Osteoporos Int
DOI 10.1007/s00198-014-2840-0
ORIGINAL ARTICLE
Bone mineral density and vitamin D status in ambulatory
and non-ambulatory children with cerebral palsy
A.-K. Finbråten & U. Syversen & J. Skranes &
G. L. Andersen & R. D. Stevenson & T. Vik
Received: 29 March 2014 / Accepted: 1 August 2014
# International Osteoporosis Foundation and National Osteoporosis Foundation 2014
Abstract
Summary This study assessed distal femur and lumbar spine
bone mineral density (BMD) Z-scores in children with cere-
bral palsy. BMD z-score was lower in non-ambulatory than in
ambulatory children. Somewhat surprisingly, among ambula-
tory children, those with better walking abilities had higher
BMD z-score than those with more impaired walking ability.
Introduction Children with cerebral palsy (CP) have in-
creased risk for low bone mineral density (BMD). The aim
was to explore the difference in BMD at the distal femur and
lumbar spine between ambulatory and non-ambulatory chil-
dren with CP and the relationship between vitamin D status
and BMD.
A.<K. Finbråten (*) : J. Skranes : G. L. Andersen : T. Vik
Department of Laboratory Medicine, Children’s and Women’s
Health, Norwegian University of Science and Technology, Olav
Kyrres gt.11, 7489 Trondheim, Norway
e-mail: ane.finbraten@gmail.com
U. Syversen
Department of Cancer Research and Molecular Medicine,
Norwegian University of Science and Technology, Trondheim,
Norway
A.<K. Finbråten
The Department of Pediatrics, St. Olav’s University Hospital,
Trondheim, Norway
U. Syversen
The Department of Endocrinology, St. Olav’s University Hospital,
Trondheim, Norway
G. L. Andersen
The Cerebral Palsy Register of Norway, Habilitation Center, Vestfold
Hospital Trust, Tønsberg, Norway
R. D. Stevenson
Department of Pediatrics, Division of Developmental Pediatrics,
Kluge Children’s Rehabilitation Center and Research Institute,
University of Virginia, Charlottesville, VA, USA
Methods Fifty-one children (age range 8–18 years; 20
girls) with CP participated. Their BMD Z-scores were
measured in the lumbar spine and the distal femur using
dual X-ray absorptiometry, and 25-hydroxy-vitamin D
(25-OHD) concentrations were measured in serum. Chil-
dren with GMFCS level I–III were defined as ‘walkers’
while children with level IV–V were defined as ‘non-
walkers.
Results Non-walkers had lower mean BMD Z-scores (range
−1.7 to −5.4) than walkers at all sites (range −0.8 to −1.5).
Among walkers, BMD Z-scores at the distal femur were lower
in those with GMFCS level II than with level I (p values<
0.004). A similar difference was found between the affected
and unaffected limb in children with hemiplegia. Mean 25-
OHD concentration was 45 nmol/L (SD = 18); lower in
walkers (mean=41 nmol/L; SD=18) than in non-walkers
(mean=53 nmol/L; SD=19; p=0.041). There were no corre-
lations between 25-OHD and BMD z-scores.
Conclusions The main predictor of low BMD Z-scores in the
distal femur was the inability to walk, but the results suggest
that the degree of the neuromotor impairment may also be a
significant predictor. Vitamin D status did not correlate with
BMD z-scores.
Keywords Bone mineral density . Cerebral palsy . Dual X-ray
absorptiometry . Motor function . Vitamin D insufficiency
Abbreviations
AED antiepileptic drug
BMD bone mineral density
CP cerebral palsy
DXA dual x-ray absorptiometry
GMFCS Gross Motor Function Classification System
25-OHD 25-hydroxy-vitamin D

