Three children with lower limb fractures and a mineralization defect: a novel bone fragility disorder?

(1)

Three children with lower limb fractures and a mineralization defect:

a novel bone fragility disorder?

Craig F.J. Munns, Frank Rauch, Rose Travers, Francis H. Glorieux*

Genetics Unit, Shriners Hospital for Children, Montre´al, Que´bec, Canada H3G 1A6

McGill University, Montre´al, Que´bec, Canada H3G 1A6 Received 12 March 2004; revised 21 June 2004; accepted 2 August 2004

Available online 28 September 2004

Abstract

In this report, we describe three unrelated children with an apparently novel bone fragility disorder that is associated with an idiopathic mineralization defect. Recurrent lower limb fractures started with weight bearing. The patients had none of the phenotypic, radiological, or histomorphometric features classically associated with known bone fragility disorders such as osteogenesis imperfecta (OI), idiopathic juvenile osteoporosis (IJO), or mild autosomal dominant osteopetrosis. Radiologically, there was increased metaphyseal trabeculation, normal to increased cortical thickness, and no evidence of rickets or osteomalacia. Areal and volumetric bone mineral density (BMD) of the lumbar spine did not show any major alteration. Peripheral quantitative computed tomography of the radius showed elevated cortical thickness and total and trabecular volumetric bone mineral density in one patient. Qualitative histology of iliac bone biopsy specimens showed a paucity of the birefringent pattern of normal lamellar bone. Quantitative histomorphometric analysis demonstrated osteomalacia with a prolonged mineralization lag time in the presence of a decreased mineral apposition rate. There was no biochemical evidence of abnormal calcium or phosphate metabolism. Type I collagen mutation analysis was negative. We conclude that this is a bone fragility disorder of moderate severity that tends to cause fractures in the lower extremities and is associated with the accumulation of osteoid due to an intrinsic mineralization defect. The pathogenetic basis for this disorder remains to be elucidated.

Keywords:Bone density; Bone fragility; Histomorphometry; Mineralization; Osteomalacia; pQCT

Introduction

Bone fragility is rare in otherwise healthy children but may cause severe clinical problems. Bone fragility may result from either an intrinsic abnormality of the skeleton (primary fragility) or from other diseases or their treatment (secondary fragility)[23]. The known primary bone fragility disorders in children fall into two categories: conditions with low bone mass and conditions with high bone mass. The two major primary bone fragility disorders associated with low bone mass are osteogenesis imperfecta (OI) and idiopathic juvenile osteoporosis (IJO)[23]. Fragility with high bone mass is seen in autosomal dominant osteopetrosis [3]. Despite our

improved understanding of pediatric bone disorders, children with apparent primary bone fragility often do not fit the pattern of any of the recognized disorders.

In this report, we describe three children with multiple lower limb fractures who were otherwise healthy and had no major alterations in bone mass. Bone histology revealed signs of a mineralization defect in bone tissue, but there was no radiographic evidence of rickets. We suggest that these three patients suffer from a novel primary bone disorder.

Subjects and methods

Patients

doi:10.1016/j.bone.2004.08.004

* Corresponding author. Genetics Unit, Shriners Hospital for Children, 1529 Cedar Avenue, Montre´al, Que´bec, Canada H3G 1A6. Fax: +1 514 842 5581.

(2)

All had been referred to this institution because of frequent fractures. The birth and fracture histories were obtained from the parents. All patients underwent clinical examina-tion, biochemical measurements, lumbar spine densitometry, and transiliac bone biopsy for bone histomorphometry. No patient received drugs that are known to interfere with bone metabolism before the transiliac bone biopsy. All patients were counseled on age appropriate calcium and vitamin D requirements [24]. Unless otherwise stated, the growth, radiological, densitometric, and biochemical data presented were obtained at the time of transiliac bone biopsy. The study protocol was approved by the Ethics Committee of the Shriners Hospital.

Histomorphometric results in these patients were com-pared to those of a group of 58 children and adolescents (age 1.5–22.9 years) without bone disease, as described earlier [5]. These individuals had undergone iliac bone biopsies during minor orthopedic procedures. Data were also compared to those of six children with X-linked hypophosphatemic rickets (age 3–15 years) who had undergone bone biopsy before starting medical therapy. Methods

Growth data were converted to chronological age- and sex-specific z scores using standard growth curves [4,7]. Biochemical parameters of bone and mineral metabolism were measured on fasting samples as recently described in detail[17]. Radiographic surveys of the entire skeleton were performed using standard techniques.

