Introduction

Top Abstract Introduction Participants and methods Results Discussion References

Idiopathic growth hormone deficiency is the indication for treatment in 50% of children receiving growth hormone, as reported for 100 000 children worldwide in 1999.¹ Growth hormone treatments aim to normalise growth, correct health problems associated with growth hormone deficiency, and help patients achieve an adult height in the normal range for the general population and for familial genetic potential.^2-4

Growth hormone has been used for four decades, initially as an extract and now in recombinant form, but we still know little about its long term effects on adult height.¹

We presented here data for all French children who were diagnosed with isolated idiopathic growth hormone deficiency whose treatments began between 1 July 1987 and 31 December 1992 and who had attained adult height by September 1999.

	Participants and methods

Top Abstract Introduction Participants and methods Results Discussion References

Participants
We identified patients as having growth hormone deficiency according to the criteria used at the time, which included data on height, two growth hormone stimulation tests, or assessment of spontaneous growth hormone secretion.⁵ Sex steroid priming was used before growth hormone testing in 2% of patients.

Data collected
The Association France Hypophyse decided annually whether the treatment should be continued. Criteria for discontinuation of treatment (scored as completion) were growth velocity <3 cm/year, bone age 13 years (girls) or 15 years (boys), or height 160 cm (girls) or 170 cm (boys). The third criterion applied to 30 of the 2852 patients (1.1%).

We prospectively collected follow up data in 1998-9 from doctors or from patients who provided "self reported" values for height and weight. We considered that adult height had been attained if growth velocity was 1 cm/year or if bone age was 16 years (girls) or 18 years (boys). We obtained adult heights for 2165 patients (76% of 2852).

Analysis of growth and statistical methods
We calculated standard deviation (SD) scores of height and weight for age, sex, or gestational age, and target height.⁵ Age at onset of puberty was expressed in standard deviations. ^{6 7}

We constructed, in several stages, a model for predicting adult height (for details see bmj.com). ^{8 9} The model was adjusted for baseline height, and describes adult height gain  that is, the difference between adult and baseline height expressed in standard deviation  and adult height itself (in standard deviations).⁸

	Results

Top Abstract Introduction Participants and methods Results Discussion References

Characteristics of participants at baseline and treatments
At a mean age of 13.3 years (boys) and 11.6 years (girls), 1178/1836 (64%) boys and 677/1016 (66%) girls were prepubertal, indicating pubertal delay. These initially prepubertal patients entered puberty late, at 14.1 (SD 1.5) years (boys) and 12.5 (1.3) years (girls) (0.9 (1.3) SD score). The mean dose of growth hormone at start of treatment was 0.42 IU/kg/week (0.29 to 0.62 IU/kg/week in 90% of the patients).

Changes in height
We classified patients according to whether treatment was completed (1524, 53.4%) or stopped early (table 1). When treatment was completed, the height gain was 1.0 (0.7) SD in a total of 3.6 (1.4) years; most of the height was gained during the first two years (figure). If treatment was stopped early, the height gain was significantly smaller (P=0.0003). Normal results in retests of growth hormone secretion were the reason for non-completion in 14%.

View larger version (20K): [in this window] [in a new window]   Changes in height in standard deviation (SD) score in patients treated with growth hormone, relative to beginning of treatment (top) and in absolute values. Solid line indicates treatment; broken line indicates end of treatment to attainment of adult height

Adult height was recorded in 81% and 70% (1232/1524 and 933/1328) of patients who completed treatment and patients who did not, respectively. After growth hormone treatment, patients who completed treatment gained 2.8 (2.8) cm, and those who did not complete treatment gained 12.3 (8.0) cm. Mean adult height was therefore similar (1.6 SD), 0.4 SD below target height.

