A randomized, controlled dose-ranging study of risedronate in children with moderate and severe osteogenesis imperfecta

Abstract

Moderate to severe osteogenesis imperfecta is associated with multiple fractures in childhood. There are no published data regarding the effects of third-generation bisphosphonates in these children. This randomized study investigated which of three different doses of risedronate was most effective in reducing fracture incidence. We randomly assigned 53 children with moderate to severe osteogenesis imperfecta to receive 0.2, 1, or 2 mg/kg per week of risedronate. We assessed safety, fracture incidence, and bone measurement outcomes at 3, 6, 12, 18, and 24 months. At 24 months, 69% of children assigned 0.2 mg/kg per week had had new fractures compared with 44% receiving 1 mg/kg per week and 75% receiving 2 mg/kg per week. Poisson regression with age and prior fracture as covariates showed that there was no difference in incident nonvertebral fracture between groups. Fracture rate diminished in each group during the trial compared with previous the 2 years (p = .005). Lumbar spine bone mineral density increased significantly (p = .009) only in the 2 mg/kg per week group. Long bone bowing deformities reduced more in children receiving 1 or 2 mg/kg per week of risedronate [odds ratio (OR) 0.67, 95% confidence interval (CI) 0.48–0.93 per unit increase in risedronate dose, p = .015]. There were no serious adverse events. Bone mass increased and bowing deformities reduced with increasing risedronate dose. Children suffered fewer fractures irrespective of risedronate dose. The most appropriate dose of risedronate for children with moderate to severe osteogenesis imperfecta in this study was 2 mg/kg per week. © 2010 American Society for Bone and Mineral Research

Introduction

Children with osteogenesis imperfecta (OI) suffer recurrent fractures resulting in pain, deformity, and disability. Without treatment, severely affected children suffer chronic bone pain and repeated fractures of their limbs and vertebrae, resulting in progressive bony deformity and impaired mobility. Some die in childhood.1 Randomized, controlled studies have shown that olpadronate2 given orally each day and pamidronate3 and neridronate4 given intravenously at 3-month intervals increase bone mass and reduce fracture incidence in affected children.

Absorption of bisphosphonates given orally is poor.5 Weekly dosing has improved compliance in adults and reduced adverse effects.6 There have been no studies comparing the efficacy of different doses of an oral bisphosphonate given weekly in children with osteogenesis imperfecta.

The apparent success of intravenous pamidronate in improving both clinical course and quality of life for children with osteogenesis imperfecta has made placebo-controlled trials difficult for families to accept. We therefore conducted a randomized, double-blind, dose-ranging intervention study using the bisphosphonate risedronate. The study tested the hypothesis that children with moderate to severe OI receiving the highest dose of risedronate, 2 mg/kg per week, would have fewer fractures than children receiving the lowest dose, 0.2 mg/kg per week. The doses were chosen to span the anticipated acceptable range of safety and efficacy.

Methods

Patients were recruited through clinics in Sheffield Children's Hospital (n = 47), Birmingham Children's Hospital (n = 2), and Glasgow Yorkhill Hospital (n = 4) from 2001 to 2004. Ethical review and permission for the study were obtained from the appropriate review boards. Written informed consent was obtained in all cases. The study International Standard Randomized Controlled Trial Number (ISRCTN) is ISRCTN67376467.

Children were recruited if they had phenotypic OI with one or more of the following features: recurrent fractures affecting mobility, two or more crush-fractured vertebrae, or fractures causing bony deformity and requiring surgical correction. Children were excluded if they had had previous bisphosphonate therapy or had another chronic disease likely to affect bone metabolism. Children were randomized to receive 0.2, 1, or 2 mg/kg per week risedronate for 2 years, dose prescribed to the nearest 5 mg. Doses administered were multiples of the same tablet of sizes of 2.5, 5, 15, 30, and 35 mg; for instance, a child weighing 50 kg randomized to 2 mg/kg per week would receive 3 × 35 mg, i.e., 105 mg. Randomization was by multiple permuted blocks stratified according to age (up to 10.99 years or above) and allocated by a remote telephone system. Medication was provided through the pharmacy at the Sheffield Children's Hospital. Medication bottles were returned to the pharmacy at each study visit, and pill counts were undertaken to assess compliance. Children, their families, and medical staff were blinded to the treatment assignment.

