Positive Linear Growth and Bone Responses to Growth Hormone Treatment in Children With Types III and IV Osteogenesis Imperfecta: High Predictive Value of the Carboxyterminal Propeptide of Type I Procollagen


  • The authors have no conflict of interest


Extreme short stature is a cardinal feature of severe osteogenesis imperfecta (OI), types III and IV. We conducted a treatment trial of growth hormone in children with OI and followed linear growth velocity, bone metabolism markers, histomorphometrics, and vertebral bone density. Twenty-six children with types III and IV OI, ages 4.5–12 years, were treated with recombinant growth hormone (rGH), 0.1–0.2 IU/kg per day for 6 days/week, for at least 1 year. Length, insulin-like growth factor (IGF-I), insulin-like growth factor binding protein (IGFBP-3), bone metabolic markers, and vertebral bone density by DXA were evaluated at 6-month intervals. An iliac crest biopsy was obtained at baseline and 12 months. Approximately one-half of the treated OI children sustained a 50% or more increase in linear growth over their baseline growth rate. Most responders (10 of 14) had moderate type IV OI. All participants had positive IGF-I, IGFBP-3, osteocalcin, and bone-specific alkaline phosphatase responses. Only the linear growth responders had a significant increase in vertebral DXA z-score and a significant decrease in long bone fractures. After 1 year of treatment, responders' iliac crest biopsy showed significant increases in cancellous bone volume, trabecular number, and bone formation rate. Responders were distinguished from nonresponders by higher baseline carboxyterminal propeptide (PICP) values (p < 0.05), suggesting they have an intrinsically higher capacity for collagen production. The results show that growth hormone can cause a sustained increase in the linear growth rate of children with OI, despite the abnormal collagen in their bone matrix. In the first year of treatment, growth responders achieve increased bone formation rate and density, and decreased fracture rates. The baseline plasma concentration of PICP was an excellent predictor of positive response.


OSTEOGENESIS IMPERFECTA (OI) is a genetic disorder of connective tissue, with bone fragility as the hallmark feature. Almost all cases of OI are caused by mutations in type I collagen,(1) the most abundant protein component of the extracellular matrix of bone, skin, and tendons. The severity of the OI phenotype according to the Sillence classification(2) correlates broadly with the biochemical consequences of the collagen mutations. The mild OI type I phenotype is associated with mutations that cause haploinsufficiency of the α1(I) chain of type I collagen. These individuals produce a reduced amount of structurally normal collagen. The moderate, severe, and lethal OI, types IV, III, and II, respectively, are associated with structural defects of type I collagen. In these individuals, the phenotype is associated with the dominant negative effect of the mutant collagen present in the bone matrix.

Growth deficiency is one of the cardinal features of OI. Although there is no clear height demarcation between patients with different OI types, the adult stature of patients with nonlethal OI type III is generally in the range of prepubertal children, while patients with OI type IV may have the final stature of a young teenager.(3) Even mild OI type I patients are shorter than same-gender siblings and may fall below the 5th percentile on standard growth curves. Antoniazzi et al.(4) examined growth rate and bone metabolism in a controlled growth hormone treatment trial of children with biochemically verified type I OI. In these children with a quantitative deficiency of matrix, recombinant growth hormone (rGH) resulted in a significant increase in linear growth velocity and vertebral bone density. Serum levels of carboxyterminal propeptide of type I procollagen rose in the treatment group.

We have previously examined the growth hormone axis in children with OI types III and IV.(5, 6) Generally, they had a normal response to GH provocative tests and normal results on serial sampling of spontaneous growth hormone secretion. Some children had a blunted response to growth hormone-releasing hormone or failed to double their serum insulin-like growth factor-I (IGF-I) in a 5-day somatomedin generation test. However, no consistent relationship between these responses or between the responses and type of OI emerged. The purpose of this study was to examine the ability of growth hormone treatment to stimulate linear growth in children with OI and to define the best predictors of a positive response. We conducted a prospective treatment trial of rGH in children with types III and IV OI to determine its effect on linear growth rate, bone metabolic markers, bone density, and histology.



