Recombinant human growth hormone for the treatment of growth disorders in children: a systematic review and economic evaluation

Growth hormone deficiency

Growth hormone deficiency (GHD) occurs when the pituitary gland fails to produce sufficient levels of growth hormone (GH).

There is some debate about the diagnostic criteria for GHD: the diagnosis of GHD includes short stature, growth velocity (GV) below the 25th percentile for at least 1 year, and delayed bone age. 1 Rosenfeld2 suggests other criteria: height > 3 standard deviations (SDs) below the mean, < –2 SD to –3 SD for age and deceleration in growth (such as GV < 25th percentile for age), GV < 5th percentile where there is no other explanation, a predisposing condition along with growth deceleration or other signs of pituitary dysfunction. Juul and colleagues3 found ‘large heterogeneity in the current practice of diagnosis and treatment of childhood GHD’. Their survey of European paediatricians found that the cut-off points of GH peak response used for diagnosis of deficiency clustered around 10 ng/ml or 20 mU/l.

The primary goals of recombinant human growth hormone (rhGH) treatment for children with GHD are to normalise height during childhood, for the treated child to reach a ‘normal’ adult height (AH) as defined by the parental target and for mature somatic development to be reached around age 25. 4 The British Society for Paediatric Endocrinology and Diabetes (BSPED) recommends 3- or 6-monthly growth monitoring, annual insulin-like growth factor-1 (IGF-1)/ insulin-like growth factor building protein-3 (IGFBP-3) monitoring, and compliance assessment at each appointment. 5

Turner syndrome

Prader–Willi syndrome

Prader–Willi syndrome (PWS) is a genetic disorder characterised by short stature, abnormal body composition, hypogonadism, obesity, dysmorphic features, hyperphagia (compulsive overeating), hypotonia (diminished muscle tone), and specific learning and behavioural issues. 18

Chronic renal insufficiency

Chronic renal insufficiency (CRI) is defined as a persistent elevation of serum creatinine and/or urea level. It can be caused by a variety of conditions, including congenital disorders, glomerular disorders and infections. Growth failure associated with CRI can be caused by acidosis, rickets, GH resistance, inadequate nutrition and anorexia. 24 Children with CRI experience impaired growth once their glomerular filtration rate (GFR) falls to 50% of normal, with increasing problems once the GFR falls below 25%. 25 Following kidney transplantation, chronic graft rejection and treatment with steroids can restrict growth and development. 26 Patients undergoing haemodialysis or peritoneal dialysis can be considered for rhGH treatment, as well as those who have received kidney transplantations.

Small for gestational age

There are various thresholds for defining a child as being born ‘small for gestational age’ (SGA), the most commonly used being where the birth height or weight is ≤ 2 SDs below the population average, or is below the 10th centile for birthweight. 38 However, this group is heterogeneous in composition. Between 50% and 70% of these babies are ‘constitutionally small’ but otherwise healthy. The other babies in the group are those who have not reached their height or weight potential, having possibly experienced fetal growth restriction (FGR). 38 For this reason, the terms intrauterine growth restriction/retardation (IUGR) and SGA are not synonymous: a child born SGA has not necessarily undergone IUGR or FGR, and a child who has IUGR or FGR may not necessarily be born SGA.

SHOX deficiency

Impact of health problem

Severe short stature may be physically debilitating in untreated children,62 with children being at greater risk of bullying at school and social isolation. 63 Some children with short stature may also have difficulties with emotionally immature behaviour, anxiety and poor school performance. 64 However, not all children who are shorter than their peers will experience problems. For example, the Royal College of Obstetricians and Gynaecologists states that the majority of children born SGA do not have any appreciable morbidity or mortality. 38 However, others indicate that children born SGA who remain short may suffer from alienation, low self-esteem, impaired social dynamics, behavioural problems, lower educational achievement and professional success. 39,43

Children with short stature can also be at increased risk of morbidity and mortality in later life. For example, the risk of cardiovascular morbidity is increased in patients with GHD,65 TS,66 and PWS,67 while some patients with growth disorders may also be at increased risk of type 2 diabetes and metabolic syndrome. 67,68 Low birth weight is also associated with future increased risk of coronary heart rate and stroke. 69

Outcome measures

The main parameter used to measure the efficacy of rhGH treatment is growth. This reflects the main goals of therapy, which are physiological catch-up growth if possible, achievement of normal height during childhood, timely and normal growth during puberty and normal height in adulthood. In children with PWS, treatment with rhGH aims to improve body composition as well as boosting growth.

Measures of growth include:

• Final height (FH) or adult height (AH) Measured either in centimetres or expressed as a standard deviation score (SDS), this is the best measure of how rhGH treatment affects growth. FH has been achieved when the growth rate has slowed to less than some specified amount (e.g. 1–2 cm/year), and radiographs of the wrist and hand show that the epiphyses have closed (often expressed as a bone age of more than 14–15 years). 6 Ideally, FH would be calculated in comparison with an untreated control group in a randomised controlled trial (RCT). Some non-RCT designs use historical controls, which may overestimate the effects of rhGH treatment. Similarly, database studies may not include all relevant factors or be representative samples of treated patients. 6

• Near-final height (NFH) Sometimes reported where it is assumed that FH has been reached using the above criteria, but it is acknowledged that growth may not yet be quite complete. 6

• Height Usually measured standing, using a wall-mounted Harpenden stadiometer or a similar device. For very young children, supine length is measured.

• Height standard deviation score (HtSDS) This expresses height relative to norms for children of the same age, allowing comparisons that are independent of age or gender. The normal population mean is zero and a normal SD score will lie between –2 and +2 SDs. Increase over time in SDS or upward centile crossing implies catch-up growth and a decrease implies growth failure. Calculation of SDS depends on the reference data used, i.e. normal height for children in the same country.

• Growth velocity (GV) Also referred to as height velocity, this is the change in height over a specified period, e.g. cm/year. Although the overall effectiveness of rhGH in treating short stature is to be found in measures of FH, velocity may be a better interim growth measure than height attained at a particular age, as it is independent of growth in previous years.

• Growth velocity standard deviation score (GVSDS) This is the GV relative to norms for children of the same age.

• Bone age (BA) A measure of skeletal maturity, usually determined by examining the relative positions of the bones in the left hand and wrist from a radiograph. The measurement of BA relative to chronological age is important in height-prediction models. In addition, BA assessments are used to evaluate when the epiphyses have closed and growth is complete. The interim assessment of BA is important in determining whether treatment is advancing bone maturity, such that short-term GV might come at the expense of early closure of the epiphyses. Clinical trials often measure BA to monitor whether this is accelerating undesirably fast in rhGH-treated patients compared with control patients. Height for BA can also be used as an estimate of improved height potential in response to rhGH therapy, especially in short-term studies.

Measures of body composition assess obesity and the amount of fat relative to other body tissues. Body mass index (BMI) calculates the ratio of body mass to the square of body height, expressed as kg/m². The National Institute for Health and Clinical Excellence (NICE) recommends BMI as providing a practical estimate of overweight in children, although mentions that it needs to be interpreted with caution as it is not a direct measure of adiposity. 70 Dual-energy X-ray absorptiometry (DEXA) can be used to measure lean mass (fat-free mass) and percentage body fat, which can be used to indicate body composition.

Physiological outcomes reported in studies of rhGH may include assessments of the concentrations of hormones, glucose, cholesterol, and markers of bone and general metabolism. Such measures are important for assessing the biochemical, metabolic and adverse effects of rhGH, and can have implications for long-term health. IGF-1 is an endocrine hormone that is produced by the liver, and its production is stimulated by GH. Lower than normal levels are therefore seen in people with growth disorders. The insulin-like growth factor building proteins (IGFBPs) act as carrier proteins for IGF-1. There are six IGFBPs, with IGFBP-3 being the most abundant. 71 IGF-1 is monitored during rhGH therapy as there is a theoretical concern that persistently elevated levels may predispose the patient to other diseases later in life. Monitoring levels also helps to tailor the dose to the individual. As IGFBP-3 binds IGF-1, monitoring this gives an indication of the levels of ‘free’ IGF-1 in circulation. High levels of IGF-1 with low levels of IGFBP-3 may be linked with breast, colorectal and prostate cancer. 72,73

Current service provision

GHD

Treatment with rhGH is currently recommended by NICE to help increase the growth of children with GHD. 75 For children with congenital GHD, rhGH therapy is not generally started before the child is 4 years old. 6 However, if there is profound growth failure or evidence of recurrent hypoglycaemia, which may occur in infants under the age of 1, treatment may be started earlier. For children who acquire GHD at an older age, treatment can start at a time that is appropriate to their condition and stage of growth. Treatment is discontinued after the first year if there is a poor response, i.e. < 50% increase in growth rate, or if compliance or growth rate remains poor thereafter. Otherwise treatment can continue until GV is < 2 cm/year, assessed over 6–12 months, when FH is achieved. Other clinical advice suggests that treatment is necessary for the patient to attain peak bone mass, which may not be until the age of 25 or 26 in some people. A recent survey of paediatric endocrinologists (56 responses out of 72 questionnaires) found that 56% of clinics provide transfer clinics for patients ending paediatric treatment and transferring to the care of an adult endocrinologist. Of the 56 respondents, 80% retest for GHD prior to transfer, 55% transfer all rhGH-treated patients and the remainder transfer only those who are still GH deficient on retesting. 74

Turner syndrome

Current NICE guidance recommends that rhGH treatment for girls with TS should begin at the earliest age possible, to boost growth. 75 Some patients with profound growth retardation and failure to thrive may commence treatment earlier than those who are diagnosed later. A Belgian study16 found that median age at diagnosis of 242 girls was 6.6 (range 0–18.3) years, although the survey found that 22% of girls were diagnosed after the age of 12 years. Some clinical expert advice suggests that the mean age for starting treatment is 8–9 years of age as many girls are not diagnosed until later in childhood, although there has been a recent trend towards earlier diagnosis.

Prader–Willi syndrome

NICE guidance currently recommends the use of rhGH for children with PWS to improve height, body composition and bone mineral density. For children with PWS, treatment with rhGH is intended to improve body composition and metabolism as well as increase FH. Its place in the treatment pathway depends on age at diagnosis. Children with PWS are assessed for obesity, potential for obstructive sleep apnoea and ongoing respiratory illness before treatment is considered. Low muscle tone and its impact on the child’s development are also considered.

Chronic renal insufficiency

Treatment with rhGH is currently recommended by NICE to help increase the growth of prepubertal children with CRI. 75 The guidance recommends that treatment should be stopped after a renal transplantation, and re-established after only 1 year if it has been ascertained that catch-up growth has not occurred. 75 The place of rhGH in the treatment pathway for children with CRI depends on age at diagnosis, and on clinical factors related to management of the child’s condition. rhGH treatment can take place either before or after renal transplant, although allograft rejection can be a concern if rhGH treatment is given post transplant.

Small for gestational age

Previous NICE guidelines did not consider children born SGA, as rhGH was not licensed for this indication at the time. 82 Children born SGA but with no comorbidities may not be diagnosed until they fail to achieve catch-up height by the age of 2–4 years,39 or when they start school. The International SGA Advisory Board indicated that SGA children aged 2–4 years who show no evidence of catch-up with a height of –2.5 SD should be eligible for rhGH treatment. They also recommended that treatment should be considered in children older than four years who show no catch up at a height –2 SD or less. 39 The European licence for rhGH is for children aged 4 years and over.

SHOX deficiency

Currently, there is no NICE guidance available for the use of rhGH in children with SHOX-D. Initiation of rhGH treatment for children with SHOX-D depends on age at diagnosis. Clinical evaluation is used to assess growth failure, but GH provocation tests are not required once SHOX-D has been established via a positive SHOX DNA blood test.

Anticipated costs associated with intervention

The costs associated with rhGH therapy interventions comprise:

• the drug (dose adjusted for body weight)

• self-therapy training of the patients and their parents (involving home visits by specialist and community nurses)

• monitoring of treatment effectiveness (involving paediatric endocrinology outpatient visits for blood tests, a test of pituitary function, and an assessment of BA by hand radiograph).

The costs of training patients and their parents are limited to the first year of treatment. During each year of treatment, until they stop growing, patients would typically attend two outpatient consultations. Estimates of the current costs of these components of the rhGH interventions for patients with GHD, TS, PWS, CRI and SGA are provided in Chapter 4 (see Estimation of costs).

Interventions

The intervention is rhGH, also known as somatropin. It is marketed as the following products: Humatrope (Eli Lilly & Co.); Zomacton (Ferring Pharmaceuticals); NutropinAq (Ipsen); Norditropin SimpleXx (Novo Nordisk); Genotropin (Pfizer); Omnitrope (Sandoz) and Saizen (Merck Serono).

Population, including subgroups

The population consists of children with one of the following conditions: GHD, TS, PWS, CRI, SHOX-D, being born SGA. No age-specific definition of a child was given during the scoping process for this review. Possible subgroups could be children with different causes of GHD, and children with CRI who are either pretransplant or post transplant. However, analysis of the effectiveness of rhGH treatment for any of these subgroups of patients is limited by the available data and the statistical power of the identified trials.

Transition of care from paediatric to adult endocrine services of young people requires patients to have repeat testing of their GH axis to be sure that they need to continue treatment. This transition period is only considered within this review where evidence from the identified studies allows for patients whose linear growth is not complete.

Relevant comparators

The standard comparator for this review is management strategies without rhGH. This includes placebo injections and no treatment.

Outcomes

Clinical outcomes of interest include: FH gained, HtSDS, GV, GVSDS, body composition, biochemical/metabolic markers, AEs of treatment; health-related quality of life (HRQoL). Direct costs include estimates of all health-care resources consumed in the provision of the intervention, including diagnostic tests, administration and monitoring costs – as well as consequences of those interventions, such as treatment of adverse effects.

Search strategy

An experienced information specialist developed and tested search strategies for this review. Separate searches were carried out to identify studies reporting clinical effectiveness, cost-effectiveness, HRQoL, resource use and costs, and epidemiology/natural history of the conditions. The search strategy for MEDLINE, shown in Appendix 2, was adapted as appropriate for a number of other electronic databases. We searched: The Cochrane Database of Systematic Reviews (CDSR); The Cochrane Central Register of Controlled Trials; NHS Centre for Reviews and Dissemination (NHS CRD, University of York) Database of Abstracts of Reviews of Effectiveness (DARE) and the NHS Economic Evaluation Database (NHS EED); MEDLINE (OVID); EMBASE (OVID); National Research Register (NRR); Current Controlled Trials; ISI Proceedings; Web of Science; and BIOSIS. For all disease areas we searched the databases from their inception to June 2009. This meant there was some duplication of earlier work for the previous review, but this was necessary as the present review required searches for additional outcomes, such as biochemical and metabolic markers. Searches were limited to the English language.

Relevant conferences (European Society for Paediatric Endocrinology, The Endocrine Society, American Association of Endocrinologists, Paediatric Academic Societies) were searched for recent abstracts (up to June 2009) to assess against the inclusion criteria. Bibliographies of related papers were screened for relevant studies, and we contacted experts to identify any additional published or unpublished references. We also assessed the MSs to NICE for any additional studies that met the inclusion criteria.

Inclusion and data extraction process

Titles and abstracts of studies identified by the search strategy were assessed for potential eligibility by two reviewers. The full text of relevant papers was then obtained, and inclusion criteria were applied by two independent reviewers. At both stages of the screening process, any differences in opinion on inclusion of a particular study were resolved through discussion. Data from included studies were extracted by one reviewer using a standard data extraction form and checked by a second reviewer. Any discrepancies were identified and resolved through discussion.