Children took the medication after an overnight fast using a dosing spoon (a type of screw-top test tube) in which the medication was dissolved in water. The dosing spoon was rinsed with water that was then swallowed (twice), followed by a large glass of water [typically 250 mL (∼8 oz.)].

We did not formally assess calcium intake but rather milk and dairy product intake (using the diary kept by each child) and provided calcium supplementation as required when the calculated calcium intake fell below the UK Recommended Nutrient Intake (RNI)7 for calcium. Children were reviewed after 3, 6, 12, 18, and 24 months for safety and clinical and radiologic purposes.

The primary outcome for the study was the number of nonvertebral fractures in each group.

Skeletal outcomes

Despite encouragement, many families did not take their children for radiologic evaluation of possible fracture because of concerns over radiation exposure and because of their familiarity with the procedures needed for treatment, i.e., analgesia and immobilization. The numbers of fractures recorded here were those reported by families as resulting in severe pain and consequent immobilization, confirmed, where possible, by examination of their relevant radiographs.

Fracture history prior to entering the study was recorded in detail (i.e., site, degree of trauma, and timing). In particular, we noted the period of 2 years leading up to study entry as a baseline against which to assess the subsequent response. During the period of the study, a pain diary was kept, fracture episodes were noted, and copy X-ray films were obtained to ascertain fracture numbers.

A skeletal survey was performed annually. This provided vertebral morphometry using separate films of the lumbar and thoracic vertebrae in lateral projection, metacarpal cortical width at the midpoint of the second metacarpal bone of the left hand using vernier calipers to the nearest 0.1 mm, bone age of the left wrist by Tanner Whitehouse 3, and scoliosis noted on anteroposterior (AP) spine views.

Bowing of the long bones of the legs was assessed visually by all three radiologists at baseline and study end. Bowing was recorded as improved, no change, or worse for each child based on his or her combined opinion. No child underwent operative procedures to correct bowing during the period of the study.

Vertebrae were classified according to their size and shape. The height loss was assessed as no loss (0% to 10%), moderate (10% to 50%) or severe (>50%). The shape was assessed as normal or showing single-end-plate, double-end-plate, or anterior-wedge deformity, with the anterior wedging being the most severe deformity and single-end-plate deformity the least severe. The films were assessed in pairs for each child.

Numbers of other fractures were noted. X-rays were assessed “blind” for fractures by a panel of three expert pediatric radiologists (SC, CH, and AS).

We measured total-body and lumbar spine bone outcomes by dual-energy X-ray absorptiometry (DXA; GELunar Prodigy, software version 4.0) at each visit. We estimated lumbar spine (L2–4) bone area, bone mineral content, areal and volumetric bone density and total-body (less head) bone area, bone mineral content, and areal bone density. Age-adjusted Z-scores for areal bone density were recorded. The assessment of bone mass in children using DXA is confounded by the areal nature of the technique, which fails to adjust adequately for changes in bone size during growth. The outcomes thus are assessed after adjusting the values for each measurement at baseline. An approximation to volumetric bone density in the lumbar spine using the approach of Kroger8 was employed. The assessment of the total-body outcomes after removal of the head follows the procedure recommended by the Pediatric Development Conference of the International Society for Clinical Densitometry (www.iscd.org/Visitors/positions/OfficialPositionsText.cfm#PEDIATRIC).

Height was measured to the next succeeding 1 mm (wall-mounted Holtain stadiometer); where disability precluded standing, measurements were made lying. Weight was measured wearing underwear to the nearest 0.1 kg (Seca 770 scale, Birmingham, UK).

Functional outcomes

The Pediatric Evaluation of Disability Inventory (PEDI) and the Gross Motor Function Measure (GMFM) were used to provide estimates of overall function. Grip strength was assessed using a hand dynamometer (Biometrics, Ltd., Gwent, UK).