Twenty-six children with types III and IV OI were recruited to this study from the National Institute of Child Health and Human Development (NICHD) OI study population under an Institutional Review Board (IRB)-approved protocol (Table 1). Subjects were excluded by (1) femur angulation exceeding 60°, because bone was likely to require surgical correction during study period and distort growth rate; (2) scoliosis exceeding 40°, because of concern that growth spurt might worsen scoliosis; or (2) prior spinal fusion, because of inability to grow trunk height.

Table Table 1.. Patient Clinical Profile and Responses to GH Therapy
original image

Study design

All participants were treated with rGH (Somatropin Nutropin, generously donated by Genentech, Inc., South San Francisco, CA, USA) for at least 1 year and were monitored at the National Institutes of Health every 3 months. An accurate baseline growth rate was available for 2 years pretreatment from the NICHD OI Study. Total height was measured using a recumbent stadiometer, as the average of 10 measurements (5 measurements per side).

All children received 0.1 IU/kg per day rGH for 6 days/week for the first 6 months of treatment and were maintained on the initial dose if they achieved at least a 50% increase over baseline growth rate. Those children who did not achieve or maintain the increased rate after 6 months were subsequently treated with 0.2 IU/kg per day.

Measurement of bone metabolic peptides

Bone metabolic peptide levels were measured at baseline and every 6 months thereafter. IGF-I(7) and carboxyterminal propeptide (PICP) of type I collagen(8) were measured by Quest Laboratories (San Juan Capistrano, CA, USA). insulin-like growth factor binding protein-3 (IGFBP-3)(9) and bone specific alkaline phosphatase(10) values were determined by Endocrine Sciences (Calabasas Hills, CA, USA). Osteocalcin was measured by a homologous equilibrium radioimmunoassay developed in the laboratory by one of the authors (CMG) and used extensively in clinical studies.(11)

Bone density measurement

Lumbar (L1-L4) bone density was measured using a Hologic QDR 2000 (Bedford, MA, USA). Scans obtained over a 6-month period gave a 0.51% CV for the anterior-posterior (AP) spine. DXA scans were obtained by single beam acquisition using Hologic low-density software version 4.74A:1. Both automatic and manual bone mapping was done was done by one of the authors (JCR) without knowledge of the treatment response status of the patients.

Bone biopsy and histomorphometry

Each child underwent an iliac crest bone biopsy at baseline and after 1 year of treatment, using the contralateral iliac crest for the second procedure. Periosteum to periosteum cores were obtained 2 cm below and behind the anterior spine of the iliac crest by one operator (CMR) using a 6-mm Bordier trochar. Sections were processed and analyzed by one of the authors (FHG) without knowledge of the treatment status of the patients, as previously described.(12) Results pre- and posttreatment were compared separately for responders and nonresponders. The results were also compared with an historical normal control group from the Shriners Hospital in Montreal.

Statistical analysis

Paired t-tests were used to evaluate raw values of bone metabolic peptide (Excel, version 4.0; Microsoft, Redmond, WA, USA). Logistic regression was used to calculate the positive predictive value of PICP baseline measurements. Bone density measurements were converted to z-scores, using the Hologic pediatric reference database, before statistical evaluation. A p value less than 0.05 were considered significant. For the histomorphometric data, repeated measures ANOVA were used to evaluate results (StatView, SAS, Cary, NC, USA).


Study population

The clinical profile of the 26 participants is summarized in Table 1. Seventeen children had the moderately severe type IV OI and 9 had progressive deforming type III. Classification is based on radiographic features and skeletal and nonskeletal clinical features, including extent of short stature. Children were ages 4.5–12 years at baseline, with average and median ages of 8 and 8.5 years, respectively. All children over age 10 years were Tanner stage 2 or 3. All participants received growth hormone treatment.