Introduction
Cerebral palsy (CP) is the most common cause of severe
physical disability among children, and children with CP are
at increased risk for reduced bone mineral density (BMD) [1].
Several factors may contribute to low BMD in this population,
including immobility, low nutritional status, low vitamin D
status, and the use of anticonvulsant drugs [2, 3]. As a conse-
quence of low BMD, children with CP may sustain painful
fractures often following minimal trauma [4–7].
Vitamin D is important for calcium metabolism and accre-
tion of bone mass during growth [8]. The skin produces
vitamin D when exposed to sunlight, and the vitamin is also
found naturally in some foods, in particular fatty fish. Children
with CP may be less exposed to sunlight than children without
CP [4]. Moreover, a significant proportion has feeding prob-
lems [9] and insufficient intake of calcium and vitamin D has
been reported [10], even in tube-fed children [11]. Conse-
quently, their vitamin D status may be deficient or insufficient,
and this may contribute to poor mineralization of the bone.
Furthermore, it has been shown that vitamin D supplementa-
tion reduces rate of fractures in children with CP [12].
Dual X-ray absorptiometry (DXA) is the optimal method
for clinical assessment of BMD in children and adults [13].
BMD results in children are reported as z-scores, which rep-
resent the standard deviation from the mean BMD of age- and
sex-matched peers. A BMD z-score less than minus two is
labeled “low BMD” [14]. In adults, BMD is normally
assessed at the lumbar spine and hip. However, Henderson
et al. [15] found that BMD in the lumbar spine was not
predictive of fracture risk in children with spastic quadriplegic
CP. Instead, he and his coworkers demonstrated that BMD
could be measured at the distal femur [16], which may be
clinically more useful since this is a common site of fractures
in non-ambulatory children with CP [17]. Henderson et al. [4]
used this region of interest to measure BMD in children with
CP within Gross Motor Function Classification System
(GMFCS) levels III–V. They observed that the children unable
to walk had reduced BMD at the distal femur. However, many
of the risk factors for low BMD may also be present in CP
children who are able to walk, such as lack of exposure to
sunshine, use of anticonvulsants, and poor nutritional status.
Nevertheless, measurements of BMD at the distal femur are
rarely reported in children with CP with GMFCS levels I and
II [18, 19]. Furthermore, utilizing the measurement at the
distal femur site has to our knowledge only been applied in
one study of BMD in European children with CP, and they
only included children with GMFCS levels IV and V [20].
In this study, the aim was to evaluate the BMD in the
lumbar spine and distal femur by examining a heterogeneous
group of children with CP across all levels of gross motor
function. We hypothesized that ambulatory children (walkers)
would have higher BMD than non-ambulatory children (non-
Osteoporos Int
walkers) and that there would be a positive correlation be-
tween vitamin D concentration and BMD.
Material and methods
Study design and population
A sample of 64 children with CP across all functional ability
levels who lived close to the city of Trondheim in the counties
of Sør- and Nord-Trøndelag, Norway was invited to participate
in this study. The inclusion criterion was a diagnosis of CP
according to the Surveillance of Cerebral Palsy in Europe
(SCPE) [21] and age between 6 and 18 years. There were no
exclusion criteria. Fifty-one (80 %) children (20 girls) accepted
to participate and were examined by one of the authors (AKF)
from January to May 2010 (n=25) or during the same months
in 2013 (n=26). The clinical examination, DXA-scanning, and
blood analyses were all performed on the same day.
Exposure variables
The Gross Motor Function Classification System (GMFCS)
[22] was used to describe gross motor function on a five level
scale, where children at level I and II are able to walk without
support, children in level III often need assistive mobility
devices and frequently orthoses to walk, while children at
level IV function in sitting, but have very limited self-mobil-
ity. Finally, children in level V are completely unable to walk
and need support in the sitting position. The classification in
GMFCS levels was indicated by the parents and confirmed by
data recorded in the Cerebral Palsy Register of Norway
(CPRN). The CP subtypes were recorded as dyskinetic, atax-
ic, spastic unilateral, and spastic bilateral CP according to the
criteria proposed by the SCPE [21]. In some of the analyses,
we divided the study population into walkers (GMFCS levels
I–III) and non-walkers (GMFCS levels IV–V).
Outcome variables
BMD (g/cm2) was assessed by DXA (Hologic Discovery)
performed by a trained technician. Standard scanning proce-
dures were used for the lumbar spine and whole body, and
BMD Z-scores calculated. In addition, we measured BMD z-
score at the distal femur, where patients were placed in a
lateral position with the non-scanned limb flexed at the hip
and knee so that it did not overlie the limb that was scanned
[23]. The mean of the right and left distal femur z-score was
calculated and used in the analysis. In cases where we only
had measurement from one limb (n=4), we used the data from
the measured limb.
BMD at the distal femur was assessed in three separate
regions (Fig. 1), where region 1 (R1) is just proximal to the
Osteoporos Int
Fig. 1 The three regions of the distal femur; region 1 (R1) containing
trabecular bone, region 2 (R2) containing a mixture of cortical and
trabecular bone, and region 3 (R3) containing cortical bone
growth plate and is considered to contain mainly trabecular
bone. The in-between located region 2 (R2) covers the transi-
tion from metaphysis to diaphysis and is considered to contain
a mixture of cortical and trabecular bone. Finally, proximal
from R2 is region 3 (R3) that consists mainly of cortical bone.
The technician mapped the regions of interest manually on the
scanned distal femur. The percent coefficient of variation of
BMD in the three regions has been reported to be less than
3 %; (R1, 2.9 %; R2, 2.3 %; and R3, 2.6 %) based on duplicate
scans in five healthy children [23].
Since the measurement in distal femur has not been used in
Norway earlier, we used age-, gender-, and race-normalized
values obtained from a healthy United States (US) pediatric
population to calculate Z-scores [24]. Standard height (cm)
measurements were available in walkers, and for comparisons
within this group, whole body and lumbar spine BMD Z-
scores were adjusted for height according to Zemel et al.
[25]. In non-walkers, standard height was not obtainable,
and adjustment for height was not performed in the compar-
isons of BMD Z-scores of walkers with non-walkers. Mean
BMD Z-scores below −2 SD in region 3 were defined as low
BMD for age, since it has been reported that this region has the
highest relative risk for fractures [26].
Covariates
Weight was measured with an electronic chair-weight
(M40010, ADE, Germany) to the nearest 100 g. Standing
height was measured in all children able to stand in an upright
position using a fixed stadiometer and recorded to the nearest
0.1 cm. Knee-heel length (knee height) was measured in both
legs in all the children to the nearest 0.1 cm with a calliper
(Holtain LTD, Crosswell, UK) for the purpose of estimating
height in children unable to stand applying the equations pro-
posed by Stevenson et al. [27]. We used both standing height
and estimated height to calculate body mass index (BMI)
(weight (kg)/height2 (m)) and height z-score. Based on the
previous LMS models for national references for weight,
height, and BMI for age, z scores were estimated using the
formula: sds = ((Measurement/M)L-1)/(L × S), where M repre-
sents the median, L the power to remove skewness, and S the
coefficient of variation [28]. Current use of antiepileptic drugs
(AED) and a previous history of fractures were also recorded.
Pubertal state was reported by the adolescents themselves based
on exposed pictures showing the different Tanner stages [29].
The parents indicated if their child had feeding problems.
A peripheral venous blood sample was obtained from 43
(84 %) of the 51 study participants, with the remainder de-
clining, and was analyzed for serum concentrations of 25-
hydroxy-vitamin D (25-OHD), total and ionized calcium,
phosphate, albumin, alkaline phosphatase, and parathyroid
hormone (PTH) at the Department of Clinical biochemistry
at St. Olav’s University Hospital, Trondheim. In 2010, a
radioimmunoassay (RIA), Immunodiagnostic Systems (IDS)
was used, while in 2013 25-OHD serum concentration was
analyzed with the Roche Cobas e601 immunology
analyser™. Serum concentrations of 25-OHD of 75 nmol/L
or higher indicates optimal vitamin D status, concentrations
between 50 and 75 nmol/L adequate levels, between 30 and
50 nmol/L insufficient, and levels of less than 30 nmol/L
indicate profound vitamin D deficiency [30].
Statistical methods
Statistical analyses were performed using IBM SPSS Statistics
for Windows (version 19.0). Differences in mean values be-
tween groups were analyzed using Mann–Whitney U test and
for the comparison of BMD Z-scores within patients (e.g.,
lumbar spine with distal femur) related-samples Wilcoxon
Signed Rank test was used. A general linear model was
applied to adjust for the potential confounding effect of use
of AED and body size (height and BMI z-scores). In the
comparisons of walkers with non-walkers, we used estimated
height for all children, while when we compared children with
GMFCS level I with II, we used Z-scores calculated from
standing height. Pearson’s correlation coefficients were calcu-
lated to study the association between 25-OHD concentration
and BMD z-scores. As proposed by Portney [31], the strength
of a correlation may be described as 0.00 to 0.25=no relation-
ship, 0.26 to 0.50=low degree of relationship, 0.51 to 0.75=
moderate to good relationship, and 0.76 to 1.00=good to
excellent relationship. Finally, chi-squared statistics were
 Osteoporos Int

