Vitamin D Levels and Bone Turnover in Epilepsy Patients Taking Carbamazepine or Oxcarbazepine


Address correspondence and reprint requests to Dr. S. Mintzer at 900 Walnut Street, Suite 200, Philadelphia, PA 19107, U.S.A. E-mail:


Summary: Purpose: Evidence suggests that enzyme-inducing antiepileptic drugs (AEDs) may decrease serum 25-hydroxyvitamin D (25-OHD) levels and increase bone turnover. We sought to determine whether these are affected by treatment with carbamazepine (CBZ) or oxcarbazepine (OXC).

Methods: We measured serum levels of 25-OHD, parathyroid hormone (PTH), osteocalcin (OCLN), bone alkaline phosphatase (BAP), and urinary N-telopeptides of type I collagen cross-links (NTX) in normal controls (n = 24) and in epilepsy patients taking CBZ (n = 21) or OXC (n = 24) in monotherapy. CBZ patients were subsequently switched overnight to OXC monotherapy, and after 6 weeks, the tests were repeated.

Results: 25-OHD levels were lower in each drug-treated group (OXC, 19.4 ± 2.3 pg/ml; CBZ, 20.4 ± 2.4) than in the controls (27.5 ± 2.8) (ANOVA, p = 0.052). This difference was significant for the OXC group (p < 0.05). PTH, BAP, and NTX did not differ significantly among groups. OCLN levels were somewhat elevated in the OXC group (2.79 ± 0.47 ng/ml) and more clearly and significantly elevated in the CBZ group (3.63 ± 0.36) compared with controls (2.38 ± 0.41) (p = 0.053). Because the data were very similar between OXC and CBZ groups, they were combined to increase statistical power. The combined drug-treatment group had significantly higher BAP (p = 0.02) and lower 25-OHD (p = 0.015) than did controls. The latter remained significant even after accounting for the confounding effects of age on 25-OHD levels (p < 0.05). No significant differences were found after CBZ patients were switched to OXC.

Conclusions: Epilepsy patients taking OXC or CBZ have significantly lower 25-OHD than do normal controls, with a pattern of changes in other bone biomarkers suggestive of secondary hyperparathyroidism. It may be prudent for patients taking CBZ or OXC to be prescribed 25-OHD replacement.

 Patients with epilepsy often express concerns about the potential for chronic side effects with the use of antiepileptic drugs (AEDs). One area of considerable interest is that of the effects of AEDs on bone density. Questions about possible bone loss and increased fracture risk have been present for more than three decades (1). The prevailing notion has been that those drugs that are inducers of the hepatic cytochrome P450 system (CYP450) promote the metabolism of 25-hydroxyvitamin D (25-OHD) to less biologically active analogues, resulting in decreased bone mineralization, decreased intestinal calcium absorption, increased calcium mobilization from the skeleton to maintain eucalcemia, and decreased bone density (2). Evidence to support this mechanistic hypothesis is mixed, however. The inducing AED phenytoin (PHT) has been shown rather convincingly to be associated with decreased bone density (3). One study of PHT-treated patients found changes on bone biopsy resembling the mineralization deficits of osteomalacia, suggesting that 25-OHD insufficiency was the pathologic substrate (4). No comparable study has been done of carbamazepine (CBZ), also a CYP450 inducer, but what studies have been performed have not clearly demonstrated decreased bone density (5–7). Even studies that suggest that these drugs may have osteopenic effects do not necessarily implicate decreased 25-OHD as a cause. Increased bone turnover, which leads to clinically significant bone loss over time, has been found with the use of these drugs in some studies, but often it is not correlated with 25-OHD levels (5,8,9). In one report, bone biopsy changes in AED-treated patients were found to differ from those of intestinal bypass patients, suggesting that mere 25-OHD deficiency was not the sole cause of the former (10). Direct effects of the drugs on bone cells has been proposed as another possible mechanism for bone loss in this population (11). Furthermore, aside from a single study which included patients taking lamotrigine (LTG) (5), the effects of the newer-generation AEDs on bone metabolism remain completely unknown.

We reasoned that if CYP450 induction were the root cause of bone disease—via decreased 25-OHD, increased bone turnover, or both—then the effects of CBZ, a potent broad-spectrum CYP450 inducer, might differ from those of the closely related newer-generation drug oxcarbazepine (OXC), which is only a limited enzyme inducer. We examined this by using serum bone-specific alkaline phosphatase (BAP) and osteocalcin (OCLN), markers of bone formation, and urinary crosslinked N-telopeptides of type I collagen (NTX), a marker of bone resorption, to assess bone turnover indirectly. We also reasoned that any effects due to enzyme induction would be rapidly apparent, because the processes of CYP450 induction and deinduction have a half-life of <4 days (12).


