| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Other Original Articles |
Göteborg Pediatric Growth Research Center (C.L., L.G., S.R., K.A.W.), Institute for the Health of Women and Children and Department of Statistics (E.A.), The Sahlgrenska Academy at Göteborg University, S-416 85 Göteborg, Sweden; and University of Giessen and Lilly Research Laboratories (W.F.B.), D-35 385 Bad Homburg, Germany
Address all correspondence and requests for reprints to: Chatarina Löfqvist, University of Göteborg, Institute for the Health of Women and Children, Göteborg Pediatric Growth Research Center, The Queen Silvia Children’s Hospital, S-416 85 Göteborg, Sweden. E-mail: chatarina.lofqvist{at}vgregion.se
Abstract
We have constructed a reference model to facilitate comparison of serum IGF-I values among children, and thereby to improve the value of IGF-I measurements for diagnosis. The data set consists of serum values measured in 969 samples from 468 healthy children and adolescents (232 males, 236 females; ages, 1.1–18.3 yr). One sample per child was used for the model, each being selected so as to provide sufficient observations for each stage of puberty. The samples not selected were used to validate the reference data. The IGF-I values were log transformed, and multiple regression analysis was used in the model-building process. The best linear model, which converts serum IGF-I concentrations into SD scores and explains 66% of the variation in logIGF-I values, includes the variables of age, gender, and puberty, and takes the interactions among these variables into account. In prepubertal and early pubertal children, the relationship between age and logIGF-I was positive, with greater effect in girls older than 8 yr. In mid-puberty, logIGF-I values were higher in girls than in boys of the same age, up to 16 yr of age. Among boys, the most pronounced positive relationship between age and logIGF-I occurred in mid-puberty, whereas the relationship between age and logIGF-I among girls in mid-puberty is fairly constant. In late puberty, logIGF-I values were higher than earlier in puberty, and there was a negative relationship with age in both boys and girls. Instead of separate models for each combination of puberty and gender, estimating a single regression model permits simultaneous estimation of all explanatory variables and uses all observations in the data set, thereby making it easier to select those variables that have a significant effect on logIGF-I. Our model shows that IGF-I levels are related to age during each stage of puberty. The model also accounts for the fact that serum IGF-I concentrations during puberty are different for boys and girls.
THE CLINICAL USE of measurements of IGF-I has been focused primarily on diagnosing or excluding GH deficiency, and on monitoring GH therapy. IGF-I SD scores have been shown to relate to the GHmax in serum during provocative tests (1, 2). Although measurements of IGF-I are useful for diagnosis and follow-up of patients with acromegaly (3), the sensitivity and specificity of IGF-I or IGFBP-3 in the diagnosis of GH deficiency in children is a matter of controversy (4, 5). Because the liver is the principal source of IGF-I in the circulation (6), and because hepatic production of IGF-I is highly influenced by nutritional factors (7), it is possible that decrements in IGF-I expected with GH deficiency are modified by nutritional status and other factors, such that only severe GH deficiency produces a clear segregation of children who are deficient from those who are not (8). Also problematic is the marked variation found in healthy children, where IGF-I values may double or be reduced by half from one month to another (9).
Numerous reports indicate that the level of IGF-I is higher in older than in younger children (10, 11, 12, 13, 14, 15, 16). Also, puberty produces significant increments in IGF-I (17, 18, 19, 20, 21, 22, 23, 24, 25). These likely result from increased GH secretion mediated by sex hormone (i.e. estrogen) (10, 11, 12, 13, 14, 15, 16). In boys, aromatization of T to E2 appears to be important for the pubertal increase in GH secretion and IGF-I (24).
The objective of our study was to develop a childhood and adolescent model for converting serum concentrations of IGF-I to SD scores. We sought to develop a model that would explain as much of the variation in IGF-I as possible, and to include in the model the explanatory variables that have a significant effect on IGF-I concentrations. The variables selected were those shown in a preliminary exploratory study and in previously published studies to effect IGF-I concentrations. It was also important that the model be robust, as well as simple to use.
Subjects and Methods
Subjects
Our data originally consisted of 1022 measurements of IGF-I in serum from 663 children. Some children had IGF-I measurements only once; a large group had two measurements and some children had several measurements. Because we wished to use only one IGF-I measurement per child, we determined whether any systematic differences in IGF-I could be detected between the children with one measurement and those with several measurements. No such difference was observed (data not shown).