applied to study differences in proportions, and binary logistic
regression was applied to calculate the odds ratios (OR) with
95 % confidence intervals (CI). The OR for low mean BMD
Z-scores for age at the distal femur R3 in non-walkers
(GMFCS levels IV–V) was calculated using walkers
(GMFCS levels I–II) as the reference, and among walkers,
the OR that children with GMFCS level II would have low
BMD for age was calculated using children with GMFCS
level I as the reference. Two-sided p values below 0.05 were
considered significant.
Ethics
The Regional Ethical Committee for Medical Research in
mid-Norway approved the study protocol (Reference number:
2009/1842-2). Informed consent was obtained from each par-
ticipant’s parent or legal guardian, and from the participants.
classified with GMFCS level I and four with level II. Twelve
had right and ten had left hemiplegia. The table also shows that
12 children (24 %) were currently using AED, and 11 children
(22 %) had experienced a previous bone fracture. Two partici-
pants used multiple AED. In all, four patients used valproate,
five lamotrigine, three used carbamazepine, one used
vigabatrine, and information was missing in one. The mean
age of the children when they had their bone fractures was 6.4
(SD=3.5), and the fracture sites were femur (1), tibia (1), fibula
(1), foot (4), forearm (3), and multiple sites (1). The parents of
eight children reported that their child had feeding problems;
however, only five children had a gastrostomy. None of the
participants had a growth hormone deficiency or received endo-
crine therapy. The median age of the participants was 13 years,
2 months (range 8 to 18 years). Table 2 shows that non-walkers
were significantly shorter than walkers.

Bone mineral density

Results
The distribution of pubertal status, CP subtypes, and gross motor
impairments according to GMFCS is reported in Table 1. There
were 22 children in this study with hemiplegia. Eighteen were
At the distal femur, BMD was measured on the left side in 50
(98 %) participants, on the right side in 46 (90 %) participants,
and in the lumbar spine of 49 (96 %) participants. The reasons
for missing BMD measurements at the distal femur were
metallic implants (n=3) or lack of cooperation (n=2). The

Table 1 Characteristics of the 51

children with cerebral palsy (CP) GMFCS I GMFCS II GMFCS III GMFCS IV GMFCS V Total
included in the present study
n value 20 11 5 9 6 51
Sex
Boy 13 7 1 4 6 31
Girl 7 4 4 5 0 20
Pubertal status*
Prepubescent 8 6 3 6 3 26
Pubescent 4 2 1 1 1 9
Postpubescent 7 3 1 2 1 14
CP subtype
Spastic
Unilateral 18 4 1 0 0 23
Bilateral 2 5 4 8 6 25
Dyskinetic 0 1 0 1 0 2
Ataxia 0 1 0 0 0 1
Feeding difficulties**
Yes 0 2 2 1 3 8
No 18 9 3 7 3 40
Anticonvulsant therapy*
Yes 2 6 0 2 2 12
No 16 5 5 7 4 37
Fracture History**
Yes 4 4 1 2 0 11
*Missing two cases
No 15 6 4 6 6 37
**Missing three cases
Osteoporos Int

