AcIu

Padiutr Scatid [Suppl] 370: 122-130, 1990

Growth Response in Prepubertal Children with Idiopathic Growth Hormone Deficiency during the First Year of Treatment with Human Growth Hormone. Analysis of the Kabi International Growth Study. M.B. RANKE and 0. GUILBAUD on behalf of the Executive Scientific Committee of the Kabi International Growth Study From the University Children 's Hospital. Tiibingen. FRG ond the Srurisrics Department, Kubi Peptide Hortnones. Stockholtn. Sweden

ABSTRACT. Ranke, M.B. and Guilbaud, 0. on behalf o f the Executive Scientific Committee o f the Kabi international Growth Study (University Children's Hospital, Tiibingen. FRG and the Statistics Department, Kahi Peptide Hormones, Stockholm, Sweden). Growth response in prepubertal children with idiopathic growth hormone deficiency during the first year o f treatment with human growth hormone. Analysis o f the Kahi International Growth Study. Acta Paediatr Scand [Suppl] 370; 122-130. 1990.

In order that children w i t h growth hormone deficiency (GHD) reach the goal o f normal adult stature, treatment modalities need to be optimized. From the large database o f patients enrolled in the Kahi International Growth Study (KIGS), 257 prepuhertal patients with idiopathic G H D undergoing their first year o f growth hormone (GH) substitution therapy were selected. A multiple regression analysis was performed to determine both auxiological factors characterizing the patients and the factors related to the chosen treatment modalities which are of significance for the observed magnitude o f the growth response. Due to the structure o f the data, pretreatment height velocity and bone age-derived auxiological data were not considered. I t was observed that the magnitude o f the growth response was inversely correlated with chronological age and relative height (HT SDS) at the start o f G H treatment but was positively correlated with mid-parental height. The growth response was also positively correlated with the G H dose (IUlkglweek) and the frequency o f G H injections per week. A regression equation using these five parameters was derived, allowing the growth response o f these patients to be predicted. The extension of this analytical approach in the future will allow the treatment o f patients with GHD to he tailored to individual requirements. Key words: Idiopathic growth hormone deficiency, growth hormone therapy, growth response prediction.

Since the first report of successful treatment of patients with human pituitary growth hormone (GH) more than 30 years ago ( I ) , there have been many papers on the treatment of growth hormone deficiency (GHD) (for reviews, see references 2, 3). Although most authors have studied growth only during the first year of treatment of GHD, there have been several publications on the effect of treatment on adult height in these patients (4).The results of these analyses show that only about 50% of children with GHD reach an adult height within the normal range. This result may be due to the diagnostic and therapeutic inadequacies of former years, not least the scarcity of pituitary hGH. Treatment regimens should be optimized so that normal height becomes the rule for a child with GHD. A study such as the Kabi International Growth Study (KIGS) is important because it allows analysis of the pattern of growth during the total phase of development to be based on a large database and on homogeneous groups of individuals. The analysis of the growth response to hGH during the first year of treatment in patients with idiopathic G H D (IGHD) is a first step in this complex evaluation. PATIENTS AND METHODS The analysis was based on 257 prepubertal patients (178 boys, 79 girls) with IGHD. The diagnosis was primarily established by the physicians contributing to the KIGS database on the basis of conventional GH

