Acta Pediatr Scad [Suppl]
370: 131- 137. 1990
Growth Response in the First Year of Growth Hormone Treatment in Prepubertal Children with Organic Growth Hormone Deficiency: a Comparison with Idiopathic Growth Hormone Deficiency D.A. PRICE, M.B. RANKE and 0. GUILBAUD on behalf of the Executive Scientific Committee of the Kabi International Growth Study From the Depurtmenr of Child Health. University of Manchester. Royal Manchesfer Children 's Hospital. Manchester. UK.the University Children 's Hospital, Tilbingen. Federal Republic of Germany and Kabi Peptide Hormones, Research and Development, Stockholm, Sweden.
ABSTRACT. Price, D.A., Ranke, M.B., Guilbaud, 0. (Department o f Child Health, University o f Manchester, Royal Manchester Children's Hospital, Manchester, UK. the University Children's Hospital, Tiibingen, West Germany and Kabi Peptide Hormones, Research and Development, Stockholm, Sweden). Growth response in the first year o f growth hormone treatment in prepubertal children with organic growth hormone deficiency: a comparison with idiopathic growth hormone deficiency. Acta Paediatr Sand [Suppl] 3 7 0 131-137, 1990. Patients in the Kabi International Growth Study (KIGS) up to 1st January 1990 who had organic growth hormone deficiency (OGHD) were identified. They accounted for 21% of all patients with growth hormone deficiency (GHD). Diagnostic categories within the OGHD group included septo-optic dysplasia, postnatal trauma, craniopharyngioma, other cranial tumours, and following acute leukaemia. Features at presentation and during the first year o f h G H treatment were compared with those o f children with idiopathic growth hormone deficiency (IGHD). Ninety prepubertal children with OGHD were selected for comparison with predicted values based on those observed in o f ohserved first-year height velocity (Hv) 257 children with IGHD. Those with septo-optic dysplasia, postnatal trauma and craniopharyniioma responded as predicted, whereas those with other cranial tumours appeared to grow less well than predicted. Glucocorticoid treatment did not affect response, hut previous cranial o r craniospinal irradiation was found to be associated with an observed H V which was significantly less than predicted. Key words: Orgunic gmwrh hormone deficiency, growfh hormone treatment, regression analysis, glucocofticoids, irradiation.
There is evidence that children with organic growth hormone deficiency (OGHD) respond less well in the short and long term to exogenous human growth hormone (hGH) compared with children who have idiopathic growth hormone deficiency (IGHD) ( I -4). Although there may have been differences in presenting Characteristics between OGHD and IGHD patients, such as age and height deficit, which might have intluenced response to treatment. the poor response in OGHD may equally well be due to features associated with the primary disease. In this paper. the term OGHD is used conventionally to describe deficiency of growth hormone ( G H ) other than IGHD. In the Kabi international Growth Study (KIGS), 21 % of cases o f G H D were of organic aetiology (p. 107). The characteristics of the children with OGHD and their response to hGH rherapy were compared with those of children with IGHD. KlGS is an international. multicentre study of the safety and efficacy of recombinant hGH: as such, it is an observational and not an experimental study and thus selection biases must be expected. However, its broad Sase may neutralize small biases and it does reflect clinical practice in many countries. PATIENTS AND METHODS Patient selection. The definition of GHD was based on the response to conventional provocation tests (other ihan the GHRH stimulation lest). If the GH response was less than 20 rnU/I. GHD was confirmed. The referring
physicians could then make a diagnosis from the following list: idiopathic aetiology, septo-optic dysplasia.
132
D.A. Price et al.