Table 2 Results of the anthropometric measurements (mean Z-scores level II had significantly lower mean BMD Z-scores than
with standard deviations (SD)) and age in children with CP dichotomized
children with GMFCS level I at the distal femur (Table 4).
in walkers (GMFCS* level I–II) and non-walkers (GMFCS levels III–V)
Neither whole body nor lumbar spine BMD Z-scores differed
Gross Motor Classification System between the two groups. The difference in mean BMD Z-
scores in R3 at the distal femur between children with
Total Level I-III Level IV-V
n value 51 36 15 GMFCS level I (estimated mean = −0.60; CI = −1.13 to
mean (SD) Mean (SD) Mean (SD) p-value* −0.07) and II (estimated mean=−2.1; CI=−2.83 to −1.39)
was reduced when we adjusted for height z-score, but was
Age 13.21 (2.81) 13.24 (2.81) 13.13 (2.89) .891 still highly significant (p=0.003). Adding use of AED or BMI
Weight Z-score −0.73 (2.14) −0.64 (2.22) −0.93 (1.98) .668 Z-scores to this model did not change the estimated means
Height1 Z-score −1.55 (1.81) −1.23 (1.92) −2.36 (1.20) .047 further. No differences in mean BMD Z-scores were observed
BMI1 Z-score 0.22 (1.91) 0.04 (1.95) 0.66 (1.80) .310 between children with GMFCS level II and III.
When we compared mean BMD Z-scores in all three
*t test statistics between GMFCS level I–III and IV–V
1 regions of the distal femur between the affected and
Based upon estimated height from knee-height according to equations
by Stevenson et al. unaffected limb in children with hemiplegia (n=22), the
affected leg had significant lower mean BMD Z-scores
(Table 5).
missing BMD measurements in the lumbar spine were due to
the position of an intrathecal baclofen pump in two patients.
Table 3 shows that in the total population mean BMD z- Low-bone mineral density for age in region 3 of the distal
score in the lumbar spine was −1.1 (SD=1.3; range −4.2 to + femur
1.3). At the distal femur, the mean z-score ranged from −2.7
(SD=2.8) to −1.7 (SD=1.9). BMD Z-scores were significant- Twenty-one (42 %) children had mean BMD Z-scores below
ly lower in all three regions of the distal femur than in the −2 at this site, defined as low BMD for age. In the group of
lumbar spine in the total population (all p values below 0.001). non-walkers, 11 (71.4 %) out of 14 participants had low BMD
When we reanalyzed our data separately within each GMFCS for age, while in the group of walkers 11 (30.6 %) of the 36
level, lower Z-scores at the distal femur compared with the children had low BMD (p=0.009; Table 6). The OR for low
lumbar spine were found in all children except for in children mean BMD was 5.7 (CI=1.5–22.1) in children unable to walk
with GMFCS level I (data not shown). using children able to walk as reference.
Non-walkers (GMFCS IV-V) had lower mean BMD Z- Within the group of walkers, seven (64 %) of 11 children
scores at the distal femur (p values=<0.005) compared with with GMFCS level II had low BMD for age, compared with
walkers (GMFCS I–III) (Table 3). This difference was unaf- only two (10 %) of 20 children with GMFCS level I (p=
fected when adjusted for potential confounders (height Z- 0.002). Using children with GMFCS level I as reference,
scores, BMI z-scores, and use of AED) (data not shown). children with GMFCS level II had an OR for low BMD for
Moreover, within the group of walkers, children with GMFCS age of 15.8 (CI=2.3–106).

Table 3 Mean bone mineral density Z-scores with standard deviations (SD) in the lumbar region and at the distal femur in children with CP in the total
population and dichotomized in walkers (GMFCS level I–III) and non-walkers (GMFCS levels IV–V)

Gross Motor Classification System

Total population Level I–III Level IV–V p value*

Number of children 50 36 14**

mean (SD) mean (SD) mean (SD)

Lumbar −1.1 (1.3) −0.8 (1.2) −1.7 (1.2) 0.085

Distal femur
R1 −1.7 (2.1) −1.2 (1.8) −3.0 (1.8) 0.007
R2 −2.7 (3.1) −1.6 (1.8) −5.2 (3.2) .000
R3 −1.9 (1.9) −1.2 (1.5) −3.4 (1.7) .000

*Mann–Whitney U test between walkers and non-walkers

**One missing distal femur scan in GMFCS level IV
 Osteoporos Int

Table 4 Mean bone mineral density (BMD) Z-scores with standard Table 6 Cross-table between Gross Motor Function (GMFCS) and low
deviation (SD) in children with CP with gross motor function classifica- mean bone mineral density (BMD) for age (i.e., Z-score below −2) at the
tion (GMFCS) level I and II at the total body less head and in the lumbar distal femur region 3
spine adjusted for height and in three regions of the distal femur not
adjusted for height GMFCS level Distal femur region 3 Total

Gross motor function classification <−2 >=−2

(GMFCS)
Level I n value 2 18 20
Level I Level II p value* % 10.00 90.00 100.00
Number of children 20 11 Level II n value 7 4 11
% 63.60 36.40 100.00
BMD Z-scores: mean (SD) mean (SD) Level III n value 2 3 5
% 40.00 60.00 100.00
Adjusted total body less head −1.5 (0.9) −2.2 (1.1) 0.824
Level IV n value 4 4 8
Adjusted lumbar spine −0.4 (1.1) −1.4 (1.3) 0.834
% 50.00 50.00 100.00
Distal femur
Level V n value 6 0 6
R1 −0.2 (1.3) −2.3 (1.1) 0.001
% 100.00 0.00 100.00
R2 −0.7 (1.2) −2.8 (1.2) 0.000
Total n value 21 29 50
R3 −0.5 (1.2) −2.4 (0.9) 0.001
% 42.00 58.00 100.00
*Independent sample Mann–Whitney sample test
Chi-squared statistics: p<0.001

Table 8 shows the relationship between 25-OHD concen-

Vitamin D status
trations and BMD Z-scores in the total population as well as
among walkers (GMFCS I–III) and non-walkers (GMFCS
As shown in Table 7, the mean 25-OHD concentration was
IV–V). In the total population, as well as in non-walkers the
45 nmol/L (range 10–96), in the total population. Seven
coefficients were negative and statistically significant in re-
children (16 %) had 25-OHD serum concentrations below
gion 1 at the distal femur. In walkers, the correlation coeffi-
30 nmol/L, while 24 children (56 %) had concentrations
cients suggested no relationship.
between 30 and 50 nmol/L. Thus, a total of 31 (72 %) children
had a vitamin D level lower than recommended. Non-walkers
2010 cohort compared with 2013 cohort
had significantly higher mean 25-OHD values than walkers.
None of the subjects had abnormal PTH, albumin, alkaline
When we reanalyzed our data separately within each year
phosphatase, or phosphate. Two individuals had low total
cohort, the results within the 2-year cohorts were essentially
calcium concentrations (1.29 and 2.14 mmol/L), and these
the same as in the total population (data not shown).
children had 25-OHD concentration of 47 and 39 nmol/L,
respectively. In addition, one child had low total calcium
(2.09 mmol/L) and ionized calcium (1.10 mmol/L), and also
a low 25-OHD concentration (29 nmol/L).
Discussion

