1 Introduction

Epilepsy is a common chronic neurological disorder with more than half of cases beginning in childhood (Begley et al., 2000). Most patients require long-term, and sometimes lifelong, therapy with antiepileptic drugs (AEDs). AEDs are associated with significant side effects including, but not limited to, radiological evidence of rickets, decreased bone mineral density (BMD), altered bone turnover, and increased risk of fracture (Souverein et al., 2005; Karceski, 2007; Gniatkowska-Nowakowska, 2010; Verrotti et al., 2010; Meier and Kraenzlin, 2011). AEDs increase catabolism of 25-hydroxyvitamin D by induction of the hepatic P-450 enzyme system, leading to relative hypocalcemia, increased levels of parathyroid hormone (PTH), and subsequent low BMD (Dent et al., 1970; Keck et al., 1982; Chung and Ahn, 1994). However, some studies have suggested a significant reduction in BMD with non-enzymeinducing AEDs (Oner et al., 2004; Petty et al., 2005). Other environmental and epidemiological parameters including age, sex, diet, and mobility may also affect mineral metabolism (Pack and Morrell, 2001), suggesting a multifactorial etiology.

A meta-analysis, pooling data from multiple cross-sectional studies showed only a very limited decrease in BMD with a decline of 0.4 Z-scores in the spine and −0.6 Z-scores in the hip (Vestergaard, 2005). Most of the relevant data are derived from adults. Data on bone metabolism in children receiving AEDs are scarce, with conflicting results. Some studies found a significant difference in BMD and bone metabolism between children treated with AEDs and healthy controls (Oner et al., 2004; Pack et al., 2008; Salimipour et al., 2013); other studies found no significance (Akin et al., 1998; El-Hajj Fuleihan et al., 2008). However, these studies may be limited, as most included a small sample, and numbers were too few to enable comparison between specific AEDs.

We, therefore, undertook a meta-analysis of studies that had analyzed the effects of AEDs on BMD and bone metabolism in children.

2 Materials and methods

2.1 Search strategy and inclusion criteria

We searched PubMed and Web of Science for all studies up to July 2014 that reported on BMD or bone metabolism and the use of AEDs in children using the keywords “epilepsy” or “antiepileptic drugs” in combination with “bone mineral density” or “bone density” or “bone metabolism” or “bone turnover”. This meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Liberati et al., 2009). We included studies reporting the effects of AEDs on BMD and bone metabolism in children, regardless of sample size, if they met the following criteria: (1) inclusion of at least one marker of bone metabolism; (2) BMD measured by dual-energy X-ray absorptiometry (DXA); (3) studies involving mean BMD (or bone markers) and standard deviations (SDs) or standard errors (SEs), or where these numbers could be calculated; (4) child subjects (<18 years old); (5) studies assessing patients with epilepsy who received AED monotherapy or polytherapy; (6) studies with a healthy control group; (7) observational studies or randomized controlled trials (RCTs). We excluded animal studies, reviews, and letters.

2.2 Data extraction and quality assessment

Two reviewers independently evaluated studies for inclusion. Discrepancies were resolved by arbitration and consensus following discussion. The following information was extracted from each study: name of the first author, year of publication, country where the study was performed, design of the study, drug dosage, age and gender of individuals, number of cases, mean BMD (or serum concentration of bone markers), and SDs (or SEs). Study quality was independently assessed by two reviewers according to the Newcastle-Ottawa Scale for quality assessment of cohort and case-control studies (Liberati et al., 2009). Each study was allocated up to nine stars, the criteria being quality of selection (maximum, 4 stars), comparability (maximum, 2 stars), and outcome (maximum, 3 stars).

2.3 Statistical analysis

When the outcome of interest was measured using the same scale in every study, weighted mean differences (WMDs) with 95% confidence interval (CI) were used as summary measures. However, when studies used different scales to measure the effect of AED treatment, the standardized mean difference (SMD) with 95% CI was used.

All data were initially analyzed by a fixed-effects model if there was no significant heterogeneity, and a random-effects model otherwise. Heterogeneity was assessed by Q statistics, with P<0.10 indicating significant heterogeneity. We quantified heterogeneity using the I2 statistic. I2 values of 25%, 50%, and 75% correspond to cut-off points for low, moderate, and high degrees of heterogeneity, respectively. To explore heterogeneity, we performed subgroup analyses according to the drug patients used.

Publication bias was checked through the use of funnel plots with an asymmetric funnel plot indicating publication bias. All analyses were conducted using Review Manager (RevMan) v.5.2 statistical software and Microsoft Excel. P<0.05 was considered statistically significant.