Bone densitometry was performed in the anteroposterior direction at the lumbar spine (L1–L4) using a Hologic QDR 2000W or 4500A device (Hologic Inc., Waltham, MA). Areal bone mineral density (BMD) results were transformed to age-specificzscores combining reference data from Salle et al. [21] and data provided by the densitometer manu-facturer. Volumetric BMD was derived mathematically as described[18].

Patient 1 underwent peripheral quantitative computed tomography of her nondominant forearm using an XCT-2000 device (Stratec Inc., Pforzheim, Germany), as described[11,12]. The radius was analyzed at a metaphyseal and a diaphyseal site, and results were converted tozscores using reference data established by Neu et al.[11,12].

Full-thickness bone biopsy specimens were obtained with a Bordier trephine under general anesthesia, from a site located 2 cm dorsal to the anterior superior iliac spine. Biopsy specimens were collected on the 4th or 5th day after dual tetracycline labeling (Declomycin; Wyeth-Ayerst Can-ada, Inc., Montreal, Canada). Biopsy specimens were processed and analyzed as previously described [5]. Histomorphometric results were compared to reference data, as established in our laboratory[5].

Genomic DNA from peripheral blood leukocytes was analyzed as described by Korkko et al.[8]. All exons of the COL1A1 and COL1A2 genes and their respective exon–

intron boundaries, with the exception of the six exons encoding the N-propeptides, were amplified by polymerase chain reaction. This was followed by heteroduplex mutation screening. Patient 1 also had full-length sequencing of COL1A1 and COL1A2 performed.

Results

All three patients presented with histories of fractures that started at about the age of weight bearing (Table 1). Ninety-four percent (44 out of 47) of all fractures occurred in the lower limbs, the only exceptions being fractures of a finger in Patient 1, the 12th thoracic vertebra in Patient 2, and the left distal radius and ulna in Patient 3. All fractures were documented radiologically and resulted from minimal (e.g., fall from standing height) or no trauma (spontaneous). Fracture repair was normal. At 8 years of age, Patient 1 had an intramedullary rod placed in her left tibia due to anterior tibial bowing. This rod was removed at 17.1 years of age and there was no subsequent bone deformity. Patient 3 underwent bilateral tibial and femoral rodding procedures at 3.3 years of age due to recurrent fractures and anterior tibial bowing. Patient 2 did not undergo any rodding procedures. Height was normal for Patients 1 and 2 and was slightly below the reference range in Patient 3 (Table 1). Other than the 12th thoracic vertebra in Patient 2, there were no vertebral compression fractures and no scoliosis evident on spinal radiographs. All had normal dentition, white sclera, and normal joint laxity. History and physical examination did not reveal evidence of a cardiovascular, respiratory, gastrointestinal, or endocrinological disorder. At presenta-tion, all patients were consuming their recommended daily allowance of calcium and vitamin D[24].

The three patients were from nonconsanguineous, unre-lated French-Canadian families. Family histories were negative for bone disorders. There were no fractures at birth, and birth weight and length were normal for all three patients. Table 1

Demographic and clinical data at the time of bone biopsy

Patient 1 2 3

Sex F M F

Age (years) 17.1 9.4 3.4

Birth weight (grams) 3175 3280 3345

Height (zscore) 1.74 0.33 2.18

Age at first fracture (years) 0.8 2.5 0.7

Total number of fracturesa ₁₇ ₁₅ ₁₅

Number of lower limb fracturesa ₁₆ ₁₄ ₁₄

Lumbar spine areal BMD (g/cm2₎ _1.34 _0.692 _0.337

Lumbar spine areal BMD (zscore) 2.93 0.02 3.03 Lumbar spine volumetric BMD (g/cm3₎ _0.20 _0.12 _0.08

Lumbar spine volumetric BMD (percentage of age-matched controls)

148 124 91

Lumbar spine/volumetric BMD (zscore)

5.00 3.36 N/A

Age at last follow-up (years) 18.6 12.1 11.7

(3)

