| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Original Article |
Sektion Pädiatrische Endokrinologie, Universitätsklinikum Tübingen, Eberhard Karls Universität (M.B.R.), D-72076 Tubingen, Germany; Pharmacia, Inc. (A.L., P.W.), S-11287 Stockholm, Sweden; Robert H. Vines Growth Research Center, Ray Williams Institute of Pediatric Endocrinology, Diabetes and Metabolism, Children’s Hospital (C.T.C.), Westmead, New South Wales 2145, Australia; Pediatric Growth Research Center, Department of Pediatrics, Queen Silvia Children’s Hospital (K.A.W.), Sahlgrenska Academy of Goteborg University, S-416 85 Goteborg, Sweden; Baystate Medical Center Children’s Hospital, Tufts University of Medicine (E.O.R.), Springfield, Massachusetts 01199-1001; and Department of Pediatrics, St. Mary’s Hospital (D.A.P.), M27 1HA Manchester, United Kingdom
Address all correspondence and requests for reprints to: Prof. M. B. Ranke, Sektion Pädiatrische Endokrinologie, Universitätsklinikum Tübingen, Eberhard Karls Universität, Hoppe-Seyler Strasse 1, D-72076 Tubingen, Germany. E-mail: michael.ranke{at}med.uni-tuebingen.de.
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The majority of children born SGA experience catch-up growth by 2 yr of age. In about 10%, however, catch-up growth does not occur. Without treatment, these children remain short and constitute some 20–25% of adults whose final height is below -2 SD scores (2). Although these individuals are not GH deficient, recent long-term studies have shown that treatment with recombinant human GH is successful in promoting catch-up growth (3).
Treatment with GH is effective in increasing height velocity in children with a variety of conditions resulting in short stature, although individual patient responsiveness varies. Prediction models have therefore been developed for children with idiopathic GH deficiency (4) and Turner syndrome (5) and for short children with a range of GH secretory capacities (6) as tools for optimizing GH therapy in individual patients. These prediction models enable physicians to calculate expected height velocities, to determine putative treatment modalities, to identify discrepancies between observed and predicted height velocities, and to provide the rationale for continuation or discontinuation of treatment. Most importantly, they enable a rational discussion between physicians and patients or parents based on a realistic expectation of the benefits of treatment.
The aim of the present study was to develop and validate a model with which to predict individual responsiveness to GH therapy of short children born SGA.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The patients included in this analysis were receiving recombinant human GH (Genotropin, Pharmacia Corporation) during follow-up in a large pharmacoepidemiological survey, the KIGS (Pharmacia International Growth Database), or had participated in clinical trials to evaluate the safety and efficacy of Genotropin in patients born SGA (3).
Diagnosis was made according to the KIGS etiology classification list: codes 3.1 (idiopathic short stature), 3.4 (intrauterine growth retardation with persisting short stature without stigma), and 3.5 (intrauterine growth retardation with persisting short stature with minor dysmorphic stigma) (7).
Additional inclusion criteria for both the patients in KIGS and those included from the clinical trials were a birth weight for gestational age below -1.28 SD score (approximately equal to the 10th percentile) and a gestational age of at least 30 wk. Furthermore, the maximum GH response to one to three GH stimulation tests had to be over 5 µg/liter to exclude patients with additional severe GH deficiency, and the patients had to be prepubertal (mean testes volume, 
3 ml; Tanner breast stage B1) at the onset of GH treatment and less than 12 and 10 yr of age at the end of the analyzed treatment period for boys and girls, respectively. Patients also had to be receiving 6 or 7 injections of GH per week. Those patients (accounting for 8% of the original cohort) who missed GH injections for a total of more than 14 d during the first year of treatment were not included in the analysis. Only 6 of these patients (1%) were excluded because of unscheduled breaks in treatment. These inclusion criteria resulted in an original cohort of 682 patients (448 from KIGS and 234 from clinical trials). Height measurements, recorded at intervals of 9–15 months, were used to calculate height velocity (centimeters per year).
Data were available for 613 patients (408 boys) treated for 1 yr. Of these, about 10% (n = 68; 42 boys) were randomly assigned to the validation group, as were about 10% (n = 43; 26 boys) of the 432 patients treated longitudinally for 2 yr. All patients for these validation groups were from the KIGS cohort.