3 Results

3.1 Study selection and characteristics

The information flow for the search and selection of studies is shown in Fig. 1. The initial search yielded 1603 research reports, of which 228 were excluded due to identical title or authors; 1056 were excluded due to ineligible study design (including non-human studies, review articles, case reports, comment, letter, experimental study, and/or fracture-only outcome). After full-text screening, 22 studies were included (Chung and Ahn, 1994; Baer et al., 1997; Akin et al., 1998; Kafali et al., 1999; Erbayat Altay et al., 2000; Verrotti et al., 2002; Voudris et al., 2002; Ecevit et al., 2004; Oner et al., 2004; Babayigit et al., 2006; Kumandas et al., 2006; Nicolaidou et al., 2006; El-Hajj Fuleihan et al., 2008; Nettekoven et al., 2008; Sheth and Hermann, 2008; Sheth et al., 2008; Gniatkowska-Nowakowska, 2010; Rauchenzauner et al., 2010; Babacan et al., 2012; Dimić et al., 2013; Razazizan et al., 2013; Turan et al., 2014). Table 1 summarizes the characteristics of the included studies published between 1994 and 2014. Nine were conducted in Turkey, six in Europe, one in Australia, one in Korea, one in Iran, one in Lebanon, and three in the United States. Twenty-one studies were written in English, and one in Serbian. Six provided data only on BMD, eight reported results on bone markers, and the remaining eight reported both. Study quality scores (range, 0–9) averaged 7.54, and 63.6% were high-quality studies (score>8) (Table 1).

Fig. 1
figure 1

Flow diagram of the study-selection process

Table 1
figure Tab1

Characteristics of the included studies

3.2 Pooled effect of AED treatment on BMD

A total of 11 studies comprising 23 datasets, which included 645 subjects and 579 controls, evaluated the effect of AEDs on lumbar spine BMD (Chung and Ahn, 1994; Akin et al., 1998; Kafali et al., 1999; Erbayat Altay et al., 2000; Oner et al., 2004; Babayigit et al., 2006; Kumandas et al., 2006; El-Hajj Fuleihan et al., 2008; Gniatkowska-Nowakowska, 2010; Babacan et al., 2012; Dimić et al., 2013). The results suggested an association between AED treatment and decreased BMD (SMD=−0.30, 95% CI [−0.61, 0.01]; Fig. 2), with high heterogeneity (I2= 85%). When data were pooled for absolute changes in BMD (WMD=−0.03, 95% CI [−0.06, 0.00]) or BMD Z-scores (WMD=−0.87, 95% CI [−1.25, −0.48]), decreased BMD was found. No single study influenced the pooled effect significantly. Duration of follow-up and sex ratio did not affect the pooled effect, either. Subgroup analyses showed that the difference in AED type was a source of heterogeneity. As the results suggested, valproic acid (VPA) (SMD=−0.48, 95% CI [−0.95, −0.02]) could significantly decrease BMD of lumbar spine in epileptic children, while carbamazepine (CBZ) (SMD=0.32, 95% CI [−1.08, 1.72]) had no significant effect. The funnel plots indicated no obvious publication bias.

Fig. 2
figure 2

Meta-analysis of datasets involving AED treatment and BMD of lumbar spine

The changes in BMD of trochanter and femoral neck were assessed by three datasets from two studies (Oner et al., 2004; Ecevit et al., 2004) with 66 subjects and 64 controls. The pooled effect showed that AED treatment resulted in a significant decrease of BMD in trochanter (mean difference (MD)=−0.07, 95% CI [−0.10, −0.05]) and femoral neck (MD=−0.05, 95% CI [−0.09, −0.02]) (Table 2). Subgroup analyses revealed that VPA was significantly associated with decreased BMD of trochanter (WMD=−0.08, 95% CI [−0.11, −0.05]) and femoral neck (WMD=−0.07, 95% CI [−0.11, −0.02]). Sensitivity analysis showed that no single study affected the summary estimates significantly. The funnel plots showed no obvious publication bias.

Similarly, meta-analysis of two studies (Sheth and Hermann, 2008; Sheth et al., 2008), which included 198 subjects and 78 controls, and evaluated the effect of AED treatment on total body BMD, obtained a significant result (WMD=−0.33, 95% CI [−0.51, −0.15]). Sensitivity analysis showed that no single study affected the summary estimates significantly. The funnel plots showed no obvious publication bias.

Table 2
figure Tab2

Main analysis for BMD and bone markers

3.3 Pooled effect of AED treatment on bone markers

The effects of AED treatment on bone markers, including 25-hydroxyvitamin D, serum alkaline phosphatase (ALP), serum calcium, phosphorus, and PTH, were assessed in our study.