Safety data were recorded at each visit. Adverse events were reported centrally and followed up by the study coordinator within 48 hours. Pain and diet diaries were kept; pain was recorded using a Wong-Baker faces scale, and episodes of fracture-like pain with consequent immobilization were specifically noted.

Safety and adverse events

Biochemical data were used to assess the degree of suppression of bone turnover in order to ensure safety. We measured bone-specific alkaline phosphatase (BSALP; in-house wheat germ lectin precipitation assay) and urinary cross-linked N-terminal telopeptide of type I collagen (NTx; Johnson & Johnson, Ortho Clinic Diagnostics ECi, New York, USA) at each visit, together with renal and liver function and a full blood count. Serum 25-hydroxyvitamin D (HPLC) and parathyroid hormone (PTH; immunochemiluminometric assay; Roche Elecsys) were measured annually. Any child with a serum 25-OH vitamin D level below 25 nmol/L received vitamin D treatment with 6000 (up to age 10 years) or 10,000 IU/d (10 years old and older) of vitamin D for 2 months.

Statistics

We based our sample-size calculation on available data from the Montreal study of 30 children receiving pamidronate,9 where the annualized fracture rate fell from 2.2 ± 2.2 to 0.6 ± 0.5 fractures per year. The reported reduction in the number of untreated children experiencing a fracture after 1 year is 11%.10 This study was powered to show a reduction in incident nonvertebral fractures in the highest dose group compared with the lowest dose group, where, if treatment were ineffective, fracture rate would be similar to that in untreated children. The size of the expected difference was 1.5 fractures per child averaged over each group (i.e., 24 fewer fractures in the highest as opposed to lowest dose group). Data were analyzed as intention to treat. The per-child frequency of fracture was analyzed by Poisson regression. Age at study entry and prior fracture frequency were used as covariates in the analyses. Secondary outcomes were total-body and lumbar spine bone area, bone mineral content, areal bone mineral density, and volumetric bone mineral density (Kroger method). Further exploratory analyses were undertaken of the effects of giving risedronate on vertebral morphometry, bowing of long bones, metacarpal cortical width, and functional outcomes.

Results

We recruited 53 subjects over the period from April 2001 to July 2004. Five subjects withdrew during the 2-year period of the study and were lost to follow-up (Fig. 1). One child did not like the taste of the tablets, one developed an inflammatory bowel condition not thought to be related to the medication, and three withdrew without giving a reason. The demographic data are shown in Table 1 together with the baseline anthropometry and growth data over the 2-year period. There was no alteration of growth in any of the groups.

Figure 1. Consort diagram for the study.

Table 1. Demographic and Anthropometric Data

	Dose group
	0.2 mg/kg/week	1 mg/kg/week	2 mg/kg/week
Note: Anthropometric data are expressed as raw data, and Z-scores are adjusted for age. No significant differences between groups for any anthropometric measurement or change in measurement. There were no differences between the groups for age, gender, or OI type.
Age at entry (median, range) (years)	10.8, 3.8–17.0	10.8, 5.2–16.5	11.0, 5.8–17.0
Gender (M/F)	6/11	8/10	4/14
OI type (I/III/IV)	0/4/13	0/2/16	1/1/16
Actual risedronate dose (mg; median, range)	0.18, 0.16–0.26	0.99, 0.89–1.16	1.99, 1.73–2.37
Anthropometry
At trial entry	n = 17	n = 18	n = 18
Height (cm, mean ± SD)	122.4 ± 30.5	125.0 ± 20.4	129.9 ± 23.2
Height Z-score (mean ± SD)	−3.03 ± 3.32	−2.77 ± 2.59	−2.41 ± 2.21
Weight (kg, mean ± SD)	32.9 ± 19.9	31.4 ± 13.1	31.9 ± 10.6
Weight Z-score (mean ± SD)	−1.39 ± 2.43	−1.41 ± 1.88	−1.28 ± 1.04
At 2 years	n = 16	n = 16	n = 16
Height (cm, mean ± SD)	126.7 ± 26.7	135.0 ± 22.2	139.6 ± 24.3
Height Z-score (mean ± SD)	−3.28 ± 2.70	−2.72 ± 2.78	−2.13 ± 2.24
Median change in height range (cm)	9.3, 17.8	10.5, 17.0	9.9, 18.6
Weight (kg, mean ± SD)	37.1 ± 24.1	37.0 ± 14.8	38.7 ± 12.2
Weight Z-score (mean ± SD)	−1.41 ± 2.21	−1.54 ± 2.02	−1.04 ± 1.19