Acta Paediatr &and [Suppll 370
trauma (postnatal trauma, including postnatal infection), craniopharyngioma. other cranial tumours, and leukaemia (patients in remissioy after radiotherapy for acute lymphoblastic leukaemia). Target height SDS was calculated as the mean of the fathdr’s and mother’s height SDSs. The weight-for-height index was calculated as the weight expressed as a,percentage of the median weight for height. The height velocity (HV) in patients with OGHD during the first year of hGH treatment was compared with the predicted values derived from multiple regression analysis of prepubertal IGHD patients (pp. 122-30). The patients had to fulfil the same criteria for selection as those met by IGHD patients (p. 123): they had to be prepubertal, 2-10 years of age (for girls) or 2-12 years of age (for boys), and to have completed 1 year of treatment (actually 9-15 months). Furthermore, information on all of the following five ‘predictors’ had to be available: age and height at onset of treatment, target height SDS (mid-parental height SDS), hGH dose at onset of treatment (IU/kg) and frequency of injections. Srarisrical mcfhods. The estimated regression equation (a), derived by Ranke and Guilbaud (p. 127) from 257 prepubertal IGHD patients in the KIGS database, can be used to predict the HV (cdyear) during the first year of hGH therapy for a prepubertal OGHD patient who is behaving in the same way as the IGHD patients. Five predictors are used in regression equation (a), and so all must be available for the predicted HV to be calculated. The residual (i.e. actual - predicted) response can then be calculated for each OGHD patient. If a group of OGHD patients responds to treatment in the same way as the IGHD group, the OGHD residuals will be scattered around zero in a similar way to the original 257 IGHD residual values. More importantly. the way in which the OGHD residuals differ in their behaviour from the IGHD residuals gives an indication of the way in which the OGHD responses differ from the IGHD responses, controlling for the five predictors. For example, if most of the OGHD residuals are below zero, this indicates that the OGHD responses are smaller than the expected corresponding IGHD responses, given the same constellation of predictor values. One advantage with this approach is that even a very small OGHD group can be compared with the IGHD group. As is the case for the IGHD residuals, it is convenient to express the OGHD residuals in a common standardized way (SDS); that is, to consider Studentized residuals. Each Studentized residual is obtained by dividing the residual by its standard error (which is slightly larger than the error SD given in connection with regression equation (a)). The standard errors used in this analysis were based on the variability information from the IGHD group only (and not from the OGHD group itself), so it is more precise to say that externally Studentized residuals were considered in this OGHD group (5). Approximately 5% of such Studentized residuals were expected to be outside the limits - 2 and +2 for a comct regression model (assuming normal errors) if this OGHD group behaved like the IGHD group. The procedure REG in the program package SAS (Statistical Analysis Systems), mainframe version 5.18, was used for these calculations.
RESULTS Patient characteristics in KIGS database. By 1 January 1990. there were 2261 subjects with GHD available for analysis; 1785 (79%) of the children had IGHD and 476 (21 %) had OGHD. The numbers of children in each diagnostic category were: 5 1 (1 1%), septo-optic dysplasia; 62 (13%), trauma; 106 (22%), craniopharyngioma; 196 (41 %), other cranial tumours; 61 (13%), leukaemia. Fig. l(a-e) gives a descriptive comparison of OGHD patients with IGHD patients. The median age of onset of hGH treatment is earlier in children with septo-optic dysplasia, but later in those with leukaemia, when compared with IGHD. The median height SDS at onset of hGH treatment is greater in OGHD children with craniopharyngioma, other cranial tumours or leukaemia than in IGHD children (i.e. the height deficit in these groups is less, and the median target height SDS is normal) (Fig. lb). The target height is reduced in IGHD, septo-optic dysplasia and, to a lesser extent, trauma. Weight-for-height index is similar in OGHD and IGHD patients. The median weekly dose of hGH was 0.55 IU/kg in IGHD and 0.53 IU/kg in OGHD children. Subdividing the OGHD patients into diagnostic groups, the median weekly doses were 0.55,0.53,0.45,0.56 and 0.57 IU/kg for septo-optic dysplasia, trauma, craniopharyngioma, other cranial tumours and leukaemia, respectively. The frequency of injections was similar in IGHD and OGHD patients, with half receiving 1-3 injections per week and half receiving 4-7 injections per week. The HVs (mean f SD) during the first year of hGH treatment for prepubertal IGHD and
GH treatment of OGHD
Acta Paediatr Scand [Suppl] 370
- 5
%$:.=.
