreatment year, parental heights, and GHmax at AITT. Replacing the GHmax at AITT by the IGF-I sd score and the IGFBP-3 sd score at the start of treatment, the IGF-I sd score was selected, and the level of explanation could be maintained. As much as 58% of the variance in the 2-yr growth response could be explained using the variables age at the start of treatment, GHmax at AITT, IGF-I sd score at the start of treatment, and change in the IGF-I sd score during the first 3 months of treatment. Thus, by increasing the included information about the child, the explained variation in growth response increased from 33% to 58%. According to these results, there is the possibility of using

distance after 1 and 2 yr of treatment. Box and whisker plots indicate the median, lower, and upper quartiles, and the whiskers show the 1st and 99th percentiles. The mean values are depicted as solid circles in each plot.

2894

¨ M ET AL. KRISTRO

JCE & M • 1997 Vol 82 • No 9

TABLE 2. Variables correlated (P , 0.10) to 1 or 2 yr growth response to GH treatment Numbers of children, Pearson correlation coefficient, P-values obtained by Pitman’s permution test. DHt

PRETREATMENT VARIABLES Parental ht Maternal ht SD score Paternal ht SD score Midparental ht SD score Diff-SD scorea Sex Birthb Wt SD score Length SD score 1 yr of ageb Wt SD score 2 yr of ageb WHSD score SD score Difference between 2 yr of age and birthb Ht SD score WHSD score SD score Start of treatment Age Ht SD score WHSD score SD score Difference between start of treatment and 1 yr of ageb Wt SD score Ht SD score WHSD score SD score Difference between start of treatment and 2 yr of ageb Wt SD score WHSD score SD score Difference during pretreatment year Ht SD score WHSD score SD score GHmax at AITT (log) Start of treatment IGF-I SD score IGFBP-3 SD score IGF-I SD score/IGFBP-3 SD score SHORTTERM VARIABLES At 10 and/or 30 days of treatment IGF-I SD score IGFBP-3 SD score IGF-I SD score/IGFBP-3 SD score At 3 months of treatment IGF-I SD score IGFBP-3 SD score IGF-I SD score/IGFBP-3 SD score Difference between 10 and/or 30 days of treatment and start DIGFI SD score DIGFBP-3 SD score Difference between 3 months of treatments and start DIGF-I SD score DIGFBP-3 SD score Dratio IGF-I SD score/IGFBP-3 SD score a b

SD

score 0 –1 yr treatment

DHt

score 0 –2 yr treatment

r

P

n

r

P

193 193 193 193 193

0.20 0.32 0.33 20.42 20.15

,0.01 ,0.001 ,0.001 ,0.001 ,0.05

128 128 128 128 128

0.24 0.37 0.37 20.45 20.20

,0.01 ,0.001 ,0.001 ,0.001 NS

193 193

0.08 0.15

NS ,0.05

128 128

0.18 0.28

,0.05 ,0.01

182

0.02

NS

119

0.17

,0.10

173

0.10

NS

115

0.20

,0.05

174 173

20.21 20.11

,0.01 NS

116 115

20.24 0.16

,0.05 ,0.10

193 193 193

20.36 20.13 0.13

,0.001 ,0.10 ,0.10

128 128 128

20.30 20.08 0.17

,0.001 NS ,0.10

180 180 180

20.07 20.26 0.27

NS ,0.01 ,0.01

118 118 118

20.22 20.28 0.15

,0.05 ,0.01 NS

171 171

20.09 0.02

NS NS

113 113

20.21 20.16

,0.05 ,0.10

188 188 191

20.21 0.17 20.43

,0.01 ,0.05 ,0.001

128 126 126

20.20 0.19 20.47

,0.05 ,0.05 ,0.001

193 193 193

20.45 20.35 20.34

,0.001 ,0.001 ,0.001

128 128 128

20.49 20.39 20.36

,0.001 ,0.001 ,0.001

152 153 152

20.17 20.14 20.15

,0.05 ,0.10 ,0.10

102 103 102

20.34 20.25 20.29

,0.001 ,0.05 ,0.01

143 143 143

20.14 20.08 20.13

,0.10 NS NS

97 97 97

20.29 20.20 20.23

,0.01 ,0.05 ,0.05

152 152

0.39 0.35

,0.001 ,0.001

102 102

0.32 0.32

,0.01 ,0.01

143 143 142

0.34 0.37 0.19

,0.001 ,0.001 NS

97 97 96

0.34 0.37 0.10

,0.01 ,0.001 NS

Difference in height SD score of the individual child at the start of treatment and the midparental height Summarized in text as early growth variables.