Linear growth response and fracture rates

Participants were defined as “responders” or “nonresponders” based on linear growth rates (Table 1). Participants were considered pre hoc to be responders if they achieved a 50% or more increase in their linear growth rate during growth hormone treatment compared with their pretreatment growth rate. Fourteen children had a sustained response to rGH treatment, whereas 12 children were unresponsive or failed to sustain a growth response even after an increased dose of rGH for the second 6 months of treatment. The treatment growth rate of the responders was 6.4 ± 2.0 cm/year, whereas that for the nonresponders was 4.0 ± 1.7 cm/year. Type IV OI children predominated among the responders, with 10 of the 14 responders having type IV OI. The groups were not distinguishable by baseline growth rate (responders, 3.8 ± 1.6 cm/year; nonresponders, 3.7 ± 1.6 cm/year) or age at initiation of treatment (responders, 8.3 ± 2.4 years; nonresponders, 8.5 ± 2.1 years).

To estimate incremental height achieved from treatment, the height gain expected from the baseline growth rate was subtracted from the measured height gain in the treatment year. Incremental growth gains averaged 3.3 cm, ranging from 1.0 to 8.0 cm. These are minimum estimates, because several children had severe postoperative contractures. Five children achieved treatment growth rates that exceeded the normal average rate for their age.

Treatment side effects were minimal. Progression of scoliosis was unchanged compared with the NICHD OI population. Responders had a significant decrease in long bone fractures, with an average of 1.21 ± 0.7 fractures/year during the treatment year compared with 2.0 ± 1.1 fractures/year in the prior year (p = 0.030). Nonresponders fractures were unchanged (1.08 ± 1.0 fractures/year during rGH; 1.25 ± 0.6 fractures/year before rGH, p = 0.6).

Responses of growth factors and bone metabolites to treatment

Serum levels of factors in the growth hormone axis were measured at baseline and after 6, 12, and 18 months. IGF-I increased significantly (Fig. 1A) with treatment for all patients, from an average of 131.8 ± 16.3 ng/ml at baseline to 303.15 ± 28.3 (p = 2.24 × 10−8) and 378.66 ± 35.04 ng/ml (p = 4.61 × 10−8) at 6 and 12 months of treatment, respectively. IGFBP-3 also increased significantly for all patients (Fig. 2B), from 2.45 ± 0.16 mg/liter at baseline to 3.14 ± 0.18 (p = 0.00012) and 3.4 ± 0.24 mg/liter (p = 0.00012) at 6 and 12 months of treatment, respectively. There was no significant difference between responders and nonresponders in IGF-I and IGFBP-3 levels.

Figure FIG. 1..

Serum levels of growth axis hormones and bone factors in children with OI treated with gGH at baseline and after 6, 12, and 18 months of treatment. Values are shown for all participants. (A) Insulin-like growth factor-I levels. (B) Insulin-like growth factor binding protein-3 levels. (C) Osteocalcin levels. (D) Bone specific alkaline phosphatase levels.

Figure FIG. 2..

Serum levels of PICP in children with OI treated with rGH. (A) PICP levels in all participants at baseline and after 6, 12, and 18 months of treatment. (B) PICP values in responders (solid line) and nonresponders (dashed line); the groups had significantly different PICP values at baseline and after treatment.

The bone-derived proteins, osteocalcin and bone-specific alkaline phosphatase, also increased significantly in response to rGH treatment. Osteocalcin (Fig. 1C) increased from an average of 21.72 ± 3.6 to 33.8 ± 4.9 (p = 0.003) and 26.6 ± 3.9 ng/ml (p = 0.05), at 6 and 12 months of treatment, respectively. Bone-specific alkaline phosphatase (Fig. 1D) was significantly increased at 12 months, from an average of 87.0 ± 12.8 μg/ml at baseline to 112.7 ± 14.7 μg/ml at 1 year (p = 0.02). Responders and nonresponders did not differ significantly in osteocalcin and alkaline phosphatase levels.

Only the terminal peptide of type I collagen was significantly different in growth responders versus nonresponders. For all participants (Fig. 2A), PICP increased from 104.4 ± 13.9 μg/ml at baseline to 139.96 ± 17.2 μg/ml at 6 months of treatment (p = 0.0028) and remained stable at 1 year of treatment (134.7 ± 16.3 μg/ml). The PICP values of responders were higher than those of nonresponders at all three time points.