Quality assessment

The quality of included studies was assessed using NHS CRD (University of York) criteria. 83 Quality criteria were applied by one reviewer and checked by a second reviewer, with differences in opinion resolved by discussion. The criteria used are shown in Appendix 3. Publication bias was not assessed.

Inclusion criteria

Data synthesis

• Clinical effectiveness studies were synthesised through a narrative review with tabulation of results of included studies. Key outcome measures are reported in tables in the text, and other outcomes are shown in the full data extraction forms in Appendix 4. For conciseness, where a study reported outcome measures after 1 and 2 years, only the final year’s outcomes are included in the table, as these show the longest duration of treatment effect.

• Where data were of sufficient quality and homogeneity, a meta-analysis of the clinical effectiveness studies was considered using review manager 5.0 software.

• Quality-of-life studies were synthesised using the same methods as above, i.e. narrative review and meta-analysis only if feasible.

Quantity and quality of research available

The number of references considered at each stage of the review is shown in Figure 1. Of the 674 references identified, 560 were excluded on inspection of their titles and abstracts. The full papers of 114 references were retrieved and assessed against the inclusion criteria. A total of 77 of the retrieved full papers were rejected at this stage, mostly due to the patient group not meeting the inclusion criteria (n = 40) or due to a non-RCT study design (n = 27). A list of papers excluded at this stage is included in Appendix 5, together with reasons for exclusion. A total of 28 RCTs in 34 publications were included in the systematic review of clinical effectiveness. Appendix 6 lists conference abstracts that were identified as being of interest, but which contained insufficient information to be included in the review of clinical effectiveness.

FIGURE 1.

Flow chart of identification of published studies for inclusion in the systematic review of clinical effectiveness. ^aOne of the systematic reviews was the previous Health Technology Assessment report written for the National Institute for Health and Clinical Excellence, so this was not data extracted. It is discussed briefly (in Summary of previous systematic reviews).

An overview of the included studies is given in Table 2. Only one SGA paper and one TS paper reported FH; none of the other conditions’ studies reported FH as an outcome measure. None of the papers reported specific QoL measures. All disease areas included at least one paper which reported outcomes on height gained, body composition, biochemical markers and AE. The characteristics and quality assessment of the included studies are discussed in each of the relevant disease-specific results chapters.

Quantity and quality of research available

One study met the inclusion criteria for this review, and the key characteristics are presented in Table 3. The full data extraction form in Appendix 4 has further details.

Soliman and Abdul Khadir84 recruited two groups of GH-deficient children and one group of children who were not GH deficient. These groups were then subdivided into treatment groups: group 1a received 30 units (U)/m²/week of rhGH and group 1b received 15 U/m²/week. Group 2a received 15 U/m²/week and group 2b received no treatment. Group 3 (non-GHD short children) was subdivided in the same way as group 2. Group 2 was the only group in this study with GHD and with children randomised to either rhGH or no treatment, and, as such, is the only group considered in this report. The treatment groups’ baseline characteristics were similar. The study used a dose of 15 U/m²/week, and it is not clear how this corresponds to the licensed dose as neither milligrams (mg) nor international units (IUs) are used.

Overall the quality of the reporting of the included study was mixed (Table 4). No details were given on randomisation or allocation to treatment groups. For example, Soliman and Abdul Khadir84 recruited children into specified groups according to peak GH response to provocation, and these groups were then divided at random into two subgroups. No further details were given. The low patient numbers will affect interpretation of results from this trial.

The comparator group did not receive placebo: this could mean that both care providers and patients would have been aware of whether they were receiving treatment, which, in turn, can affect reporting of some outcomes. Soliman and Abdul Khadir84 appear to have carried out an intention-to-treat analysis (ITT), which can protect against attrition bias.

Growth outcomes

The Soliman84 study reported GV and HtSDS, and these are presented in Table 5. The data extraction forms in Appendix 4 list further outcome measures, such as BA.

Children in the treated group in the Soliman study grew an average of 2.7 cm/year faster than those receiving no treatment in the 12 months of the study, and the difference between groups was statistically significant (p < 0.05). Similarly, children in the treated group had a statistically significantly higher HtSDS: –2.3 ± 0.45 versus –2.8 ± 0.45 in the untreated group (p < 0.05).

Body composition outcomes

Soliman and Abdul Khadir84 did not report body composition outcomes.

Biochemical markers

The results reported for IGF-1 levels in the Soliman study84 are shown in Table 6. Further biochemical markers, such as insulin, are included in the data extraction tables in Appendix 4.

The IGF-1 levels at 12 months are statistically significantly higher in the treated than in the untreated group: 91.2 ± 30.4 versus 49.4 ± 19.

Quality of life

Soliman and Abdul Khadir84 did not report QoL results.

Adverse events

Adverse events were not reported by Soliman and Abdul Khadir. 84

Summary

One trial examining the effectiveness of rhGH for GHD met the inclusion criteria for the review.

• The quality of the included study was mixed. It was an unblinded study, which can have an impact on outcome reporting but did report an ITT analysis.

• Children in the rhGH group grew 2.7 cm/year faster than children in the untreated group during the 1-year study, and had a statistically significantly higher HtSDS: –2.3 ± 0.45 versus –2.8 ± 0.45.

• The IGF-1 levels were statistically significantly higher in the treated group than in the untreated group.

• The included study did not report QoL or AE.

Quantity and quality of research available

Six studies assessing the effectiveness of GH for growth restriction in TS met the inclusion criteria for the review. 12,85,86,88–90 The key characteristics of these studies are presented in Tables 7–12. Appendix 4 has further details.

Two of the included studies were of a crossover design,88,89 and these compared doses of 0.1 IU/kg/day88 and a mean of 1.3 ± 0.3 mg/day (alone or in combination with oestradiol)89 with placebo. The group receiving oestradiol is not discussed further here. Of the remaining studies, two compared rhGH with no treatment,85,86 one with low-dose estrogen,90 and one with placebo. 12 Stephure and colleagues86 administered a rhGH dose of 0.30 mg/kg/week, with a maximum weekly dose of 15 mg. The dose of 50 µg in the Davenport study85 is comparable with that of Stephure and colleagues. Those in the Quigley study12 were slightly different: group 1 received 0.27 mg/kg/week and group 2 received 0.36 mg/kg/week. Johnston and colleagues90 gave a dose of 28–30 IU/m²/week. All studies included at least one treatment arm with a dose that was broadly comparable with the licensed dose of 45–50 µg/kg/day or 1.4 mg/m²/day.

Four of the six included studies reported growth outcomes, including height gain and change in HtSDS. 12,85,86,90 The remaining two studies reported body composition and biochemical marker outcomes. 88,89

The trials varied considerably in size. The two crossover trials were small, with 1288 and nine89 participants. The Stephure86 and Quigley12 studies were larger, with 154 and 232 participants, respectively. Johnston and colleagues90 recruited 58 patients, and Davenport and colleagues recruited 89. 85 The included trials also ranged in length. The groups in Quigley and colleagues12 remained randomised for 18 months, the Davenport study85 for 2 years and the Johnston study lasted for 1 year. 90 Protocol completion in the Stephure86 study was defined as annualised GV less than 2 cm/year and BA of 14 years or greater, which we have interpreted to mean FH. In contrast, the two Gravholt studies88,89 were short crossover trials, with rhGH treatment for 2 months.

Five12,86,88–90 of the six trials recruited broadly similar age groups, whilst the sixth by Davenport and colleagues85 specifically targeted very young girls with TS. As a result their girls have much younger mean ages of 1.98 ± 1.01 and 1.97 ± 1.01 for treatment and control groups, respectively.

Four of the included studies reported baseline characteristics that were similar between groups. 12,85,86,90 However, none reported p-values for between-group differences, so there may have been small differences at baseline. For example, in the study by Stephure and the Canadian Growth Hormone Advisory Committee (CGHAC) 2005,86 girls in the rhGH group were on average 3 cm shorter than those in the control group. The SD values indicate overlapping CI, suggesting there is no statistically significant difference between the two groups. However, the 3-cm difference could have an impact on end of study height. The other two studies, reported by Gravholt and colleagues, were of crossover design. One reported baseline characteristics for the whole study group89 and the other did not appear to report any baseline conditions. 88

The six included trials were generally of poor methodological quality, and poorly reported (Table 8). Only one reported adequate methods of randomisation to treatment groups. 85 Davenport and colleagues85 stratified their participants by age and then randomised them using a blinded phone-in process. Four of the six trials did not describe randomisation techniques. 12,86,88,89 Johnston and colleagues90 reported that five participants were reallocated from the oestrogen group to receive rhGH: it is unclear when this occurred and therefore method of randomisation was judged inadequate.

Concealment of treatment allocation was also judged to be adequate in the Davenport trial, and ‘unknown’ in the remaining five. In the Gravholt89 study it is unclear how allocation to treatment groups had taken place. The study had only nine participants, and these were simply reported to have been given the treatment regimen sequentially and in random order.

Blinding of participants, those who provide care and those who assess outcomes can protect against the reporting of some outcomes being affected by the knowledge of which treatment is being received. Blinding of outcome assessors, care providers and patients was judged ‘unknown’, ‘inadequate’ or ‘partial’ in five out of the six trials; Gravholt and colleagues89 adequately blinded their patients by administering placebo in place of both rhGH and the oestradiol.

None of the six studies included here used an ITT analysis. This kind of analysis can protect the study from attrition bias, where, for example, participants withdrawing from the treatment arm could represent AE or treatment failure.

Growth outcomes

Four out of the six included studies reported growth outcomes, and key measures are shown in Table 9. Please see Appendix 4 for additional outcomes. Neither of the studies by Gravholt and colleagues88,89 reported growth outcomes.

Two studies reported height at the end of the study: both found a statistically significant difference between the treated and untreated groups (p < 0.0001). 85,86

Children in the treated group in the Stephure study86 were 6.5 cm taller on average than the untreated group at protocol completion. However, there was a 3-cm difference between the groups’ mean heights at baseline. Mean change from baseline was therefore 9.3 cm more in the rhGH than in the untreated group at the end of protocol completion (28.3 ± 8.9 vs 19.0 ± 6.1).

The Stephure study86 also reported an addendum follow-up (approximately 10 years since randomisation), which included 66% of rhGH patients and 44% of the control group. The treated group’s mean FH was 149.0 ± 6.4 cm compared with 142.2 ± 6.6 cm in the untreated group (p < 0.001), i.e. a difference of 6.8 cm. Mean change from baseline to FH was 8.7 cm more in the rhGH than in the untreated group.

In the Davenport study85 the mean difference was 7.6 cm (height at study end was 99.5 ± 7.6 cm in the treated group vs 91.9 ± 7.2 cm in the untreated group, p < 0.0001).

Height standard deviation score is also reported by Davenport and colleagues85 and Stephure and CGHAC. 86 Both authors report statistically significant differences between groups for this outcome, with the treated groups both achieving higher HtSDS. In the Stephure study86 the HtSDS is reported for the age-specific Turner population and for the adult Turner population.

The difference in change in height was statistically significant between groups in the two studies that reported it. Stephure and colleagues86 report a change in height at protocol completion of 28.3 ± 8.9 cm versus 19 ± 6.1 in the untreated group, p < 0.001. Davenport and colleagues85 reported a 2-year height gain of 20.4 ± 3.3 cm (treated group) versus 13.6 ± 3.5 cm (untreated group), p < 0.001 (not shown in table). Change in HtSDS in both the Stephure86 and Johnston90 studies was higher in the treated than untreated group: 1.6 ± 0.6 (treated) versus 0.3 ± 0.4 (untreated), p < 0.001, at protocol completion in the Stephure study; 0.7 (0.7) versus 0.4 (0.9), p < 0.05, in the Johnston study90 after 1 year.

Growth velocity was statistically significantly greater in the treated groups in the Stephure,86 Davenport85 and Quigley12 studies. Davenport and colleagues85 reported GV at the end of the first and second year. Although this was greater in the treated groups at both times, GV fell in the second year in both groups: 8.4 ± 1.6 cm/year (treated group) versus 5.5 ± 1.8 (untreated). Additionally, Davenport and colleagues85 measured GV SDS at the end of the first and second years. Again, this was greater in the treated group at the end of the first year: 1.75 ± 1.25 versus 0.8 ± 0.95, p < 0.001, but was reduced by the end of the second year in both groups: 0.70 ± 1.11 (treated) versus –1.63 ± 1.29 (untreated), p < 0.001. Quigley and colleagues reported GV after 18 months. This was broadly similar in both the lower- and higher-rhGH-dose groups: both were significantly higher than that in the placebo (Pla) group: 6.6 ± 1.1 (GH 0.27/Pla group) versus 6.8 ± 1.1 (GH 0.36/Pla group) versus 4.2 ± 1.1 (Pla/Pla group), p < 0.001 compared with placebo.

Bone age differences for the younger participants in the Davenport study were statistically significant:85 the GH-treated group at 2 years had a mean BA of 4.24 ± 1.35 versus 3.38 ± 1.11 in the untreated group, p = 0.0033. Davenport and colleagues85 also reported BA/chronological age; this is lower in the treated group, and the difference was statistically significant: 0.64 ± 0.80 versus 0.21 ± 0.96, p < 0.001.

Body composition outcomes

Three of the TS studies reported body composition outcomes, and these are presented in Table 10. One of the studies reported weight, weight standard deviation score (WtSDS) and BMI,85 whereas the remaining two reported FM, bone mineral content (BMC) and LBM for arms, legs, trunk and head, and as a total. 88,89 Please see Appendix 4 for BMC results.

Weight and WtSDS were significantly greater in the group receiving rhGH than in the untreated group in the Davenport study,85 reported as 16.62 kg ± 2.86 versus 13.81 kg ± 2.50, and 0.20 ± 1.06 versus –1.37 ± 1.36, respectively (p < 0.0001 for both comparisons).

Two studies considered FM, BMC and LBM. 88,89 In both studies the total FM was greater in the untreated group than in the treated group, and LBM was slightly higher in treated than in untreated patients (Table 10). The differences between groups were of borderline statistical significance in one study88 but no p-values were presented in the other study. 89

Biochemical markers

Three of the studies85,88,89 reported biochemical outcomes. Key results are shown in Table 11 – other outcomes are in Appendix 4.

Two studies reported mean levels of IGF-1 at end of treatment. In both studies IGF-1 levels were statistically significantly higher in the group receiving rhGH. One study88 reported values of 380.5 ± 116.3 versus 179.8 ± 79.4 in the treated and untreated groups, respectively (p < 0.0005). The other89 reported 661 ± 192 versus 288 ± 69 (p-value not reported) for treated and untreated patients, respectively.

Davenport and colleagues85 reported that IGF-1 SDS was significantly greater in the treated group (1.26 ± 0.72 vs –0.69 ± 0.84, p < 0.0001). Change in IGF-1 SDS from baseline to year 2 was 1.53 ± 0.93 versus –0.09 ± 0.87 in the treated and untreated groups, respectively.

One Gravholt study88 reported that IGFBP-3 levels were statistically significantly higher in the treated group than in the untreated group (5982 ± 1557 vs 4344 ± 787, respectively, p = 0.002). The other study by Gravholt and colleagues89 reported higher IGFBP-3 SDS values in treated patients, but no clear p-value was reported. 89 Davenport and colleagues85 found that IGFBP-3 SDS was higher in their treated group (0.97 ± 0.94 vs –1.12 ± 1.13, p < 0.0001).

