Epidemiology of low bone mineral density and fractures in children with severe cerebral palsy: a systematic review


    The authors would like to acknowledge ASVZ and Phelps-Stichting as sources of support.

Sandra Mergler at Erasmus MC, Intellectual Disability Medicine, Department of General Practice, PO Box 2040, 3000 CA Rotterdam, the Netherlands. E-mail: s.mergler@erasmusmc.nl


Aim  Children with severe cerebral palsy (CP) are at risk for developing low bone mineral density (BMD) and low-impact fractures. The aim of this study was to provide a systematic literature review of the epidemiology of fractures and low BMD in children with severe CP, with an emphasis on risk factors. Gross Motor Function Classification System (GMFCS) levels IV and V were criteria for severe cerebral palsy.

Method  The literature (PubMed) was searched and eligible studies were given a level of evidence score using the Scottish Intercollegiate Guidelines Network criteria.

Results  Seven studies were found concerning epidemiology of fractures, 11 studies described epidemiology of low BMD, and 14 studies concerned risk factors. The methodological quality of most of these studies was poor. Five studies were considered well-conducted with low risk of confounding and bias. In these studies, the incidence of fractures in children with moderate to severe CP approached 4% per year, whereas the prevalence of low BMD in the femur was 77%. Limited ambulation, feeding difficulties, previous fractures, anticonvulsant use, and lower body fat mass were associated with low BMD z-scores.

Interpretation  There is only a limited amount of high-quality evidence on low BMD and fractures in children with severe CP.

Children with severe cerebral palsy (CP) may have an intellectual disability in addition to severe motor impairment. These children frequently experience health problems such as epilepsy, recurrent pulmonary infections, gastro-oesophageal reflux, and constipation.1,2 In addition to the above, children with multiple disabilities are prone to develop low bone mineral density (BMD).3,4 The underlying pathophysiology in these patients is complex, and several risk factors for the development of low BMD, such as immobility, low calcium intake, low vitamin D status and anticonvulsant use, are frequently observed in this group.5 Most of these risk factors for developing low BMD are present from early childhood, so osteopenia primarily results not from true losses in bone mineral, but from a diminished growth rate of bone mineral compared with healthy children.6

Earlier studies in children with primary osteoporosis or osteoporosis associated with chronic disease or its treatment have shown a relationship between decreased bone density and increased fracture incidence.7,8 In children with a profoundly low BMD, these fractures can occur without significant trauma.9 Moreover, the lack of verbal communication in severely disabled children may lead to diagnostic delay of fractures and, therefore, increased morbidity.

To acquire a greater insight into fracture risk in children with severe CP it is necessary to evaluate not only fracture incidence and risk factors, but also the prevalence of low BMD and the determinants causing this condition.

The aim of this review is to provide a structured and comprehensive overview of the current literature on the prevalence of low BMD and the incidence of fractures in children with severe CP, with an emphasis on the risk factors.



For this survey, we divided our search into the following two main subjects: fractures and low BMD. Studies concerning low BMD were further divided into prevalence studies and studies concerning determinants. Studies concerning both prevalence and determinants were reviewed for both subjects separately.

Articles were identified through the Medline database using PubMed by combining search terms for cerebral palsy (‘generalized cerebral palsy’, ‘cerebral palsy’, ‘mental retardation’, ‘multiple disability’, ‘intellectual disability’, ‘learning disability’) with keywords for fracture (‘fractures’, ‘fracture’) or keywords for bone mineral density (‘bone mineral density’, ‘bone density’, ‘osteoporosis’, ‘osteopenia’). Search results were limited to human and paediatric studies published in the English language. Studies described in this review were published between 1950 (start of Medline) and February 2009.


Types of studies

All types of study design (e.g. cross-sectional, cohort, or case–control), except case reports and case series, were considered for inclusion in this review.

Studied patients

Children (aged 0–18y) with severe CP were the subject of our search. Severe CP was defined as level IV or V according to the Gross Motor Function Classification system and a history of clinically diagnosed CP.10

Outcome measures

Incidence rate of fractures, prevalence of low BMD (z-score below −2) and predictive or associative factors for fractures and low BMD were our primary outcome measures. Statistical significance was set at p<0.05.

The mean z-score for BMD was a secondary outcome measure. The individual z-score was calculated by comparing BMD values with age- and sex-related reference values.

Validity assessment

The abstracts of the studies identified by the literature search were read to identify relevant studies for full review. Studies needed to concern children with moderate to severe CP and also had to provide data on fractures or bone density. In addition, we scrutinized the reference lists of the identified publications to find additional studies.

After review, we recorded information about the year of publication, study design, number of included participants, representativeness of the study population, applied diagnostic methods, and outcome measures.

Quality assessment

The quality of the eligible studies was assessed by taking into account the study design, the size and representativeness of the study population (i.e. the presence of selection bias), the validity of outcomes (risk of confounding or bias), and the quality of the statistical analysis. Two of the authors evaluated the levels of evidence of the articles independently using previously published criteria developed by the Scottish Intercollegiate Guidelines Network (SIGN; Table SI, supporting information published online).11 The highest levels of evidence in the SIGN grading system are accorded to randomized controlled trials and meta-analyses of randomized controlled trials. In the current review, concerning observational nonintervention studies, the highest possible score according to this rating system for this purpose was 2++, which is given for high-quality case–control or cohort studies with a very low risk of confounding, bias, or chance and a high probability that the relationship is causal.