either the GHmax at AITT or the IGF-I sd score at the start of treatment to predict the growth response, but the explanation level is still limited. Both means of investigation have their disadvantages; GHmax at AITT may be falsely low because of refractoriness of the somatotrophs at the time of provocation (25, 26), beside its clinical disadvantage, and IGF-I is markedly dependent not only on GH but also on nutritional factors. We found that the transformation of IGF-I and IGFBP-3 into the sd score, relevant for age and pubertal stage, was

SD

n

SD

score.

very important, as the level of explanation was much lower when absolute values were used (data not shown). In our patient group, both IGF-I- and IGFBP-3-values were available at each occasion that blood was sampled. The values were strongly correlated with each other, and like results in earlier studies, where IGFBP-3 has been shown to be a reliable reflector of the GH secretion (13, 27), this was also found in this study group, with the IGFBP-3 sd score having a higher correlation coefficient to GHmax at AITT than the

GROWTH RESPONSE AND IGF-I AND IGFBP-3

FIG. 4. Individual values for IGF-I SD score and IGFBP-3 at the start of GH treatment (n 5 128; top) and the individual short term (10 and/or 30 days) changes in IGF-I SD score and IGFBP-3 (n 5 102; bottom) plotted against individual 2-yr growth responses, expressed as change in height SD score/ chronological age.

FIG. 5. Individual pretreatment variables plotted against the individual 2-yr growth response on GH treatment, expressed as change in height SD score/ chronological age.

2895

1

2

2

1

1

2

2

1

1

1

1

1

2

2

Analysis 1

Analysis 2a

Analysis 2b

Analysis 3a

Analysis 3b

Analysis 4a

Analysis 4b

Early GH at growth AITT

1

1

1

1

1

1

2

1

2

1

2

1

2

2

DIGF IGF-I SD score and SD score IGFBP-3 SD 1–3 months score at start

Variables R2

78 0.43

129 0.39

69 0.58

119 0.46

70 0.55

122 0.41

119 0.41

n

0.20

0.21

0.18

0.20

0.18

0.21

0.21

SD of the residuals

Intercept GHmax at AITT Diff-SD score Wt SD score at 1 yr of age Age at start Intercept IGF-I SD score at start Age at start Wt SD score at 1 yr of age Diff-SD score Intercept IGF-I SD score at start Age at start Wt SD score at 1 yr of age DWHSD score SD score pretreatment yr DIGF-I SD score 3 months of treatment to start Intercept IGF-I SD score at start Age at start GHmax at AITT Wt SD score at 1 yr of age Diff-SD score Intercept IGF-I SD score at start Age at start Wt SD score at 1 yr of age DWHSD score SD score pretreatment yr GHmax at AITT DIGF-I SD score 3 months of treatment to start Intercept IGF-I SD score at start Age at start MPH SD score DWHSD score SD score pretreatment yr Intercept IGF-I SD score at start Age at start DWHSD score SD score pretreatment yr DIGF-I SD score 3 months of treatment to start

Variable

0.0922 0.0261 0.00954 0.0571 0.0300

0.734 20.146 20.0375 0.137 0.0650

0.0719 0.0159 0.00762 0.0222 0.0417

0.0698 0.0292

20.143 0.0600 0.780 20.0879 20.0282 0.0609 0.101

0.113 0.0178 0.00817 0.0555 0.0192 0.0211 0.135 0.0222 0.00994 0.0227 0.0534

0.0283

0.0715 0.968 20.0607 20.0253 20.176 0.0644 20.0490 1.11 20.0765 20.0415 0.0924 0.150

0.117 0.0561 0.0192 0.0120 0.00842 0.0905 0.0176 0.00806 0.0195 0.0215 0.100 0.0222 0.00965 0.0228 0.0542

SE(b)

1.02 20.255 20.0844 0.0710 20.0207 0.742 20.0770 20.0295 0.0696 20.0538 0.942 20.0835 20.0464 0.0985 0.161