Bone outcomes

The frequency of nonvertebral fractures is documented for each group in Table 2. The overall number of nonvertebral fractures during the 2 years of the study was reduced compared with the number of such fractures reported for the 2 years prior to entry in the study (Wilcoxon sign rank test, p = .005). The variance of the number of new fractures was 3.1, in comparison to a mean of 1.5 indicating overdispersion. A Poisson regression model with adjustment for age, sex, number of previous fractures, OI type, and baseline Z-scores for both height and weight still showed overdispersion (χ² = 89.5, df = 39, overdispersion index = 2.29, p < .001). However, there was no significant relationship between risedronate dose and the incidence of new fractures whether or not overdispersion was taken into account (no overdispersion assumed, t = 0.61, df = 1, p = .65; correcting for overdispersion, t = 0.40, df = 1, p = .76). The scatterplots in Fig. 2 indicate the change in fracture frequency by child in each group.

Table 2. Fracture Outcomes by Group

Numbers of fractures suffered by subject, by group, over the 2-year period before the study and during the 2-year study period
	0.2 mg/kg/week		1 mg/kg/week		2 mg/kg/week
Numbers of fractures	2 years prior to study; subject n = 17	During study; subject n = 16	2 years prior to study; subject n = 18	During study; subject n = 16	2 years prior to study; subject n = 18	During study; subject n = 16
0	5	5	4	9	4	4
1	3	6	2	1	1	3
2	4	2	3	3	4	5
3	2	2	2	1	3	3
4	0	0	4	0	3	0
5	1	1	2	0	3	1
6	0	0	0	1	0	0
7	1	0	1	0	0	0
8	1	0	0	1	0	0
Total fractures	37	21	47	24	45	27
Dropouts' prior fractures	0		2, 3		2, 0

Figure 2. Graphs show incidence of nonvertebral fractures during the course of the study versus fractures during the last 2 years prior to entering the study for each of the dose groups.

Table 3 summarizes the analyses for each bone outcome at study end after adjusting for the baseline value of each outcome and age. Positive effects of risedronate on the lumbar spine and total-body bone mineral content, areal bone mineral density, age-related areal density, and volumetric bone mineral density were seen in the 2 mg/kg per week group. The largely negative coefficients for the 0.2 mg/kg per week group that approach significance (p = .06 for bone mineral content, areal density, and age-adjusted areal density) suggest that this dose is ineffective in the treatment of children with OI and may be associated with bone loss; the 1 mg/kg per week dose was not associated with a positive response at any site.