Predictive models for adult height
We constructed a multivariate model of factors explaining adult height. In the final model, target height and birth weight and regression towards the mean accounted for 33% of outcome variance (table 2). Variables determined at baseline that predicted a good outcome were age, bone age delay, and prepubertal status. Thus, older patients presenting no signs of puberty and with marked bone age delay had better outcomes. A negative association with male sex reflected sex dependent differences in pubertal age. Severe growth hormone deficiency was associated with better outcome. Whether or not treatment was completed, and length of treatment, were independent predictors. Patients who did not complete the study grew 0.3 SD more than those who did; conversely, duration of treatment was positively associated with outcome (0.2 SD per year of treatment). Growth hormone dose did not predict adult height. Together, all variables associated with treatment accounted for 4.5% of outcome variance.

	Discussion

Top Abstract Introduction Participants and methods Results Discussion References

Strengths and weaknesses of the study
Studies generally assess change in height and assume that all improvement results directly from treatment. We did not select our sample according to outcome. Instead, we studied all children who started treatment; in this population, growth continued in children who stopped treatment before the end of growth. Patients who completed treatment generally had more severe growth hormone deficiency and seemed to benefit from growth hormone in a time dependent fashion. Our results are consistent with completion bias, analogous to indication bias, in which more severely affected patients receive heavier treatment.¹⁰

We should also consider methodological aspects, such as whether the diagnosis of growth hormone deficiency was valid in our study population. The main criterion for a diagnosis of growth hormone deficiency in short children in the 1990s was a peak growth hormone value, measured in two stimulation tests, of <10 µg/l. ^11-14 However, this cut off has recently been questioned.¹⁴ Sex steroid priming before growth hormone testing increases growth hormone secretion and may prevent the incorrect diagnosis of growth hormone deficiency, especially if puberty is delayed.¹⁵ Only 2% of our patients were primed with sex steroids, and priming would have increased growth hormone secretion in many of the others. Growth velocity is an important diagnostic criterion^16-18 but was only slightly reduced in our patients compared with normative values for age and sex. However, these patients are typical of patients treated worldwide for growth hormone deficiency. All data were obtained from routine examination in daily practice and various growth hormone tests and assays were used, therefore their reliability may be questioned.

Finally, we selected a subgroup of the patients treated for growth hormone deficiency; patients with non-idiopathic growth hormone deficiency or abnormalities on pituitary magnetic resonance imaging were excluded, and patients with early onset growth hormone deficiency were excluded by the design of the study focusing on adult height. Therefore, our findings cannot be generalised to other patient populations.

Comparison with other studies
Our patients' data are similar to patients in other studies in terms of age and height standard deviation scores at the start of treatment. The growth hormone doses used were 20% lower than those used in other European countries at the time but are unlikely to explain the differences found because growth hormone dose did not predict outcome.¹⁹ We followed 76% of our target population, whereas other reports focused on a smaller proportion (1.9% to 3.5%) of the patient sample. ^{20 21} Such selection may focus on patients who responded well to treatment, providing an overoptimistic view of the results (figure). This probably explains the 15% to 30% difference from other studies.

Influence of pubertal delay
Overall, the onset of puberty was delayed considerably in our patients, as in the Pharmacia International Growth Database,²² and variables linked to pubertal delay positively were associated with adult height. This strongly suggests that many had constitutional delay in growth and puberty, which should not be confused with growth hormone deficiency. ^{2 15 23}

Conclusion
Long term treatment with growth hormone has no clearcut benefit in a large proportion of patients treated for so called idiopathic isolated growth hormone deficiency. Most of the patients actually have pubertal delay and a potential for spontaneous catch up, which must be taken into account when measuring the effect and cost effectiveness of growth hormone treatments. The diagnosis of idiopathic isolated growth hormone deficiency should be restricted to a small minority of patients with severely and permanently altered growth hormone secretion: our results support the use of growth hormone in such patients. We propose that peak growth hormone values should be below 2-4 µg/l, that sex steroid priming is used before growth hormone testing, and that more attention is paid to the causes of hypopituitarism.

Long term controlled trials to evaluate the effects of growth hormone treatment in patients who do not have growth hormone deficiency are needed, given the number of children treated worldwide. We should try to identify predictive markers for short stature in adults and focus intervention on patients at higher risk.