Meta-analysis of nine studies including 616 subjects and 421 controls (Verrotti et al., 2002; Babayigit et al., 2006; Nicolaidou et al., 2006; El-Hajj Fuleihan et al., 2008; Nettekoven et al., 2008; Rauchenzauner et al., 2010; Babacan et al., 2012; Razazizan et al., 2013; Turan et al., 2014), revealed a significant association between AED treatment and decreased 25-hydroxyvitamin D (MD=−3.37, 95% CI [−5.94, −0.80]), as shown in Fig. 3. Sensitivity analysis showed that no single study affected the summary estimates significantly. The funnel plots showed no obvious publication bias.

Fig. 3
figure 3

Meta-analysis of datasets involving AED treatment and 25-hydroxyvitamin D

The changes in serum ALP were assessed for 449 subjects and 397 controls from eight studies (Akin et al., 1998; Kafali et al., 1999; Voudris et al., 2002; Oner et al., 2004; Babayigit et al., 2006; Kumandas et al., 2006; Babacan et al., 2012; Razazizan et al., 2013). The pooled effect showed that AED treatment resulted in a significant elevation of serum ALP (SMD=0.71, 95% CI [0.38, 1.05]), as shown in Fig. 4. Sensitivity analysis showed that no single study affected the summary estimates significantly. The funnel plots showed no obvious publication bias.

Fig. 4
figure 4

Meta-analysis of datasets involving AED treatment and BMD of serum ALP

No significant effects of AED treatment on serum PTH, serum calcium, or phosphorus were found. The detailed results are shown in Table 2.

4 Discussion

Childhood is a critical time for bone mineralization. During periods of high mineralization, children are prone to osteoporosis and bone fractures (Sheth et al., 1995; Oner et al., 2004). AEDs are one of the most important factors that may affect bone health, but there is no agreement about the effect of AEDs on BMD and bone metabolism of children with epilepsy. Our meta-analysis of studies into the effects of treatment with AEDs on epileptic children shows: (1) a reduction in BMD at lumbar spine, trochanter, femoral neck and total body BMD; (2) a reduction in 25-hydroxyvitamin D and an increase in serum ALP; (3) no significant changes in serum PTH, calcium, or phosphorus.

Although the effect of AED treatment on the lumbar spine BMD was of borderline significance (P=0.06), the omission of studies using BMD Z-scores made the effect significant (P=0.03). The heterogeneity in the analysis was high (I2=85%). The potentially confounding factors of polytherapy and difference in seizure type may be a source of heterogeneity. El-Hajj Fuleihan et al. (2008) demonstrated that polytherapy patients were associated with a significantly lower BMD level than patients treated with monotherapy. Sheth and Hermann (2008) found that children with symptomatic generalized epilepsy appear to be at highest risk for BMD. However, the two factors were not analyzed in this paper because of lack of data, and future studies are warranted to explore this issue further.

CBZ and VPA are the frontline treatments of partial and generalized seizures in children and adults. With conflicting results about the effects of the two drugs on BMD, the mechanism by which AEDs decrease BMD has yet to be fully established. A classical theory holds the view that enzyme-inducing antiepileptic drugs (EIAEDs) decrease BMD by reducing vitamin D levels secondary to the therapeutic activation of specific cytochrome P450 isoenzymes. Our meta-analysis partially supports this theory by proving that CBZ was associated with lower levels of 25-hydroxyvitamin D (Fig. 3), and no significant difference in BMD among children treated with CBZ compared with controls (Fig. 2). The results also showed that non-EIAEDs, presented as VPA, could result in a significant decrease in BMD. Cerveny et al. (2007) suggested that VPA may interfere with bone metabolism by activation of the pregnane X receptor, which promotes the expression of vitamin D-responsive genes. Feldkamp et al. (2000) observed a decrease in bone cell proliferation, suggesting that the direct effects of anticonvulsant drugs on bone cells may contribute to skeletal damage. Furthermore, both EIAEDs and non-EIAEDs may also contribute to bone loss by inhibiting intestinal absorption of calcium and activation of vitamin D (Lee et al., 2010).

The newer antiepileptics are believed to have similar efficacy to the older AEDs, but fewer side effects. Babayigit et al. (2006) examined CBZ, VPA and oxcarbazepine use in children, and reported no significant differences on lumbar spine BMD between the groups, but 25-hydroxyvitamin D in oxcarbazepine-treated children was significantly higher than that in the VPA group. Rauchenzauner et al. (2010) found no significant difference in 25-hydroxyvitamin D between VPA and the newer AEDs. More studies are needed to make direct comparisons between newer and older AEDs in epileptic children.