Because the reference model is cross-sectional, using only one observation per child, not all of the original observations could be used to produce the model. To construct a reference model for a normal population, the only children included were those with height and weight within ± 3 SD of the population mean (26). Because puberty is an important explanatory variable for IGF-I and the total number of measurements available within each pubertal stage differed considerably, we selected observations in such a way as to provide sufficient information for each pubertal stage. The selection process was as follows:
Girls. The pubertal stage having the smallest number of children was early puberty (25 girls measured). If these girls had any measurements in pre-, mid-, or late puberty, these measurements were omitted from the data. The same procedure was thereafter applied to mid-puberty; that is, those girls who had a measurement in mid-puberty were omitted from prepuberty and late puberty. Thereafter, this procedure was applied to the prepubertal stage.
Boys. The omission procedure described above was applied to data from the boys, in the following order for the pubertal stages: mid, early, and late.
Pubertal stage. Puberty in girls was classified into four stages according to breast development: pre- (breast stage 1), early (breast stage 2), mid- (breast stage 3–4) and late (breast stage 5). Puberty in boys was also classified into four stages according to testicular volume: pre- (testis 1–3 ml), early (testis 4–8 ml), mid- (testis 9–19 ml), and late (testis > 20 ml).
The characteristics of the group of children used to produce the model
are given in Table 1a
.
|
|
When a preliminary exploratory analysis was performed on the children in prepuberty, the distribution of height SD was found to be highly skewed, with a mean height of -1.45 SD, i.e. the majority of children were below average height. Because the aim was to construct a reference model for a normal population, a portion of IGF-I measurements from prepubertal children with a height SD between -1 SD and -3 SD had to be excluded from the model. One hundred seventy-one IGF-I values were therefore omitted in randomized manner. Thus, the cross-sectional data used for estimating and validating the reference model was obtained from 468 children (236 girls, 232 boys) from a cohort of 969 serum samples (Table 1).
The data used come from children involved in three different studies, here denoted groups A, B, and C
Group A. 1–18 yr of age (n = 176). These were siblings of children with short stature, relatives or friends of employees at our research center, or children seen in pediatric endocrine outpatient clinics and subsequently regarded as healthy, and as normal in terms of height and growth.
Group B. 8–13 yr of age (n = 80). These were children from one primary school who had participated in a study to determine the mean intraindividual monthly coefficient of variation of serum IGF-I concentrations (9).
Group C. 15–18 yr of age (n = 212). These were children from several schools in Göteborg who had participated in a follow-up assessment of a study involving iron supplementation in flour.
When data from several studies are combined, the results sometimes produce biased estimates. Therefore, tests were performed to uncover systematic differences between groups. No evidence of such deviations was found (data not shown).
Cross-sectional validation group. More than 50% of the 468 children whose samples were used in the reference model had more than one IGF-I measurement. For these children, one of their serum samples that was not used in the reference model was randomly selected (n = 268) and used to validate the reference model (Table 1b).
Longitudinal validation group. Factors that must be accounted for in a longitudinal reference model are intraindividual variation and dependence. For 153 children (79 boys and 74 girls) who had at least 2 IGF-I samples measured in the same pubertal stage, an exploratory analysis was performed to determine their change in logIGF-I over time. The mean changes for each combination of gender and pubertal stage were then compared with the regression coefficients given by the multiple regression (see Results—Estimated regression model). To determine the effect of using the cross-sectional reference model on longitudinal data, we used IGF-I data from a small group of children with measurements in at least three pubertal stages. The reference model was used with the assumptions that the intraindividual development was the same as the over-time development, and that observations from the same person can be regarded as approximately independent. This group consisted of 21 children: 10 girls and 11 boys with height and weight within ± 2 SD.
Residual analysis. Residual analysis was performed, the residuals were tested for normality (skewness), and exploratory analysis was used to check the assumption of homoscedastisity (i.e. constant variance of the residuals).