Table 5 Mean bone mineral density Z-scores with standard deviation Key results
(SD) at the distal femur in the affected and unaffected leg in children with
hemiplegia We found that a significant proportion of children with CP had
low BMD for age at the distal femur and that in particular
Affected limb Unaffected limb p value*
children unable to walk had very low BMD z-scores. Some-
Number of children 22 22 what surprisingly, we found that among children able to walk
those with GMFCS level II had significantly lower BMD Z-
BMD Z-scores: mean (SD) mean (SD) scores than children with GMFCS level I. Also, among chil-
Distal femur dren with hemiplegia BMD Z-scores were lower in the affect-
R1 −1.1 (1.8) −0.4 (1.4) 0.001 ed than in non-affected limb.
R2 −1.3 (1.4) −0.8 (1.0) 0.002
Finally, approximately two thirds of the children had insuf-
R3 −0.9 (1.3) −0.5 (1.0) 0.005
ficient vitamin D status and every sixth child had a deficient
status. In contrast to our hypothesis, we did not find a positive
*Paired Wilcoxon Signed Rank test correlation between 25-OHD status and BMD z-scores.
Osteoporos Int

Table 7 Result of biochemical analyses in the total population as well as in walkers (GMFCS I–III) and non-walkers (GMFCS IV–V)

Total population GMFCS I-III GMFCS IV-V p value*

Number of children 43 29 14

mean (SD) mean (SD) mean (SD)

25-OHD (nmol/L) 45 (18) 41 (16) 53 (19) 0.041

Calcium (mmol/L) 2.33 (0.18) 2.31 (0.22) 2.38 (0.07) 0.223
Ionized calcium (mmol/L) 1.25 (0.04) 1.24 (0.04) 1.26 (0.03) 0.040
Albumin (g/L) 45.9 (1.9) 45.9 (2.33) 45.8 (1.8) 0.949
Phosphate (mmol/L) 1.28 (0.22) 1.28 (0.25) 1.29 (0.18) 0.999
Alkaline phosphatase (U/L) 198 (96) 211 (111) 185 (73) 0.220
Parathyroid hormone (pmol/L) 3.59 (1.18) 3.70 (1.18) 3.39 (1.19) 0.453

*t statistics

Strength and limitations
Strengths of the present study are that the scans were analyzed
by one technician, who was unaware of the children’s vitamin
D status and GMFCS levels, and that we were able to include
children across all GMFCS levels. Moreover, the distribution
of GMFCS levels and CP subtypes in our study population
was similar to the population of children with CP in Norway
[32], suggesting that our population is representative for the
general population of children with CP in Norway.
The main results regarding BMD Z-scores are unlikely to
be due to chance as demonstrated by the low p values. How-
ever, in some of the subgroup analyses, the numbers are low
and lack of statistical significance should therefore be
interpreted with caution. This applies in particular for the
comparison of BMD Z-scores between children with GMFCS
level II and III.
Our estimates of the proportions of children with low BMD
for age at the distal femur were based upon reference data
obtained in the USA [23], and might have been slightly
different if we had access to Norwegian reference values.
However, this potential bias should not affect the observed
differences across GMFCS levels, between walkers and non-
walkers, or between children with GMFCS level I and II.
Another potential bias could be that the study population
included participants recruited and examined during the
spring of two different years. However, the number of partic-
ipants recruited in each year was similar, and when we
reanalyzed our data separately for the 2-year cohorts, the
results within each year cohort were essentially the same as
for the whole population (data not shown).
The differences in height Z-scores between the various
groups could be a potential confounder of the association
between low femur BMD Z-scores in non-walkers compared
with walkers and in children with GMFCS level I compared
with level II. In children with abnormal stature, it is recom-
mended to adjust BMD Z-scores for height z-score following
an equation proposed by Zemel et al. [25, 33]. However, such
equations have to our knowledge not been developed for the
distal femur, and in children with CP, accurate height is
difficult to obtain, in particular in non-walkers. Finally, the
proposed equations for height adjustment have not been val-
idated in children with CP. Nonetheless, the results of the
multivariable analyses where we calculated estimated means
for BMD Z-scores adjusting for height Z-scores did not sug-
gest that our main findings were confounded by the differ-
ences in height between walkers and non-walkers, or between
participants with GMFCS level I and II.
Other potential confounders could be age, sex, and the use
of AED. However, the BMD Z-scores are adjusted for sex and

Table 8 The correlation (Pearson correlation coefficient, r) between 25-OHD and mean BMD Z-scores in the lumbar spine and in the distal femur in the
total population, as well as in walkers (GMFCS I–III) and non-walkers (GMFCS IV–V)

GMFCS Lumbar Distal femur R1 Distal femur R2 Distal femur R3

I–III 25-OHD (nmol/L) r value −0.021 −0.086 −0.024 −0.027

Sig. (2-tailed) 0.915 0.657 0.904 0.889
IV–V 25-OHD (nmol/L) r value −0.339 −0.574* −0.270 −0.284
Sig. (2-tailed) 0.281 0.040 0.373 0.347
Total 25-OHD nmol/L r value −0.176 −0.352* −0.296 −0.280
Sig. (2-tailed) 0.271 0.022 0.057 0.073