Fasting glucose and fasting insulin were reported in the two studies by Gravholt and colleagues,88,89 both of which were raised in the groups receiving GH in each study. Mean glucose (nmol/l) was 4.28 ± 0.5988 and 4.46 ± 0.4089 in the treated groups, versus 4.02 ± 0.4488 and 4.04 ± 0.4789 in the untreated groups. This difference reached statistical significance in the first study,88 p = 0.046. Mean fasting insulin levels in the first Gravholt study88 were 17.17 ± 8.30 versus 8.58 ± 4.27, p = 0.007.

Quality of life

None of the TS studies reported QoL as an outcome.

Adverse events

Adverse events were reported by only four of the studies. 12,85,86,90 Details presented by three of the studies are shown in Table 12 (the fourth study did not present figures90).

The group receiving GH in the Stephure study86 experienced a statistically significantly greater level of all AEs (where statistical significance was reported), with the exception of goitre, and one instance of death from ruptured aortic aneurysm, which occurred in the untreated group. The one case of elevated transamine levels in the treated group led to withdrawal from the study.

Davenport and colleagues85 report the same level of serious adverse events (SAEs) for both the treated and untreated groups. For treatment-emergent AEs, defined as ‘events or conditions that began or worsened after study entry’, the results were similar. There were 42 (93%) in the treated group and 43 (98%) in the untreated group. Most treatment-emergent AEs were ear disorders.

Quigley and colleagues12 found a significant difference in levels of occurrence or worsening of otitis media between the treated group (29%) and the control group (13%), p = 0.037. Ear pain and ear disorder were reported as not differing between groups. Three girls discontinued rhGH due to hypertension, ulcerative colitis and brain tumour. The authors stated that these were not directly related to GH. Overall, AEs were not presented separately for the groups; however, five were reported to have accidentally overdosed on the study drug. Five further events described as possibly related to the study drug were hypertension (two), surgical procedures (two) and scoliosis (one).

Five participants were reallocated from the group receiving estrogen to rhGH after concerns over early breast development in the study by Johnston and colleagues. 90 Seven patients developed ‘coincidental disorders’ not severe enough to warrant treatment discontinuation. The authors reported that compliance problems led to the withdrawal of four patients, but no details were given. It is unclear which treatment groups these latter events occurred in.

Summary

Six trials examining the effectiveness of GH for growth disturbance in patients with TS met the inclusion criteria for the review.

The reporting and methodological quality of the studies was poor. Of the six included studies, one reported adequate randomisation to treatment groups,85 one study described adequate concealment of treatment allocation85 and one adequately blinded the patient to treatment by administering placebo. 89 None of the included trials used an ITT analysis.

Children in the rhGH group in the Stephure86 study grew an average of 9.3 cm more from baseline than those in the untreated group. In a study of younger children85 the difference was 7.6 cm. Both of these were statistically significant results. In the same two studies85,86 the groups receiving rhGH achieved a significantly higher HtSDS.

Change in height and change in HtSDS were statistically significantly greater in the groups treated with rhGH. 85,86,90

Growth velocity was greater in the treated groups in three studies that reported this outcome,12,85,86 although this was greater in the first year and fell in the second year in both treatment groups where this was reported separately. 85

One study86 found a significant difference in BA between groups, being higher in the treated patients.

Fat mass and LBM were reported in two studies. 88,89 In both, the total FM was at a lower level in the treated groups, compared with those untreated, and LBM was higher in the treated groups compared with untreated. There was no statistically significant difference in BMI between treated and untreated girls in one study. 85

The IGF-1 levels were substantially higher in the treated groups in the studies reporting this outcome. 88,89 IGF-1 SDS was also significantly higher in the group receiving GH. 85 Levels of IGFBP-3 and IGFBP-3 SDS were also found to be higher in children treated with GH. 85,88,89

Levels of fasting glucose and fasting insulin were both raised in the treated groups in two studies. 88,89

There were variable levels of detail in the reporting of AEs across the six studies. Two studies did not discuss these. 88,89 In those studies that did, no clear picture emerges. One found greater levels of AEs in the treated group,86 one found similar levels across groups,85 one found significantly higher levels of or worsening of otitis media, and one reported seven patients with ‘coincidental disorders’ and four withdrawals due to compliance problems, but gave no further details.

Quantity and quality of research available

Eight RCTs in 13 publications of the clinical effectiveness of rhGH in patients with PWS met the inclusion criteria for this review. 22,91–102 Their key characteristics are shown in Table 13 – see Appendix 4 for further details.

It was not possible to perform any meta-analysis of outcomes from the PWS studies due to variation in the trials’ participants’ ages, dosing calculations, and methods of presenting results. The included studies had well-matched patient groups, whose baseline characteristics were generally similar in the treated and untreated groups. Median baseline HtSDS was lower in the rhGH group than in the untreated group in the study reported by both Festen and colleagues94 and by de Lind van Wijngaarden and colleagues,93 although the interquartile ranges (IQRs) were similar [–2.0 (–3.1 to –1.7) versus –2.5 (–3.3 to –1.9), respectively]. Other exceptions were the crossover study by Haqq and colleagues,102 which presented baseline characteristics for the study population as a whole, and the study by Lindgren and colleagues,100,101 which reported slightly lower baseline GV SDS in the rhGH group [–1.9 ± 2.0, range –6.4 to –0.9, vs –0.1 (SD not reported) range –1.7 to –2.71].

Five of the studies were RCTs, which compared 1 mg/m²/day rhGH with no treatment for 122,92–98 or 291,93,94 years. The study by Haqq and colleagues102 was a crossover RCT, which compared 0.043 mg/kg/day of rhGH with placebo injections, with patients spending 6 months in each treatment arm. There does not appear to have been a washout phase between the two treatment phases, which could affect the generalisability of results.

The doses used in the included studies reflect the various marketing authorisations for this drug (0.035 mg/kg body weight or 1.0 mg/m² BSA), with 1 IU of rhGH being equivalent to approximately 0.33 mg/kg. The study reported by both de Lind van Wijngaarden and colleagues93 and Festen and colleagues94 reported results separately for infants and children. Two RCTs reported results for infants and toddlers aged between 1 and 2.5 years. 22,92,97,98 The five remaining trials were in children aged between approximately 6 and 10 years old. The studies were generally small, randomising between 14102 and 5495,96 children. The study reported by both de Lind van Wijngaarden and colleagues93 and Festen and colleagues94 had a total of 91 participants, but as children and infants were randomised separately, the randomised comparisons were of rhGH versus no treatment within two smaller groups (42 infants and 49 children). This was the only study to report a sample size/power calculation,93,94 and it is not clear whether the other studies were adequately powered to detect a difference between treatment groups.

With the exception of the two RCTs by Festen and colleagues,91,92 the studies did not clearly state which of their reported outcomes were primary or secondary measures of effect. Seven of the eight trials reported measures of body composition. The two RCTs by Festen and colleagues91,92 focused on body composition and biochemical markers, and did not report any measure of change in height. The other six studies all reported GV SDS or an indicator of linear GV. 102 IGF-1 and other biochemical markers were reported by five RCTs. 22,92–99

One RCT was reported in three papers, by Carrel and colleagues,22 Myers and colleagues97 and Whitman and colleagues. 98 The most complete data were reported by Carrel and colleagues,22 and these data are included in the tables in this section.

The included studies were generally poorly reported (Table 14) and lacked information on method of randomisation or concealment of allocation. It is possible that selection bias could have affected the trials if they were not properly randomised, but there is insufficient information provided on which to make such a judgement. The trial by Haqq and colleagues102 was a crossover study, and did not report baseline characteristics separately for the two groups. The other studies reported baseline characteristics, which indicated that patients in the two treatment groups were comparable at the start of the study. With the exception of the crossover trial by Haqq and colleagues,102 which had a placebo-injection group, the studies were open label, with the comparator groups receiving no treatment. Although this could have allowed a degree of bias in reporting and assessing results, measurement of objective outcomes, such as height gained, is less likely to be open to bias. Only two of the studies reported results on an ITT basis,91,96 so attrition bias could have affected the remaining studies.

The outcome measures for the included studies are shown in Tables 15–17. The p-values in the tables refer to between-group differences, as this is the comparison of interest for this report. Some of the studies reported statistical significance in change from baseline for each of the treatment groups individually, but not for between-group comparisons. To avoid confusion with the between-group comparison p-values, such results have not been included in the tables below and are not discussed in the text. The full data extraction tables in Appendix 4 include any statistical significance for change from baseline for individual treatment groups without between-group comparisons.

Growth outcomes

Changes in height and other growth outcome measures are shown in Table 15. The infants in the study by Carrel and colleagues22 who received rhGH for a year grew an average of 6.2 cm more than those in the untreated group (p < 0.001). None of the other studies reported change in height as an outcome measure.

Two studies reported a statistically significant difference in HtSDS at end of treatment between treated and untreated patients. 93–96 Treated patients in the study reported by both Carrel and colleagues95 and Myers and colleagues96 had a mean HtSDS of –0.6 ± 1.2 compared with –1.6 ± 1.2 in the untreated group (p < 0.01). The studies reported by de Lind van Wijngaarden and colleagues93 and by Festen and colleagues94 also reported statistically significant improvements in height for rhGH-treated infants and children compared with unmatched controls. The rhGH-treated infants in their study had a median HtSDS of –0.9 compared with –1.8 in the untreated patients (p = 0.003). This reflected a change from baseline HtSDS of +1.2 for treated infants and –0.2 for untreated infants (p < 0.0001). After 2 years of treatment with rhGH, children had a median HtSDS of –0.5 compared with –2.6 in untreated children (p < 0.001). 93

Festen and colleagues91 reported that the difference between the two groups was statistically significant at year 1 (year 1 HtSDS –1.3 vs –2.8, p < 0.01). At year 2, the difference between the two groups was even greater (–0.6 compared with –3.0 in the treated and untreated groups, respectively), but no p-value was reported. 91 The other five studies all reported that HtSDS values were higher in treated than in untreated children, but did not report whether or not differences between groups were statistically significant.

The five studies that used GV as an outcome measure all reported faster growth in the treated group than in the untreated group, although statistical significance for differences between groups was only reported for three of these. The mean GV in the studies reported by Carrel and colleagues95 and by Myers and colleagues96 was twice as fast in the treated children as in the untreated children (10.1 vs 5.0, p < 0.01). The corresponding mean GV SDS values were 4.6 in the treated children and –0.7 in the untreated children (p < 0.01), indicating faster than average growth in the treated group and slower than average growth in the untreated patients. Similarly, Hauffa and colleagues99 reported a positive GV SDS for treated children and a negative one for untreated children (5.5 vs –2.3, p = 0.0012). Haqq and colleagues102 calculated GV that was 3 cm/year faster in patients receiving rhGH than in patients in the placebo arm (7.5 vs 4.5, p < 0.05).

Two of the included studies reported BA as an outcome measure. There was no statistically significant difference in BA at follow-up between patients in the treated and untreated groups in the study reported by both Carrel and colleagues95 and by Myers and colleagues. 96 Lindgren and colleagues100,101 reported similar change from baseline in both groups (1.4 in the treated group, 1.5 in the untreated group), but did not report whether or not there was any statistical significance to their results.

Body composition

Seven of the trials reported changes in body composition, as shown in Table 16. 22,91–94,100–102 The trial by Hauffa and colleagues99 did not report any results but stated that there were no significant within- or between-group changes for BMI, skinfold thickness, waist or hip circumference.

Four of the trials reported a statistically significantly lower percentage of body fat in patients treated with rhGH than in those with no treatment or placebo. In the trial reported by Carrel and colleagues22 mean percentage body fat was 10% lower for treated patients than for untreated patients (p = 0.03). On average, treated patients in this trial experienced an approximately 5% reduction in body fat, compared with an average 4% increase in the untreated patients’ body fat (p = 0.001). The other two trials that found a statistically significant difference reported that treated patients had approximately 4% (Haqq and colleagues102) or 7% (Carrel95 and Myers96) less body fat than those in the comparator group. De Lind van Wijngaarden and colleagues93 did not report percentage body fat for infants, but did report this outcome for the children in their study who were over 4 years of age (n = unclear). Children who received rhGH for 1 year had a median percentage body fat SDS of 1.5, compared with 2.3 in the control group (p < 0.001). After 2 years of treatment, the SDS values were 1.9 versus 2.4 for the treated and untreated groups. respectively (p < 0.001).

Four trials reported that patients treated with rhGH had statistically significantly higher LBM93,95,96,102 or a larger improvement in LBM than untreated patients. 22 In the trial reported by Carrel and colleagues,22 treated patients’ LBM increased by 1.8 kg more than the improvement seen in the untreated group (3.6 vs 1.8 kg, p < 0.001). Treated patients in the other two studies had approximately 2 kg102 or 4 kg95,96 more LBM than their untreated counterparts (p < 0.05 and p < 0.01, respectively). De Lind van Wijngaarden and colleagues93 reported that change in trunk LBM was statistically significantly better for treated than for untreated infants (1.7 vs 0.7, respectively). For children, they reported SDS for LBM adjusted for age and height, as well as change in trunk LBM. All of these outcomes were statistically significantly better for treated children than for untreated children after both 1 and 2 years of treatment.

Six of the studies reported BMI, with mixed results. Festen and colleagues91 reported a BMI of 16.1 at year 1 for treated patients and 18.5 for untreated patients (p < 0.05), with similar results at year 2. Haqq and colleagues102 also reported a statistically significant difference of 1.6 in BMI (31.2 vs 32.8 for treatment phase vs placebo phase in a small crossover RCT, p < 0.05). By contrast, the RCTs reported by Carrel95 and Myers96 and by de Lind van Wijngaarden93 found no statistically significant difference between treated and untreated patients. Neither of the other RCTs that reported BMI gave a value for between-group statistical significance, and both treated and untreated patients had similar values. 92,100,101

There was no statistically significant difference in bone mineral density between treated and untreated patients in the study reported by Carrel and colleagues. 22 No statistically significant differences in progression of scoliosis or onset of scoliosis in either infants or children were reported by de Lind van Wijngaarden. 93

Biochemical and metabolic markers

The included studies reported a range of biochemical and metabolic markers, and key results are included in Table 17 – see Appendix 4 for further outcomes. For conciseness, only the key outcomes of IGF-1, IGFBP-3, insulin and glucose are discussed in the narrative summary below.

All of the RCTs reported IGF-1 values or IGF-1 SDS as an outcome measure, and found that levels were higher in rhGH-treated patients than in untreated children. Three studies reported that IGF-1 values were statistically significantly higher in rhGH-treated patients than in untreated patients. 22,95,96,102 Three studies reported that IGF-1 SDS values were statistically significantly higher in treated than in untreated patients. 91–94

The included studies had well-matched patient groups, whose baseline characteristics were similar in the treated and untreated groups. The only exception was the crossover study by Haqq and colleagues,102 which presented baseline characteristics for the study population as a whole, and the study by Lindgren and colleagues,100,101 which reported slightly lower baseline GV SDS in the rhGH group [–1.9 ± 2.0, range –6.4 to –0.9, vs –0.1 (SD not reported) range –1.7 to –2.71].

Three of the RCTs reported IGFBP-3 values,93,95,96 and these were higher in treated patients than in untreated patients. In the trial reported by Carrel95 and Myers,96 patients treated with rhGH had a mean level of 3.5 mg/ml compared with 2.07 in the untreated patients (p < 0.01). Haqq and colleagues reported mean values of 6029 ng/ml in the treated patients and 4247 ng/ml in the untreated patients (p < 0.01). 102 Treated children and infants in the study reported by de Lind van Wijngaarden and colleagues93 had higher IGFBP-33 values than untreated children, although no p-values were reported for between-group comparisons.