Biochemical parameters of bone and mineral metabolism were within normal limits (Table 2). There was no evidence of a mutation in either of the two genes encoding collagen type I chains (COL1A1 and COL1A2) in any of the patients. Despite the frequency of the lower extremity fractures, radiographic evaluation in all patients revealed only subtle abnormalities. After the initial lower limb rodding proce-dures described above, the extremities of all patients remained straight and there were no signs of rickets, pseudofractures, or osteomalacia. The metaphyseal trabecu-lae appeared somewhat prominent, however, and the cortical width of long bones appeared to be increased in Patients 1 and 2 (Fig. 1). These qualitative observations were confirmed by peripheral quantitative computed tomographic analysis of the radius in Patient 1. At the age of 18 years, trabecular volumetric BMD at the distal radius was slightly elevated (268 mg/cm3; corresponding to azscore of +2.3), and cortical thickness at the radial diaphysis was increased (3.51 mm;zscore +2.7). The diaphyseal stress strain index, a measure of bone strength that combines bone geometry and cortical volumetric BMD, was within normal limits (259 mm3; z score 0.16). As to lumbar spine areal and volumetric BMD (Table 1), Patients 1 and 2 had results that were above or within the reference range of age-matched controls, whereas values were below the reference range in Patient 3. However, this patient had been immobilized for a total of 28 weeks (6 weeks in traction, 12 weeks in spica casts, and 10 weeks in long leg casts) during the 11 months preceding her densitometric assessment, which may have affected results.

Histologic analyses of iliac bone specimens (Fig. 2) demonstrated a conspicuous paucity of lamellae under

polarized light in both trabecular and cortical bone. There was an abundant amount of osteoid and diffuse tetracycline label uptake, suggesting a mineralization defect. This impression was confirmed by quantitative histomorphom-etry (Table 3), which consistently revealed osteoid thickness N9Am and mineralization lag timeN25 days, thus fulfilling the histomorphometric criteria for osteomalacia that we had proposed earlier[22]. The bone surface-based parameters of bone formation (mineralizing surface per bone surface, bone formation rate per bone surface) were elevated, indicating high remodeling activity. All three patients had a normal to elevated trabecular thickness. There was no evidence of hyperosteocytosis in any of the bone samples.

To investigate the mineralization defect in these patients in more detail, we compared their results to those of children without bone disorders and to patients with hypophosphate-mic rickets, a typical and well-characterized mineralization disorder. In healthy children and adolescents, there was a positive correlation between the rate of osteoid deposition (as estimated by adjusted apposition rate) and osteoid thickness (Fig. 3), which corresponds to observations made in healthy adults [14]. It is well established that in adults with osteomalacia, the relationship between adjusted apposition rate and osteoid thickness is perturbed, as osteoid thickness is elevated in the presence of low adjusted apposition rate[14]. The same is evident in children with untreated hypophos-phatemic rickets and in the three patients described in this report (Fig. 3). However, it is also obvious fromFig. 3that adjusted apposition rate in these three patients is higher than in patients with hypophosphatemic rickets, whereas osteoid thickness is similar between the two groups. It appears therefore that the mineralization defect in the three patients Table 2

Biochemical data at the time of bone biopsy

Patient 1 2 3

Serum

Calcium (mmol/l) 2.29 2.32 2.48

Phosphate (mmol/l) 1.08 1.27 1.54

Alkaline phosphatase (U/l) 124 252 228

Parathyroid hormone (39–84) (pmol/l) 8.2 8 6 25-hydroxy vitamin D (nmol/l) 60 57 45 1,25-dihydroxy vitamin D (pmol/l) 167 108 89 Creatinine (Amol/l) 61 78 60 Urine N-telopeptide

(pmol BCE/nmol creatinine)

155 399 818

Calcium/creatinine ratio (mmol/mmol)

0.34 0.28 0.38

Reference data[1,10,13,17]: Calcium: 2.20–2.70 mmol/l; phosphate: 1–3 years 1.25–2.10 mmol/l, 4–11 years 1.20–1.80 mmol/l, adult 0.90–1.50 mmol/l; creatinine: child 27–62Amol/l, adolescent 44–88Amol/l; alkaline phosphatase: 1–9 years 145–420 U/l, 16–19 years 50–130 U/l; PTH (37– 84): 2.6 – 10.0 pmol/l; 25-hydroxy-vitamin D: 34–91 nmol/l; 1,25 dihydroxy-vitamin D: 65–134 pmol/l; uNTX/uCr: Ref[1]; uCa/uCr: Ref [10].