age, and the results of the multivariable analyses suggested
that the excess risk for low BMD in children with GMFCS
level II compared with children with level I was only partly
confounded by the use of AED, and in a model excluding
adjustment for height and BMI z-scores. Nevertheless, the
difference in BMD Z-scores between the two groups persisted
after adjustment for these confounders. In the comparison of
walkers with non-walkers, the multivariable analyses sug-
gested that the difference in BMD was not affected by use of
AED or by height z-scores. Thus, confounding is unlikely to
explain our main findings.
Regarding the children’s vitamin D status, we collected the
blood samples during the winter/spring months when vitamin
D concentrations are at the lowest making confounding by
season unlikely. Unfortunately, the method for analyzing 25-
OHD changed between 2010 and 2013, and this may explain a
slightly, albeit not statistically significant higher 25-OHD
concentrations in the 2013 cohort compared with the 2010
cohort (data not shown). However, this change in methodol-
ogy did not affect the results when we explored our data for
the 2-year cohorts separately.
Comparison with other studies
Our findings that the children’s GMFCS levels were important
predictors of low BMD are consistent with the previous stud-
ies [4, 16, 34], although the lower mean BMD Z-scores at the
distal femur of non-walkers compared with walkers have not
been previously described. Henderson et al. reported that non-
walkers had lower BMD Z-scores in the proximal femur
compared with walkers [16], and Wren et al. using quantita-
tive computed tomography found lower volumetric BMD in
the tibia of non-walkers [35]. Finally, our results may be
consistent with Chen et al. who described diminished areal
BMD at the distal femur in walkers compared with non-
walkers [36].
In line with Henderson et al. [26], the lowest distal femur
mean BMD z-score in our study population was in region 2 of
the distal femur. The finding of lower mean BMD Z-scores at
the distal femur compared with the spine is also consistent
with Henderson et al. [4, 26]. However, in contrast to Hen-
derson et al. [16], we were not able to demonstrate a statisti-
cally significant difference in BMD z-score of the lumbar
spine between walkers and non-walkers. Since the mean Z-
scores at this site in walkers and non-walkers in our study
were very similar to the values reported by Henderson et al.,
we consider that this may be a question of statistical power.
A noteworthy and new finding of the present study is that
within the group of walkers, children with GMFCS level II
had lower mean BMD Z-scores in all three regions of the
distal femur and more often had low BMD for age than
children with GMFCS level I. This finding is in contrast to
Chen et al. who reported that the distal femur areal BMD did
Osteoporos Int
not differ between children with CP with GMFCS levels I and
II [18]. There are several potential explanations for the differ-
ent findings. In addition to the differences in ethnicity and
potential exposure to sunshine, we applied z-score taking into
consideration gender, ethnicity, and age, while they used
absolute data (g/cm2). Moreover, our study population was
on average 4 years older, and it has been shown that BMD Z-
scores at the distal femur decrease with age in non-ambulatory
children with CP [37]. Nonetheless, their interpretation that
the BMD in children with CP may also be affected by de-
crease in muscle strength is consistent with our study, since it
may be assumed that children at GMFCS level II have reduced
muscle strength compared with children at GMFCS level I.
The lack of a positive correlation between serum 25-OHD
concentrations and BMD Z-scores in our study is in line with
Henderson et al. [16]. The proportion of children (31 of 43
(72 %)) with insufficient vitamin D status is in keeping with a
previous Norwegian study on nutritional status of children
with CP [10]. However, the proportion is higher than reported
by Henderson et al. [38] and compared with a recently pub-
lished study on vitamin D status in Tasmanian children with
CP (i.e., 13 (34 %) of 38 children had insufficient vitamin D
status) [39]. These discrepancies may partly be explained by
the climate differences and thereby different exposure to sun-
shine in the two populations.
Interpretation
The differences in BMD Z-scores between the lumbar spine
and the distal femur found in children with CP within GMFCS
levels II–V most likely reflect the differences in weight bear-
ing between the two sites. This assumption is further strength-
ened by the lack of difference between the two sites in children
with the least motor impairments (GMFCS level I). It could be
assumed that the children in the latter group have an activity
level and a weight bearing that may not be very different from
children without CP. Finally, BMD at the lumbar spine was
not found to be predictive of fracture risk in a population of
children with CP with severe gross motor function [15], while
BMD z-score at the distal femur has been found to be strongly
correlated with fracture risk [26]. Thus, using lumbar BMD Z-
scores to assess bone health in children with CP with GMFCS
level II–V may be inappropriate.
The very high ORs for low BMD z-score in non-walkers
compared with walkers and among walkers for those with
GMFCS level II compared with level I are consistent with a
causal relationship. Whereas it is reasonable to assume that the
difference in BMD Z-scores between walkers and non-
walkers can be attributed mainly to weight bearing [40, 41],
this may not be a sufficient explanation of the lower BMD at
the distal femur of children with GMFCS level II compared