The three studies that reported IGFBP-3 SDS found positive values in the treated children, with SDS of 0.492,93 and 0.5 (year 1) or 0.6 (year 2). 91,93 In comparison, untreated patients’ median scores were between –2.491,93 and –3.192 in year 1 and between –1.793 to –1.891 in year 2. Differences between treated and untreated patients were statistically significant in all three studies (p < 0.05,92 p < 0.001,93 p < 0.00191).

The RCT reported by Carrel and colleagues22 reported that there was no statistically significant difference in fasting insulin levels between the treated and untreated infants in their study (5.6 vs 5.7 μIU/ml, respectively). Two other studies91,95,96 reported slightly higher insulin levels in treated patients, but did not report p-values. The study by Haqq and colleagues102 reported very similar levels in both treated and untreated patients. Glucose levels appeared to be similar in both treated and untreated patients in the two studies that presented this as an outcome, but neither study reported any p-values. 91,102

Quality of life

None of the included studies reported a measure of HRQoL.

Adverse events

None of the studies reported AEs in any detail. Neither of the studies reported by de Lind van Wijngaarden and colleagues93 and by Festen and colleagues94 nor the one reported by Festen and colleagues91 reported on AEs at all. In the other study by Festen and colleagues,92 the paper stated that rhGH treatment did not induce disadvantageous effects on carbohydrate metabolism, sleep-related breathing disorders or thyroid hormone levels. Hauffa and colleagues99 reported that one patient in the rhGH group developed pseudotumour cerebri after increasing the starting dose to the final dose, but their symptoms resolved on discontinuation. No abnormalities of glucose regulation were observed in either group. None of the patients in the study reported by Carrel and others95,96 experienced pseudotumour cerebri. Two of their patients who received rhGH experienced headaches within the first 3 weeks, but these resolved with temporary stoppage and gradual reinstitution of treatment.

Carrel and colleagues22 commented that there was no evidence of changes in the prevalence of scoliosis with rhGH treatment, although another paper reporting the same study reported that there was progression of scoliosis in one patient. 97 Lindgren and colleagues100,101 and Haqq and colleagues102 reported that there was no severe progression of scoliosis (angle ≥ 20°) during their RCTs.

Lindgren and colleagues100,101 noted that one child in their study developed low levels of thyroxine without any change in TSH levels. He received substitution with l-thyroxine during the rhGH treatment. Carrel and colleagues22 commented that no child in their RCT required thyroid hormone therapy. Haqq and colleagues102 reported that only one patient required thyroid hormone replacement while receiving rhGH treatment.

Summary

The evidence for the clinical effectiveness of HGH as a treatment for PWS comes from eight small RCTs (one crossover trial and seven parallel group trials), reported in 13 publications. The included studies were generally poorly reported and only two91,96 presented results on an ITT basis.

Only one of the studies reported changes in height. Infants who received rhGH for 1 year grew an average of 6.2 cm more than those in the untreated group (p < 0.001). 22 Two studies reported a statistically significant difference in HtSDS between treated and untreated patients. The difference was 1 SDS (favouring rhGH treatment) in one study,95,96 and > 2 (year 2) in the other. 93

Treated patients grew 3 cm/year faster than untreated patients in one RCT102 and 5 cm/year faster in another. 95,96 Another study reported a positive GV SDS for treated patients and a negative one for untreated children (5.5 vs –2.3). 99 The differences between groups were statistically significant in all three studies.

Two of the included studies reported BA as an outcome measure, and this was similar in both treatment groups. 95,96,100,101

Four trials reported a statistically significantly lower percentage of body fat (between 1%93 and 10%22 lower) in patients treated with rhGH than in those with no treatment or who were given placebo.

Three trials reported that patients treated with rhGH had statistically significantly higher LBM,95,96,102 or a larger improvement in LBM, than untreated patients. 22 One study reported that LBM SDS was significantly better in treated than in untreated children. 93

Two studies found that BMI was statistically significantly lower in treated patients than in untreated patients. 91,102 However, another RCT95,96 found no statistically significant difference between the two groups, and three more studies did not report a p-value for between-group statistical significance. 92,93,100,101

Insulin-like growth factor-1 values were statistically significantly higher in patients treated with rhGH than in untreated patients in three studies.

Two RCTs reported IGFBP-3 values that were statistically significantly higher in treated patients than in untreated patients. 95,96,102 Three studies91–93 reported positive IGFBP-3 SDS values in treated patients and negative values in untreated children; differences between the groups were statistically significant.

Four of the studies reported insulin levels, with varying results. One study22 reported that there was no statistically significant difference between treated and untreated infants. Insulin levels in another study95,96 appeared to be considerably higher in treated patients than in untreated patients. Another study91 reported higher insulin levels in treated patients at year 1 but lower levels than in untreated patients at year 2. Similar values in both groups were also reported. 102

None of the included studies reported a measure of HRQoL.

None of the studies reported AEs in any detail.

Quantity and quality of research available

Six RCTs of patients with CRI met the inclusion criteria for this review,103–108 and their key characteristics are shown in Table 18 – further details are shown in Appendix 4. The inclusion criteria for this systematic review specified that children should be prepubertal. Five of the studies stated in their inclusion criteria that patients should be prepubertal/Tanner stage 1, but one study included both prepubertal and pubertal patients. 107 However, we have included outcome measures from this study where data were presented separately for prepubertal children and pubertal children.

The included RCTs were of different designs (two crossover and four parallel group). Three of the parallel-group RCTs were open label, with the comparator groups receiving no treatment,103,106,107 and one was placebo controlled. 108 The two crossover studies104,105 had placebo and treatment phases. There does not appear to have been a washout phase in either of the crossover trials, so a carry-over effect could have affected results. The doses all appeared to correspond to those specified in the marketing authorisation, but dosages were reported differently, with some using IUs and others using mgs, and some using doses based on weight, whereas others used surface area. Randomised treatment duration was 6 months in the two crossover trials,104,105 2 years in one study108 and 12 months in the other studies.

Three of the studies investigated rhGH treatment in children who had received a kidney transplant at least 1 year before starting the study103,105,107 and the other three studied children who had CRI. 104,106,108 There was considerable variation in the age of children in the included studies, ranging from 5.6106 to 12.6107 years old. Two of the studies were relatively large (n = 203107 and n = 125108), one was of medium size (n = 69106), and the remaining three were rather small (n = 23,103 n = 20104 and n = 11105).

Only one study107 specified a primary outcome. The Pharmacia and Upjohn Study Group107 designed their study to test GFR, with GV and HtSDS being used as secondary outcomes. The other studies reported various outcomes relating to growth, body composition and biochemical/metabolic markers, but did not specify which were primary outcomes. Only Sanchez and colleagues103 mentioned a power calculation, and this appears to have been based on bone formation rates in a previous study, so it is not clear what the primary outcome was for the included study. The lack of clarity around primary outcomes and power calculations, together with the small size of three of the studies,103–105 suggests that the trials may have been underpowered to detect differences in outcomes relating to growth and body composition.

The included studies had well-matched patient groups, whose baseline characteristics were similar in the treated and untreated groups.

None of the included RCTs provided clear information on method of randomisation or concealment of allocation (Table 19), so it is not possible to say whether or not selection bias may have affected these studies. The studies all reported eligibility criteria, and presented baseline characteristics that indicated that groups (within trials) were similar at the start of the studies.

The studies gave little information on whether or not outcome assessors were blinded to patients’ treatment groups, although Sanchez and colleagues103 did comment that skeletal radiographs were reviewed by a single observer who had no information about patients’ clinical condition or treatment status. In addition, three of the trials gave patients in the comparator group no treatment, so it would have been clear to patients and their care providers whether or not they were receiving rhGH. In three trials, patients in the comparator group had placebo injections. It is not clear whether or not their care providers were also blinded to treatment group. Lack of blinding could have led to performance bias in measuring treatment effect, but the objective nature of outcomes such as height change and GV would have protected against bias to a certain degree.

All the studies presented results as mean values with SDs or standard errors to give a measure of variability. The studies all provided adequate details of any patients who withdrew from the study, but only one study105 presented results on an ITT basis (no patients withdrew from this study). Attrition bias could therefore have affected the results of the non-ITT studies, i.e. if there had been unbalanced and selective withdrawal from different treatment groups within a study, or if particular patients were more likely to withdraw or be excluded from the analysis.

There was a statistically significant difference between treated and untreated children’s birth length SDS in one study,111 but baseline height was the same in both groups. The very small study by de Zegher and colleagues113 reported slightly lower baseline GV in treated compared with untreated children [5.1 (range 4.0–6.8) vs 6.4 (range 5.3–7.5) cm/year, respectively]. Otherwise, the studies’ treatment groups were generally comparable at baseline, with no discernible differences between treated and untreated patients.

The outcome measures for the included studies are shown in Tables 20–22 below. The p-values in the tables refer to between-group differences.

Growth outcomes

Key growth outcome measures are shown in Table 20 – see Appendix 4 for other outcome measures. Only one of the included studies reported height gain. Powell and colleagues106 found that treated children grew an average of 3.6 cm more than their untreated counterparts after 1 year of treatment (9.1 cm vs 5.5 cm, p < 0.0001). All children in the study by The Pharmacia and Upjohn Study Group107 experienced an improvement in HtSDS, but this was statistically significantly higher in the children treated with rhGH than in the untreated children (0.6 vs 0.1, p < 0.0001). RhGH-treated children in the study by Powell and colleagues106 had a statistically significantly higher HtSDS at end of 12 months than untreated children (0.8 vs 0.0, p < 0.0001).

One of the six studies reported change in GV, and this was statistically significantly faster in treated than in untreated children. 107 Four studies reported GV at end of treatment, all reporting statistically significantly faster growth in children who received rhGH treatment than in untreated children. 103–105,107,108 The 2-year study by Fine and colleagues108 reported that rhGH-treated patients’ GV in the first year was 4.2 cm/year faster than the untreated patients’ (p < 0.00005). The difference between the two groups was less in the second year (2.3 cm/year faster in rhGH-treated children) but the difference between groups was still statistically significant (p < 0.00005) when comparing the difference in change from baseline in those patients who completed 2 years of the study. A statistically significant difference in GV between groups of just over 3 cm/year was reported by both The Pharmacia and Upjohn Study Group107 (3.4 cm/year difference, p < 0.0001) and by Sanchez and colleagues103 (3.2 cm/year difference, p < 0.01).

The two crossover studies by Hokken-Koelega and colleagues104,105 also reported statistically significantly faster growth velocities in patients during the rhGH phase compared with the placebo phase, with an average of 2.9 cm/6 months difference in velocity. In the study of children with CRI, patients who received rhGH followed by placebo grew at an average velocity of 5.2 cm/6 months during treatment compared with 1.5 cm/6 months in the placebo phase. Patients who received placebo followed by rhGH grew 2.4 cm/6 months during the placebo phase compared with 4.4 cm/6 months in the treatment phase. The overall mean effect of rhGH was statistically significant (p < 0.0001). Statistical tests showed that there was no significant carry-over effect (–0.04 cm/6 months, p = 0.94). The crossover study in children who had received a renal transplant had similar results. Patients grew, on average, 3.8 cm/6 months faster during the active treatment phase in the group who received rhGH followed by placebo, and 2 cm/6 months faster in the active treatment phase for patients who received placebo followed by rhGH (p < 0.0001 for overall effect of rhGH vs placebo). 105 Hokken-Koelega and colleagues reported that there was no significant carry-over effect (0.5 cm/6 months, p = 0.30). 104,105

The two crossover trials,104,105 but none of the parallel-group RCTs, reported GVSDS. Both trials reported positive SDS values during the active treatment phases and negative scores during the placebo phases. The reported difference in scores between active treatment and placebo phases in the trial of children with chronic renal failure (CRF) was 7.7 (p < 0.0001),104 and in the trial of children who had received a renal transplant the difference was 8.0 (p < 0.0001). 105

Bone age was reported by five of the six studies. The studies by Powell and colleagues106 and Sanchez and colleagues103 reported that there was no statistically significant difference in BA between the treated and untreated patients. The two crossover studies by Hokken-Koelega and colleagues reported small differences with slightly lower mean ages for rhGH overall compared with placebo (mean differences –0.01 years104 and –0.5 years105) but did not present any p-values for these comparisons. Fine and colleagues108 reported that the change in BA between baseline and 2 years was greater in patients treated with rhGH than in untreated patients for those who completed both years of the study (2.3 vs 1.6 years, p = 0.0001).

Body composition

Measures of body composition were reported by three of the studies, and selected outcomes are shown in Table 21. 103,106,108 Other outcomes are tabulated in the data extraction forms in Appendix 4. Children treated with rhGH gained statistically significantly more weight than those in the control groups in the studies reported by Fine and colleagues108 (2.1 kg more in 2 years, p = 0.0004) and by Powell and colleagues106 (1.3 kg more in 1 year, p = 0.007). However, there was no statistically significant difference between groups in change in weight for HtSDS. Sanchez and colleagues103 did not report actual weight gain, but reported a statistically significant difference in change in SDS for weight that favoured treatment with rhGH (0.2 vs –0.3, p < 0.01). Although Powell and colleagues106 reported a statistically significantly greater weight gain in treated patients, the weight for HtSDS was the same for both groups (0.4, p = 0.8703).

Biochemical markers

The included studies reported a range of biochemical and metabolic markers, and these are included in Table 22. For conciseness, only the key outcomes of IGF-1, IGFBP-3, insulin and glucose are discussed in the narrative summary below. In addition, the studies reported a range of markers related to liver function. These are not reported in Table 22 or discussed in the narrative summary below, but are included in the data extraction forms in Appendix 4. No data from Sanchez and colleagues are included in Table 22 as their results focussed on liver function and did not report IGF, insulin or glucose.

Four studies reported IGF-1 as an outcome measure,104–106,108 and levels were higher in treated patients than in untreated patients. IGF-1 values were statistically significantly higher in treated patients at both years 1 and 2 in the study by Fine and colleagues108 (p = 0.0004 and p = 0.0001, respectively), but only approximately one-half of the randomised patients were included in this analysis. Powell and colleagues106 also reported that IGF-1 and IGF-1 SDS values were statistically significantly higher for treated patients than untreated patients (p < 0.006). 106 The two crossover studies by Hokken-Koelega and colleagues104,105 reported that IGF-1 SDS for BA was statistically significantly higher for treated than for untreated patients (2.7 higher in treated children with CRF104 and 3.7 higher in treated children who were post transplant,105 p < 0.0001 for both).

Three studies reported IGFBP values,104–106 and in all three the IGFBP-3 values were higher in the treated patients. Powell and colleagues106 reported that IGFBP-3 and corresponding SDS values were statistically significantly higher in treated patients than in untreated patients (p < 0.011). Hokken-Koelega and colleagues104 reported that the IGFBP-3 SDS for bone age was statistically significantly higher for treated patients (p < 0.0001).

Fine and colleagues108 reported that fasting insulin levels were statistically significantly higher in rhGH patients than in untreated patients after 2 years (p = 0.03). Similarly, Hokken-Koelega and colleagues105 reported slightly higher insulin values in treated children, but did not present p-values.

Quality of life

Five of the included studies did not report QoL as an outcome measure. One study107 reported QoL but did not present data for prepubertal patients (the licensed patients) separately from pubertal patients, so it is not discussed here.

Adverse events

Hokken-Koelega and colleagues105 reported that no patients in their study had an acute rejection episode, and that there were no SAEs. Sanchez and colleagues103 reported that two patients with normal rates of bone formation experienced acute rejection episodes after 3 and 12 months of rhGH therapy. One of these episodes was associated with non-compliance to immunosuppressive medications and both reversed after treatment with methylprednisolone. There were no rejection episodes in untreated patients.