Fig. 1. (a) Hand radiograph from Patient 2 at the age of 7 years. The growth plate is normal with no evidence of rachitic changes. There are increased trabecular markings in the proximal radius and ulna. (b) Femoral radiograph from Patient 1 at the age of 18.4 years. The bone is straight with marked cortical thickening. There is no evidence of Looser zones. No fracture had occurred to this bone.

(4)

described in this report is milder than in hypophosphatemic patients.

Discussion

In this report, we describe an apparently new bone fragility disorder that occurred in otherwise healthy children

of normal stature with minimal bone deformities. The clinical course was characterized by recurrent lower limb diaphyseal fractures, starting at the age of weight bearing. The patients were grouped together on the basis of common bone histological findings, notably a large amount of osteoid, blurred tetracycline labels, and a paucity of bone lamellation. Despite the histological evidence for a mild mineralization defect in bone tissue, rickets was absent and Fig. 2. Iliac bone specimen from Patient 2 and a healthy age-matched control.

Table 3

Histomorphometric results in iliac bone specimen

Patient 1 Patient 2 Patient 3

Age at Biopsy (years) 17.1 9.4 3.4

Structural parameters

Cortical width (Am) 1512 (1006F195) 650 (974F367) 490 (703F277)

Bone volume/tissue volume (%) 33.2 (27.8F4.5) 18.7 (22.7F4.2) 10.9 (17.7F2.6)

Trabecular thickness (Am) 208 (153F24) 130 (129F17) 117 (101F11)

Formation parameters

Osteoid thickness (Am) 10.4 (6.9F1.2) 14.0 (5.9F1.1) 15.9 (5.8F1.4)

Osteoid surface/bone surface (%) 46 (17F5) 52 (29F13) 90 (34F7)

Osteoid volume/bone volume (%) 4.8 (1.6F0.7) 11.4 (2.6F1.0) 22.1 (4.0F1.2)

Osteoblast surface/bone surface (%) 15.6 (5.3F2.7) 31.2 (8.2F4.4) 42.2 (8.5F4.1)

Single label surface/bone surface (%) 21.4 (7.1F3.7) 25.9 (9.1F2.5) 31.9 (8.7F2.7)

Double label surface/bone surface (%) 7.1 (5.1F2.4) 16.9 (10.3F3.6) 29.8 (8.2F2.4)

Mineralizing surface/bone surface (%) 17.8 (7.9F2.7) 29.8 (14.9F4.5) 45.8 (12.5F4.5)

Mineralizing surface/osteoid surface (%) 39 (58F14) 57 (50F21) 51 (38F13)

Mineral apposition rate (Am/day) 0.46 (0.75F0.09) 0.81 (0.95F0.1) 0.51 (1.04F0.17)

Adjusted apposition rate (Am/day) 0.18 (0.43F0.12) 0.46 (0.47F0.18) 0.26 (0.40F0.16)

Mineralization lag time (days) 58 (17.3F6.5) 30 (14.1F4.3) 61 (16.7F6.4)

Bone formation rate/bone surface (Am3Am 2year 1) 30 (22F9) 88 (51F16) 86 (48F19) Resorption parameters

Eroded surface/bone surface (%) 7.3 (18F6) 27.9 (17F6) 7.9 (15F4)

Osteoclast surface/bone surface (%) 0.8 (1.0F0.4) 1.8 (1.3F0.6) 1.5 (1.1F0.8)

(5)

parameters of bone and mineral metabolism were within normal limits. This suggests an intrinsic abnormality in the incorporation of mineral into the osteoid.

Whereas a large number of disorders with primary bone fragility in children have been described, most of these have easily distinguishable characteristics[23]. Three conditions have been delineated where primary bone fragility occurs in children that can otherwise appear healthy—OI, IJO, and autosomal dominant osteopetrosis[23]. The most common of these disorders is probably OI, a heritable disease with increased bone fragility, low bone mass, and variable other connective tissue manifestations. Seven types of OI have been described [16]. The disorder described in our three patients resembles OI type I with regard to fracture rates and growth pattern. However, most children with OI type I have osteoporosis, which was not a consistent feature in our patients. OI type I is typically associated with blue sclera and mutations in collagen type I, whereas our patients had white sclera and there were no detectable mutations in the COL1A1 and COL1A2 genes. Bone histology usually reveals hyperosteocytosis in OI type I[20], which was not found in our patients. In addition, there is no evidence for a mineralization defect in OI type I[20].