Development of the prediction model
Growth responses (annualized height velocities) during the initial 2 yr of GH therapy were correlated, by multiple regression analysis, with potentially relevant variables. These variables are reported as the median and range as well as the mean ± SD.
The variables tested were 1) status at birth: sex, weight SD score, length SD score, ponderal index, mode of delivery, and Apgar score; 2) genetic background: height SD score of the mother, height SD score of the father, and midparental height (MPH) SD score; 3) treatment modality: GH dose [per kilogram of body weight and per kilogram of ideal body weight (weight for height)], frequency of GH injections, and accumulated years of GH treatment; 4) patient variables at the beginning of the treatment period: age, bone age, height SD score, weight SD score, height SD score minus MPH SD score, and the peak serum GH concentration during stimulation testing. SD scores were calculated as follows: SD score = (patient value - mean value for age- and sex-matched normal subjects) ÷ SD of the value for age- and sex-matched normal subjects. Predictive growth models based on the above variables were derived from the analysis for each of the initial 2 yr of therapy.
To be consistent with previous similar analyses (4, 5), the height standards used for normal children were those of Tanner et al. (8), and the weight standards were those of Freeman et al. (9). Birth weight for gestational age was transformed to an SD score based on the standards of Niklasson et al. (10). The MPH SD score was calculated as: (father’s height SD score + mother’s height SD score) ÷ 1.61 (8, 11). Bone ages, calculated according to the method of Greulich and Pyle (12), were determined by the treating physician.
Statistical analysis
The prediction models were developed by means of multiple linear regression analysis fitted by least squares and the REG procedure in the SAS computer program (version 6.12, SAS Institute, Inc., Cary, NC) A hierarchy of predictive factors was derived by the all-possible regression approach, using Mallow’s C(p) criterion for ordering predictive factors, as described previously (13, 14). Differences between observed and predicted height velocities were expressed in terms of Studentized residuals. The residual is calculated as the observed height velocity minus the predicted height velocity for each observation, and the Studentized residual is the residual divided by its SE.
Validation of the model
From the group of patients in KIGS originally identified for inclusion in the study, approximately 10% were randomly assigned to a validation group and were not used to construct the prediction model. The actual growth responses over 1 and 2 yr of GH treatment in this validation group were then compared with the growth responses predicted from the model.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The characteristics at the start of GH treatment for the 613 children treated for 1 yr are given in Table 1
. The corresponding data for the 385 children who were treated longitudinally for 2 yr are shown in Table 2. The children were typical for short children born SGA and started GH treatment at a mean age of 6.6 yr after failing to achieve spontaneous catch-up growth (all were 
2 SD below the mean for height). The mean maximum GH peak during a stimulation test exceeded 9 µg/liter, indicating that these children were not GH deficient.
|
|
Prediction models
The variables found by multiple linear regression analysis to be predictive of the growth response over 1 and 2 yr are given in Tables 3 and 4. These also give the rank order of importance of the variables as predictors, the overall correlation coefficients of the prediction models (R2), the contribution of each variable to R2 (partial R2), and the error SD of the prediction in centimeters. Two models have been constructed for the second year growth response (Table 4). Model A is based on the same four predictors as the first year model, whereas model B is a three-parameter model that includes height velocity in the previous year of treatment, age at the start of treatment, and GH dose. All single predictors were significant (P < 0.0001).
|
|
) is as follows: PHV (cm/yr) = 8.0 + [-0.31 x age at start (years)] + [0.30 x weight SD score at start] + [56.51 x GH dose (mg/kg·d)] + [0.11 x MPH SD score] ± 1.3.