10
c c
(a)
mal a E
qj 5s
5
*I
a 0
=5
0
;i;E
0
;gE z
-l
IGHD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
IGHD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
''ObiD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
0
(b)
QT-
z2
S6 5 a
=:
8
-2 -3
u)
+0.5r
2 1
IGHD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
IGHD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
7 6 5 4
3 2 1
Fig. 1. Characteristics of children with OGHD and IGHD in the total KIGS database. Median age and height SDS at onset of hGH treatment, target height SDS, sex ratio and ratio of isolated hGH deficiency to multiple pituitary hormone deficiencies.
133
134
D.A. Price et al.
Acta Paediatr Scand (Suppl] 370
*
OGHD patients were 8.6 f 3.0 cm/year and 8.1 3.0 cm/year. respectively. For IGHD children without or on glucocorticoid treatment, the values were 8.5 f 2.5 cm/year and 9.6 f 2.8 cm/year, respectively. Equivalent values for OGHD children without or on glucocorticoids were 8.3 f 2.5 cm/year and 9.1 f 2.3 cm/year, respectively. There were no significant differences in any of these comparisons using two-sided 5 % level Wilcoxon rank tests. Observed versus predicted growth response. The observed first-year HVs of prepubertal OGHD children are shown in Fig. 2. For the total OGHD group ( n = 90). the predicted HV was subtracted from the observed H V (measured in cmlyear) for each individual. Each value was divided by the standard error (which is approximately equal to the error SD given in connection with regression equation (a)) to obtain the Studentized residual. As described above in the statistical methods section, if the OGHD group responded in the same way as the IGHD group (n = 257). these Studentized residuals would be scattered around zero. with about 5% outside the limits -2 and + 2 (assuming normal errors). The first plot in Fig. 3 shows these Studentized residuals (vertical axis) plotted against their corresponding predicted responses (horizontal axis). As can be seen. the vertical scatter seems to be centred about the zero line for all levels of predicted response. with a tendency to greater spread at certain levels of predicted response. Four observations are greater than + 2 and 7 are less than -2 (SDS). Thus. the responses in the OHGD group as a whole do not differ very much from what might be expected from the responses in the IGHD group, though there are indications of some differences. Some subgroups of OGHD patients were therefore considered in the same manner. The OGHD patients on glucocorticoid treatment (n = 15) behaved in a similar way to the IGHD group, and to those in the OGHD group not receiving glucocorticoids (Fig. 3). The children with a diagnosis of septo-optic dysplasia, trauma or craniopharyngioma (n = I I , 16 and 21. respectively) behaved in a similar manner to the IGHD group (Fig. 3).
25 20
c r
Male n = 49
I I
Female n = 41
i
0;
h 16
A
l b l b 1; Age (years)
16 18
io Age (years)
Fig. 2. The HVs during the first year of hCH treatment in prepubenal children with OGHD (n = 90)selected for comparison of growth response with prepubertal children with IGHD (n = 257). The dotted lines give the normal limits for H V , according to Tanner.
GH treatment of OGHD
Acta Paediatr Scand [Suppll 370
In patients with other cranial tumours (n = 38), there appeared to be a tendency for lower
HV responses (than predicted) at the higher predicted levels (Fig. 3). An approximate F test of whether the regression line differed significantly from the zero line gave a p value of 0.07. If all patients who had received cranial or craniospinal irradiation, regardless of diagnosis (n = 43), are selected for comparison, the regression line (n = 43) (Fig. 4) has a negative slope, indicating relatively poorer responses (than predicted) at the higher predicted response levels. An approximate F test of whether this line differs significantly from the zero line gives a p value of 0.04.