with level I, since differences in weight bearing may not differ
significantly between these two groups. The considerably
Osteoporos Int
lower BMD z-score in participants with GMFCS level II was
therefore somewhat unexpected. On the other hand, the dif-
ference could be explained by differences in the activity level
during walking and running [42]. However, evidence is accu-
mulating that the causes of low BMD may be much more
complex and may include factors that we were not able to
examine in this study. Such factors may include genetics,
hormones, and growth factors [40]. We also found lower
BMD Z-scores in the affected than in the non-affected limb
in children with hemiplegia, where differences in weight
bearing and activity must be marginal. Thus, the lower
BMD Z-scores in children with GMFCS level II than with
level I may in part be the result of the more severe motor
impairment per se (neurological signals and lower muscle
mass). This explanation would be in line with Frost’s
mechanostat hypothesis that active motion and muscle stress-
es on bone may play an important role in bone formation [18,
43].
The lower vitamin D concentration in walkers than non-
walkers is probably explained by extra vitamin supplements to
the latter group, which we unfortunately did not record, and
extra supplementation may also explain the negative correla-
tion between 25-OHD and BMD seen in this group.
Implications
Low BMD in children with CP may lead to bone fragility and
bone fractures, consequently reducing their quality of life.
Since the risk factors for low BMD are present from early
childhood in children with CP, optimal peak bone mass will
probably not be achieved [41]. It is, therefore, important to
identify children at risk for low BMD early, perhaps in infancy
in order to optimize bone growth. Although normal bone
growth probably cannot be expected, securing appropriate
nutrition with emphasis on calcium and vitamin D intake,
and early mobilization may optimize bone acquisition.
We should understand more about the differences in BMD
within the levels of walking skill, as we may find different
strategies of management for children at GMFCS level I that
can manage high intensity physical versus children at level
GMFCS III who need modified exercise programs due to
more motor limitations. To obtain more valid data on bone
quality as well as prevention of low BMD in children with CP
across all levels of motor function, longitudinal studies may be
required [44].
Conclusion
In this study, a significant proportion of children with CP had
low BMD for age at the distal femur. We did not find a
positive correlation between vitamin D status and BMD.
Instead, non-walkers had higher 25-OHD concentrations than
walkers, but this was most likely due to reverse causality.
Limited ambulation was a significant determinant of low
BMD. However, in children able to walk, we found a signif-
icantly lower mean BMD z-score at the distal femur of chil-
dren with GMFCS level II than in children with level I.
Moreover, in children with hemiplegia, we found that BMD
Z-scores at the distal femur were significantly lower on the
affected than on the unaffected side. These findings strongly
support that not only weight bearing and activity level, but
also the degree of the neuromuscular impairment per se, play a
significant role in bone modeling in this population.
Acknowledgments We thank Ellen Gjerløw and Ruth Kari Holmli at
the Department of Endocrinology, St. Olav’s University Hospital, Trond-
heim, Norway for performing and Ellen Gjerløw for analyzing the DXA
scans, and Heidi Kecskemethy, RD, CSP, CBDT, Nemours/A. I. duPont
Hospital for Children, Departments of Biomedical Research and Medical
Imaging, Wilmington, Delaware, USA, for teaching the distal femur
technique. We thank Bente Brannsether, Section for Pediatrics, Depart-
ment of Clinical Medicine, University of Bergen, Norway, for calculating
the height, weight, and BMI z-scores.
Funding source Liaison Committee between the Central Norway Re-
gional Health Authority (RHA) and the Norwegian University of Science
and Technology (NTNU).
Conflicts of interest None
References
1. Sheth RD (2004) Bone health in pediatric epilepsy. Epilepsy Behav
5(Suppl 2):S30–S35
2. Zacharin M (2004) Current advances in bone health of disabled
children. Curr Opin Pediatr 16:545–551
3. Helin I, Landin LA, Nilsson BE (1985) Bone mineral content in
preterm infants at age 4 to 16. Acta Paediatr Scand 74:264–267
4. Henderson RC, Lark RK, Gurka MJ, Worley G, Fung EB, Conaway
M, Stallings VA, Stevenson RD (2002) Bone density and metabolism
in children and adolescents with moderate to severe cerebral palsy.
Pediatrics 110:e5
5. Stevenson RD, Conaway M, Barrington JW, Cuthill SL, Worley G,
Henderson RC (2006) Fracture rate in children with cerebral palsy.
Pediatr Rehab 9:396–403
6. Bischof F, Basu D, Pettifor JM (2002) Pathological long-bone frac-
tures in residents with cerebral palsy in a long-term care facility in
South Africa. Dev Med Child Neurol 44:119–122
7. Uddenfeldt Wort U, Nordmark E, Wagner P, Duppe H, Westbom L
(2013) Fractures in children with cerebral palsy: a total population
study. Dev Med Child Neurol 55:821–826
8. Boot AM, Krenning EP, de Muinck Keizer-Schrama SM (2011) The
relation between 25-hydroxyvitamin D with peak bone mineral den-
sity and body composition in healthy young adults. J Pediatr
Endocrinol Metab 24:355–360
9. Dahlseng MO, Finbraten AK, Juliusson PB, Skranes J, Andersen G,
Vik T (2012) Feeding problems, growth and nutritional status in
children with cerebral palsy. Acta Paediatr 101:92–98
10. Hillesund E, Skranes J, Trygg KU, Bohmer T (2007) Micronutrient
status in children with cerebral palsy. Acta Paediatr 96:1195–1198