Fine and colleagues108 reported that there were no differences between groups in year 1. In the second year, eight of 55 rhGH patients experienced asthma or wheezing, but all episodes were preceded by upper respiratory tract infections. Fine and colleagues108 reported that there were no clinically significant side effects associated with rhGH treatment. Hokken-Koelega and colleagues104 reported that serum alkaline phosphate was significantly increased during rhGH treatment, but returned to pretreatment levels when rhGH therapy was replaced by placebo (p < 0.0001). There was no significant change in parathyroid hormone concentration during either treatment schedule, and thyroid function was reported to have been normal. The Pharmacia and Upjohn Study Group107 did not present AEs separately for prepubertal and pubertal children, so no data are reported here. Powell and colleagues106 did not report AEs from their study. 106

Summary

The evidence for the clinical effectiveness of rhGH as a treatment for short stature owing to CRI comes from six RCTs, two of which were crossover trials. The trials were generally poorly reported, and only one105 presented ITT results. Three of the studies had fewer than 25 participants, which suggests that the trials may have been underpowered to detect differences in outcomes relating to growth and body composition.

One study reported that rhGH-treated patients grew an average of 3.6 cm more than their untreated counterparts after 1 year of treatment. Two studies reported that HtSDS was statistically significantly better in treated children than in untreated children.

Five studies reported that change in GV or GV SDS was statistically significantly faster for children who received rhGH treatment than for untreated children, with between-group differences in velocity ranging from 3.2 cm/year103 to 4.2 cm/year108 in the parallel-group trials.

Two studies reported that there was no statistically different difference in BA between the treated and untreated patients. Two reported small differences with slightly lower mean ages for rhGH overall compared with placebo, but did not present any p-values for these comparisons. One study reported that the change in BA between baseline and 2 years was greater in patients treated with rhGH than in untreated patients for those who completed both years of the study.

IGF-1 levels were statistically significantly higher in treated patients than in untreated patients in two of the four studies that reported this outcome.

Three studies reported that IGFBP-3 values were higher in the treated patients. Only one of these reported that differences between groups were statistically significant.

Insulin levels were statistically significantly higher in children receiving rhGH than in those receiving placebo injections or no treatment.

Four studies presented data on AEs. Two rhGH-treated patients in one study experienced acute rejection episodes (one associated with non-compliance to immunosuppressive medications) but both reversed after treatment with methylprednisolone. There were no SAEs reported.

Quantity and quality of research available

In the UK, rhGH is licensed for use in children born SGA who are over 4 years of age, have a current HtSDS of < 2.5, with a parental adjusted HtSDS of –1, had a birth weight and/or length SDS of < –2, and have failed to show catch-up growth during the previous year (HV SDS < 0). No RCTs meeting these criteria were identified. Following discussion with NICE, the criteria were amended in order to include evidence from RCTs on rhGH. As discussed above (see Inclusion criteria), the following amended criteria were agreed: growth disturbance (current height < –2.5, no reference to parental height), birth weight and/or length < –2 SD and failure to show catch-up growth (no stated criteria) by the age of 3 years.

Six studies109–114 met the amended inclusion criteria for this review, and their key characteristics are shown in Table 23 – see Appendix 4 for further details. In the UK, the licensed dose of rhGH for SGA children is 0.035 mg/kg/day, which equates to 0.105 IU/kg/day. Only the study by Phillip and colleagues114 included a treatment arm with the licensed dose; the other studies all used approximately two or three times the UK licensed dose.

Treatment duration was comparable across five of the six included studies. Four of the trials stated a treatment duration of 2 years. 109,110,112,113 Carel and colleagues111 administered GH for an average of 2.7 ± 0.6 years, until the participants reached AH. The children in the study by Phillip and colleagues114 received treatment for 2 years, but only the first year allowed a randomised comparison between GH and no treatment.

The mean age of participants was similar both across groups within studies and across five of the six trials included. 109,110,112–114 The mean ages of groups in these trials ranged from 4.7 (2.3–6.3)113 to 6.3 (4.0–8.0) years. The Carel study111 included older children with mean ages of 12.7 ± 1.4 in the rhGH group and 12.8 ± 1.6 in the control group.

The six included trials were generally of poor methodological quality (Table 24).

Phillip and colleagues114 reported that a centralised computer-controlled system was used to randomly assign children to groups. In the other five trials it was unclear whether the assignment to treatment groups was really random. This was reflected in the assessment of whether treatment allocation was concealed, with one exception being the study by Carel and colleagues,111 which reported that group assignment was not masked and this was therefore judged to be inadequate.

The blinding of outcome assessors can defend against bias affecting the measurement of some outcomes. In two trials112,114 outcome assessors for BA were blinded to chronological age and treatment allocation. It was not stated whether this extended to assessors of other outcomes. In the remaining four trials it was not stated whether the outcome assessors were blinded.

Performance bias, where knowledge of treatment can potentially lead to differences in care provided can be protected against by blinding care givers and patients. The care provider was not blinded to treatment in the studies by Carel and colleagues111 or Phillip and colleagues,114 and in the four remaining trials this was unknown. In each of the six trials blinding of the patient was inadequate as no placebo was used. Only one of the trials conducted an ITT analysis. 113 This guards against bias arising where for example only the results of patients who did not experience AE or compliance issues are included in the analysis.

Growth outcomes

All six studies109–114 reported growth outcomes, and these are presented in Table 25.

Carel and colleagues111 reported AH for the 70% of control patients and 89% of treated patients for whom these data were available. They reported a mean gain in AH of 26 ± 7 cm in the treated group compared with 22 ± 6 cm in their untreated group (p = 0.005). They also reported AH SDS, which was statistically significantly higher in the rhGH treated group (–2.1 ± 1.0) compared with the untreated group (–2.7 ± 1.0), p = 0.005. Similarly, the SDS for AH total gain was statistically significantly higher in treated patients than with untreated patients (1.1 ± 0.9 vs 0.5 ± 0.8, p = 0.002). Carel and colleagues111 also reported the difference from target HtSDS. This was statistically significantly lower in the group receiving GH, compared with the control group (–0.9 ± 1.2 vs –1.7 ± 1.2, p = 0.005).

Children who received the licensed dose of 0.033 mg/kg/day for 1 year in the study by Phillips and colleagues114 gained an average of 3.3 ± 0.2 cm in height compared with children in the untreated control group. Those receiving the higher dose of 0.1 mg/kg/day rhGH gained an average of 6.5 ± 0.2 cm compared with untreated children. No p-values were presented for between-group comparisons, although the CIs suggest a statistically significant difference.

de Zegher and colleagues112 found that gain in HtSDS at the end of the study was higher in the group receiving a higher dose [2.1 ± 0.1 (0.2 IU/kg/day) vs 2.5 ± 0.1 1 (0.3 IU/kg/day) vs 0.2 ± 0.1 (untreated), p < 0.001 treated vs untreated groups]. The other study by de Zegher and colleagues113 reported higher HtSDS in treated patients, but did not present p-values. However, there were only four patients in the no-treatment group, so between-group comparisons are difficult.

Phillips and colleagues114 found that HtSDS was higher in the two rhGH-treated groups than in the untreated groups (–2.3 ± 0.6, –1.8 ± 0.8 and –3.0 ± 0.6 for the 0.033 mg/kg/day (licensed dose), 0.1 mg/kg/day and untreated groups, respectively). These scores reflected changes of 0.8 and 1.4 in SDS for the licensed- and high-dose groups, respectively, compared with a change of only 0.1 in the untreated patients’ mean SDS value.

Three109,110,113 of the included studies that used higher doses of rhGH reported that HtSDS was higher in the treated groups than in the untreated groups. De Schepper and colleagues109 and de Zegher and colleagues113 reported HtSDS at the end of the first and second years of treatment. In each of these studies, at both time points, the SDS was higher in the treated group, and this difference between groups increased in the second year. In De Schepper and colleagues’109 study at the end of year 1, HtSDS in the treated group was –2.1 ± 0.7 versus –3.1 ± 1 in the untreated group (p < 0.0001). In year 2, HtSDS in the treated group was –1.7 ± 0.7 compared with 3.1 ± 1 in the untreated group (p < 0.0001). At the end of 2 years’ treatment, the treated group in the Lagrou110 study had a statistically significantly higher mean HtSDS (–1.9 ± 0.7) than the untreated group (–3.1 ± 0.9), p < 0.001.

Two studies109,110 were suitable for meta-analysis of the HtSDS outcome because they were sufficiently homogeneous in terms of dose, duration of treatment, and the children’s mean age at start of treatment. However, both trials were small (≤ 20 girls in each treatment group), which affects the validity of tests for heterogeneity, and both used twice the licensed dose, so a meta-analysis of these was considered unlikely to add to the evidence base.

Growth velocity (cm/year) was greater at the end of year 2 in the groups receiving rhGH in the two studies that presented results for this outcome. 112,113 de Zegher and colleagues 1996112 found an increased GV in their group receiving a higher dose of GH, and a greater GV for their treated participants overall: 10.2 ± 0.2 (0.2 IU/kg/day) versus 11.0 ± 0.4 (0.3 IU/kg/day) versus 5.7 ± 0.3 (untreated), p < 0.001 untreated versus treated. The de Zegher 1996 study112 also found that GV SDS was statistically significantly higher at the end of treatment in the treated groups [4.3 ± 0.3 (0.2 IU/kg/day) and. 5.2 ± 0.4 (0.3 IU/kg/day)] compared with –0.9 ± 0.3 in the untreated group (p < 0.001 for untreated vs treated groups).

de Zegher and colleagues 1996112 reported BA. The gain in BA (years) was statistically significantly greater in the groups receiving GH than in those who were untreated. The 0.2 IU/kg/day rhGH group had a mean gain of 1.35 ± 0.16, compared with 1.33 ± 0.24 in the 0.3 IU/kg/day rhGH group and 0.84 ± 0.07 in the untreated group (p < 0.001 treated vs untreated groups). This is reflected in the gain in HtSDS for BA: 1.0 ± 0.2 (0.2 IU/kg/day) versus 1.2 ± 0.4 (0.3 IU/kg/day) versus 0.0 ± 0.3, p < 0.05, treated versus untreated groups.

Body composition outcomes

Four of the included studies reported body composition outcomes. 109,110,112,113 These results are shown in Table 26. It should be noted that all of these studies used higher doses of rhGH than the UK licensed dose.

De Schepper and colleagues reported a WtSDS for treated patients that was almost half that for untreated patients (–1.8 vs –3.4; p < 0.0001). Lagrou and colleagues110 found that WtSDS at the end of year 2 was statistically significantly higher in their treated group (–2.3 ± 1.2) than in their untreated group (–3.7 ± 1.5; p < 0.01). Similar values were reported by de Zegher and colleagues,113 although no p-values were given.

de Zegher and colleagues 1996112 also reported gain in WtSDS and weight gain (kg). For both of these outcomes the difference was statistically significant and higher in the groups treated with GH. Mean weight gain (kg) was 6.9 ± 0.6 (0.2 IU/kg/day) versus 7.8 ± 0.5 (0.3 IU/kg/day) versus 3.6 ± 0.4 in the untreated group (p < 0.001 treated vs untreated groups). This pattern was reflected in the gain in WtSDS, which was 1.3 ± 0.1 in the 0.2 IU/kg/day group, 1.8 ± 0.1 in the 0.3 IU/kg/day group and 0.4 ± 0.1 in the untreated group (p < 0.001 untreated vs treated groups).

Lean mass and FM were reported in kilograms and as a percentage by De Schepper and colleagues. 109 Lean mass (kg) increased from year 1 to year 2 in both groups, and was greater in the group receiving GH at both times (13.2 ± 3.4 vs 10.9 ± 2.4 and 15.5 ± 3.4 vs 12.2 ± 2.5 for years 1 and 2, respectively). The p-value was reported as p < 0.0001, but it is unclear at which time point this p-value refers to. Lean mass (%) remained virtually unchanged from year 1 to year 2, but was higher in the rhGH group (82 ± 3 vs 77 ± 5 at year 2). The difference between treated and untreated groups was statistically significant (p < 0.05), but it is unclear whether this refers to the year 1 or year 2 data.

The difference in FM (%) between the two groups was statistically significant: 15 ± 2 versus 20 ± 5, p < 0.05.

Two studies reported BMI SDS. 110,113 One of these reported that there was no statistically significant difference between treated and untreated children,110 and the other reported similar values but gave no p-value. 113

Biochemical markers

Two of the included studies, both of which used higher doses than the UK licensed dose, reported biochemical markers. 112,114 These results are shown in Table 27.

Serum IGF-1 levels were statistically significantly higher in rhGH treated groups at the end of treatment. In one study,112 children receiving 0.2 IU/kg/day rhGH had values of 332 ± 29, compared with 655 ± 69 in the 0.3 IU/kg/day group and 168 ± 46 in the untreated group (p < 0.01, 0.2 IU/kg/day vs untreated group) after 2 years’ treatment. Phillip and colleagues114 reported similar IGF-1 values as de Zegher and colleagues112 after 1 year’s treatment, and, in addition, reported that IGF-1 SDS was higher in rhGH treated patients than in untreated patients. Values were 0.9 ± 1.9 and 3.3 ± 2.1 in the low- and high-dose groups, respectively, and 0.9 ± 1.2 in the untreated group.

Serum IGFBP-3 levels were also greater in the groups receiving rhGH. In the 1-year study,114 values were lowest in untreated patients (3.9 ± 1.1 µg/l) and higher in the two rhGH groups (4.8 ± 1.1 and 6.1 ± 1.4 for the low- and high-dose groups, respectively). No p-values were reported. At the end of year 2 in the second study, mean values were 6.10 ± 0.35 in the 0.2 IU/kg/day rhGH group, 6.50 ± 0.52 in the 0.3 IU/kg/day rhGH group and 4.00 ± 0.58 in the untreated group (p < 0.001, untreated vs 0.2 IU/kg/day rhGH group). 112

Quality of life

None of the included studies reported QoL outcomes.

Adverse events

Four of the included studies discussed AEs in varying detail. 109,111,112,114

Carel and colleagues111 found that 44% of patients reported AE, with 10% of these reporting four or more. It was not stated whether these patients were from the treated or untreated group. The authors described two AEs that they believed to be causally related to treatment; one slipped capital epiphysis after 1.5 years of treatment and one simple seizure episode 10 minutes after first injection. The authors do not state if these led to withdrawal. Sixteen severe AEs in 14 patients were reported. These were not thought by the authors to be related to treatment, and included trauma, psychiatric symptoms, abdominal symptoms, otitis, asthma, variocele, striae and migraine. De Schepper and colleagues109 stated only that no participants ‘had a noteworthy adverse event during the two years of study’. No further details were given.

de Zegher and colleagues 1996112 reported four SAEs. The authors suggested that these might not be linked to GH, but gave no further details. The authors described two treated children versus one untreated child hospitalised as a result of viral disease (group/dose not reported). There was one case of aggravated cutaneous eczema reported in group 1 (0.2 IU/kg/day). Three treated children (group/dose not given) reported possible increase in size or number of pigmented naevi. Treatment was not interrupted in any of these cases.

Phillip and colleagues114 reported AEs only for the 2-year study overall, so it was not possible to compare the treated and untreated children. The majority (349/358) of AEs in the study were of mild to moderate severity, the most common events (57%) being childhood infections. Of 16 SAEs reported, three were described as likely to be related to rhGH. Two of these (convulsions and papilloedema) resolved on discontinuation of treatment, and the third (epilepsy) stabilised when treatment was withdrawn.