-
z p : i
(a) Organic- total
(b) OGHD- glucocorticoid treatment
3-
n = 90
w V
1
-9c
o
0,-1 i i -2
2 -4 0, u -5 3
rn -,
-
370: 131- 137. 1990
Growth Response in the First Year of Growth Hormone Treatment in Prepubertal Children with Organic Growth Hormone Deficiency: a Comparison with Idiopathic Growth Hormone Deficiency D.A. PRICE, M.B. RANKE and 0. GUILBAUD on behalf of the Executive Scientific Committee of the Kabi International Growth Study From the Depurtmenr of Child Health. University of Manchester. Royal Manchesfer Children 's Hospital. Manchester. UK.the University Children 's Hospital, Tilbingen. Federal Republic of Germany and Kabi Peptide Hormones, Research and Development, Stockholm, Sweden.
ABSTRACT. Price, D.A., Ranke, M.B., Guilbaud, 0. (Department o f Child Health, University o f Manchester, Royal Manchester Children's Hospital, Manchester, UK. the University Children's Hospital, Tiibingen, West Germany and Kabi Peptide Hormones, Research and Development, Stockholm, Sweden). Growth response in the first year o f growth hormone treatment in prepubertal children with organic growth hormone deficiency: a comparison with idiopathic growth hormone deficiency. Acta Paediatr Sand [Suppl] 3 7 0 131-137, 1990. Patients in the Kabi International Growth Study (KIGS) up to 1st January 1990 who had organic growth hormone deficiency (OGHD) were identified. They accounted for 21% of all patients with growth hormone deficiency (GHD). Diagnostic categories within the OGHD group included septo-optic dysplasia, postnatal trauma, craniopharyngioma, other cranial tumours, and following acute leukaemia. Features at presentation and during the first year o f h G H treatment were compared with those o f children with idiopathic growth hormone deficiency (IGHD). Ninety prepubertal children with OGHD were selected for comparison with predicted values based on those observed in o f ohserved first-year height velocity (Hv) 257 children with IGHD. Those with septo-optic dysplasia, postnatal trauma and craniopharyniioma responded as predicted, whereas those with other cranial tumours appeared to grow less well than predicted. Glucocorticoid treatment did not affect response, hut previous cranial o r craniospinal irradiation was found to be associated with an observed H V which was significantly less than predicted. Key words: Orgunic gmwrh hormone deficiency, growfh hormone treatment, regression analysis, glucocofticoids, irradiation.
There is evidence that children with organic growth hormone deficiency (OGHD) respond less well in the short and long term to exogenous human growth hormone (hGH) compared with children who have idiopathic growth hormone deficiency (IGHD) ( I -4). Although there may have been differences in presenting Characteristics between OGHD and IGHD patients, such as age and height deficit, which might have intluenced response to treatment. the poor response in OGHD may equally well be due to features associated with the primary disease. In this paper. the term OGHD is used conventionally to describe deficiency of growth hormone ( G H ) other than IGHD. In the Kabi international Growth Study (KIGS), 21 % of cases o f G H D were of organic aetiology (p. 107). The characteristics of the children with OGHD and their response to hGH rherapy were compared with those of children with IGHD. KlGS is an international. multicentre study of the safety and efficacy of recombinant hGH: as such, it is an observational and not an experimental study and thus selection biases must be expected. However, its broad Sase may neutralize small biases and it does reflect clinical practice in many countries. PATIENTS AND METHODS Patient selection. The definition of GHD was based on the response to conventional provocation tests (other ihan the GHRH stimulation lest). If the GH response was less than 20 rnU/I. GHD was confirmed. The referring
physicians could then make a diagnosis from the following list: idiopathic aetiology, septo-optic dysplasia.
132
D.A. Price et al.