Table 3. DXA-Estimated Outcomes

	0.2 mg/kg/week			1 mg/kg/week			2 mg/kg/week
	Baseline mean (SD)	Coeff. (SE)	p Value	Baseline mean (SD)	Coeff. (SE)	p Value	Baseline mean (SD)	Coeff. (SE)	p Value
Note: DXA estimated outcomes; BA, bone area; BMC, bone mineral content; aBMD, areal bone mineral density; aBMD, age adjusted BMD; zBMD, Z-score BMD; LS, lumbar spine; TB, total body; BMAD, calculated volumetric BMD in the lumbar spine only; head excluded from TB estimates. Estimates adjusted for age and baseline value.
LSBA, cm^2	19.8 (7.5)	−0.29 (0.68)	0.67	21.7 (7.4)	−0.57 (0.66)	0.40	21.7 (6.1)	0.86 (0.66)	0.20
LSBMC, g	12.0 (8.7)	−2.4 (1.3)	0.06	12.3 (6.8)	−0.4 (1.2)	0.75	12.7 (6.3)	2.8 (1.3)	0.03
LSaBMD, g/cm^2	0.479 (0.186)	−0.040 (0.023)	0.10	0.531 (0.125)	−0.023 (0.022)	0.31	0.558 (0.135)	0.063 (0.023)	0.009
LSzBMD	−3.8 (2.4)	−0.56 (0.29)	0.06	−3.0 (2.1)	−0.06 (0.28)	0.82	−3.1 (1.4)	0.62 (0.28)	0.03
LS BMAD, g/cm^3	0.107 (0.029)	−0.011 (0.005)	0.04	0.116 (0.017)	−0.004 (0.005)	0.46	0.121 (0.020)	0.015 (0.005)	0.005
TBBA, cm^2	814 (527)	−41 (37)	0.26	873 (414)	−45 (36)	0.21	885 (429)	86 (36)	0.02
TBBMC, g	566 (475)	−45 (52)	0.40	621 (383)	−67 (50)	0.19	621 (377)	112 (50)	0.03
TBaBMD, g/cm^2	0.621 (0.131)	−0.028 (0.019)	0.15	0.671 (0.155)	−0.016 (0.018)	0.39	0.662 (0.103)	0.044 (0.018)	0.02
TBzBMD	−2.2 (2.3)	−0.37 (0.19)	0.06	−1.9 (2.7)	−0.18 (0.19)	0.35	−2.3 (2.0)	0.55 (0.19)	0.006

The differences between the highest and lowest groups for each outcome were significant. The size of the difference can be assessed by subtracting the coefficient for each of the outcomes in the 0.2 mg/kg per week group from the respective 2 mg/kg per week outcomes. Thus, for lumbar spine bone mineral apparent density (LSBMAD), the difference in Z-score between the groups was 0.026 g/cm³, equivalent to a difference over mean baseline of 22% in volumetric bone mineral density (BMD).

The positive outcomes shown for the 2 mg/kg per week group could reflect altered body size. We undertook additional exploratory analyses to pursue this possibility (not shown). The coefficients in the 2 mg/kg per week group remained positive for each of the bone outcomes, although the size of each coefficient decreased slightly. For example, for total-body BMD, the coefficient (SE) reduced from 0.044 (0.019) to 0.043 (0.018) g/cm² but remained significant at p = .026. Other bone size and mass outcomes changed to a similar degree.

We did not demonstrate any difference between the groups for improvement in size or shape of vertebrae or the resolution or worsening of scoliosis (Tables 4a–4d). Twelve children had loss of height in one or more vertebrae that was classified initially as normal during the course of the study. Metacarpal cortical thickness did not change significantly with treatment (see Tables 4a–4d).

Table 4a. Initial Shape and Degree of Height Loss of Vertebrae and Change in Appearance After Two Years of Treatment by Dose Group

	0.2 mg/kg/week	1 mg/kg/week	2 mg/kg/week
SEPD, single-end-plate deformity; DEPD, double-end-plate deformity; AW, anterior wedging. Some vertebrae had more than one deformity. Values are mean (SD) percentage. Total number of vertebrae assessed for change 500; 15 excluded as changes were equivocal (shape improved/size reduced; size improved/shape worsened).
Initial shape
Normal	12 (17)	23 (26)	19 (28)
SEPD	12 (14)	19 (17)	19 (18)
DEPD	73 (33)	55 (35)	56 (35)
AW	59 (26)	45 (28)	48 (29)
Initial size
No loss	25 (26)	34 (30)	27 (27)
10–50% loss	21 (34)	7 (16)	11 (28)
51–100% loss	54 (28)	59 (27)	60 (29)
Appearance after 2 years
Improved	31 (19)	32 (21)	36 (20)
No change	50 (23)	45 (27)	35 (26)
Worse	19 (13)	23 (20)	29 (19)

Table 4b. Number of Children Where Scoliosis Developed, Remained Unchanged, or Progressed

	0.2 mg/kg/week	1 mg/kg/week	2 mg/kg/week
AP films not available for 4 children at 2 years.
Appearance after 2 years
Worse	4	3	5
No Change	11	9	10
Improved	0	2	0