Using this simple four-parameter model, 52% of the variability of the growth response could be explained, with an error SD of 1.3 cm. The dose of GH was the most important predictor of the four identified in the first year model, accounting for 35% of the variability, followed by age (11%), weight SD score (5%), and MPH SD score (1%). Height SD score was not included in the model because it was highly correlated with weight SD score (R2 = 0.93; P < 0.0001) and was of less predictive value. The GH dose, weight SD score, and MPH SD score were positively correlated, and age was negatively correlated with the response to treatment. Thus, the greatest first year response to treatment occurs in younger children on higher doses of GH. The positive linear correlation between the GH dose and height velocity is shown in Fig. 1.
|
Validation of prediction models
The children used to validate the prediction models were not used in model development, but were taken at random from the original KIGS cohort. The demographic characteristics of the model and validation groups were therefore similar (Tables 5 and 6). There were no statistically significant differences between the predicted and observed growth responses for the validation groups in either the first or second year models. This was demonstrated by the fact that the Studentized residual values are not significantly different from zero (Tables 5 and 6).
|
|
. Studentized residual plots are used to identify outliers, nonlinearity, and nonconstant error variance in prediction models and are a part of their mathematical validation. The random distribution indicates that there is no heterogeneity in the group with respect to the relative importance of the different predictors. Importantly, the corresponding plot of the KIGS cohort used to validate the model shows a similar random distribution of values (Fig. 2B). The same concordance between the original model-generation cohorts and validation groups was found for the two second year growth response models (Fig. 3, A–D).
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
). Thus, 52% of the variation in the first year response was explained by the present model. At the same time, the error SD (degree of accuracy) was only 1.3 cm. This level of predictive accuracy was based on the readily available variables of dose of GH, age, weight SD score, and MPH SD score, making the model suitable for use in clinical practice. In addition to providing an accurate prediction of treatment outcome, by including the GH dose as a variable, the model has an important role in individualizing treatment to achieve the required target height.
|
). In the second year models, although the R2 values (predictive power) were lower than in the first year model, explaining between 30–35% of the variation in response, the degrees of accuracy were higher. This may indicate a stabilization of the growth response after initial catch-up growth in a group of heterogeneous patients born SGA due to various causes. Similar trends in R2 and error SD values after the first year of treatment have also been found in previous prediction models for children with idiopathic GHD (4) and girls with Turner syndrome (5). As with these two previous studies, height velocity during the first year of treatment in the present model was the most important predictor of subsequent growth, suggesting that the final height outcome may be indicated by the initial response to GH.
In the only previous prediction model for children born SGA (n = 135) from KIGS with different inclusion criteria, several variables were used that did not feature in the present model (15). These included birth weight SD score, number of GH injections per week, and target height SD score minus height SD score, contributing to a predictive power of 23% and an error SD of 1.6 cm. In the present model, birth weight SD score was not found to be a predictive variable, nor was target height SD score minus height SD score. With regard to birth weight SD score, this is probably because one of the inclusion criteria in the present study was a gestational age at birth of at least 30 wk, thereby reducing the variability of birth weights and excluding cases with less certain gestational ages. The frequency of injections is also no longer predictive, as GH treatment is now standardized at six or seven injections per week. MPH SD score and weight SD score in the present model mirror, to a certain extent, target height SD score minus height SD score.
In the present study the prediction model was validated using random samples of patients from KIGS. The lack of a significant difference between the predicted and actual growth responses in these groups supports the validity of the prediction equations.
It is clear that simple, robust, and accurate prediction models, which enable treatment to be tailored to an individual patient’s requirements, will become increasingly important in the era of evidence-based medicine. The present study shows how the dose of GH can be calculated and adjusted to obtain the optimum balance between efficacy and cost of treatment. This may have implications beyond the normalization of stature, as intrauterine growth retardation is thought to produce a permanent resetting of normal development (16) and is possibly predictive of a range of conditions in later life, including hypertension, coronary heart disease, stroke, and type 2 diabetes (17, 18). It is currently not known whether GH treatment, whether effective or not in terms of growth promotion, will ameliorate any of the long-term sequelae of intrauterine growth retardation.
In conclusion, we have developed an accurate model that can be used in normal clinical practice to predict the response to GH treatment in individual short patients born SGA. Such a model could provide the basis for a rational discussion between the treating physician and the patient and/or guardian concerning the expectation of treatment. In addition, it would alert physicians to differences between predicted and expected outcomes and may help to identify compliance problems or other underlying causes of treatment failure. Most importantly, this model will assist physicians in tailoring GH therapy to individual patients.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received July 1, 2002.