Acta Paediatr &and [Suppll 370
trauma (postnatal trauma, including postnatal infection), craniopharyngioma. other cranial tumours, and leukaemia (patients in remissioy after radiotherapy for acute lymphoblastic leukaemia). Target height SDS was calculated as the mean of the fathdr’s and mother’s height SDSs. The weight-for-height index was calculated as the weight expressed as a,percentage of the median weight for height. The height velocity (HV) in patients with OGHD during the first year of hGH treatment was compared with the predicted values derived from multiple regression analysis of prepubertal IGHD patients (pp. 122-30). The patients had to fulfil the same criteria for selection as those met by IGHD patients (p. 123): they had to be prepubertal, 2-10 years of age (for girls) or 2-12 years of age (for boys), and to have completed 1 year of treatment (actually 9-15 months). Furthermore, information on all of the following five ‘predictors’ had to be available: age and height at onset of treatment, target height SDS (mid-parental height SDS), hGH dose at onset of treatment (IU/kg) and frequency of injections. Srarisrical mcfhods. The estimated regression equation (a), derived by Ranke and Guilbaud (p. 127) from 257 prepubertal IGHD patients in the KIGS database, can be used to predict the HV (cdyear) during the first year of hGH therapy for a prepubertal OGHD patient who is behaving in the same way as the IGHD patients. Five predictors are used in regression equation (a), and so all must be available for the predicted HV to be calculated. The residual (i.e. actual - predicted) response can then be calculated for each OGHD patient. If a group of OGHD patients responds to treatment in the same way as the IGHD group, the OGHD residuals will be scattered around zero in a similar way to the original 257 IGHD residual values. More importantly. the way in which the OGHD residuals differ in their behaviour from the IGHD residuals gives an indication of the way in which the OGHD responses differ from the IGHD responses, controlling for the five predictors. For example, if most of the OGHD residuals are below zero, this indicates that the OGHD responses are smaller than the expected corresponding IGHD responses, given the same constellation of predictor values. One advantage with this approach is that even a very small OGHD group can be compared with the IGHD group. As is the case for the IGHD residuals, it is convenient to express the OGHD residuals in a common standardized way (SDS); that is, to consider Studentized residuals. Each Studentized residual is obtained by dividing the residual by its standard error (which is slightly larger than the error SD given in connection with regression equation (a)). The standard errors used in this analysis were based on the variability information from the IGHD group only (and not from the OGHD group itself), so it is more precise to say that externally Studentized residuals were considered in this OGHD group (5). Approximately 5% of such Studentized residuals were expected to be outside the limits - 2 and +2 for a comct regression model (assuming normal errors) if this OGHD group behaved like the IGHD group. The procedure REG in the program package SAS (Statistical Analysis Systems), mainframe version 5.18, was used for these calculations.
RESULTS Patient characteristics in KIGS database. By 1 January 1990. there were 2261 subjects with GHD available for analysis; 1785 (79%) of the children had IGHD and 476 (21 %) had OGHD. The numbers of children in each diagnostic category were: 5 1 (1 1%), septo-optic dysplasia; 62 (13%), trauma; 106 (22%), craniopharyngioma; 196 (41 %), other cranial tumours; 61 (13%), leukaemia. Fig. l(a-e) gives a descriptive comparison of OGHD patients with IGHD patients. The median age of onset of hGH treatment is earlier in children with septo-optic dysplasia, but later in those with leukaemia, when compared with IGHD. The median height SDS at onset of hGH treatment is greater in OGHD children with craniopharyngioma, other cranial tumours or leukaemia than in IGHD children (i.e. the height deficit in these groups is less, and the median target height SDS is normal) (Fig. lb). The target height is reduced in IGHD, septo-optic dysplasia and, to a lesser extent, trauma. Weight-for-height index is similar in OGHD and IGHD patients. The median weekly dose of hGH was 0.55 IU/kg in IGHD and 0.53 IU/kg in OGHD children. Subdividing the OGHD patients into diagnostic groups, the median weekly doses were 0.55,0.53,0.45,0.56 and 0.57 IU/kg for septo-optic dysplasia, trauma, craniopharyngioma, other cranial tumours and leukaemia, respectively. The frequency of injections was similar in IGHD and OGHD patients, with half receiving 1-3 injections per week and half receiving 4-7 injections per week. The HVs (mean f SD) during the first year of hGH treatment for prepubertal IGHD and
GH treatment of OGHD
Acta Paediatr Scand [Suppl] 370
- 5
%$:.=.