Table 4c. Metacarpal Cortical Outcomes After Two-Years Treatment by Dose Group

	0.2 mg/kg/week			1 mg/kg/week			2 mg/kg/week
	Baseline mean (SD)	Coeff. (SE)	p Value	Baseline mean (SD)	Coeff. (SE)	p Value	Baseline mean (SD)	Coeff. (SE)	p Value
Outcomes are adjusted for starting value and age.
External width (mm)	5.86 (1.10)	0.10 (0.06)	0.12	5.94 (1.07)	−0.02 (0.06)	0.69	6.19 (1.02)	−0.07 (0.06)	0.21
Medullary width (mm)	3.12 (0.65)	−0.06 (0.12)	0.63	3.17 (1.32)	0.02 (0.11)	0.86	3.07 (0.55)	0.04 (0.11)	0.75
Cortical width (mm)	2.73 (1.13)	0.15 (0.11)	0.17	2.80 (0.81)	−0.04 (0.10)	0.71	3.12 (1.09)	−0.11 (0.10)	0.28

Table 4d. Bone Age Parameters Showing Bone Age and Chronologic Bone Age by Dose Group

	0.2 mg/kg/week		1 mg/kg/week		2 mg/kg/week
	Bone age	Chronologic bone age	Bone age	Chronologic bone age	Bone age	Chronologic bone age
Values are mean (SD).
First visit	10.5 (4.3)	0.45 (1.82)	10.3 (3.8)	0.12 (0.43)	10.9 (3.7)	0.56 (1.28)
Study end	11.1 (4.2)	0.78 (1.37)	12.0 (3.3)	0.58 (1.58)	12.1 (3.0)	0.72 (1.04)

Bowing deformities of long bones, as assessed qualitatively by the radiologists, were reduced according to dose of treatment (see Tables 4a–4d). The odds ratio by logistic regression for the relationship between bowing and each unit increase in risedronate dose was 0.48 [95% confidence interval (CI) 0.36–0.65, p < .0001] unadjusted and 0.67 (95% CI 0.48–0.93, p = .015) after adjustment for age, sex, previous fracture rate, and OI type.

Two children had tibial rods inserted, and one child had a femoral rod inserted; these children were excluded from the analysis of bowing deformity. No adjustment was made with respect to the rods in analyzing the total-body scans.

Bone age was similar in all three groups at study start and slightly but not significantly reduced; bone age was reduced further compared with chronologic age at study end, but again, the difference between groups did not reach significance (see Tables 4a–4d).

Functional outcomes

Assessment of the children was hampered by the lack of an OI-specific tool that could account for contemporaneous pain or recent fracture. There were no significant differences in Pediatric Evaluation of Disability Inventory and Gross Motor Function Measure scores between the groups. Grip strength in the right and left hands, respectively, was 33.6 ± 23.9 and 31.8 ± 21.6 lb at study initiation and 38.4 ± 20.9 and 39.3 ± 21.8 lb at study end. Neither difference was significant. There were no treatment-related differences. We showed no difference between the groups in pain scores.

Biochemical changes

Bone-specific alkaline phosphatase activity declined by 1% from 81 ± 28 to 80 ± 39 IU/L (maximum decline for any child was 63%) and NTx by 21% from 2575 ± 1618 to 2044 ± 1031 nmol/mmol bone collagen equivalents (BCE); (maximum decline for any child 70%) over the 2-year period of the study. The decline in both markers was very similar between dose groups. There was no relationship between change in bone markers and any bone size or mass outcome either within or between the dose groups.

One child was persistently vitamin D deficient with a raised PTH (>65 ng/L) despite receiving vitamin D treatment (10,000 IU/day vitamin D for 2 months), suggesting noncompliance. PTH was elevated in two other children at baseline but fell into the normal range with vitamin D therapy. There was no relationship between any bone mass outcome and either PTH or 25 hydroxyvitamin D in serum at baseline or study end.

Adverse events

The adverse events are shown in Table 5. There were no significant differences between the groups for any type of event. There were no serious adverse events during the study, with the exception of fracture.