GH

Acta Paediatr Scand [Suppl] 370

treatment

of IGHD

stimulatory testing. with all patients in the present analysis showing GH levels of less than 10 pg/l (20 mUII). Patients below the age of 2 years were excluded, because it was thought that such an early presentation was probably indicative of forms of GHD which were not representative of the total population with IGHD. An upper age limit was set (12 years for boys and 10 years for girls) to allow direct comparison with normal growth standards for prepubertal children ( 5 , 6). Puberty was defined on the basis of clinical features. Boys were assumed to have entered pubeny, and hence were excluded from the study. if testicular size exceeded 4.0 ml (7). Girls were excluded if breast development had reached Tanner's stage 2 or more ( 5 ) . Patients were included in the analysis only if all of the following parameters - in addition to age and height a1 start of treatment and patient's sex - were available: weight, mid-parental height (MPH. calculated as the mean of the mother's and father's height), gender mean target height of the patient (MPH, expressed as specific SDS. see below). total GH dose per week (calculated per kg body weight or per m' body surface area). number of GH injections per week, and type of hypothalamic/pituitary deficit (isolated GHD versus multiple pituitary hormone deficiency). After these restrictions were imposed. the number of cases included i n the analysis was 257, 25% of the initial 1042 patients with prepubertal idiopathic GHD in the KlGS database. Bone ages were not used in the analysis because of the assumed inaccuracy resulting from multicentre evaluation. Height velocity IHV) was calculated if two consecutive height measurements were available for a minimum of 9 months and a maximum of 15 months apart. However. an observation phase of 9-15 months before the initiation of treatment was not always available. so pretreatment HV was not included in the present analysis. Height parameters and HV were analysed after transformation into SDS (SDS = ( x - i i )/SD. where x is the individual measurement. ii the mean at the relevant chronological age. and SD is the standard deviation a! that age). The weight-for-height index was calculated as the patient's weight expressed as a percentage of the median weight of the normal population for individuals of the same height as the patient). Three variables were used to describe growth response: HV expressed in crnlyear; HV SDS for chronological age (based on the chronological age mid-way between the ages at the two height measurements); change i n height SDS for chronological age (A HT SDS for CA; i.e. HT SDS for CA after 1 year on GH treatnien! minub HT SDS for CA at start of treatment).

STATISTICAL ANALYSIS A regression analysis was performed. based on the two response variables HV and A HT SDS, and the available parameters. which could be potential predictors for response. For each response variable, this analysis consisted of two steps: selection of predictors that are most important, and detailed analysis using the selected predictors. The selection step was based on the 'all possible regression approach'. using Mallow's C(PJ criterion to choose among the various subsets of predictors, as described by Weisberg (8). The second, more detailed. pan of the analysis included the calculation of Studentized residuals (the term residual refers to the differencc between the actual response and the response predicted by the estimated regression equation. while the corresponding Studentized residual is the residual divided by its standard error). Approximately 5 % of such Studentized residuals are expected to be outside the limits 2 for a correct regression model (assuming normal errors). The procedures RSQUARE and REG in the program package SAS (Statistical Analysis System). mainframe version 5.18. were used for the selection and regression analysis steps, respectively.

*

RESULTS Some characteristics of the group of patients investigated are given in Table 1 . The relative deviation in height in relation to chronological age (HT SDS for CA) at the start of treatment is given in Fig. 1 . It shows that there is not much association between age and the magnitude ofgrowth deviation. (However. this is probably a selection bias, since the patients are brought to the attention of a physician only when shortness has reached a certain degree.) The response to treatment with hGH during the first year. expressed in terms of HV (in cmiyear). and HV SDS or increment in height (A H T SDS) is given in Table 1 and in Fig. 2. The illustrations show that there are different trends with age, depending on which variable is used to express the response to treatment. The result of the regression analysis is given in qualitative terms in Table 2 . It can be seen that the five important predictors are: age and height SDS at start of treatment, mid-parental height SDS (i.e. target height SDS),total weekly GH dose. and frequency of GH injections. While the growth response is negatively correlated with chronological age and HT SDS, it

123

124

M . B . Ranke et al.

Acta Paediatr Scand [Suppl] 370 -

Table 1. Characteristics of the group of patients investigated. Number of patients (n)

Characteristic

251 119 251 257 251 251 251 251 251 257 251

Weight at birth (g) Length at birth (cm) Target height (SDS) Age at start of hGH treatment (years) Height at start o f hGH treatment (cm) Weight-for-height index Dose of hGH (IU/kg/week) Dose of hGH (IU/m*/week) Height velocity (cm/year) Height velocity (SDS) Height increment (SDS)

Median

3070 49.0 -0.8 6.5 101

16.4 0.6 14.3 8.1 2.5 0.6

10th centile

9 0 1h centile

2302 45.0 -2.1 3.3 82.9 11.6 0.4 10.3 5.9 0.2 0.2

3850 52.0 0.8 10.7 124 23.8 0.9 21.8 12.3 6.8 I .4

is positively correlated with the remaining three parameters. The individual relationship between HV and each of the five important predictors is illustrated in Figs. 2(a) and 3(a-d). The 'all possible regression approach' led to the selection of the following five predictors: target height SDS, chronological age, injection frequency, height SDS and total weekly GH dose per kg. Thus, this subset of five predictors was good for each of the two response variables considered (i.e. H V and A height SDS). For each of the two response variables, the best single predictor was target height SDS. Table 3 shows the r2 values associated with the best I-, 2-, 3-, 4- and 5-predictor(s) subsets of the five predictors selected, as well as the r2 value when all of the potential predictors were considered. The selected subset of five predictors is seen to explain most of the variability that can be explained by all potential predictors considered.