The histologic features of the present disorder bear some resemblance with OI type VI. Both diseases are characterized by histologic signs a mineralization defect in the absence of rickets or disturbances in bone and mineral metabolism[6]. However, bone tissue in OI type VI has a characteristicd fish-scale patternTunder polarized light, which was not present in the bone samples from our patients. Also, OI type VI is clinically much more severe, with long bone deformities and multiple vertebral crush fractures. Nevertheless, it cannot be ruled out that OI type VI and the condition described here

represent the extreme ends of the spectrum of a single disease of uncertain etiology.

Although poorly characterized, IJO is commonly described as a transient, nonhereditary bone fragility disorder of uncertain etiology associated with metaphyseal long bone fractures and wedge compression fractures of the spine[9,19]. Children with IJO typically present a few years before puberty and there is spontaneous improvement after puberty [9,19]. Thus, fractures occur at a later age in IJO than in our patients and they are associated with evident osteoporosis. Further, bone histomorphometry in IJO is characterized by low cancellous bone volume and bone formation rate, with no evidence of a mineralization defect

[19].

Children with autosomal dominant osteopetrosis may develop bone pain and sustain lower limb fractures[3]. Two of our patients had slightly elevated bone density readings, which however were nowhere near the results seen in osteopetrosis. Radiographic findings also were not sugges-tive of osteopetrosis[2].

In the most common forms of disordered bone mineralization, there is a slowing of mineral incorporation into osteoid due to calcium or phosphorus substrate deficiency, as occurs in vitamin D deficiency or hereditary hypophosphatemia [15]. None of the three patients in this report had an abnormality of phosphate or calcium metabolism or renal function that could explain their mineralization disorder. Serum alkaline phosphatase, which is often elevated in association with mineralization defects, was also normal. During growth, disorders that are associated with impaired mineralization of bone tissue usually lead to a problem with mineralization of growth plate cartilage as well, which then becomes clinically and radiologically evident as rickets. The normal serum biochemistry and the lack of rachitic changes in this disorder suggest that the mineralization defect is limited to the bone matrix and distinguishes it from previously described forms of osteomalacia.

The disorder described in this report is probably rare, as it was found in only 3 out of 128 children who had undergone bone biopsy at our institution as part of a work-up to investigate bone fragility. At present, it is not established that the disorder is hereditary, as none of the affected individuals in this report have affected parents or offspring.

In summary, we describe a bone fragility disorder in otherwise healthy children that can be diagnosed on the basis of unique bone histological findings. The etiology of this condition remains to be determined.

Acknowledgments

This study was supported by the Shriners of North America. CM is a Royal Children’s Hospital Foundation/ Woolworths Scholar.

Fig. 3. Relationship of osteoid thickness to adjusted apposition rate in the three patients described in this report (black circles). For comparison, the same relationship is shown for children without bone disorders (grey squares) and in six children with untreated hypophos-phatemic rickets (grey triangles, age 3–15 years). The drawn grey line corresponds to the regression line in children without bone disorders. The dotted oblique line through the origin corresponds to a mineraliza-tion lag time (Mlt) of 25 days. The horizontal line represents an osteoid thickness (O.Th) of 9Am.

(6)

References

[1] Bollen AM, Eyre DR. Bone resorption rates in children monitored by the urinary assay of collagen type I cross-linked peptides. Bone 1994;15:31 – 4.

[2] Bollerslev J. Autosomal dominant osteopetrosis: bone metabolism and epidemiological, clinical, and hormonal aspects. Endocr Rev 1989;10:45 – 67.

[3] Bollerslev J. Osteopetrosis. A genetic and epidemiological study. Clin Genet 1987;31:86 – 90.

[4] Buschang PH, Tanguay R, Demirjian A. Growth instability of French-Canadian children during the first three years of life. Can J Public Health 1985;76:191 – 4.

[5] Glorieux FH, Travers R, Taylor A, Bowen JR, Rauch F, Norman M, Parfitt AM. Normative data for iliac bone histomorphometry in growing children. Bone 2000;26:103 – 9.