Trial flow

In Figure 1 we present a flow diagram, according to the Quorom guidelines,12 that summarizes the results of our literature search.

Figure 1.

 Flow diagram (Quorom guidelines)12 modified for epidemiological studies.

Study characteristics

Incidence and determinants of fractures

The characteristics and levels of evidence of the seven publications on fractures are presented in Table SII.

Prevalence and determinants of low bone mineral density

In Table SIII the characteristics of the 11 remaining articles on prevalence of low BMD are presented. The 14 articles on determinants of low BMD are more extensively described in Table SIV.

Methodological quality

We did not find any high-quality systematic reviews, cohort, case–control, or cross-sectional studies. Four studies4,6,13,14,19 were considered well conducted with a low risk of confounding or bias and a level of evidence score of 2+. In four of these studies4,13,14,19 the data presented had been collected as part of the North American Growth in Cerebral Palsy Project (NAGCPP), a project that started in 1996 at six separate sites in the USA and Canada. To obtain a representative population-based sample, multiple recruitment methods have been used to identify potentially eligible children including hospitals, newspapers, special education teachers, physical therapists and regional United Cerebral Palsy newsletters. All other reviewed studies had used selected populations (children at orthopaedic clinics or institutionalized children) rather than population-based samples, and, therefore, were more prone to selection bias. The five well-conducted studies also differed from the other studies in including a more extensive description of the statistical methods that had been used to analyze the data.

In 16 studies, the study design was not reported, while in two studies an incorrect design was formulated.15,16 Determining the applied study design was difficult in some cases because of unclear descriptions of methods or because of the use of multiple designs in one study. For example, sometimes it was not clear whether the determinants had been assessed at a distinct point in time or over a period of time, and some cross-sectional studies included small intervention studies or follow-up studies.13,15,17

None of the articles reported a power calculation to determine the population size necessary to answer the research question. As 95% confidence intervals for prevalences of low BMD were not reported in any study (Table SIII), these were calculated by us using the reported size of the study population because they provide valuable additional information in estimating the validity of the outcomes.

In the group of studies concerning fractures, three studies were file studies and two studies had used a questionnaire to collect data. Such studies are sensitive to information bias.

In eight out of the 14 studies in which determinants (for fractures or low bone mineral density) were described and analysed, a multiple regression analysis had been performed. The other studies used linear regression analyses or t-tests only, so little could be said about the interference between different determinants in these studies.

Study outcomes

Incidence rate and determinants of fractures

The prevalence of fractures reported in two studies was 12% (95% CI 8.9–15.1%)18 and 23% (95% CI 14.0–31.5%).15 The incidence of fractures was reported in two other studies and varied between 2.7% (95% CI −0.3 to 5.7%) and 4.5% (95% CI 1.0–7.9%).13,19 The most frequently reported determinants were use of antiepileptic drugs, immobilization, fracture in history and use of a feeding tube (Table SII).13,15,16,18–20

In a study by Stevenson et al.13 (level of evidence 2+), the incidence of fractures in children with moderate to severe CP (n=261) was 4% per year (95% CI 1.5–6.2%). A higher percentage of body fat and the presence of a gastrostomy catheter were significantly associated with a higher number of reported fractures. Determinants that did not have a significant relation were sex, Gross Motor Function Classification System (GMFCS) level, race, anticonvulsant use, height z-score, and weight z-score.

Prevalence of low BMD

The prevalence of low BMD of the distal femur, defined as a z-score lower than −2, was 77% (95% CI 65.0–87.1%) in a study by Henderson et al.4 (level of evidence score 2+). The mean z-score for BMD of the distal femur was −3.1 SD 0.2 and for BMD of the lumbar spine was −1.8 SD 0.1 (Table SIII).

In three other studies, a prevalence of 27%,21 58%3, and 66%22 was found. In seven other studies only mean BMD z-scores were calculated for children with different levels of CP and mobility and of different ages. Mean BMD z-scores varied in these studies between −3.4 (distal femur in children with moderate to severe CP aged 12–19 y)6 and −0.8 (lumbar spine in children with spastic CP).23 In a second study by Henderson et al.,6 which we assigned a level of evidence 2+, mean BMD z-scores in the distal femur ranged from −2.4 to −3.4 with increasing age.