Table 5. Adverse Events by Group

Body system	Treatment dose			Total
	0.2 mg/kg/week	1 mg/kg/week	2 mg/kg/week
Note: One patient in group 2 reported 24 AEs over a 13-month period; it was not thought that any of the events were study drug-related. a Neurologic AEs were 7 headaches, 2 blank episodes/appearing to lose conciousness,1 dizzy/light-headed, 1 fall, hit head (last 4 from patient gp 2 with 24 AEs).
GI	4	11	4	19
Musculoskeletal	6	13	7	26
Cardiovascular	0	2	0	2
Respiratory	0	6	1	7
Skin	1	1	0	2
Allergy	1	0	0	1
Dental	0	1	0	1
Neurologica	0	9	2	11
Pale	1	0	0	1
Tongue ulcers	0	0	1	1
Infection	0	1	1	2
Urogenital	0	5	1	6
Psychological	0	1	0	1
Not categorized	0	1	1	2
Total	13	51	18	82

Discussion

This is the first study to report the effects of a third-generation oral bisphosphonate in children with moderate or severe OI. We failed to show any effect of risedronate dose on the incidence of nonvertebral fracture, although risedronate reduced fracture frequency overall compared with pretreatment fracture rate. We showed that 2 mg/kg per week of risedronate increased spine and total-body bone size, mass, and density and that long bone bowing was improved with both 1 and 2 mg/kg per week of risedronate.

We based our sample size calculation on the reported reduction in fracture frequency during open-label studies of intravenous pamidronate.9 Risedronate doses used in adult diseases range from 5 mg/day for osteoporosis11 to 30 mg/day for 60 days, repeated once, for Paget's disease12 (approximately 25 to 50 mg/kg per year). Bone turnover is higher in children than in adults and greater still in children with OI.13, 14 We therefore decided to study the efficacy of risedronate at doses ranging from 10 to 100 mg/kg per year. We hypothesised that the highest dose of risedronate might reduce fracture frequency more than the two lower doses.

The reduction in total fracture numbers from 129 in the 2 years before treatment to 72 during our study was similar to that shown by Sakkers.2 In Sakkers' study, the placebo group of 18 children suffered 50 fractures over 2 years and the treatment group 18. The average number of fractures in each of our groups before treatment was 43, reducing to 24.

Our failure to show an effect of risedronate dose on fracture frequency may reflect entry into a system that reduced fracture risk because of specialist clinical care. Alternatively, the antifracture effect of risedronate could be achieved at low doses. Recent data from in vitro studies suggest that low concentrations of bisphosphonates can have positive effects on osteocyte survival.15 This mechanism would enhance mechanical strain sensing and strengthen bone. Fracture data were recorded prospectively once children were in the study but retrospectively with respect to prior fracture. Given that recall may be imperfect, it may be that the number of prior fractures for the children was underestimated, blunting the apparent effect of treatment overall. However, this problem likely would affect the groups equally, so the likely difference between groups would remain nonsignificant.

Bone mass, adjusted for covariates, increased significantly only in children receiving 2 mg/kg per week risedronate. The increase over baseline after 2 years for lumbar spine bone density was 11% and for total body less head was 7% after adjusting for age. Essentially, this means that the increase in spine bone density was 11% greater and for total body less head 7% greater than that expected for age alone. The coefficients for bone outcomes were close to zero in the 1 mg/kg per week group, suggesting that the change over 2 years was that expected for age but without catch-up. The significant negative coefficient for volumetric BMD in the 0.2 mg/kg per week group suggests a falling away from the expected bone mass accretion trajectory.