Hormone measurements. Concentrations of IGF-I were measured in duplicate by an IGF binding-protein blocked RIA, without extraction and in the presence of an approximately 250-fold excess of IGF-II (Mediagnost GmbH, Tübingen, Germany) (27). The intraassay coefficients of variations (CVs) were 8.1, 4.4, and 4.5% at concentrations of 55, 219, and 479 µg/liter, respectively, and the interassay CVs were 10.4, 7.7, and 5.3% at concentrations of 55, 219, and 479 µg/liter, respectively.
Auxology. Heights and weights were transformed into SD scores for sex and age according to Swedish reference values (26).
Statistical evaluation. The statistical package SPSS (SPSS, Inc., Chicago, IL) was used for the analysis.
When selecting the parameters to include in the reference model, the significance and the variables found to be biologically relevant in our exploratory study and in earlier studies were taken into account. Only variables that are significant (gender, age, and pubertal stage) are included in the final model.
Results
The analysis was performed after transformation of IGF-I values
(Fig. 1, left panel) to
logarithmic values (right panel). R2
measures the proportion of the variation in outcome that is explained
by the combination of the explanatory variables in the model. In
multiple regression analysis, an adjusted R2 is
used, with a correction for degrees of freedom. The adjusted
R2 was found to be 0.66 for the reference group.
An ANOVA was performed, and the resulting F test showed a
highly significant regression relationship between logIGF-I and at
least one of the explanatory variables (P < 0.001).
The estimated regression model for the normal reference group is shown
in Fig. 2 and Table 2.
|
|
|
for the prepubertal children reveals
that the relationship between their age and logIGF-I is positive, with
the regression coefficient b (bAge*Pre) =
0.0325 (P < 0.001). This means that older prepubertal
children had higher logIGF-I values compared with younger prepubertal
children. For girls above 8 yr of age, the relationship between age and
logIGF-I is somewhat more pronounced (bSex*Age*Pre (>8 yr)
Early = 0.0089; P < 0.001) than that
observed in boys of the same age. Also in early puberty the
relationship between age and logIGF-I is positive and again a bit more
pronounced in girls than in boys. In mid-puberty, the relationship
between age and logIGF-I is different for girls and boys. Virtually no
age-effect is observed in girls, whereas an age-effect is present in
boys. The logIGF-I values are higher for mid-pubertal girls, compared
with boys. Finally, in late puberty there is a negative relationship
between age and logIGF-I, somewhat more pronounced in girls than in
boys. Estimated regression model
The estimated multiple regression model in Table 2 can be used for
estimating the value of logIGF-I, given the age, sex, and puberty. In
the multiple regression model, the effects of all of the explanatory
variables included are taken into account simultaneously. To simplify
the comparison between gender and pubertal stage, the estimated
regression equations for each combination of gender and puberty are
presented below, using the appropriate coefficients from Table 2. Use
of these equations facilitates visual illustration of the reference
ranges for each gender and pubertal stage, as seen in Fig. 3.
|
Boys. Prepubertal 1–14 yr: logIGF-I = 1.8010 + 0.0325*Age; early 9–15.5 yr: logIGF-I = 1.8010 + 0.0401*Age; mid 10–18 yr: logIGF-I = 1.8010 + 0.0476*Age; late 13–19.5 yr: logIGF-I = 3.3140 - 0.0464*Age.
To calculate the SD score, the SE of the predicted value of logIGF-I must also be estimated. The SE varies with the values of the explanatory variables. In the reference group, the mean of the SE was 0.1373 and the range was 0.1366–0.1425. There was very little variation among the mean of the SE for the different combinations of gender and pubertal stage.
To illustrate how to use an equation: A 9-yr-old prepubertal girl has
an IGF-I measurement of 219 µg/liter. To calculate her SD
score, the following formula should be used.
 |
is calculated from the estimated
regression equations: that is = 1.8010 + (0.0325*9)
+ (0.0089*9) = 2.1736 and the mean SE value (0.1373) is
used. Validation of the model
Distribution of SD scores. The distribution of
SD scores for the reference group and the validation group
are compared in Fig. 4, top
and bottom. The graphs show that the reference model gives a
distribution of SD scores for the validation group that is
similar to that for the reference group.
|
. This shows that
the assumption of a constant variance for logIGF-I was fulfilled.
|
The exploratory analysis for the 153 children in the longitudinal validation group showed that the change in IGF-I with age was positive in prepuberty and early puberty among both boys and girls. In late puberty, the change in IGF-I with age was negative for boys. In mid- (for girls above 15 yr of age) and late puberty, the change in IGF-I with age was negative for girls. These observations agree with the results from the reference model. Taking the data from the 21 healthy subjects (10 girls and 11 boys) followed longitudinally and using it in our cross-sectional reference model showed mean range variation of 0.9 IGF-I SD scores with all logIGF-I SD scores within the range of ± 2 SD.