Regression coefficient

Estimated regression algorithm for the two year growth response

score/chronological age

0.41

0.42

0.36

0.40

0.37

0.41

0.41

Mean of predictors

60.42

60.43

60.39

60.40

60.39

60.41

60.42

Mean 6 1 SD

60.44

60.45

60.46

60.42

60.45

60.43

60.43

Mean 6 2 SD

¨ M ET AL. KRISTRO

0.0336

0.0001 0.0001 0.0002 0.0192

0.0001 0.0001 0.0003 0.0069 0.0169

0.0446 0.0446

0.0001 0.0009 0.0025 0.0019 0.0011 0.0218 0.0001 0.0010 0.0001 0.0001 0.0065

0.0140

0.0001 0.0001 0.0001 0.0006 0.0157 0.0001 0.0001 0.0004 0.0005 0.0135 0.0001 0.0004 0.0001 0.0001 0.0042

P

SD

95% prediction interval for a patient with values at

TABLE 3. Estimated regression algorithms for the 2 yr growth response to GH treatment (0.1 U/kgzday), expressed as yearly change in height

2896 JCE & M • 1997 Vol 82 • No 9

GROWTH RESPONSE AND IGF-I AND IGFBP-3

2897

FIG. 6. The growth chart from one of the boys before and after the start of GH treatment (arrow at 7.1 yr) with the prediction interval after 2 yr of treatment (mean 6 1 and 2SD), according to analysis 3a, with his individual pretreatment values included into the algorithm.

IGF-I sd score. However, regarding the growth response to GH treatment, although IGF-I and IGFBP-3 levels at the start of treatment were each negatively correlated with the 1- and 2-yr growth responses, in this study the IGF-I value was selected as the most informative variable of the two. In a more heterogeneous group of children or in a clinical setting, it can be wise to measure both the IGF-I sd score and the IGFBP-3 sd score, because of the stability over time of the latter. If the levels are discordant, additional evaluation may be indicated. The serum IGF-I concentrations for the whole group were low at the start of GH treatment, but were within the reference range after 1 and 2 yr of treatment in most children. This indicates that the dose of GH given was not too high for most of the children, as we did not find high IGF-I concentrations during treatment. However, the reference ranges of IGF-I and IGFBP-3 are wide, and some children in our study attained high values during treatment, whereas others still had low values. This variability illustrates the need for using individual GH treatment regimens that take account of the variability in individual sensitivity of the GH/IGF-I axis, as reflected by individual IGF-I levels. The variability in the growth response in the present heterogeneous group of short children was expected and reflects the individual differences in tissue responsiveness to GH. Because all children were treated with the same GH regimen, we did not have to consider the effects of different GH doses and injection frequencies, which have been shown to be important contributory factors to the growth response to GH treatment (28, 29). Ranke and Guilbaud (30) and Blethen et al. (10) have shown that the dose and frequency of injections are important predictors of growth in children with GHD and certainly in children with idiopathic short stature (31). Some of the short children in our study group were probably partially GH resistant (32, 33), although we were unable to assess GH-binding protein levels in a sufficient number of children to include this variable in the analysis. However, Attie et al. (34) showed that in prepubertal chil-

dren, the growth response to GH treatment did not differ between groups with low or normal GH-binding protein levels. We conclude that this group of short prepubertal children with a wide range of GH status has reduced serum concentrations of IGF-I and IGFBP-3 compared with those in healthy children of normal height. Treatment with GH (0.1 IU/kgzday) normalized the levels of IGF-I and IGFBP-3 for most, but not all, children. There was a broad range in the growth response to GH treatment, which could be best explained by chronological age (the younger the better) and diffsd score (the higher the better). These variables together with the IGF-I sd score at the start of treatment and the GHmax at AITT (the lower the better for both) accounted for about 41– 46% of the variability in the 2-yr growth response. The IGF-I sd score and AITT GHmax were interchangeable in the prediction model, as were GHmax at AITT and short term changes in IGF-I sd score during treatment. Although both IGF-I and IGFBP-3 were highly correlated, and both correlated with the growth response, IGF-I was the more informative in explaining the variation in growth response. Thus, there are two alternatives, to use either the GHmax at AITT or the IGF-I sd score, when predicting the 2-yr growth response before possible GH treatment of a short child. With all of the variables used now, as much as 58% of the variance in the growth response could be explained. This is, to our knowledge, the highest level of explanation presented to date for a study group such as this. However, when using our estimated regression algorithms to construct the prediction intervals, the range in yearly growth response is still wide (about 60.4 sd) even given the now available combinations of tests and baseline variables. In clinical work, when selecting children for GH treatment, it is valuable to know the calculated estimate of the predictive value of different combinations of variables, although it is low. Thus, the search for better indicators of the individual growth responsiveness to GH treatment should continue.