We recruited 21 patients receiving CBZ monotherapy and 24 patients receiving OXC monotherapy from the epilepsy clinics of the Northwestern Medical Faculty Foundation to participate in the study. All patients were older than 18 years and had been taking their respective drugs in monotherapy for a minimum of 8 weeks. Both well-treated and intractable patients participated. No restrictions were based on etiology or localization of the seizures. Patients who had any condition known to affect bone metabolism (e.g., renal disease, recent fracture, hyperparathyroidism, Paget disease, osteoporosis) or taking any drug known to cause or treat osteoporosis (bisphosphonates, raloxifene, miacalcin, corticosteroids) were excluded. Estrogen-containing preparations were permitted in study participants. Patients who were taking any preparation containing calcium or vitamin D, even a multivitamin, also were excluded. Twenty-four healthy individuals, with no history of seizures and otherwise subject to the aforementioned exclusion criteria, served as controls.

After providing informed consent and undergoing history and physical examination, each patient filled out a questionnaire reporting his or her intake of calcium-containing foods from the previous day as a spot check for calcium intake. The data from this questionnaire, developed by the Canadian Osteoporosis Study Group, were then entered into the group's website (13) to calculate calcium intake for each subject. Laboratory tests were then obtained, including routine serology, urinalysis, a serum pregnancy test for women of childbearing age, CBZ levels (for those taking that drug) and the study parameters of interest: serum 25-OHD, parathyroid hormone (PTH), BAP, OCLN, and urinary NTX. All patients taking CBZ were then switched overnight to OXC at a total daily dose ratio of 2:3, with the OXC dose divided in half and given BID. At the follow-up visit 6 weeks later, the same laboratory studies were repeated. The routine laboratory work was performed at Northwestern Memorial Hospital. Intact PTH was measured by immunochemiluminescent assay (ICMA) performed on a DPC Immulite analyzer (Intact PTH; Diagnostic Products Corporation, Los Angeles, CA, U.S.A.). Assays for 25-OHD, also measured by ICMA, were performed on a Liaison analyzer (Diasorin, Stillwater, MN, U.S.A.). The bone metabolic markers were obtained from Specialty Laboratories (Santa Monica, CA, U.S.A.). Because of clerical and laboratory errors, not every test was obtained for every patient. The precise n for each test in each patient group is noted in Table 2.

Table 2. Vitamin D levels and indices of bone turnover in normal controls and in epilepsy patients receiving CBZ or OXC monotherapy
  1. All values are expressed as mean ± SD, with the number in parentheses representing the n for each cell. All comparisons between the CBZ patients before and after switch to OXC were nonsignificant.

  2. 25-OHD, 25-hydroxyvitamin D; PTH, parathyroid hormone; BAP, bone-specific alkaline phosphatase; OCLN, osteocalcin; NTX, crosslinked N-telopeptides of type I collagen; OXC, oxcarbazepine; CBZ, carbamazepine; CBZ → OXC, carbamazepine patients after switch to oxcarbazepine.

  3. ap = 0.052 for comparison of control, OXC, and CBZ groups, with OXC significantly lower than control (p < 0.05) and a slight trend for decreased levels in the CBZ group compared with controls (p = 0.10).

  4. bp = 0.07 for comparison of control, OXC, and CBZ groups.

  5. cp < 0.05 for comparison of control, OXC, and CBZ groups, with CBZ significantly higher than controls (p < 0.05).

Calcium (mg/dl)9.37 ± 0.35 (24)9.33 ± 0.36 (24)9.29 ± 0.3 (21)    9.32 ± 0.31 (19) 
25-OHDa (ng/ml)27.5 ± 13.0 (21)  19.4 ± 10.8 (22)a20.4 ± 10.5 (19) 18.7 ± 9.3 (15)
PTH (pg/ml)45.7 ± 22.7 (23)55.6 ± 30.7 (22)55.6 ± 28.4 (21)  56.3 ± 27.0 (19)
BAP (U/L)b22.4 ± 5.0 (20) 28.5 ± 11.0 (24)27.7 ± 9.1 (17)   24.2 ± 8.3 (15) 
OCLN (ng/ml)c2.4 ± 1.6 (15)2.8 ± 2.2 (22) 3.6 ± 1.6 (20)c  3.7 ± 1.5 (18)
NTX (nM/mM creatinine)40.5 ± 24.0 (23)35.9 ± 22.0 (22)35.7 ± 25.8 (19)    35 ± 15.9 (18) 