Bone density response

Vertebral (L1-L4) bone density was measured by DXA scan at baseline and after 6 and 12 months. All scans were osteoporotic and were converted to z-scores. Average z-scores improved 5% by 6 months of treatment (−5.02 ± 0.27 at baseline; −4.73 ± 0.25 at 6 months, p = 0.01) and were stable after 12 months of treatment (−4.81 ± 0.35, p = 0.05).

The improved bone density was preponderantly in the responders. Responders had a significant 5–7% improvement in bone density z-score (−4.98 ± 0.43 at baseline, −4.62 ± 0.47 at 6 months (p = 0.01), and −4.71 ± 0.33 at 12 months; p = 0.04). Nonresponders experienced a 3–5% improvement in z-score, which was not statistically significant.

Response of bone histomorphometrics

We examined the effect of rGH treatment on bone structure and formation by comparing histomorphometry from pre- and posttreatment dual tetracycline-labeled iliac crest biopsy specimens (Table 2). Parameters of bone structure were significantly lower in the two OI patient groups than in the historical normal control group,(13) as would be expected from their bone disorder. The responder group had nearly a doubling of cancellous bone volume (BV/TV) after 1 year of treatment, while this was unchanged in nonresponders. The increased responder bone volume was caused by increased trabecular number (Tb.N), a direct consequence of the growth process. Trabecular thickness (Tb.Th) and cortical width (Ct.Wi) were not significantly changed in either group.

Table Table 2.. Histomorphometric Results in OI Patients Treated With GH
original image

The baseline surface-based parameters of cancellous bone formation and resorption were higher in the total OI patient group than in controls. Of the formation parameters, the surface extent of mineralization (MS/BS) increased in responders, leading to a 50% increase in the surface-based bone formation rate (BFR/BS) and reflecting increased bone turnover. Formation parameters did not increase in nonresponders. Nonresponders had an 80% increase in the percentage of bone surface covered by osteoclasts (Oc.S/BS).

Predictors of outcome

The pretreatment PICP values were strongly predictive of a positive response to growth hormone. PICP at baseline was 129.7 ± 22.3 μg/ml in responders compared with 78.8 ± 11.2 μg/ml in nonresponders (p = 0.047). The positive predictive value of this measurement was calculated by logistic regression to be 73% with a cut-off PICP value of 86 μg/ml. Thus, a positive growth response correlates with higher synthesis of type I collagen.

Analysis of the pretreatment growth axis evaluation done on all participants did not reveal additional predictors for positive response to growth hormone. Analyses included serial 24-h GH sampling, growth hormone releasing hormone (GHRH) and clonidine tests, and IGF-I stimulation.


We have shown that the linear growth rate of some children with structurally abnormal bone matrix can be significantly increased by treatment with rGH. Most of the responders had OI type IV, with moderate bone fragility, ambulation potential, and final adult stature in the range of a young teenager. Most nonresponders had OI type III, with severe bone fragility, physical handicap, and final stature in the pre-teen range.

Responsiveness of linear bone growth to exogenous growth hormone also resulted in significant changes in direct bone measurements.(14) There was increased vertebral bone density in the responder group, probably caused in part by the increased vertebral height of responders, but possibly also reflecting changes in trabecular structure. Responders also experienced a small but significant decrease in long bone fractures during rGH treatment. We view this positive result cautiously, because fractures in OI tend to occur in cycles and the children were 1 year older during the treatment year than during the baseline year.

The pretreatment histomorphometry results were very similar to those reported previously for OI bone, with decreased bone and high cancellous bone turnover as the tissue-level hallmarks of OI.(15) Growth hormone treatment increased cancellous bone volume in responders, caused by an increase in the number of trabeculae. In the ilium of growing children, new trabeculae arise either by endochondral bone formation or by trabecularization of cortical bone.(16) Either or both of these mechanisms could be responsible here, but our data do not distinguish them. The increased bone formation rate in responders did not increase cortical width or trabecular thickness. Nor did the bone gain of trabecular number decrease cortical width. Finally, nonresponders had no increase in bone formation rate (BFR) or bone volume. Because both endochondral bone formation and cortical trabecularization are linked to bone growth, it is consistent that the nonresponders did not have an increase in trabecular number. Nonresponders, in fact, have a more unbalanced pattern after therapy than before, with an increase in osteoclast surface. The nonresponder pattern after treatment is indicative of increased bone turnover, which exceeds the capacity of osteoblasts to match it with increased matrix synthesis. This uncoupling is in contrast to the balance of high BFR and high Oc.S that is classic of OI and is seen in all our OI patients at the start of treatment. Consistent with rGH treatment tipping the balance in nonresponders toward bone resorption, we saw a trend toward decreased trabecular thickness in this group. Thus, there is no indication for rGH use in OI children who are linear growth nonresponders.