¨ M ET AL. KRISTRO

2898 Acknowledgments

We thank Dr. Werner Blum for the IGF-I and IGFBP-3 assays, Ms. Birgitta Svensson and Ms. Lisbeth Larsson for technical assistance, NilsGunnar Pehrsson for statistical support, and all participants in the National Registry for GH treatment and clinical trials on GH treatment in short children.

References 1. Lee P, Wilson D, Rountree L, Hintz R, Rosenfeld R. 1990 Efficacy of insulinlike growth factor I levels in predicting the response to provocative growth hormone testing. Pediatr Res. 27:45–51. 2. Hindmarsh P, Smith PJ, Brook CGD, Matthews DR. 1987 The relationship between height velocity and growth hormone secretion in short prepubertal children. Clin Endocrinol (Oxf). 27:581–591. 3. Albertsson-Wikland K, Hall K. 1987 Growth hormone treatment in short children: relationship between growth and serum insulin-like growth factor I and II levels. J Clin Endocrinol Metab. 65:671– 678. 4. Schwartz D, Hu C-S, Shulman D, Root A, Bercu B. 1990 Linear growth response to exogenous growth hormone in children with short stature. Am J Dis Child. 144:1092–1097. 5. Moore K, Donaldsson D, Ideus P, Gifford R, Moore V. 1993 Clinical diagnosis of children with extremely short stature and their response to growth hormone. J Pediatr. 122:687– 692. 6. Cacciari E, Tassoni P, Cicognani A, et al. 1994 Value and limits of pharmacological and physiological tests to diagnose growth hormone (GH) deficiency and predict therapy response: first and second retesting during replacement therapy of patients defined as GH deficient. J Clin Endocrinol Metab. 79:1663–1669. 7. Allen D, Brook C, Bridges N, Hindmarsh P, Guyda H, Frazier D. 1994 Therapeutic controversies: growth hormone (GH) treatment of non-GH deficient subjects. J Clin Endocrinol Metab. 79:1239 –1240. 8. Wit J, Boersma B, Muinck Keizer-Schrama S, et al. 1995 Long-term results of growth hormone therapy in children with short stature, subnormal growth rate and normal growth hormone response to secretagogues. Clin Endocrinol (Oxf). 42:365–372. 9. Laron Z. 1995 Prismatic cases: Laron syndrome (primary growth hormone resistance) from patient to laboratory to patient. J Clin Endocrinol Metab. 80:1526 –1531. 10. Blethen S, Compton P, Lippe B, Rosenfeld R, August G, Johansson A. 1993 Factors predicting the response to growth hormone (GH) therapy in children with GH deficiency. J Clin Endocrinol Metab. 76:574 –579. 11. Spagnoli A, Spadoni GL, Cianfarani S, Pasquino AM, Troiani S, Boscherini B. 1995 Prediction of the outcome of growth hormone therapy in children with idiopatic short stature. J Pediatr. 126:905–909. 12. Kristro¨m B, Karlberg J, Albertsson-Wikland K. 1995 Prediction of the growth response of short prepubertal children treated with growth hormone. Acta Paediatr Scand. 84:51–57. 13. Jackson Smith W, Nam T, Underwood L, Busby W, Celnicker A, Clemmons D. 1993 Use of insulin-like growth factor-binding protein-2 (IGFBP2), IGFBP-3, and IGF-I for assessing growth hormone status in short children. J Clin Endocrinol Metab. 77:1294 –1299. 14. Thalange N, Price D, Gill M, Whatmore A, Addison G, Clayton P. 1996 Insulin-like growth factor binding protein-3 generation: an index of growth hormone insensitivity. Pediatr Res. 39:849 – 855. 15. Niklasson N, Ericson A, Fryer J, Karlberg J, Lawrence C, Karlberg P. 1991 An

16. 17.