.

I

-*t -1

:

.

--,-'--.-

.

I -6 -7

t

. .

_

. _ .. . . . . . . . ,

... ..

.

. .. . .. . . . ... , .. .... .. ., .. . . " - . . . . ... ..

t

Fig. 1. Height SDS at start of hGH treatment in relation to chronological age of patients; n = 257.

GH treatment of IGHD

Acta Paediatr Scand [Suppl] 370

Female n = 79

2 SD

i 4 ;; ;lb 1;

\

:1

1's

lk Age (years)

Age (years)

-1

1

1

1

1

1

1

1

1

1

1

1

1

1

'

0 1 2 3 4 5 6 7 8 91011 121314

Age (years) Fig 2. Response of prepubertal children with IGHD to 1 year's treatment with hGH (n = 257). (a) HV (cm/year) in relation to the age-related normal standards; (b) Change in height SDS during I year of hGH therapy. Dotted line in (b) is the calculated regression line.

The estimated regression equation based on the five selected predictors in this analysis of 257 isolated GHD patients are as follows: = 6.12 - 0.19 x CA (years) - 0.54 X HT SDS (a) HV (cm/year) 1.58 X dose hGH (IU/kg/week) 0.43 X injection frequency per week + 0.82 x target height SDS

+ +

(r2 = 0.48; error SD = 1.84 (cm/year)) = 0.00 - 0.04 X CA (years) - 0.17 X HT SDS (b) A height SDS + 0.32 X dose hGH (IU/kg/week) + 0.08 X injection frequency per week + 0.15 X target height SDS (r2 = 0.51; error SD = 0.35 (SDS))

For each subject, the residual (i.e. actual - predicted) response can be calculated using the appropriate estimated regression equation, (a) or (b). The corresponding Studentized

125

M. B. Ranke et ul.

I26

Acta Paediatr Scand [Suppll 370

Table 2 . Analysis of rhe relationship between gro\c,th response and certain independent tariables in prepubertal children with ICHD. Response variables I. Height velocity ( i n cmlyear) Height velocity (SDS for C A ) Change in height ( A H T SDS for C A ) 11. Independent variables

.4urologx

+

Age (years) Bone age (years) Sex (fema1e:male) Height (cm) Height SDS Height velocity before hGH treatment (cm/year) Weight (kg) Weight-for-height index Target height SDS Target height (cm)

NA -

+ NA NA NA

+

-

Hormonul dejicir isolared GHD versus mulriple piruirury hornioire rlrficietic\ liGH rherupy

+

Dose (per kg)

+

Frequency of injections (per week)

+

-

-

= selected predictor in final regression model. = considered but not selected/predictor.

N A = data not available or not considered i n present regrebsion analysis

1 1

--

2-

F l y . .? Relationship between H V and indcpcndent variahleb. durinz I year'.; hGH therapy i n prepubenal children with I G H D . (a) H V in relation t o height SDS i i t stan of hGH therapy. n = 257. r? = 0.014; fb\ H V in relation to large[ height SDS, n = 357. r l = 0.258: ( c l H V in relation to hGH dose per week. n = 257. r' = 0 140: (d) H V in relation t o the frequencv of h G H inlectlons per week. n = 257. r' = 0.196.

GH treatment of IGHD

Acta Paediatr Scand [Suppll 370

Table 3. r for the best I - , 2-, 3-, 4-, and S-predictor(s) subsets and for all nine potential predictors considered (see Table 2). Response

Predictors

rz

Height velocity (cmlyear)

Target height SDS Target height SDS + CA Target height SDS + CA + Inj. Target height SDS + C A + Inj. Target height SDS CA Inj. All predictors (see Table 2)

0.258 0.345

+

+

Target height SDS Target height SDS + HT SDS Target height SDS HT SDS Target height SDS + HT SDS Target height SDS HT SDS All predictors (see Table 2)

A height SDS

+ +

0.406

+ HT SDS

0.465 0.478 0.497

+ HT SDS + doselkg

0.213 0.342 0.447 0.494 0.508 0.521

+ Inj.