Low levels of 25-hydroxyvitamin D in AED users have been demonstrated in a number of studies, but not in all. Our results suggest that AEDs affect bone metabolism by reducing 25-hydroxyvitamin D. Low 25-hydroxyvitamin D concentration results in secondary hyperparathyroidism and accelerated bone loss, influencing absorption of calcium and phosphorus (Shellhaas and Joshi, 2010). This theory is consistent with our results. However, no significant changes in serum PTH, Ca, or P were found in our study, suggesting that there are still gaps in our knowledge of the impact of AEDs on bone metabolism.

Total ALP is considered a reliable biochemical marker of bone formation. The present meta-analysis demonstrates that children treated with AEDs have a significantly increased ALP level. However, the pooled results suggest that increased serum ALP is significantly associated with the use of CBZ and the newer antiepileptics, but not VPA, indicating that ALP is an inadequate marker for bone metabolism. It could be hypothesized that measurement of total ALP may simply reflect hepatic metabolism. Some studies consider ALP isoenzyme to be highly sensitive and specific to increased bone metabolism (Voudris et al., 2002; 2005). A longitudinal study showed that epileptic patients using CBZ can have their bone metabolism altered early in the course of treatment, as indicated by the elevated activity of serum bone ALP isoenzyme (Voudris et al., 2005). Voudris et al. (2002) reported that elevated bone ALP isoenzyme correlates with the duration of treatment in children on VPA without a concomitant significant elevation of total ALP. Bone ALP isoenzyme, but not total ALP, may therefore be used as a marker for the selection of patients who might benefit from a comprehensive evaluation of their bone metabolism profile.

There is no agreement about the relationship between BMD and vitamin D status. Hannan et al. (2008) showed that serum 25-hydroxyvitamin D and BMD were significantly related to one another in white men only. Mikati et al. (2006) conducted the only randomized, controlled trial that demonstrated no significant difference between high- and low-dose vitamin D treatments, and no change in BMD compared with healthy controls after one year of treatment. Although there is no consensus on optimal vitamin D levels, many physicians often rely on 25hydroxyvitamin D levels to evaluate bone health. However, a recent meta-analysis based on a large population indicated that vitamin D given alone was not effective in preventing fractures (Reid et al., 2014). The meta-analysis performed by Vestergaard (2005) showed that the BMD decrease in AED users was in a range consistent with the increase in fracture risk observed, if seizure-related fractures were excluded. The risk of fractures was only slightly higher compared with the general population (risk ratio (RR)=1.3, 95% CI [1.0–1.7]). Since many studies have confirmed the value of BMD for predicting fracture risk, BMD measurement may be an adequate surrogate for bone health. However, we suggest that BMD should be reserved for those with exceptionally high risk (e.g. history of fractures).

Ambulatory status can affect markers of bone metabolism and BMD. Baer et al. (1997) studied vitamin D levels in relation to ambulatory status in a large sample of children who lived at home. They found that the risk of vitamin deficiency among nonambulatory children was about twice that of ambulatory children (χ2=20.9; P<0.001). They concluded that ambulatory status correlates with abnormalities in the status of 25-hydroxyvitamin D, calcium, and bone. Another study reported no association between ambulatory status and 25-hydroxyvitamin D levels (Bergqvist et al., 2007).

There were limitations to this meta-analysis. One is that the studies included were limited to cohort studies. There have been no RCTs comparing BMD and/or bone metabolism in children treated with AEDs and healthy controls. Studies to date are of limited quality. Duration of follow-up, sex ratio, and dose of drug are not clear in some studies, making it difficult to do sensitive analysis. Important confounders, such as nutrition and season, were not always fully controlled for, which might have resulted in some over-estimation of effects due to residual confounding. The studies are grouped without adequate separation of drugs by metabolic pathway for lack of BMD or bone turnover markers corresponding to the drug used. Besides, most studies lump together a variety of drugs, some of them in monotherapy and some in polytherapy: data on specific drugs are needed. Most of the studies have used a small sample, making it difficult to compare the effects of specific AEDs. Eligible studies came from a limited number of countries, nine of 14 studies targeting BMD coming from Turkey. Duration of treatment and dose also differed widely among studies.

5 Conclusions

Findings from our meta-analysis indicate that AED treatment is associated with decreased BMD in epileptic children. BMD monitoring should be reserved for epileptic children with exceptionally high risk for abnormal bone health. In the absence of evidence to the contrary, it seems reasonable to offer supplements and optimize 25-hydroxyvitamin D levels. Further research is needed to clarify the particular subgroups, dosages, and other factors that may influence the effects of AEDs on BMD and bone metabolism.

Compliance with ethics guidelines Ying ZHANG, Yu-xin ZHENG, Jun-ming ZHU, Jian-min

ZHANG, and Zhe ZHENG declare that they have no conflict of interest. This article does not contain any studies with human or animal subjects performed by any of the authors.