18. 19.

20.

21. 22.

23. 24. 25.

26. 27. 28.

29.

30.

31.

32.

33. 34.

JCE & M • 1997 Vol 82 • No 9

update of the Swedish reference standards for birth for weight, length, and head circumference at birth for given gestational age (1977– 81). Acta Paediatr Scand. 80:756 –762. Karlberg J. 1989 On the construction of the infant-childhood-puberty growth standard. Acta Paediatr Scand. 356(Suppl):26 –37. Karlberg P, Taranger J, Engstro¨m I, Lichtenstein H, Svennberg-Redegren I. 1976 The somatic development of children in a Swedish urban community. Acta Paediatr Scand. 258(Suppl):7–76. Karlberg J, Albertsson-Wikland K. 1996 Recent developments in infant nutrition. In: Bindels JG, Goedhardt AC, Visser HKA, eds. : Kluwer; 112–127. Albertsson-Wikland K, Rosberg S, Karlberg J, Groth T. 1994 Analysis of 24-hour growth hormone profiles in healthy boys and girls of normal stature: relation to puberty. J Clin Endocrinol Metab. 78:1195–1201. Boguszewski M, Rosberg S, Albertsson-Wikland K. 1995 Spontaneous 24hour growth hormone profiles in prepubertal small for gestational children age children. J Clin Endocrinol Metab. 80:2599 –2606. Blum WF, Breier BH. 1994 Radioimmunoassays for IGFs and IGFBPs. Growth Regul. 4:11–19. Juul A, Dalgaard P, Blum W, et al. 1995 Serum levels of insulin-like growth factor (IGF)-binding protein-3 (IGFBP-3) in healthy infants, children, and adolescents: the relation to IGF-I, IGF-II, IGFBP-1, IGFBP-2, age, sex, body mass index, and pubertal maturation. J Clin Endocrinol Metab. 80:2534 –2542. Jansson C, Andersson E, Rosberg S, Albertsson-Wikland K. Proc of the 10th Int Congr of Endocrinol. 1996;P3–P49. Bradley J. 1968 Distribution of free statistical tests. London. Prentice-Hall; 68 – 86. Gelander L, Albertsson-Wikland K. 1989 Growth hormone (GH) release after administration of GH-releasing hormone to endogenous 24-h GH secretion in short children. J Endocrinol. 122:61– 68. Rosenfeld RG. 1997 Editorial: Is growth hormone deficiency a viable diagnosis? J Clin Endocrinol Metab. 82:349 –351. Blum W, Ranke M. 1990 Use of insulin-like growth factor-binding protein 3 for the evaluation of growth disorders. Horm Res. 33:31–37. Albertsson-Wikland K, Westphal O, Westgren U. 1986 Daily subcutaneous administration of human growth hormone in growth hormone deficient children. Acta Paediatr Scand. 75:89 –97. MacGillivray M, Baptista J, Johanson A. 1996 Outcome of a four-year randomized study of daily vs. three times weekly somatropin treatment in prepubertal naive growth hormone-deficient children. Genentech Study Group. J Clin Endocrinol Metab. 81:1806 –1809. Ranke M, Guilbaud O. 1991 Growth response in prepubertal children with idiopathic growth hormone deficiency during the first two years of treatment with human growth hormone. Analysis of the Kabi Pharmacia International Growth Study. Acta Paediatr Scand. 379(Suppl):109 –115. Ranke M, Guilbaud O, Lindberg A, Cole T. 1993 Prediction of the growth response in children with various growth disorders treated with growth hormone: analysis of data from the Kabi Pharmacia International Growth Study. Acta Paediatr. 391(Suppl):82– 88. Goddard A, Covello R, Luoh S-M, et al. 1995 Mutations of the growth hormone receptor in Children with idiopatic short stature. N Engl J Med. 333:1093–1098. Savage M, Woods K. 1996 The investigation of growth hormone insensitivity. Clin Endocrinol (Oxf). 45:257–260. Attie M, Carlsson L, Rundle A, Sherman B. 1995 Evidence for partial growth hormone insensitivity among patients with idiopatic short stature. J Pediatr. 127:244 –250.