Discussion
We have developed a model that can be used in children for relating serum IGF-I levels to age, puberty, and gender simultaneously, thus enabling comparisons between different groups of children without using matched controls. These reference values should improve the accuracy of diagnosis for individual children. The model is the first that can be used to convert IGF-I serum concentrations into SD scores from infancy to adulthood with a high degree of accuracy.
Our results confirm previous studies (10, 11, 12, 13, 14, 15, 16) showing that the relationship between age and serum IGF-I levels is positive in prepubertal and early pubertal children and that the relationship between age and IGF-I concentrations is negative in late puberty. For children in mid-puberty, no reference values have been available. Juul et al. (28) did not report a prediction range for children in Tanner stage 3, and for girls in Tanner stage 2 (equivalent to our early stage of puberty) and in stage 4 no significance value is given for the prediction intervals. Our study shows that in mid-puberty the relationship between age and IGF-I was different for girls and boys, with a positive age effect in boys and a fairly constant effect in girls. The IGF-I values were higher for mid-pubertal girls, compared with boys. These gender differences agree with the observation that girls have peak height velocity and concurrent elevation of serum IGF-I earlier than boys (28). Some girls do not reach stage 5 of breast development. When this occurs in a girl who has reached final height and/or had menarche more than 2 yr previously, she should be classified as being in late puberty.
In this report, one regression model was fitted to all observations, instead of dividing the sample into subgroups and fitting separate regression models for each subgroup. The advantage of using one model is that it permits simultaneous estimation of all explanatory variables and uses all observations in the estimation process. The power of a test increases with the number of observations (the number of degrees of freedom). With an increased number of observations, it is easier to obtain significant results for the exploratory variables that are important.
The reference model presented in this report was constructed from cross-sectional, not longitudinal data. A longitudinal study would be subject to the variation between children and the variation within each child. In normal children, repeated measurements of IGF-I have been shown to cluster around a level dependent on the size of the child, although the mean intraindividual monthly CV for IGF-I was found to be 16% (9). Constructing a longitudinal reference model would require repeated blood sampling from the same healthy children over a prolonged interval, for as long as 20 yr. In this study, an exploratory analysis of longitudinal data was conducted on a small number of children for the purpose of comparing the intraindividual development with the development given by the cross-sectional reference model. This analysis showed that the within-individual changes in IGF-I follow the same pattern as that exhibited by the reference model. Preliminary results using the reference model showed that the longitudinal measurements for each of the normal healthy children varied considerably, but were within the normal range given by the model (±2 SD).
Based on the very good correlation we have seen when comparing results from different IGF-I assays, we believe that this model is also suitable for use with other IGF-I assays. Nevertheless, it is important for each individual laboratory to determine whether their assays give similar measured IGF-I levels as the Mediagnost assay or whether a conversion factor is required.
Our model shows that serum IGF-I concentrations are related to age throughout puberty. The model also accounts for the difference in IGF-I for boys and girls during puberty. The results of the study provide a tool to optimize the evaluation of IGF-I in the diagnosis of GH insufficiency and to follow up on GH treatment.
Acknowledgments
The authors wish to extend their appreciation to Lisbeth Larsson for her help with the IGF-I measurements, to Lena Rossander-Hulthén for providing us with the serum samples from healthy children in the iron study, and to the staff at Ward 334T.
Footnotes
The Swedish Medical Research Council (7509), The Swedish Foundation for Health Care Sciences and Allergy Research, The Medical Faculty at Göteborg University and Pharmacia Corp. supported the study.
Abbreviations: CV, Coefficient of variation.
Received November 2, 2000.
Accepted August 28, 2001.