Determinants of low BMD

The most commonly studied determinants were GMFCS level, feeding difficulties, previous fracture and use of antiepileptic drugs (Table SIII).3,21,24,25

Two studies by Henderson et al. were assigned a level of evidence score of 2+.14,19 Based on the outcomes of the other study on predicting low BMD,19 a regression equation was developed for predicting BMD z-score from the variables weight z-score, age, GMFCS level, feeding difficulty, previous fracture, and use of anticonvulsants (R2=0.55, p<0.001).14 In the other study on bone density and metabolism, the authors reported a significant association between GMFCS score (p<0.001), feeding difficulties (p=0.003), previous fracture (p<0.001), and anticonvulsant use (p=0.003) and the BMD z-score measured in the distal femur.14 In addition, a significant relation was described between low triceps skinfold z-score and low BMD z-score measured in the lumbar spine.14 Factors that did not have a significant relation to low BMD according to both studies were age, sex, race, temporary immobilization, health status, calcium intake, serum transthyretin, serum 25-hydroxyvitamin D, serum N-telopeptides, serum osteocalcin and phosphorus, calcium, and alkaline phosphatase levels.


This review confirms that low BMD is a serious problem in children with severe CP, with mean z-scores ranging from −3.4 in the distal femur to −0.8 in the lumbar spine,4,6,23 a prevalence of BMD z-scores below −2 of 77% (95% CI 65.0–87.1%),4 and an annual incidence of fractures of 4%.13 Significant determinants of low BMD were limited ambulation feeding difficulties, previous fracture, anticonvulsant use, and lower fat mass (measured at the triceps skinfold).4,14

However, the evidence is still limited because most identified studies were either (low-quality) file studies or case reports. No high-quality study has been published, but five studies were considered well conducted with a low risk of confounding or bias. Four of them were found to be acceptable because of the representative population-based study population (no selection bias), and all four had a clear description of methods and statistical procedures.

The paucity of valid research can, to a large extent, be explained by the fact that research in children with severe CP requires complicated logistics. Study protocols are minutely examined by medical ethics committees because the studies involve a vulnerable patient group. Moreover, it is difficult to obtain a sufficiently large and representative study population if the patient group is relatively small, and parental consent rates for research are usually rather low. Indeed, most studies we reviewed used small and selected patient populations. The NAGCPP can be regarded as an example of how a representative study population can be achieved. The best way to obtain more valid data on bone quality and risk factors for low BMD in children with severe CP is by longitudinal research. A high-quality cohort study achieving a level of evidence score of 2++ on the SIGN criteria would require recruitment in a representative community setting, for example through day care centres and special schools. Bone density measurements should be performed annually over a longer period of follow-up (e.g. 5y) to determine changes in BMD over time and should be combined with repetitive assessment of determinants of low BMD. The measurements should impose a minimal burden on the children, so that parents or caregivers will be more likely to cooperate and will not readily drop out. This could be achieved by performing bone density measurement at home to avoid transportation of the child to a hospital. The research population should be towards the top of the calculated size to compensate for the drop-out of children who die or become too ill to participate in the study. Another and more convenient way to overcome the problem of small numbers might be implementing protocols for systematic data collection on low BMD and aetiological factors in hospitals and care facilities. Although this will not give a representative study population, in this way a larger group of children can be recruited over a longer period of time for data collection.

The only well-conducted study on fracture incidence in children with CP was a prospective cohort study with a median follow-up of 1.6 years. The annual incidence of fractures found in this study (4%) seems to be accurate and corresponds to results found in other studies.19 This percentage is higher than the fracture rate in healthy children, which is around 2.5%. This is notable because children without disabilities are more prone to accidents in the playground, whereas children with CP are often wheelchair bound and unable to walk.13

Significant negative associations between BMD z-score and limited ambulation, feeding difficulties, previous fracture, anticonvulsant use, and lower fat mass correspond to existing theories and clinical practice. Stevenson et al. found that higher body fat was significantly associated with the number of fractures that occurred during follow-up.13 This association is not confirmed by other studies in children with severe CP.20 A possible explanation could be that after inserting a gastrostomy catheter, which was also found to be a significant association, a rapid increase in fat mass may be observed in previously malnourished children.26 Malnutrition is a known risk factor for fractures and low BMD.

In our clinical experience, not all predictive factors for developing low BMD in this population have been studied. For example, daylight exposure and amount of exercise were not assessed in any of the studies. These factors could be assessed by using diaries in which caregivers record the amount of time the child spends daily on activities or being outdoors. To provide reliable data, this should preferably be done in different seasons and over a substantial period of time (e.g. 2wks).


Implications for practice

Children with CP who are not independently ambulant, who have had previous fractures, who have feeding difficulties, or who use anticonvulsive drugs are at a high risk for developing low BMD. We recommend monitoring BMD in such cases. If BMD is found to be low, parents and caregivers need to be cautious to avoid fractures. Furthermore, we recommend optimization of calcium intake and determination of vitamin D status in these children. Interventions that increase muscle mass are advisable.

Implications for research

Longitudinal research is required to determine predictive factors for low BMD in this group. A practical way to acquire relevant data is by implementing protocols for systematic data collection and registration of low BMD and aetiological factors in children with CP, for example by paediatricians and physicians for people with intellectual disabilities.

The feasibility and reliability of new diagnostic methods such as quantitative computed tomography (with less interference of bone shape and size) need to be tested in children with severe CP.

Strategies to prevent bone loss and optimize peak bone mass, for example by increasing muscle mass or vitamin D supplementation, should be developed and evaluated.