10
c c
(a)
mal a E
qj 5s
5
*I
a 0
=5
0
;i;E
0
;gE z
-l
IGHD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
IGHD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
''ObiD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
0
(b)
QT-
z2
S6 5 a
=:
8
-2 -3
u)
+0.5r
2 1
IGHD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
IGHD
Septo- Trauma Cranio- CNS Leukaemia optic pharyn- tumour dysplasia gioma
7 6 5 4
3 2 1
Fig. 1. Characteristics of children with OGHD and IGHD in the total KIGS database. Median age and height SDS at onset of hGH treatment, target height SDS, sex ratio and ratio of isolated hGH deficiency to multiple pituitary hormone deficiencies.
133
134
D.A. Price et al.
Acta Paediatr Scand (Suppl] 370
*
OGHD patients were 8.6 f 3.0 cm/year and 8.1 3.0 cm/year. respectively. For IGHD children without or on glucocorticoid treatment, the values were 8.5 f 2.5 cm/year and 9.6 f 2.8 cm/year, respectively. Equivalent values for OGHD children without or on glucocorticoids were 8.3 f 2.5 cm/year and 9.1 f 2.3 cm/year, respectively. There were no significant differences in any of these comparisons using two-sided 5 % level Wilcoxon rank tests. Observed versus predicted growth response. The observed first-year HVs of prepubertal OGHD children are shown in Fig. 2. For the total OGHD group ( n = 90). the predicted HV was subtracted from the observed H V (measured in cmlyear) for each individual. Each value was divided by the standard error (which is approximately equal to the error SD given in connection with regression equation (a)) to obtain the Studentized residual. As described above in the statistical methods section, if the OGHD group responded in the same way as the IGHD group (n = 257). these Studentized residuals would be scattered around zero. with about 5% outside the limits -2 and + 2 (assuming normal errors). The first plot in Fig. 3 shows these Studentized residuals (vertical axis) plotted against their corresponding predicted responses (horizontal axis). As can be seen. the vertical scatter seems to be centred about the zero line for all levels of predicted response. with a tendency to greater spread at certain levels of predicted response. Four observations are greater than + 2 and 7 are less than -2 (SDS). Thus. the responses in the OHGD group as a whole do not differ very much from what might be expected from the responses in the IGHD group, though there are indications of some differences. Some subgroups of OGHD patients were therefore considered in the same manner. The OGHD patients on glucocorticoid treatment (n = 15) behaved in a similar way to the IGHD group, and to those in the OGHD group not receiving glucocorticoids (Fig. 3). The children with a diagnosis of septo-optic dysplasia, trauma or craniopharyngioma (n = I I , 16 and 21. respectively) behaved in a similar manner to the IGHD group (Fig. 3).
25 20
c r
Male n = 49
I I
Female n = 41
i
0;
h 16
A
l b l b 1; Age (years)
16 18
io Age (years)
Fig. 2. The HVs during the first year of hCH treatment in prepubenal children with OGHD (n = 90)selected for comparison of growth response with prepubertal children with IGHD (n = 257). The dotted lines give the normal limits for H V , according to Tanner.
GH treatment of OGHD
Acta Paediatr Scand [Suppll 370
In patients with other cranial tumours (n = 38), there appeared to be a tendency for lower
HV responses (than predicted) at the higher predicted levels (Fig. 3). An approximate F test of whether the regression line differed significantly from the zero line gave a p value of 0.07. If all patients who had received cranial or craniospinal irradiation, regardless of diagnosis (n = 43), are selected for comparison, the regression line (n = 43) (Fig. 4) has a negative slope, indicating relatively poorer responses (than predicted) at the higher predicted response levels. An approximate F test of whether this line differs significantly from the zero line gives a p value of 0.04.
-
z p : i
(a) Organic- total
(b) OGHD- glucocorticoid treatment
3-
n = 90
w V
1
-9c
o
0,-1 i i -2
2 -4 0, u -5 3
rn -,
-