While we did not observe a statistically significant effect of treatment dose on growth, the increase in height Z-score was greatest in the 2 mg/kg per week group. Since this would affect bone size and hence overall mass, it might be the case that some of the positive change in bone outcomes for the 2 mg/kg per week group could reflect this. Our additional analyses did suggest that some of the effect could be explained by variation in body size, but the differences between groups remained significant. The possibility that was some of the difference between groups might reflect a positive effect on growth could be regarded as a positive outcome. Larger bones are stronger bones, and much of the antifracture efficacy of bisphosphonates is likely to be due to their effect on bone size rather than material density. The unadjusted increase in bone density reported in open-label studies of pamidronate is typically 25% to 40% at the lumbar spine3, 9, 16 and 10.4% at the total body including head13 after 1 year. Typical annual increases in healthy children are 6% to 8.5% and 2.8% to 4.1%, respectively.13 However, earlier studies of pamidronate typically enrolled more severely affected children; in our previous open-label study of pamidronate, 7 of 20 children had type III disease. Percentage increases are greater if the starting point is low. There were only 7 children with type III disease, the most severe form of OI, in this study of 53 children.

Vertebral morphometry improved but did not completely resolve over a period of 2 to 4 years in Land's observational study17 of children receiving pamidronate, consistent with our observations here. We saw an improvement in approximately one-third of vertebrae, assessed qualitatively, in each treatment group but also alterations to a worse shape or size in 20% to 30% of vertebrae. We found that scoliosis was more likely than not to progress irrespective of treatment dose. Given the overall lack of improvement in vertebral morphometry, we suggest that a better understanding of the effect of early treatment on the genesis of scoliosis is needed.

We demonstrated, for the first time, that bisphosphonate treatment improved bowing deformity of the limbs. This suggests that treatment enables bone to respond appropriately to mechanical loading.

We did not find any relationship of change in bone markers to change in bone mass. No child showed any biochemical or radiologic evidence of oversuppression of bone turnover. There was no evidence of undertubulation of bone. We thus saw no immediate skeletal ill effects of risedronate therapy in these children. Nevertheless, the bone age data suggest that we need to exercise caution when considering the longer-term effects of bisphosphonate therapy in children. Although the differences in bone age were not statistically significant either within or between the groups, the fact that they were retarded after 2 years of therapy suggests that longer-term monitoring of the effect of treatment on skeletal maturity would be warranted.

We saw no major adverse events in our study. In common with Sakkers,2 we saw no evidence of a reduction in pain assessed using a visual analogue scale.

Intravenous bisphosphonate therapy with pamidronate requires hospital admission at regular intervals, often for 2 to 3 days at a time. Hospital-based therapy disrupts schooling and requires parents to take time off work. Using risedronate, the number of hospital visits can be reduced and timing made more flexible to better suit families' needs.

Cyclic intravenous pamidronate, often continuing for many years, remains the most frequently reported treatment for children with OI. Pamidronate still may be the treatment of choice for infants and young children, in whom adherence to risedronate dosing procedures may be difficult or, for some immobile children, impossible. Risedronate may be a suitable alternative for many children because its mineral affinity is lower than that of other nitrogen-containing bisphosphonates18 such as alendronate and olpadronate. This means that in the event of any long-term ill effects, as yet unreported, risedronate would exit the skeleton more rapidly than any other bisphosphonate. This could be of particular relevance for girls who later wish to start a family.

The use of bisphosphonates in children with OI has had a major impact on the course of the disease. This study reinforces the view that bisphosphonates are effective in reducing fracture risk, increasing bone mass, and improving bone shape in children with moderately severe OI and provides evidence to support the use of risedronate at a dose of 2 mg/kg per week in such children.

Disclosures

Nick Bishop provides consultation for Proctor & Gamble, Novartis, and Roche with respect to the use of bisphosphonates in children. Nick Bishop is lead investigator, and Faisal Ahmed is a coinvestigator on the Alliance for Better Bone Health Pediatric Osteogenesis Safety and Efficacy (POISE) study that also uses risedronate. Richard Eastell consults and receives research funding from Proctor & Gamble Pharma, Sanofi-Aventis, and Novartis. All other authors have no conflicts of interest.

Acknowledgements

The Arthritis Research Campaign funded this study; Proctor & Gamble also provided financial support to the study via the Arthritis Research Campaign. Study drug was provided by the Alliance for Better Bone Health via Proctor & Gamble, Cincinnati, Ohio. The study would not have been possible without the wholehearted support of the children and their families.