11. Duncan B, Barton LL, Lloyd J, Marks-Katz M (1999) Dietary
considerations in osteopenia in tube-fed nonambulatory children with
cerebral palsy. Clin Pediatr (Phila) 38:133–137
12. Lee JJ, Lyne ED, Kleerekoper M, Logan MS, Belfi RA (1989)
Disorders of bone metabolism in severely handicapped children
and young adults. Clin Orthop Relat Res 245:297–302
13. Winzenberg T, Jones G (2011) Dual energy X-ray absorptiometry.
Aust Fam Physician 40:43–44
14. Gordon CM, Bachrach LK, Carpenter TO et al (2008) Dual energy
X-ray absorptiometry interpretation and reporting in children and
adolescents: the 2007 ISCD Pediatric Official Positions. J Clin
Densitom 11:43–58
15. Henderson RC (1997) Bone density and other possible predictors of
fracture risk in children and adolescents with spastic quadriplegia.
Dev Med Child Neurol 39:224–227
16. Henderson RC, Lin PP, Greene WB (1995) Bone-mineral density in
children and adolescents who have spastic cerebral palsy. J Bone
Joint Surg Am 77:1671–1681
17. Pritchett JW (1990) Treated and untreated unstable hips in severe
cerebral palsy. Dev Med Child Neurol 32:3–6
18. Chen CL, Lin KC, Wu CY, Ke JY, Wang CJ, Chen CY (2012)
Relationships of muscle strength and bone mineral density in ambu-
latory children with cerebral palsy. Osteoporos Int 23:715–721
19. Chen CL, Chen CY, Liaw MY, Chung CY, Wang CJ, Hong WH
(2013) Efficacy of home-based virtual cycling training on bone
mineral density in ambulatory children with cerebral palsy.
Osteoporos Int 24:1399–1406
20. Tatay Diaz A, Farrington DM, Downey Carmona FJ, Macias Moreno
ME, Quintana del Olmo JJ (2012) Bone mineral density in a popu-
lation with severe infantile cerebral palsy. Rev Esp Cir Ortop
Traumatol 56:306–312
21. (2000) Surveillance of cerebral palsy in Europe: a collaboration of
cerebral palsy surveys and registers. Surveillance of Cerebral Palsy in
Europe (SCPE). Dev Med Child Neurol 42:816–824
22. Palisano RJ, Hanna SE, Rosenbaum PL, Russell DJ, Walter SD,
Wood EP, Raina PS, Galuppi BE (2000) Validation of a model of
gross motor function for children with cerebral palsy. Phys Ther 80:
974–985
23. Henderson RC, Lark RK, Newman JE, Kecskemthy H, Fung EB,
Renner JB, Harcke HT (2002) Pediatric reference data for dual X-ray
absorptiometric measures of normal bone density in the distal femur.
AJR Am J Roentgenol 178:439–443
24. Zemel BS, Stallings VA, Leonard MB, Paulhamus DR, Kecskemethy
HH, Harcke HT, Henderson RC (2009) Revised pediatric reference
data for the lateral distal femur measured by Hologic Discovery/
Delphi dual-energy X-ray absorptiometry. J Clin Densitom 12:207–
218
25. Zemel BS, Leonard MB, Kelly A et al (2010) Height adjustment in
assessing dual energy x-ray absorptiometry measurements of bone
mass and density in children. J Clin Endocrinol Metab 95:1265–1273
26. Henderson RC, Berglund LM, May R et al (2010) The relationship
between fractures and DXA measures of BMD in the distal femur of
children and adolescents with cerebral palsy or muscular dystrophy. J
Bone Miner Res 25:520–526
Osteoporos Int
27. Stevenson RD (1995) Use of segmental measures to estimate stature in
children with cerebral palsy. Arch Pediatr Adolesc Med 149:658–662
28. Juliusson PB, Roelants M, Nordal E, Furevik L, Eide GE, Moster D,
Hauspie R, Bjerknes R (2013) Growth references for 0–19 year-old
Norwegian children for length/height, weight, body mass index and
head circumference. Ann Hum Biol 40:220–227
29. Tanner JM (1981) Growth and maturation during adolescence. Nutr
Rev 39:43–55
30. Bischoff-Ferrari HA, Giovannucci E, Willett WC, Dietrich T,
Dawson-Hughes B (2006) Estimation of optimal serum concentra-
tions of 25-hydroxyvitamin D for multiple health outcomes. Am J
Clin Nutr 84:18–28
31. Portney LG, Watkins MP (2008) Foundations of clinical research:
Applications to practice. Prentice-Hall International (UK) Limited,
London
32. Andersen GL, Irgens LM, Haagaas I, Skranes JS, Meberg AE, Vik T
(2008) Cerebral palsy in Norway: Prevalence, subtypes and severity.
Eur J Paediatr Neurol EJPN Off J Eur Paediatr Neurol Soc 12:4–13
33. Crabtree NJ, Arabi A, Bachrach LK, Fewtrell M, El-Hajj Fuleihan G,
Kecskemethy HH, Jaworski M, Gordon CM (2014) Dual-energy X-
ray absorptiometry interpretation and reporting in children and ado-
lescents: the revised 2013 ISCD Pediatric Official Positions. J Clin
Densitom 17:225–242
34. Henderson RC, Kairalla J, Abbas A, Stevenson RD (2004) Predicting
low bone density in children and young adults with quadriplegic
cerebral palsy. Dev Med Child Neurol 46:416–419
35. Al Wren T, Lee DC, Kay RM, Dorey FJ, Gilsanz V (2011) Bone
density and size in ambulatory children with cerebral palsy. Dev Med
Child Neurol 53:137–141
36. Chen CL, Ke JY, Wang CJ, Wu KP, Wu CY, Wong AM (2011)
Factors associated with bone density in different skeletal regions in
children with cerebral palsy of various motor severities. Dev Med
Child Neurol 53:131–136
37. Henderson RC, Kairalla JA, Barrington JW, Abbas A, Stevenson RD
(2005) Longitudinal changes in bone density in children and adoles-
cents with moderate to severe cerebral palsy. J Pediatr 146:769–775
38. Henderson RC (1997) Vitamin D levels in noninstitutionalized chil-
dren with cerebral palsy. J Child Neurol 12:443–447
39. Ware T, Whitelaw C, Flett P, Parameswaran V (2013) Vitamin D
status in Tasmanian children with cerebral palsy. J Paediatr Child
Health 49:E349–E350
40. Bachrach LK (2001) Acquisition of optimal bone mass in childhood
and adolescence. Trends Endocrinol Metab 12:22–28
41. Ondrak KS, Morgan DW (2007) Physical activity, calcium intake and
bone health in children and adolescents. Sports Med 37:587–600
42. Bjornson KF, Zhou C, Stevenson R, Christakis D, Song K (2014)
Walking activity patterns in youth with cerebral palsy and youth
developing typically. Disabil Rehabil 36:1279–1284
43. Frost HM (2003) Bone’s mechanostat: a 2003 update. Anat Rec A:
Discov Mol Cell Evol Biol 275:1081–1101
44. Mergler S, Evenhuis HM, Boot AM, De Man SA, Bindels-De Heus
KG, Huijbers WA, Penning C (2009) Epidemiology of low bone
mineral density and fractures in children with severe cerebral palsy: a
systematic review. Dev Med Child Neurol 51:773–778