As expected, all OI children who received rGH injections had increased IGF-I and IGFBP-3, verifying that their growth hormone axis was intact.(17) They also had significantly increased osteocalcin and bone specific alkaline phosphatase, indicating increased osteoblast synthetic activity in response to hormonal stimulation.

The OI patients also experienced a significant increase in PICP. PICP has been used as a measure of response to GH in growth hormone-deficient (GHD) and in normal short children because the metabolism of the components of extracellular matrix increases with growth.(18–21) However, the baseline values in those studies did not identify good responders. In our treatment trial, PICP was the only bone metabolic marker that distinguished growth responders and nonresponders. Responders had higher PICP than did nonresponders at baseline and after 6 and 12 months of treatment. A cut-off PICP value of 86 μg/ml is 73% predictive of OI response to rGH. The emergence of collagen peptide as a predictive factor suggests that the growth response in OI is linked to the ability of osteoblasts to secrete additional type I collagen in response to hormonal stimulation.

Other skeletal dysplasias with severe growth retardation are responsive to GH therapy, although OI is the first disorder with a discrete structural defect of matrix to be shown to be responsive. Patients with achondroplasia, an autosomal dominant dysplasia caused by defective FGFR3, and Aarskog syndrome, an X-linked recessive disorder caused by defects in Rho, have shown positive growth responses over 2–4 years.(22–24) Girls with Turner's Syndrome have been treated with GH for over a decade. In general, both childhood height and final adult stature are augmented by early treatment.(25–27)

The responsiveness of linear growth to rGH in type IV OI reopens the fundamental question of why OI, a disorder of bone matrix and not of cartilage, is associated with extreme short stature. This feature of OI may be connected to the abnormal collagen present in the bone matrix or to non-collagenous proteins. The bone matrix of OI patients contains increased bone sialoprotein and decreased decorin compared with age-matched controls.(28, 29) The bone-forming cells may also have defective signaling interactions with matrix. Using a portion of the pretreatment iliac crest biopsy, we established primary osteoblast cultures(30) from both growth responders and nonresponders. These cells will permit us to distinguish the cellular and matrix characteristics of the two groups. The type I collagen mutation is known for 13 of the children in this study; neither mutation location nor type distinguish the responder and nonresponder groups, suggesting that downstream effects of the mutations are relevant.

In this study, we have shown that treatment with exogenous rGH can override the matrix defect in OI and cause a substantial increase in linear growth rate in many OI type IV children. Our results support trial treatment of rGH in individual type IV OI patients and use of baseline serum PICP as a predictor of positive response to GH therapy. In addition, growth hormone can be viewed as part of a two-drug regimen for OI. There is currently much interest in the use of bisphosphonates to treat OI based on an encouraging uncontrolled pilot study.(31) Because the mechanism of action of bisphosphonates is the inhibition of bone resorption by osteoclasts, and growth hormone exerts its effect on bone formation, the two drugs have the potential to be synergistic.


The authors acknowledge Rose Travers and Frank Rauch, MD, for histological analyses of the bone biopsy specimens and James Troendle, PhD, for assistance with statistical analyses. Stephanie Bordenick, RN, was involved in the early stages of coordination for this protocol. Frederick Ognebene, MD, and Anthony Suffredini, MD, of the National Institutes of Health CC Intensive Care Unit supervised the sedation of patients for iliac crest biopsy specimens. JCM and EH extend appreciation to Susan R Rose, MD for pleasant and fruitful discussions of multiple aspects of growth hormone treatment.