Summary

Six109–113 trials examining the effectiveness of GH in children born SGA met the inclusion criteria for the review. The quality of the included studies was generally poor, and only one undertook an ITT analysis. 113 All but one114 of the trials used higher than licensed doses of rhGH.

One trial reported total gain in AH, and found this was approximately 4 cm higher in people who had received rhGH. The difference between groups was statistically significant (p < 0.005). 111 AH gain SDS was also statistically significantly higher in people who had received rhGH. 111 However, the study used a dose which was approximately twice the licensed dose, and it was carried out in children with a mean age of 12.7 years at start of treatment. This may limit the generalisability of the trial.

One study114 reported that patients who received 0.033 mg/kg/day of rhGH (the licensed dose) gained an additional 3.3 cm height compared with untreated children, and those who received 0.1 mg/kg/day gained 6.5 cm of additional height after 1 year’s treatment.

Height SDS was found to be statistically significantly higher in children treated with GH in two studies,109,110 and higher, but with no reported p-value, in two others. 113,114

Growth velocity (cm/year) was greater in the treated groups at the end of year 2 in the two studies that reported this outcome,112,113 but the difference was reported to be statistically significant in only one. 112

Weight standard deviation score was statistically significantly higher in children treated with rhGH in one110 of the three studies reporting this outcome.

Lean mass was reported in one study,109 and was statistically significantly greater in the treated group. Two studies reported BMI SDS. 110,113 One of these reported that there was no statistically significant difference between treated and untreated children,110 and the other reported similar values but gave no p-value. 113

One study112 reported that serum IGF-1 and IGFBP-3 levels were statistically significantly higher in patients treated with rhGH, and another114 reported similar results but did not present p-values.

Reporting of AEs was limited in detail, and only reported by four of the trials. 109,111,112 One trial111 reported two events in treated children that may have been linked to GH. They did not discuss if these led to discontinuation of the drug. A second trial109 reported only that there were ‘no noteworthy’ AEs recorded. A third trial112 reported four SAEs, which were not linked to the study drug. Three of 16 SAE in another trial114 were linked with rhGH, and these resolved/stabilised once treatment was discontinued.

Quantity and quality of research available

Only one study of patients with SHOX met the inclusion criteria for this review,49 and its key characteristics are shown in Table 28. The 2-year multicentre RCT by Blum and colleagues49 compared a daily injection of 50 µg of rhGH with no treatment in 52 prepubertal children with confirmed SHOX-D. The manufacturer’s recommended dose is 45–50 µg/kg body weight,81 but as the study did not report mean baseline weight of participants it is not possible to comment on whether or not the study reflects the licensed dose. The study also included a non-randomised rhGH-treated group of patients with TS, but this group will not be discussed further in this report.

The included study was generally poorly reported (Table 29), with little information on method of randomisation or concealment of allocation. Patients in the comparator arm received no treatment, so the patients themselves and their care providers would have been aware of whether or not they were receiving the study drug. The patients in the two groups had similar baseline characteristics, although target HtSDS was statistically significantly lower for the rhGH group (–1.3  ± 1.0 vs –1.5 ± 0.9, p = 0.013). Baseline IGFBP-3 SDS was slightly higher for the rhGH group (0.6 ± 1.3 vs 0.1 ± 1.1), although the difference was not statistically significant. The analysis was not reported on an ITT basis as one discontinuing patient was excluded from the analysis. The study did not include discussion of sample size or a power calculation, so it is not possible to determine whether or not it was adequately powered to detect a difference in the primary outcome (first-year GV).

Growth outcomes

Table 30 shows growth outcomes at the end of 2 years’ treatment. Children treated with rhGH gained approximately 6 cm more height than those in the control group (p < 0.001). Although all children remained below average height, the HtSDS was statistically significantly lower in the untreated group (–3.0 ± 0.2 vs –2.1 ± 0.2, p < 0.001). Blum and colleagues49 also commented that 41% of rhGH-treated patients reached a height within the normal range for age and gender (> –2.0 SDS), compared with only one patient in the untreated group. There was no statistically significant difference between the groups in catch-up of BA.

The difference in GV (1.9 cm/year) between the two groups during the second year of the study was statistically significant (p < 0.001). Children in the rhGH group had a positive HV SDS, i.e. their GV was above average for their age group. By comparison, those in the untreated group had a negative score, indicating slower growth than normal for their age group. Again, the difference between the groups was statistically significant (p < 0.001).

Body composition

The included study did not report body composition as an outcome measure.

Biochemical markers

Blum and colleagues49 did not report biochemical outcomes in any detail. However, they did state that IGF-1 SDS values were in the low-normal range for both groups at baseline, but increased to the upper-normal range in the rhGH-treated group. In 10 (37%) of the rhGH-treated children, IGF-1 concentrations exceeded +2 SDS at least once during treatment, whereas none of the untreated patients experienced this. Similarly, IGFBP-3 SDS values were close to the normal mean in both groups at baseline, but increased to the upper-normal range in the treated group.

Quality of life

The included study did not report QoL as an outcome measure.

Adverse events

The rate of treatment-emergent AE was higher in the rhGH group than in the no-treatment arm (Table 31), but these were reported to have mostly been common childhood illnesses.

There were no significant changes in thyroid function reported during the study, and no SAEs occurred in the patients with SHOX-D.

Summary

The evidence for the clinical effectiveness of rhGH as a treatment for short stature owing to SHOX-D comes from the single RCT that met the inclusion criteria for this review. The study was unblinded and did not report an ITT analysis.

By the end of the second year, children treated with rhGH had gained statistically significantly more height than those in the control group (approximately 6 cm more), with no statistically significant difference in catch-up of BA. HtSDS was statistically significantly higher in treated than in untreated patients.

Treatment with rhGH led to a statistically significantly greater GV in both years 1 and 2 (3.5 cm/year greater than untreated patients in year 1, and 1.9 cm/year greater in year 2). The HV SDS was positive, i.e. above the average for chronological age, during both years of rhGH treatment whereas untreated children had negative HV SDS.

Treatment with rhGH raised IGF-1 and IGFBP-3 levels to the upper normal range.

Treatment of the children with SHOX-D in this RCT was not associated with any SAE.

Methods for the systematic review of cost-effectiveness

A systematic literature search was undertaken to identify economic evaluations for rhGH in children. The details of the search strategy for the cost-effectiveness studies are in Appendix 2. The MSs were reviewed for any additional studies. Titles and abstracts of studies identified by the search strategy were assessed for potential eligibility by two health economists. Full text versions of relevant papers were retrieved and checked by two health economists. Any differences in judgement were resolved through discussion. The quality of the cost-effectiveness studies was assessed using a critical appraisal checklist based on that by Drummond and Jefferson,136 the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) checklist137 and the NICE reference case. 138

Results of the systematic review of cost-effectiveness

A total of 220 potentially relevant studies were identified in the cost-effectiveness searches and one in the QoL searches. Five full papers were retrieved with only two economic evaluations meeting the inclusion criteria. For all disease areas throughout the screening and data extraction process, differences in opinion were generally minor and easily resolved without the involvement of a third reviewer. The characteristics and results of the evaluations are discussed below.

Description of the identified studies

The literature search did not identify any economic evaluations conducted across the entire range of conditions of interest or any for the population of England and Wales. Table 32 provides a summary of the characteristics and base-case findings for the two published North American economic evaluations for human GH for children with TS135 and GHD. 139

The cost-effectiveness studies were assessed against the critical appraisal checklist (Table 33). Generally, the CADTH study135 was of a higher quality; the effectiveness of the treatment had been established through a systematic review, and the estimates for parameter values are more appropriate than the study by Joshi and colleagues. 139

Modelling approach

Both economic evaluations presented cost-effectiveness analyses using simple deterministic decision-analytical models. Both assumed that the clinical benefit achieved as a result of the rhGH treatment in the patients’ early years will last through their lifetime. Joshi and colleagues139 assumed that age-adjusted normal height was achieved after the first year of treatment. Subsequently, the benefits in terms of ‘normal height years’ and associated utility gain were assigned from the second year of treatment. Conversely, the CADTH study135 did not assume that patients experienced any improvement in HRQoL during the treatment. The utility gain is associated with the completion of treatment rather than with achieving normal height, as normal height was not achieved in the review of clinical effectiveness.

The cohorts differed with respect to age at baseline, duration of treatment and probability of achieving normal height at the end of treatment (see Table 32 above). The CADTH study135 used the characteristics and clinical effectiveness data from the TS RCT,86 whereas Joshi and colleagues139 did not provide any clinical evidence for either the baseline characteristics of the two cohorts of patients with GHD or the assumed clinical effectiveness estimates.

Joshi and colleagues139 assumed a 20% dropout rate after 12 months of treatment and related it to the slight pain experienced by patients, although no clinical evidence was presented to support this assumption. The CADTH study135 did not adjust the final outcomes for the dropout rate, effectively assuming it to be zero. As none of the TS patients achieved normal height, the CADTH study135 did not differentiate between partial and complete success of rhGH treatment. In contrast, Joshi and colleagues139 assumed that those patients who completed treatment but did not achieve normal height still acquire a partial utility gain. However, no justification for this assumption is provided.

Discounting was appropriately applied to costs and benefits in both studies, although the discounting rates were different from the 3.5% recommended by NICE138 (3% in the study by Joshi and colleagues139 and 5% in the CADTH study135).

Estimation of final outcomes (QALYs)

Both studies highlighted the difficulty of translating intermediate (clinical) outcomes to final outcomes (QALYs). There is an apparent paucity of utility-based estimates of HRQoL in rhGH patients and an absence of such estimates obtained from children eligible for rhGH treatment (see below, Review of research on quality of life). Therefore, the authors chose alternative utility estimates that, despite acknowledged shortcomings, were judged to meet the requirements of their economic models. The utility increment associated with rhGH treatment reported in the two studies ranged from 0.04135 to 0.189. 139

Joshi and colleagues139 adapted the QoL indexes presented in the Wessex Development and Evaluation Committee (DEC) report. 140 The indexes estimated in the report were not derived using one of the methodologically rigorous techniques for obtaining utility estimates, such as time trade-off (TTO) or standard gamble (SG)141 and cannot therefore be interpreted as ‘utilities’. Furthermore, the utility element of that report was a set of scenarios not based on primary or secondary data sources and thus could not be considered reliable or valid. 6 Joshi and colleagues139 used utility estimates of 0.781 for the pretreatment and no treatment groups, although this is different to the value 0.884, reported in the DEC report. Those patients who achieved success, i.e. normal height, had a utility of 0.97 applied from the start of the second year of treatment. Patients with partial success were assumed to acquire a partial utility gain defined as 35% less than the full utility gain associated with achieving normal height. The value was stated to be between 0.884 and 0.940.

The CADTH study135 did not use absolute utility values associated with each health state but applied an incremental utility value of 0.04 for patients receiving treatment with rhGH. The utility increment was estimated from a TTO survey in a small sample of adults with TS142 (see below, Review of research on qulity of life). The patients in the QoL study were asked how many years they would be willing to lose from their life to attain an average stature. The answers were translated into the incremental utility estimate of 0.04. The CADTH study135 stated that TS patients do not attain an average stature, and so this estimate is likely to be an overestimate and bias the result of economic evaluation in favour of rhGH treatment.

Estimation of costs

Joshi and colleagues139 included costs for paediatric consultations and rhGH treatment. The CADTH study135 also included costs for X-ray examination. The unit costs reported in the economic evaluations reflect the difference in clinical practices in Canada and the USA, the price difference of the unit of resources expressed in Canadian and US dollars, and the difference in methodological approach adopted in the two studies. For example, the CADTH study135 excluded the specialist visits as these do not differ between the intervention and the control groups. The total incremental cost reported varies according to the length of treatment but is consistent between the two studies.

Model results

The cost-effectiveness analysis in the CADTH study135 used an incremental difference of 6.5 cm in FH between the intervention and control groups, based on their clinical review. They calculated the undiscounted cost-effectiveness as C$26,529 per centimetre of improved FH and the discounted incremental cost-effectiveness ratio (ICER) was C$23,630 per centimetre of improved FH. They estimated an ICER of C$243,078 per QALY gained. The authors concluded that for an average patient with TS, rhGH treatment is unlikely to be cost-effective unless the payer is willing to pay more than C$200,000 to obtain a QALY.

Joshi and colleagues139 calculated the difference in ‘normal height years’ between the intervention and the control groups to estimate the incremental cost per normal height year. It was assumed that normal height was achieved by patients in the intervention group, but not in the control group. The incremental gain in ‘normal height years’ in the cohort of 5- to 16-year-olds was 17.4 (discounted). The corresponding value in the cohort of 3- to 18-year-olds was 21.1 (discounted), which translated into an incremental cost per additional year of normal height of $8900 (discounted) in the cohort of 5- to 16-year-olds and an incremental cost per additional year of normal height of $9300 (discounted) in the cohort of 3- to 18-year-olds. They estimated an ICER of about $37,000 per QALY gained for treating children with GHD from ages 5 to 16 years and an ICER of about $42,600 per QALY gained for treating children with GHD from ages 3 to 18 years. The authors concluded that the cost-effectiveness of rhGH compares favourably to accepted threshold values and represents reasonable value for money.

In both studies the deterministic one-way analyses indicated that the results were sensitive to variations in the utility estimate, the starting age of treatment, the duration of treatment and the daily dosage. The results were also sensitive to assumptions about clinical effectiveness139 and to variations in the price of rhGH. 135

The two economic evaluations arrived at opposite conclusions about the value for money of the rhGH treatment in children. The economic evaluation conducted for the CADTH study135 may provide a more reliable estimate of the cost-effectiveness as it has used clinical data from a reasonable quality RCT and TTO utility estimates. In contrast, the assumptions about clinical effectiveness of rhGH treatment by Joshi and colleagues139 did not seem to be supported by clinical evidence. Furthermore they also used indexes, interpreted as utility weights, which do not appear to be reliable or valid.

Summary and conclusion of the systematic review of cost-effectiveness studies

We undertook a systematic review of the literature in order to identify existing models in this area. The systematic review of published economic evaluations identified two North American studies relevant to the target population and no studies conducted in the UK. The results of the two identified studies produced two very different estimates of cost-effectiveness. This difference is largely due to the choice of utility estimates and assumptions on the effectiveness. As discussed below (see next section, below, Review of research on QoL), there is a paucity of reliable estimates of utility gains associated with GH treatment. Therefore, the results of both studies should be treated with caution. In particular, Joshi and colleagues139 adapted QoL indexes that were not derived according to the NICE reference case and could not be considered reliable or valid. 6 The literature study did not identify studies that we could use for this review and so a de novo independent economic model was required.

Systematic review of HRQoL studies

A systematic review was undertaken to identify HRQoL studies for rhGH for children. The HRQoL searches were undertaken to populate a lifetime economic model with utilities to calculate QALYs, so studies with adults and children were eligible for inclusion. Titles and abstracts of studies identified by the search strategy were assessed for potential eligibility by two health economists. Full text versions of relevant papers were retrieved and checked by two health economists. Any differences in judgement were resolved through discussion. The details of the search strategy for QoL are in Appendix 2.

The titles and abstract of the studies identified by the search strategy were assessed on the basis of the following criteria:

• disease condition as defined in Table 1 (Chapter 1) of this report

• primary research using a preference/utility based measure for the conditions interest

• primary research using a generic measure [i.e. Short Form questionnaire-36 items (SF-36)] that can be translated into a utility-based estimate

• primary research using a condition-/disease-specific QoL measure and an algorithm that allowed disease-specific QoL to be converted into utility values.