Accepted September 23, 2002.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  B. Raz, M. Janner, V. Petkovic, D. Lochmatter, A. Eble, M. T. Dattani, P. C. Hindmarsh, C. E. Fluck, and P. E. Mullis Influence of Growth Hormone (GH) Receptor Deletion of Exon 3 and Full-Length Isoforms on GH Response and Final Height in Patients with Severe GH Deficiency J. Clin. Endocrinol. Metab., March 1, 2008; 93(3): 974 - 980. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. A. J. de Ridder, T. Stijnen, and A. C. S. Hokken-Koelega Prediction Model for Adult Height of Small for Gestational Age Children at the Start of Growth Hormone Treatment J. Clin. Endocrinol. Metab., February 1, 2008; 93(2): 477 - 483. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Bakker, J. Frane, H. Anhalt, B. Lippe, and R. G. Rosenfeld Height Velocity Targets from the National Cooperative Growth Study for First-Year Growth Hormone Responses in Short Children J. Clin. Endocrinol. Metab., February 1, 2008; 93(2): 352 - 357. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. G Rosenfeld Pharmacogenomics and pharmacoproteomics in the evaluation and management of short stature Eur. J. Endocrinol., August 1, 2007; 157(suppl_1): S27 - S31. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Cohen, A. D. Rogol, C. P. Howard, G. M. Bright, A.-M. Kappelgaard, R. G. Rosenfeld, and on behalf of the American Norditropin Study Group Insulin Growth Factor-Based Dosing of Growth Hormone Therapy in Children: A Randomized, Controlled Study J. Clin. Endocrinol. Metab., July 1, 2007; 92(7): 2480 - 2486. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Saenger, P. Czernichow, I. Hughes, and E. O. Reiter Small for Gestational Age: Short Stature and Beyond Endocr. Rev., April 1, 2007; 28(2): 219 - 251. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Endocrinology and diabetes Arch. Dis. Child., April 1, 2007; 92(suppl_1): A13 - A15. [Full Text] [PDF]
|
|
|
|
|
|
A. Carrascosa, C. Esteban, R. Espadero, M. Fernandez-Cancio, P. Andaluz, M. Clemente, L. Audi, H. Wollmann, L. Fryklund, L. Parodi, et al. The d3/fl-Growth Hormone (GH) Receptor Polymorphism Does Not Influence the Effect of GH Treatment (66 {micro}g/kg per Day) or the Spontaneous Growth in Short Non-GH-Deficient Small-for-Gestational-Age Children: Results from a Two-Year Controlled Prospective Study in 170 Spanish Patients J. Clin. Endocrinol. Metab., September 1, 2006; 91(9): 3281 - 3286. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. J.J. Finken, F. W. Dekker, F. de Zegher, J. M. Wit, and for the Dutch Project on Preterm and Small-for-Ges Long-term Height Gain of Prematurely Born Children With Neonatal Growth Restraint: Parallellism With the Growth Pattern of Short Children Born Small for Gestational Age Pediatrics, August 1, 2006; 118(2): 640 - 643. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Binder, F. Baur, R. Schweizer, and M. B. Ranke The d3-Growth Hormone (GH) Receptor Polymorphism Is Associated with Increased Responsiveness to GH in Turner Syndrome and Short Small-for-Gestational-Age Children J. Clin. Endocrinol. Metab., February 1, 2006; 91(2): 659 - 664. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. B. Ranke, A. Lindberg, K. Albertsson-Wikland, P. Wilton, D. A. Price, E. O. Reiter, and on behalf of the KIGS International Board Increased Response, But Lower Responsiveness, to Growth Hormone (GH) in Very Young Children (Aged 0-3 Years) with Idiopathic GH Deficiency: Analysis of Data from KIGS J. Clin. Endocrinol. Metab., April 1, 2005; 90(4): 1966 - 1971. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Brodsky and H. Christou Current Concepts in Intrauterine Growth Restriction J Intensive Care Med, November 1, 2004; 19(6): 307 - 319. [Abstract] [PDF]
|
|
|
|
|
|
L B Johnston and M O Savage Should recombinant human growth hormone therapy be used in short small for gestational age children? Arch. Dis. Child., August 1, 2004; 89(8): 740 - 744. [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 |