[6] Glorieux FH, Ward LM, Rauch F, Lalic L, Roughley PJ, Travers R. Osteogenesis imperfecta type VI: a form of brittle bone disease with a mineralization defect. J Bone Miner Res 2002;17:30 – 8.

[7] Keen DV, Pearse RG. Birthweight between 14 and 42 weeks’ gestation. Arch Dis Child 1985;60:440 – 6.

[8] Korkko J, Ala-Kokko L, De Paepe A, Nuytinck L, Earley J, Prockop DJ. Analysis of the COL1A1 and COL1A2 genes by PCR amplification and scanning by conformation-sensitive gel electro-phoresis identifies only COL1A1 mutations in 15 patients with osteogenesis imperfecta type I: identification of common sequences of null-allele mutations. Am J Hum Genet 1998;62:98 – 110.

[9] Lorenc RS. Idiopathic juvenile osteoporosis. Calcif Tissue Int 2002;70:395 – 7.

[10] Matos V, van Melle G, Boulat O, Markert M, Bachmann C, Guignard JP. Urinary phosphate/creatinine, calcium/creatinine, and magnesium/ creatinine ratios in a healthy pediatric population. J Pediatr 1997;131:252 – 7.

[11] Neu CM, Manz F, Rauch F, Merkel A, Schoenau E. Bone densities and bone size at the distal radius in healthy children and adolescents: a study using peripheral quantitative computed tomography. Bone 2001;28:227 – 32.

[12] Neu CM, Rauch F, Manz F, Schoenau E. Modeling of cross-sectional bone size, mass and geometry at the proximal radius: a study of

normal bone development using peripheral quantitative computed tomography. Osteoporos Int 2001;12:538 – 47.

[13] Nicholson JF, Pesce MA. Reference ranges for laboratory tests and procedures. In: Behrman RE, Kliegmen RM, Jenson HB, editors. Nelson textbook of pediatrics. Philadelphia7 W.B. Saunders; 2000. p. 2181 – 229.

[14] Parfitt AM. Vitamin D and the pathogenesis of rickets and osteomalacia. In: Feldman D, Glorieux FH, Pike W, editors. Vitamin D. San Diego7Academic Press; 1997. p. 645 – 62.

[15] Rauch F. The rachitic bone. Endocr Dev 2003;6:69 – 79.

[16] Rauch F, Glorieux FH. Osteogenesis imperfecta. Lancet 2004;363: 1377 – 1385.

[17] Rauch F, Plotkin H, Travers R, Zeitlin L, Glorieux FH. Osteogenesis imperfecta types I, III and IV: effect of pamidronate therapy on bone and mineral metabolism. J Clin Endocrinol Metab 2003;88:986 – 92. [18] Rauch F, Plotkin H, Zeitlin L, Glorieux FH. Bone mass, size, and density in children and adolescents with osteogenesis imperfecta: effect of intravenous pamidronate therapy. J Bone Miner Res 2003; 18:610 – 4.

[19] Rauch F, Travers R, Norman ME, Taylor A, Parfitt AM, Glorieux FH. Deficient bone formation in idiopathic juvenile osteoporosis: a histomorphometric study of cancellous iliac bone. J Bone Miner Res 2000;15:957 – 63.

[20] Rauch F, Travers R, Parfitt AM, Glorieux FH. Static and dynamic bone histomorphometry in children with osteogenesis imperfecta. Bone 2000;26:581 – 9.

[21] Salle BL, Braillon P, Glorieux FH, Brunet J, Cavero E, Meunier PJ. Lumbar bone mineral content measured by dual energy X-ray absorptiometry in newborns and infants. Acta Paediatr 1992; 81:953 – 8.

[22] Terpstra L, Rauch F, Plotkin H, Travers R, Glorieux FH. Bone mineralization in polyostotic fibrous dysplasia: histomorphometric analysis. J Bone Miner Res 2002;17:1949 – 53.

[23] Ward L, Glorieux FH. The spectrum of pediatric osteoporosis. In: Glorieux FH, Pettifor J, Jueppner H, editors. Pediatric bone: biology and disease. San Diego7Academic Press; 2003. p. 401 – 42. [24] Young VR, Erdman JW, King JC, Allen LH, Atkinson SA, Dwyer

JT, et al. Dietary reference intake for calcium, phosphorus, magnesium, vitamin D and fluoride. Washington7National Academy Press; 1997.