References
This article has been cited by other articles:
 |
|
 |
|
  J. Leger, I. Mercat, C. Alberti, D. Chevenne, P. Armoogum, J. Tichet, and P. Czernichow The relationship between the GH/IGF-I axis and serum markers of bone turnover metabolism in healthy children Eur. J. Endocrinol., November 1, 2007; 157(5): 685 - 692. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P.-M. Holterhus, R. Odendahl, S. Oesingmann, R. Lepler, V. Wagner, O. Hiort, R. Holl, and the German/Austrian DPV Initiative and the German Classification of Distinct Baseline Insulin Infusion Patterns in Children and Adolescents With Type 1 Diabetes on Continuous Subcutaneous Insulin Infusion Therapy Diabetes Care, March 1, 2007; 30(3): 568 - 573. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. B. Loucks The Response of Luteinizing Hormone Pulsatility to 5 Days of Low Energy Availability Disappears by 14 Years of Gynecological Age J. Clin. Endocrinol. Metab., August 1, 2006; 91(8): 3158 - 3164. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Verkauskiene, D. Jaquet, S. Deghmoun, D. Chevenne, P. Czernichow, and C. Levy-Marchal Smallness for Gestational Age Is Associated with Persistent Change in Insulin-Like Growth Factor I (IGF-I) and the Ratio of IGF-I/IGF-Binding Protein-3 in Adulthood J. Clin. Endocrinol. Metab., October 1, 2005; 90(10): 5672 - 5676. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. H. Choe, S. Y. Song, K.-H. Paik, Y. J. Oh, S.-H. Chu, S. H. Yeo, E. K. Kwon, E. M. Kim, M. Y. Rha, and D.-K. Jin Increased Density of Ghrelin-Expressing Cells in the Gastric Fundus and Body in Prader-Willi Syndrome J. Clin. Endocrinol. Metab., September 1, 2005; 90(9): 5441 - 5445. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Tenhola, P. Halonen, J. Jaaskelainen, and R. Voutilainen Serum markers of GH and insulin action in 12-year-old children born small for gestational age Eur. J. Endocrinol., March 1, 2005; 152(3): 335 - 340. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Lofqvist, E. Andersson, L. Gelander, S. Rosberg, L. Hulthen, W. F. Blum, and K. A. Wikland Reference Values for Insulin-Like Growth Factor-Binding Protein-3 (IGFBP-3) and the Ratio of Insulin-Like Growth Factor-I to IGFBP-3 throughout Childhood and Adolescence J. Clin. Endocrinol. Metab., March 1, 2005; 90(3): 1420 - 1427. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Saukkonen, R. Amin, R. M. Williams, C. Fox, K. C. Yuen, M. A. White, A. M. Umpleby, C. L. Acerini, and D. B. Dunger Dose-Dependent Effects of Recombinant Human Insulin-Like Growth Factor (IGF)-I/IGF Binding Protein-3 Complex on Overnight Growth Hormone Secretion and Insulin Sensitivity in Type 1 Diabetes J. Clin. Endocrinol. Metab., September 1, 2004; 89(9): 4634 - 4641. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. B. Johnston, J. Dahlgren, J. Leger, L. Gelander, M. O. Savage, P. Czernichow, K. A. Wikland, and A. J. L. Clark Association between Insulin-Like Growth Factor I (IGF-I) Polymorphisms, Circulating IGF-I, and Pre- and Postnatal Growth in Two European Small for Gestational Age Populations J. Clin. Endocrinol. Metab., October 1, 2003; 88(10): 4805 - 4810. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. C. Bunt, A. D. Salbe, M. H. Tschop, A. DelParigi, P. Daychild, and P. A. Tataranni Cross-Sectional and Prospective Relationships of Fasting Plasma Ghrelin Concentrations with Anthropometric Measures in Pima Indian Children J. Clin. Endocrinol. Metab., August 1, 2003; 88(8): 3756 - 3761. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Stefan, J. C. Bunt, A. D. Salbe, T. Funahashi, Y. Matsuzawa, and P. A. Tataranni Plasma Adiponectin Concentrations in Children: Relationships with Obesity and Insulinemia J. Clin. Endocrinol. Metab., October 1, 2002; 87(10): 4652 - 4656. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
and girls, 
) whose heights and weights were
within ± 3 SD of Swedish norms (left
panel). The logarithmic transformation is shown in the
right panel, with the mean values given at 2-yr
intervals for boys (-
-) and girls (-•-).
-), late (-
-).