Statistical analyses were performed by using analysis of variance with Dunnett's posttest to compare the parameters of interest in the control, CBZ, and OXC groups. When the CBZ and OXC groups were collapsed together, unpaired t tests were used for comparison with the control group. Because age was found to be a confounding variable, multiple regression was used to account for its effects and to isolate the impact of AED treatment on the variable of interest. The data for the CBZ patients before and after switch to OXC were analyzed by using paired t tests. Logarithmic transformation was used when necessary to ensure normal distribution of data. All analyses were done by using InStat 3.0b for Macintosh (GraphPad, San Diego, CA, U.S.A.). The power analyses mentioned in the Discussion section were performed by using the same company's StatMate 2.0a for Macintosh. This study was approved by the Institutional Review Board at Northwestern University.


Baseline characteristics of the study participants are shown in Table 1. The age range of study subjects was very broad (18 to 72 years old), but the vast majority (89%) were younger than 50 years. All those older than 50 years were male patients. The three groups differed in age, with the OXC patients being significantly older than control subjects (p < 0.05). As a consequence, age was examined as a confounder in all subsequent analyses. Differences in gender between the groups were marginally significant (χ2, p = 0.057). However, gender was found to have no correlation with any of the study variables. No significant differences were found in the racial composition of the three groups, nor were there any differences in daily calcium intake. CBZ patients had been taking that AED for a median of 6 years (mean, 7.8 years; range, 0.25–32 years). OXC patients had been taking that drug for a median of 1.1 years (mean, 1.1 years; range, 0.15–3.75 years).

Table 1. Baseline characteristics of controls and epilepsy patients receiving CBZ or OXC monotherapy
AttributeControl (n = 24)OXC (n = 24)CBZ (n = 21)
  1. OXC, oxcarbazepine; CBZ, carbamazepine.

  2. ap < 0.05 for comparison between the three groups, with the OXC patients being significantly older than the control subjects.

Age (mean ±31 ± 7 41 ± 1435 ± 12
 SD, range) a(22–48)(18–72)(20–65)
Gender (M/F)9/1517/713/8
Race (% nonwhite)29%25%33%
Calcium intake (mg; mean ± SD)673 ± 456530 ± 545448 ± 416

Results of laboratory testing are shown in Table 2. The comparison of 25-OHD levels between the control, OXC, and CBZ patients revealed that the groups were different (p = 0.053), with the OXC group having significantly lower levels than the control group (p < 0.05), and the CBZ group marginally lower than controls (p = 0.10). PTH levels in the CBZ and OXC groups were higher than those in the controls, although not significantly (p > 0.1). The direction of change was consistent with the change in 25-OHD levels, however. Comparison of BAP levels showed a difference between the three groups that was marginally significant (p = 0.07), with higher levels (reflecting greater bone formation) in the CBZ and OXC groups. OCLN was significantly different among the three treatment groups (p < 0.05), with CBZ patients demonstrating significantly higher levels than controls (p < 0.05). Once again, the direction of change in BAP and OCLN levels is consistent with the direction of change in 25-OHD and PTH levels. NTX levels were not different among the three groups (p > 0.1).

Of the 21 CBZ patients, 19 were successfully switched overnight to OXC; two had exacerbation of seizures and did not complete that portion of the study. When the values in the remaining CBZ patients were compared before and after the switch to OXC, no significant alterations were seen in any of the study parameters (p > 0.1). This is consistent with the fact that, as is apparent from Table 2, the results in the CBZ and OXC groups were remarkably similar for all study measures except for OCLN. Because the values for the two drug-treated groups differed so consistently from the control group, it appeared that the drugs had very similar effects on bone and 25-OHD metabolism. As a consequence, additional analyses were done with the data from the CBZ (preswitch) and OXC patients collapsed into a single group to increase statistical power. The combined drug-treatment group had significantly lower 25-OHD levels (p = 0.015) and significantly higher BAP (p = 0.02) than the control group. Significantly more patients had 25-OHD levels <20 in the drug-treatment group (21 of 41 vs. five of 21 in controls; Fisher's exact, p < 0.05). PTH and NTX were not different from controls in the combined drug-treatment group (p > 0.1). (The OXC and CBZ groups were not combined for analysis of OCLN, because the results in these two groups appeared somewhat distinct.)