+ Inj. + CA

+ Inj. + C A + doselkg

CA, chronological age.

lnj., injection frequency.

residuals are obtained by dividing each residual by its standard error (which is approximately equal to the error SD given for each equation). Fig. 4 gives a plot of the Studentized residuals versus the predicted response for HV (cdyear). As can be seen, most of the data cluster around the zero-line (which corresponds to actual = predicted values) and stay within f2, as they should for a correct regression model. The outlier with negative residual was retained in the analysis because regression diagnostics indicated that it did not have much influence on the results and because no evident error could be found in the predictors and response variable.

4r

:cn:

3-

$

2

E

. . .. .. . . . . . .. . ...- ..... . . . . . . .. . . . . . .. .. .. . ., . . . , . - . . . . .. . .. ._ .. .. .;... . ... .. ......... .. .. . .. . . .. . .-........ . . . . . . -. . . ....... . . . . .. . . . ..:.. . . ...... ... . * . .

1-

!?

(1, c

.D r o D

-

-1-

D

-2

(1,

.-N C W

3

G

-3

..

. .

E (1,

t

-41 4

1

5

I

6

I

7

I

8

I

9

I

10

I

11

I

12

I

13

Predicted height velocity (cmlyear)

Fig. 4. Relationship between predicted HV and Studentized residuals of HV (SDS)during first-year growth in children with GHD (n = 257) (see results).

121

M.B. Ranke et a/.

j28

Acta Paediatr Scand [Suppl] 370

Table 4. Predictors of first-year growth response to hGH in patients with GHD (18). Author

Year GH dose

7 anner

1971 1976 1981 1980 1986 WII 1988 1988 eundak 1988 \herman %IGS 1989 I his study 1990

Preece I-rasier icol land Wlt

-

= =

GH injections

Chronological Bone Height age age

Weight

HV Skinfold Parental before thickness height hGH

+ +

+ +

+

+

+ + +

positive correlation with response. negative correlation with response

J>ISCUSSION k v e r a l authors have tried to determine those factors which are correlated with, and are thus Imdictive for, the response to hGH in patients with GHD (9-12). As a consequence of such attempts, regression equations have also been derived which predict the growth response from xveral independent variables. Table 4 lists the factors which have been suggested to be (,orrelated with hGH response in GHD patients. Two groups of factors appear to determine the extent of the response to hGH. One group comprises auxological characteristics of a patient at the start of GH treatment. In general, it seems that a relatively younger patient with il relatively larger delay in bone age, a greater reduction in height for a given age and a more profound reduction in basal HV is likely to achieve a better response to a chosen hGH regimen. These four characteristics are auxological expressions of the degree of GH ,w,ufficiency (9). This means that the patient with the most severe GHD will respond best h J hGH. The predictive power of bone age and HV before the start of hGH treatment (13, 14) could not be analysed in such a study. The demonstration that target height was a predictor (in fact, the most important predictor) f l J f the response to hGH is novel. Sherman et al. who derived a completely different prcdiction model for the response to hGH, had already recognized a minor influence of f1r;rtcrnal height on the response to hGH (15). It is noteworthy that MPH also has predictive p)wer, independent of that of height at start of hGH, although those two predictors are po!,itively correlated. This observation lends support to the speculation that normal variability I I I height may not only be the result of variation in GH secretion (16) but also to an inherited ./;triability in responsiveness to hGH. I t is not known whether the introduction of other auxological components into a prediction 111odel would have decreased the variability in the observed response to hGH in patients with c;liD during their first year on hGH treatment. Certainly, the phenomenon of catch-up ztowth (17), which during first-year treatment of GHD comprises an element of responsiveI I C ~ Sto hGH, is responsible for this mathematically unexplained variation. One group of factors influencing the magnitude of the response to hGH relates to the tlcatment regimen. Both the total dose of hGH and the number of injections given influence ttic biological availability of hGH and, thus, the effect of hGH treatment on growth. Increasing the injection frequency from three to seven injections per week (without increasing itlr total weekly dose) has a tremendous effect on growth which can be achieved only by doubling the total dose in a three injections per week treatment regimen (Fig. 5 ) . It is illerefore, evident that daily injections of hGH must be advocated.

GH treatment of IGHD

Acta Paediatr Scand [Suppl] 370

0) .-

a I

cn .a

6-

I

6-

5-

5-

4-

4-

3

I

I

I

I

I

I

I

I

3.

'

1

1

'

1

1

I

I

Fig. 5. HV in relation to total weekly hGH dose (IUlkg) in relation to age if total dose is divided into (a) 7 and (b) 3 injections per week.