Exclusion criteria for the systematic literature search were:

• primary research reporting QoL that could not be converted into utility values using a validated mapping algorithm

• background or discussion papers that did not report a QoL measure for the conditions of interest

• papers reported in language other than English.

The search strategy identified 391 articles that were potentially relevant. After the abstracts had been screened, 24 articles were identified and full papers were retrieved for these articles. After checking the retrieved studies, six papers met the inclusion criteria. These are summarised in Table 34. A further targeted search linking height to HRQoL is reported below (see Height and health-related QoL).

Height and health-related QoL

The NICE reference case clearly states that the measure of health outcome used in the cost-effectiveness analysis should be QALYs calculated with utilities derived from a validated, generic, preference-based measure of HRQoL. 138 The clinical effectiveness review in Chapter 3 found no RCTs that reported HRQoL measures as an outcome and the additional search for HRQoL studies (above) located only one relevant study by Koltowska-Haggstrom147 in one of the conditions of interest (GHD) that was strictly applicable to the NICE reference case. 138

Therefore, a targeted search was conducted to identify publications that reported gains and losses in utility in relation to variation in height, as height is one of the primary outcome measures of GH treatment. Details of the search are in Appendix 2. One full paper by Christensen and colleagues150 was identified.

The study used the 2003 Health Survey for England, with 14,416 observations for adults (aged > 18 years). 151 HRQoL was measured using the EQ-5D with the UK tariff. Height was converted from centimetres to HtSDS using a UK population algorithm. Inter-relationships between variables were assessed using ordinary least squares (OLS) linear regressions, controlling for age, weight and gender. All OLS analyses were controlled for multicollinearity (close interaction between explanatory variables). Where there were any highly correlated variables (weight and BMI) then one variable was omitted from the regression. The regression analyses included two-level categorical variables (‘sex’, ‘limiting longstanding illness’ and ‘social class’) to explore the relationship between height and HRQoL while controlling for these confounding factors. 150

There was a positive correlation between an increase in height and a participant’s EQ-5D score. The mean EQ-5D scores were lower in the shorter compared with taller subjects, as well as lower than the overall population mean. The authors’ report an analysis of variance (ANOVA) combined with post hoc Tukey’s Honestly Significant Difference (HSD) test for homogeneous subgroups, which showed that the sample could be split into three meaningful subgroups, each significantly different (p < 0.05) from each other in terms of their EQ-5D scores. The first subgroup ‘HtSDS ≤ –2.0’ had significantly lower EQ-5D scores than the second group ‘–2.0 > HtSDS ≤ 0’ and the third group ‘HtSDS > 0’. The second subgroup had significant lower scores than the third group. A multivariate linear analysis using the previously identified subgroups was undertaken to predict the variation in HRQoL. The full model predicted only one-third of the sample variation in EQ-5D (R² = 0.318, 0.343 and 0.290) based on 11,946 observations. 150

The model predicted that for those people shorter than –2.0 HtSDS, an improvement of 1 HtSDS will result in a change in EQ-5D score of 0.061. However, for the subgroup between –2.0 and 0 HtSDS the gain in EQ-5D is much reduced (a 1-HtSDS improvement increases EQ-5D score by only 0.010). One drawback to the Christensen study150 is that the population used to elicit QoL values are not from the conditions of interest but from the general population.

Summary and conclusions of the QoL review

The systematic review of QoL identified six studies that met the inclusion criteria. None of the studies was in a childhood population. This is to be expected, given the difficulties in conducting preference-based QoL studies in children. 152 Three studies reported the SF-36 in adults but were not useful on further examination, as they reported SF-36 scores only after rhGH treatment. One poor-quality study reported SF-36 at baseline, 6 months, 12 months and 24 months for a small cohort of adult participants with PWS, and the scores from this study were mapped to a UK-based EQ-5D preference-based utility index by a subsequent study.

There were only two studies that reported change in QoL using preference-based measures in the conditions of interest. 145,147 The first study145 used TTO methodology for people with GHD, TS and CRI. The number of years that the participants were willing to trade to reach average height was in the range of 0–4%. However, there were several limitations to this study and it was felt that it did not generally provide a robust estimate of utility gain from rhGH treatment. The second study147 used a regression model to give utility weights (based on the EQ-5D from a UK population) to the disease-specific QoL-AGHDA. The KIMS database was then used to transform patients QoL-AGHDA values into QoL-AGHDA_UTILITY values. However, it was in an adult population and it is unclear whether they had previously had rhGH treatment as children. This study was specific to patients with GHD and is unlikely to be generalisable to the other conditions of interest.

An additional targeted search was undertaken for QoL in relation to height. One study was identified by Christensen and colleagues,150 which provided utility estimates based on the EQ-5D for different HtSDS from the Health Survey for England for an adult general population. The study provides a common utility gain that could be compared across all of the conditions of interest that could be used with the clinical effectiveness outcomes from the RCTs.

Based on the review of the QoL literature, there is likely to be a small gain in utility for individuals receiving GH treatment. However, this is based on a proxy measure of gain in height from shorter people in the general population. This excludes many relevant potential benefits and disadvantages of rhGH treatment that it is not possible to capture without good-quality evidence from the conditions of interest. This is especially true for PWS as additional HRQoL gain from improved body composition is unlikely to be captured with this method. Furthermore, there is also uncertainty over the impact of extrapolating back into childhood with adult utility data.

Modelling approach

In the MSs, the base-case analyses estimated the incremental cost of rhGH per centimetre of height gained relative to no treatment (in order to compare with previous HTA report6) and the incremental cost of rhGH per QALY gained relative to no treatment. The utility scores used in the model in children with GHD, TS, CRI and children who were SGA were based upon the study by Christensen and colleagues,150 discussed above (see Description of the identified studies). A gain in height was assumed to be associated with QoL improvements, which was assessed using the EQ-5D utility scale. In patients with PWS the QoL gain was based upon a small study of adult patients with PWS, together with an estimation of the benefits associated with a reduced risk of diabetes. The assumptions used to derive QoL utility improvements are discussed above (see Review of research on QoL).

The economic evaluation of rhGH treatment in conditions such as GHD, TS, SGA and CRI is based on a single clinical effect of additional height gained as a result of treatment. This clinical effect and many of the other parameters used in the model are estimated from the Kabi International Growth Study (KIGS) database,153 which is a large-scale collaborative database developed by Pfizer for the safety and efficacy of treatment with rhGH. It includes data from more than 60,000 treated patients in over 50 countries for all licensed indications, i.e. GHD, TS, PWS, SGA and CRI. Table 35 shows the input parameters used in the manufacturers’ model that have been derived from the KIGS database. The costs used in the manufacturers’ model were based upon those used in the previous HTA report, and inflated to current prices where appropriate. 6

The cost-effectiveness analysis of rhGH treatment in PWS is based on an alternative structure of the model that estimates the utility gain based on a small study of 13 adult patients with PWS144 (see Review of research on QoL, above) who received rhGH for 2 years and a further utility gain for reduced diabetes risk. However, the PWS QoL study is for adults who have received rhGH and it is unclear how this relates to the QoL gain for a group of children, and whether this QoL benefit would be maintained throughout their lifetime. Furthermore, the two methods154,155 used by the Pfizer submission to translate SF-36 scores into utilities were not based on choice-based methods like TTO or SG, which produce utilities more rigorously. 141 The model assumes that individuals with PWS and diabetes would have a 10% lower QoL than those without. Based on Pfizer’s submission to the Pharmaceutical Benefits Advisory Committee in Australia, it was assumed that the prevalence of diabetes in patients with PWS would reduce from 8% to 2%, although it was not possible to verify these assumptions in the reference provided.

An alternative model structure that allowed for the second clinical effect (a reduction in the risk of osteoporosis) was also presented in a scenario analysis for GHD. In this model it was assumed that a proportion of GHD children continue treatment until they reach the age of 25 years.

The manufacturers’ model makes the following assumptions:

1. Patients with conditions of interest have the same life expectancy as the general population of England and Wales in the treated and untreated groups.

2. Patients can continue rhGH treatment or discontinue treatment at the end of 1 year.

3. Untreated children do not gain any utility benefit throughout the course of the lifetime of the model.

4. Treatment costs and monitoring costs are applied over the treatment years. Health benefits, as measured by QoL associated with particular attained heights, are maintained over patients’ lifetimes. The full utility value is applied after 2 years of treatment.

5. Compliance is assumed to be 90% in the base-case analysis and this was assumed to not impact efficacy.

6. Adverse events are not considered in the model for both the treated and non-treated patients.

7. In the base case, for all conditions except PWS, rhGH treatment affects only FH and does not affect the risk of morbidities, such as osteoporosis fracture or diabetes.

8. The MS estimated the average height at the end of treatment for the control group from the previous HTA report.

Appraisal of the manufacturer cost-effectiveness analysis

A summary of the MS compared with the NICE reference case requirements138 is given in Table 36 and indicates that the submission meets most of the requirements. See Appendix 9 for a tabulation of the critical appraisal of the submission against the Drummond and colleagues’ checklist. 136

Cost-effectiveness results

The mean daily per-patient cost for each of the manufacturer’s GH treatments was based upon the unit cost shown in Table 37. Merck Serono stated that there will be a reduced cost of £20.87 through the use of the Merck Serono Easypod™, which they report will reduce vial wastage and increase compliance.

The base-case analyses for Pfizer, Eli Lilly, Ipsen and Merck Serono are shown in Table 38. Merck Serono produced their own version of the model and so the health benefits differ slightly from the other models.

The base-case results for the Novo Nordisk model using KIGS data are shown in Table 39. They also reported alternative ICERs using patient level data.

Manufacturers’ conclusions

The authors suggested that many of the health benefits associated with rhGH treatment are not quantifiable and cannot be modelled easily. Many of these benefits would improve overall patient QoL and, possibly, duration of life. These benefits include self-esteem, improvements in sleep and concentration, and increased appetite as well as increases in LBM, total bone mass and muscle strength. These benefits may lead to reduced risk of diabetes, obesity and cardiovascular diseases.

The manufacturers concluded that their economic analyses demonstrated that rhGH is cost-effective for the treatment of short children with GHD, CRI and those born SGA, and borders on cost-effectiveness for the treatment of TS and PWS. They stated that the values for cost/centimetre compared favourably to those reported in the previous NICE assessment6 and supported the recommendation of rhGH for children with GHD, TS and CRI, plus its extension to include SGA children.

Summary of general concerns

• Clinical effectiveness estimates for height gain were taken from an observational cohort rather than an RCT. It is not clear whether the subset of the KIGS database chosen was representative of the UK patient population or, for example, whether the subset chosen may be more severe.

• For three of the conditions (GHD, PWS and SGA) the estimates of height gain, in centimetres, were considerably higher than those shown in the trials due to the estimates used for end height in the control group.

• All conditions, except PWS, used mortality rates from the general population. It is likely that individuals with these conditions, in particular CRI, will have increased mortality compared with the general population.

• The manufacturers have used the Christensen study150 for their HRQoL utility values but have not taken these from the regression analysis from this study. Instead, they have used the relationship between EQ-5D and height without controlling for other factors. Utility gain attributed to height is likely to be capturing the combined effects of other (unobserved) variables, such as age, longstanding illness and gender. For example, older generations generally have lower QoL because of their age. Not controlling for other factors, in particular age, results in the overestimation of the utility values. Furthermore, the group with the lowest height and QoL (< –3 SDS) had few observations and individuals in this group were generally elderly (mean age > 70 years).

• Treatment cost is calculated by rounding up to the nearest whole year of treatment.

• There is high uncertainty associated with the assumptions and sources used to estimate QoL gain in the PWS model. These were based on a small study of adult patients with PWS and it is unclear how this relates to the QoL gain for a group of children, and whether this QoL benefit would be maintained throughout their lifetime. The methods used to derive values from the SF-36 for utilities were based on rating scales and therefore did not use choice-based methods, such as the SG and TTO. QoL gain also estimated utility gain from reduced diabetes prevalence but this evidence could not be verified. There are considerable difficulties extrapolating the benefit from treating children with rhGH to their health benefits as adults.

Sandoz submission to NICE

Sandoz presented an analysis comparing Omnitrope with Genotropin. The MS contained a comparison of the annual cost of treatment with Omnitrope and with Genotropin in patients with GHD and TS. However, the MS did not comply with NICE guidance for a MTA,138 as QALYs were not estimated and a cost-effectiveness analysis was not presented. The MS attempted a cost-minimisation analysis, implicitly suggesting that treatment with Omnitrope is equally effective as treatment with Genotropin (in terms of additional height in children with GHD and TS) but is associated with less cost to the UK NHS. A critical appraisal of the Sandoz MS is given in Appendix 10.

Overview

A comparison of the costs and benefits of rhGH compared with no treatment in cohorts of children with GHD, TS, PWS, CRI and SHOX-D and children who are SGA was made using decision-analytical models. Models were constructed in Microsoft excel according to standard modelling methods. 138 To identify data to populate the model, systematic searches were conducted to locate studies on the natural history and epidemiology of the indicated conditions, HRQoL and costs.

Costs were derived from published studies (where available), and from national and local NHS unit costs. The model was from the perspective of the NHS and PSS, as only these direct costs were included. The model estimates the lifelong costs and benefits from rhGH treatment. The costs and benefits were discounted at 3.5%, as recommended by NICE. 138 The base year for the costs was 2008. The intervention effect in terms of improvement in HtSDS was derived from the systematic review of effectiveness reported in Chapter 3. The outcome of the economic evaluation is reported as cost per QALY gained and cost per centimetre gained.

Description of the model

A decision-analytical model was designed for the economic evaluation of rhGH for treatment of GHD, TS, PWS, CRI, SGA and SHOX-D, and was based upon one developed in the previous HTA report. 6 The current model compares a cohort of patients receiving rhGH during their childhood with a cohort of patients who were not treated with rhGH. The state transition Markov model has a cycle length of 1 year and a lifetime horizon. A Markov model was used as these are suitable for lifetime analyses with few health states. 156 The base-case decision-analytical model includes health states for alive and dead. The England and Wales population mortality rates are applied in each cycle for patients, with an adjustment using the SMRs for each of the conditions.

The model assumes that a daily subcutaneous injection of rhGH is administered for the duration of treatment, unless a patient from the treatment cohort drops out of treatment or dies. The parameters of the model that determine the age at the start of treatment, the duration of treatment and the annual dropout rates are estimated from the KIGS database described in the MS or based upon advice from our clinical advisory group, and vary between conditions. A daily dose is calculated according to the child’s weight. The dose regimen corresponds to the licensed indication of rhGH in children (and adults, in a scenario analysis of the GHD cohort).

Health-care resources included for the cost of patient monitoring apply to both the treatment and no treatment cohorts. The cost categories and unit costs are consistent with the costs used in the previous HTA report for rhGH. 6 The discount rate of 3.5% is applied to both costs and final outcomes.

Patients from the treatment cohort who stay in treatment receive a benefit of an additional height gain relative to patients in the no treatment cohort. Patients who drop out of treatment stop accumulating height gain, so their growth progression is no different from the height gain in the no treatment cohort. In each yearly cycle, individual HRQoL is estimated based upon their height gain. Individuals are assumed to maintain the same HRQoL after treatment has stopped for the rest of their lifetime. In each cycle, the total costs and QALYs are calculated by multiplying the individual costs and HRQoL by the number of people in the cohort still alive for the treatment and no treatment cohorts. The total lifetime costs and QALYs are calculated for the treated and non-treated groups by aggregating the costs and QALYs in each cycle. The total discounted QALY gain, and cost of treatment for the treatment and no treatment cohorts are calculated. Thus, the cost-effectiveness of rhGH is calculated:

Parameters used in the model and the data sources used to derive them are described in more detail below (see Model validation).

A list of the model assumptions is given below. Assumptions are applied to all conditions unless explicitly stated otherwise. All assumptions were tested in sensitivity analyses.

• The diagnostic costs were not included in the analysis as they were assumed to be the same for both rhGH-treated and no-treatment patients.

• The base case assumes no dropout or discontinuation of treatment. This was based upon advice from our clinical advisory group that this was likely to be a relatively rare occurrence. The base-case model therefore evaluates just rhGH treatment versus no treatment.

• There are two health states for alive or dead in the model, and the transition between them is based on age-related mortality data.

• The mortality rates were assumed to be higher than for the England and Wales general population estimates for untreated and treated cohorts for all conditions.

• It was assumed that there would be no reduction in mortality as a result of rhGH treatment. There is a lack of data to assume otherwise.

• The model time horizon is 100 years and all individuals are assumed to die by this age.

• Effectiveness estimates for the conditions were based on selection of the best-quality evidence from the clinical effectiveness review in Chapter 3. RCTs were only selected if the follow up length was at least 2 years after the start of treatment. Where there were no appropriate RCTs, long-term observational studies were considered. In the case of SGA, the most appropriate RCT was for only 1 year.

• Compliance was assumed to be 85% in the base case, with no loss of efficacy for rhGH treatment. 157

• An additional scenario was undertaken for the GHD condition where treatment continued for a transition phase into adulthood to age 25. This was only applicable for 34% of the GHD population. 158 No additional benefit, in terms of height gained, was assumed from this additional treatment.

• In the treatment and no-treatment cohorts, all children are monitored until they reach adulthood, assumed to be the age of 17 years.

Data sources

Cost-effectiveness of rhGH treatment – deterministic sensitivity analysis

One-way deterministic sensitivity analyses were performed, in which model parameters were systematically and independently varied, using a realistic minimum and maximum value. The sensitivity analysis investigated the effect of uncertainty around the model structure and for variation in parameters on the cost-effectiveness results, in order to highlight the most influential parameters. The effects of uncertainty in multiple parameters were addressed using PSA, which is reported later in this section. Where possible, the parameters were varied according to the ranges of the CIs of these parameters, based on the published estimate. Where these data were not available an alternative suitable range was chosen. The same ranges were used in the deterministic and probabilistic sensitivity analyses and these are described in Appendix 12.

Table 48 shows the results for each of the conditions using the KIGS database153 for estimate of the clinical benefit. The KIGS database, a large observational study of children treated with rhGH, was used for the effectiveness of GHD in the base case reported above. According to these results, an ICER of rhGH versus no treatment varied from an ICER of £18,980 per QALY gained for SGA to £144,050 per QALY gained for PWS. Results are of a similar magnitude to the base case with the exception of the SGA analyses. The ICER for SGA is much lower in this analysis because the incremental clinical height gain is lower in the RCT effectiveness data than in the KIGS effectiveness data.

The discount rates used for the analyses have a large effect on the results, due to the upfront costs and the health outcomes stretching over the life time of the model. Table 49 shows the results using the discount rates used in the previous HTA report, i.e. costs 6% and benefits 1.5%. Using these discount rates, rhGH treatment is more cost-effective. For all conditions, except PWS, the ICER reduces to less than £30,000 per QALY.

Tables 50–55 report the results of the deterministic sensitivity analyses for the conditions for the most influential parameters. Other variables were varied in sensitivity analyses but were found to only have a negligible effect on the results. The cost-effectiveness results are fairly sensitive to the variation in parameters included in the deterministic sensitivity analysis. For all of the conditions the model results are most sensitive to treatment start age and length, compliance and utility gain.

The deterministic sensitivity results for GHD are shown in Table 50. The results varied between £19,187 and £35,917 per QALY gained and were most sensitive to dosage.

The deterministic sensitivity results for TS are shown in Table 51. The results varied between 30,505 and £48,778 per QALY gained and were most sensitive to utility gain.

The deterministic sensitivity results for PWS are shown in Table 52. The results varied between £111,560 and £159,062 per QALY gained and were most sensitive to compliance.

The deterministic sensitivity results for CRI are shown in Table 53. The results varied between £28,080 and £54,105 per QALY gained and were most sensitive to the treatment start age and length of treatment.

The deterministic sensitivity results for SGA are shown in Table 54. The deterministic sensitivity results varied between £25,675 and £41,180 per QALY gained and were most sensitive to utility gain.

The deterministic sensitivity results for SHOX-D are shown in Table 55. The deterministic sensitivity results varied between £33,406 and £50,457 per QALY gained and were most sensitive to utility gain.

For patients with PWS there may be an additional health benefit associated with improved body composition, which may reduce the risk of diabetes and other morbidities. In addition, there is considerable difficulty estimating the magnitude of this effect and extrapolating short-term treatment in childhood to lifelong benefit. In the base case we have assumed that there is no HRQoL benefit associated with changes in body composition. In this section we present a scenario analysis for additional changes in body composition. However, there is a difficulty linking changes in lean FM to changes in utility, as there are no utility studies for lean FM. For this reason we have focused on changes in BMI.

Picot and colleagues168 conducted a targeted search to identify published utility estimates for the BMI values relevant to an adult obese population. The search aimed to identify estimates of the change in utility scores based on the unit change in BMI values. Utility estimates were considered only where they used a validated, multiattribute utility scale (e.g. EQ-5D) or appropriate methodology (e.g. SG or TTO techniques) and provided a clear definition of utility scores. They suggest the values reported by Hakim and colleagues169 represent the most methodologically sound estimates derived from subjects across a wide range of obesity levels. Hakim and colleagues169 found that a one-unit decrease in BMI, over a period of 1 year, was associated with a gain of 0.017, which was independent of age or gender.

In Chapter 3 (see Body composition), RCTs for PWS reported mixed results for changes in BMI with a maximum BMI difference of 1.8 kg/m² between treated and untreated groups after 2 years’ treatment. Assuming this change in BMI is maintained lifelong, and therefore there is an additional utility of 0.031, the cost-effectiveness of PWS would be £54,800 per QALY gained.

Probabilistic sensitivity analyses

In the probabilistic sensitivity analyses (PSAs) the main parameters were sampled probabilistically from an appropriate distribution using similar ranges as used in the deterministic sensitivity analyses. The parameters sampled were: starting age, length of treatment, dose, HtSDS at the start and end of treatment for both the rhGH and no treatment cohorts, utility increment for gains in height and all costs used in the base case excluding the cost of rhGH.

The distribution assigned to each variable included in the PSA and the parameters of the distribution are reported in Appendix 12. One thousand simulations were run for each condition of interest in this analysis. Table 56 reports the mean costs and outcomes from the PSA and the ICER for rhGH compared with no treatment, based on the mean values generated in the PSA. Table 57 shows the 2.5% and 97.5% percentiles for the PSA.

The cost-effectiveness results from the PSA are slightly lower than those from the deterministic analyses for GHD, TS, CRI, SGA and SHOX-D (which were £23,196, £39,460, £39,273, £33,079 and £40,531, respectively). The cost-effectiveness results from the PSA for PWS, however, are much lower than the deterministic estimates. This is due to non-linearity in the PWS model as a result of the baseline starting HtSDS for the treated group being at –2.0 HtSDS. This is also the height at which the utility gain per unit HtSDS changes and thus an individual with a starting height slightly lower than –2.0 HtSDS will have much higher utility gain than one with a starting height slightly higher than –2.0 HtSDS. This non-linearity results in a higher incremental QALY in the PSA results, therefore decreasing the ICER in the PSA.

Scatter plots are shown for the incremental cost and incremental QALYs for each of the conditions in Figures 2–7. The difference in the dispersion of costs and QALY data points between the six conditions reflects the different levels of uncertainty in each condition. The spread of costs and to a greater extent QALYs is more compact in GHD than for the other conditions. This is because the standard errors are much smaller due to the effectiveness data coming from the KIGS database, which has a large number of observations. This creates a tighter probability distribution around the mean value.

FIGURE 2.

Cost-effectiveness plane – incremental cost and incremental QALYs for rhGH treatment and no treatment in growth hormone deficiency.

FIGURE 3.

Cost-effectiveness plane – incremental cost and incremental QALYs for rhGH treatment and no treatment in Turner syndrome.

FIGURE 4.

Cost-effectiveness plane – incremental cost and incremental QALYs for rhGH treatment and no treatment in Prader–Willi syndrome.

FIGURE 5.

Cost-effectiveness plane – incremental cost and incremental QALYs for rhGH treatment and no treatment in chronic renal insufficiency.

FIGURE 6.

Cost-effectiveness plane – incremental cost and incremental QALYs for rhGH treatment and no treatment for the condition ‘small for gestational age’.

FIGURE 7.

Cost-effectiveness plane – incremental cost and incremental QALYs for rhGH treatment and no treatment in short stature homeobox-containing gene deficiency.

In addition, a CEAC was also derived, representing the proportion of simulations when GH treatment is cost-effective for a range of willingness-to-pay thresholds, up to £100,000, see Figure 8.

FIGURE 8.

Cost-effectiveness acceptability curve for rhGH treatment and no treatment for all the conditions.

In this analysis, rhGH treatment had the probability of being cost-effective at willingness-to-pay thresholds of £20,000, £30,000 and £50,000 per QALY as: 22%, 95% and 100% for GHD, 2%, 19% and 78% for TS, 0%, 1% and 8% for PWS, 2%, 16% and 80% for CRI, 4%, 38% and 90% for SGA, and 1%, 15% and 74% for SHOX-D, respectively.

Growth hormone deficiency

The use of rhGH as replacement therapy is well established in children who have a deficiency of the natural hormone. Therefore, most clinicians would consider it unethical to withhold treatment and there is a corresponding lack of RCT evidence in the literature. Only one trial84 met the inclusion criteria for the review of rhGH in children with GHD, and this did not report FH. No details were reported on randomisation or allocation to treatment groups or blinding. The included patients (n = 19) were part of a larger study, which was generally poorly reported. After a year’s treatment, HtSDS was statistically significantly higher in treated than in untreated children, although actual height was not reported. Children who received rhGH for 1 year had grown at a mean velocity of 2.7 cm/year faster than untreated children, which was statistically significantly faster. The low patient numbers mean that the evidence base for GHD is weak. Thus, there is very limited evidence of a slight increase in growth for children with GHD treated with GH, based on one study of mixed quality. Estimates of height gain in the previous HTA report6 suggested FH gains of approximately 1.3–1.6 SDS (i.e. within 2 SDs of the normal mean) with rhGH treatment. However, these figures were from retrospective single-cohort studies that were not included in the present review.

The cost-effectiveness estimate of rhGH treatment in GHD is about £23,200 per QALY gained or £2,800 per centimetre gained. As there were no appropriate RCTs, the KIGS database was used for the estimate of height gain from rhGH. 153 This estimate for height gain was higher than for the other conditions. The previous HTA report6 estimated a cost per centimetre gained of £6000 using 8 years’ treatment compared to the 7 years used in our analysis and a slightly lower height gain from the KIGS database. 153 The cost-effectiveness estimate for the cohort of GHD who continue rhGH treatment into adulthood was £28,200 per QALY gained and £3400 per centimetre gained.

Turner syndrome

Six trials met the inclusion criteria for the review of GH for growth disturbance in patients with TS. 12,85,86,88–90 There is some evidence of effectiveness across all reported growth outcomes for girls with TS. However, these results are reported in studies of poor reporting and methodological quality, and in some cases of short duration. Of the six included studies, none of the included trials used an ITT analysis, one reported adequate randomisation to treatment groups,85 one study described adequate concealment of treatment allocation,85 and one adequately blinded the patient to treatment by administering placebo. 89

In a large RCT that followed girls until FH, children in the rhGH group grew an average of 9.3 cm more from baseline than those in the untreated group. 86 In a study of younger children over 2 years, the difference was 7.6 cm. 85 Both of these were statistically significant results. Weight and WtSDS were found to be significantly greater in the treated group in one study of younger girls with TS. 85

The searches for this study identified a new systematic review, conducted in Canada in 2007. 135 The review concluded that rhGH is effective in improving growth and FH in girls with TS, but found no evidence available in the clinical trials to suggest that rhGH improves QoL. The evidence discussed in the present review reflects this, as we found some evidence for increased height but no RCT evidence for improvements in QoL.

In summary, there is some evidence of effectiveness across all reported growth outcomes for girls with growth disturbance as a result of TS. There is also evidence of improved body composition. These results are reported in studies of poor reporting and methodological quality, and, in some cases, short duration, issues that may affect the validity of these findings. The previous HTA report6 found that treated girls’ FH was approximately 5 cm taller than untreated controls. The full publication of the large Canadian RCT86 since the earlier HTA report6 has shown a slightly larger difference in FH of 9.3 cm, as reported in the present review.

The cost-effectiveness estimate of rhGH treatment in TS is about £39,500 per QALY gained, or £6500 per centimetre gained. The estimate of cost-effectiveness compares with the estimate of about £130,000 per QALY (at current exchange rates) from CADTH,135 which used a lower QoL benefit for rhGH of 0.042 than used in our analysis. The previous HTA report estimated a less favourable cost per centimetre gained of £16,000 as a lower estimate for height gain of 3.9 cm (compared with 9.3 cm) was used.

PWS

Eight small, rather poorly reported RCTs were included for PWS. 22,91–102 Participants’ average ages ranged from 13 months to 10 years. Only the crossover study102 used a placebo injection; the parallel-group RCTs had no treatment as the comparison arm.

Treated patients grew an average of 3–5 cm/year faster than untreated patients. Only one22 of the studies reported actual change in height, with infants treated with rhGH growing an average of 6.1 cm more than untreated patients during 1 year. HtSDS was statistically significantly greater in treated patients than in untreated patients after 1 year (1–1.5 SDS higher) or 2 years of rhGH treatment (> 2 SDSs).

Four22,95,96,102 trials reported a statistically significantly lower percentage of body fat (between 1% and 10% lower) in patients treated with rhGH than in those receiving placebo or no treatment. Three93,95,96,102 trials reported that patients treated with rhGH had statistically significantly higher LBM or a larger improvement in LBM than untreated patients. Clinical advice indicates that rhGH characteristically increases LBM and reduces FM, although weight and BMI do not always change. This is reflected in the RCTs’ findings, where changes in BMI were statistically significant in two studies,91,102 there were no statistical differences in two other studies,93,95,96 and results were similar between groups in the other two studies. 92,100,101

In summary, patients treated with rhGH grew faster than untreated patients, and tended to have lower body fat percentages. Measurements in treated patients were reported to be statistically significantly better than in untreated patients in several studies, but the included studies were rather small and did not report power calculations or specify a primary outcome, so it is not clear whether they were adequately powered. These findings were comparable with growth and body composition outcomes reported in the previous HTA review. 6 However, the previous review also reported an uncontrolled, single-cohort study of 16 children, which suggested that rhGH treatment normalised FH.

The cost-effectiveness estimate of rhGH treatment in PWS is about £135,300 per QALY gained or £5,900 per centimetre gained. The ICER values for PWS were higher due to the majority of the height gain occurring within –2 HtSDS of average height where a lower utility gain is experienced. The previous HTA report6 presented a cost per HtSDS gained of £40,815 and this compares with the current report’s estimate of £44,718.