There are some very practical consequences of this analysis. One is related to the monitoring of patients with IGHD. Deviations from the general pattern of growth during hGH treatment will motivate the investigator to search for causes and, possibly, to change therapy. Furthermore, the prediction model derived from patients with IGHD can be applied to groups with other causes of GHD or other growth disorders treated with hGH. The recognition of different or similar types of response to hGH may promote understanding of these disorders and influence general therapeutic strategies. Eventually, the data from large numbers of patients collected in KIGS will provide detailed information about the response to hGH during all stages of development up to adulthood. This will help to optimize therapy for each patient according to his or her requirements. Thus, the continuous scientific exploitation of the study will provide knowledge of how to use hGH most effectively and efficiently to normalize stature in childhood and adult life.

REFERENCES 1. Raben M. Treatment of a pituitary dwarf with human growth hormone. J Clin Endocrinol 1958; 18: 901-5. 2. Ranke MB. Bierich JR. Treatment of growth hormone deficiency. Clin Endocrinol Metab 1986; 15: 495-5 10.

3. Ranke MB, Blank B. Is there an auxological basis for growth promoting therapy with human growth hormone in short children? Acta Paediatr Scand 1990 [Suppl] 367; 4-10. 4. Bierich JR. Final height in hypopituitary patients after treatment with hGH. In: Bierich JR, Cacciari E, Raiti S, eds. Growth abnormalities. Serono Symposia Publications, Vol. 56, New York: Raven Press, 1989: 161-74. 5 . Tanner JM, Whitehouse RH, Takaishi M. Standards from birth to maturity for height, weight, height velocity and weight velocity. British Children. Arch Dis Childh 1966; 41: 454-71, 613-35. 6. Tanner JM, Whitehoux RH. Clinical longitudinal standards for height, weight, height velocity, weight velocity and stages of puberty. Arch Dis Childh 1976; 51: 170-9. 7. Tanner JM. Growth and adolescence, 2nd edn. Oxford: Blackwell Scientific, 1962. 8. Weisberg S. Applied linear regression, 2nd edn. Chichester: Wiley, 1985.

129

130

M.B.Runke

el ul.

Acta Paediatr Scand [Suppl] 370

9. Frasier SD. Costin G. Lippe BM, Aceto T Jr. Burger PF. A dose-response curve for human growth hormone. J Clin Endocrinol Metab 1981; 53: 1213-17. 10. Tanner JM, Whitehouse RH. Hughes PCR, Vince FP. The effect of human growth hormone treatment for I to 7 years on growth of 100 children, with growth hormone deficiency, low birth weight. inherited smallness, Turner’s syndrome and other complaints. Arch Dis Childh 1971; 46: 745-82.

I I . Wit J M , Faber JAJ. van den Brande JL. Growth response to human growth hormone treatment in children with partial and total growth hormone deficiency. Acta Paediatr Scand 1986; 75: 767-73. 12. Wit J M , van’t Hof MA, van den Brande JL. The effects of human growth hormone therapy on skinfold thickness in growth hormone deficiency children. Eur J Paed 1988; 147: 588-92. 13. Preece MA, Tanner JM, Whitehouse RH. Cameron N. Dose dependence of growth response to human growth hormone in growth hormone deficiency. 1 Clin Endocrinol Metab 1976; 42: 477-83. 14. Bundak R, Hindmarsh PC, Smith PJ. Brook C G D Long-term auxological effects of human growth hormone. J Pediatr 1988; 112: 875-9. 15. Sherman BM, Frane J . Johanson AJ, Kaplan SL. Predictors of response to treatment with methionyl human growth hormone. In: Underwood LE. ed. Human growth hormone. Progress and challenges. New York and Basel: Marcel Dekker. 1988: 131-44. 16. Albertsson-Wikland K, Rosberg S, Hall K. Spontaneous secretion of growth hormone and serum levels of insulin-like growth factor I and somatomedin binding protein in children with different growth rates. In: lsaksson 0, Binder C , Hall K. Hokfelt B, eds. Growth hormone. Basic and clinical aspects. Experta Medica. International Congress Series 748, 1987; 163-75. 17. Prader A, Tanner J M , von Harnack GA. Catch-up growth following illness and starvation. J Pediatr 1963; 62: 646-59. 18. Kabi International Growth Study (KIGS). Report No. 2. Stockholm: Kabi, 1989.

(M.B.R.) University Children’s Hospital Ruemelinstr. 23 7400 Tubingen Federal Republic of Germany