No notable difference was found among the groups in the number of 25-OHD levels drawn during the spring and summer months and the number drawn in the fall and winter months. We also examined whether age, gender, or calcium intake had any association with the study variables. The only significant association found was that of age on 25-OHD levels (Spearman's r, p < 0.05). This inverse association was a confound, since the CBZ and OXC groups were older than the control group. Because of this, multiple regression was used to isolate the effects of AED treatment (in the combined drug-treated group) on 25-OHD levels. The best-fit model, using both age and AED treatment to predict 25-OHD level, found drug treatment with either CBZ or OXC to be a significant predictor (p < 0.05), with age no longer being significant (p > 0.1).

The duration of treatment with the relevant AED (CBZ or OXC) was examined separately for each group and in the combined drug treatment group and found to have no association with any study variable. Dosage of CBZ and serum CBZ level were found to have no correlation with any of the study variables. However, OXC daily dose was found to have a significant correlation with both BAP (p = 0.049) and NTX (p = 0.012). OXC dose was not associated with 25-OHD, PTH, or OCLN. (Serum levels of the active metabolite of OXC were not obtained in most patients and therefore could not be examined.)

To help clarify the possible mechanism of these effects, we combined all three treatment groups for correlational analyses. We found a trend toward correlation of 25-OHD levels with PTH levels (p = 0.058) and BAP levels (p = 0.096), but the PTH levels and BAP levels were not correlated with each other, nor were 25-OHD levels correlated with OCLN or NTX levels. This analysis has the potential to overstate correlations because of the combination of different groups, however. When the same analysis was done by using solely the AED-treated patients, none of the aforementioned comparisons was significant (p > 0.1).


The results of this study indicate that patients taking OXC or CBZ in monotherapy have significant reductions in 25-OHD and significant increases in the bone-formation marker BAP (and, in CBZ patients, the bone-formation marker OCLN). These alterations would be expected to produce clinically meaningful loss of bone mass over time (14). Such bone loss has been clearly demonstrated in postmenopausal women taking PHT (3), but studies of this kind are difficult to perform, so that there has been no firm confirmation of a similar effect from taking CBZ. This merits more definitive investigation.

Our study is only the second to look at bone and 25-OHD metabolism in any of the newer-generation AEDs and the first ever to examine the effects of OXC. Our data suggest that this drug's effects on these indices are very similar to those of CBZ and indicate that the long-term effects of OXC on bone mass also merit more definitive study. Clarification of this issue will require longitudinal studies powered to detect bone-density changes over time.

Our results with regard to 25-OHD in CBZ-treated patients are consistent with some other studies in the literature (6,15) but not with others (5,7,9). The most recent of these studies yielded 25-OHD data very similar to ours for CBZ patients, and their 25-OHD data for patients treated with LTG, a noninducing, noninhibiting AED, were quite similar to the data in our control patients. It is striking that five of the six aforementioned studies (including the present one) found that 25-OHD levels were decreased in CBZ patients, although the differences did not always reach significance. The variation in results may have much to do with statistical power. The study of Tjellesen et al. (7,18) found a decrease in 25-OHD levels of ∼12% in CBZ-treated patients and had only a 20% power to detect this difference. The studies of Gough et al. (15), Hoikka et al. (6), and Pack et al. (2,5), plus the present study, each found 25-OHD to be decreased from 27 to 42% relative to other groups, with the former study having >99% power to detect the difference by using an alpha of 0.05 and a two-tailed t test, and the others (using the combined CBZ + OXC group in the present study) having ∼70% power. One would therefore expect two of the latter three studies to be positive and one negative, which is precisely the case. This lends support to the view that a true difference exists in 25-OHD levels in CBZ-treated patients. Conversely, one of these six studies (9), plus another by the same investigators performed in children (8), showed no changes at all in 25-OHD levels. Another study in normal controls showed no changes in 25-OHD when CBZ was added, but a marked and significant increase after discontinuation, a combination of findings that is difficult to interpret (16). Nonetheless, the balance of evidence clearly favors the notion that 25-OHD levels are reduced by CBZ in a significant and clinically relevant manner.

The reigning mechanistic hypothesis for bone loss with enzyme-inducing AEDs begins with the notion that they increase CYP450-mediated catabolism of 25-OHD to less biologically active metabolites, producing a decrease in vitamin D–mediated bone mineralization and intestinal calcium absorption (2). This in turn causes a compensatory increase in PTH, which stimulates the production of P450C1, the enzyme responsible for the conversion of 25-OHD to 1,25-dihydroxyvitamin D (1,25-OHD), which is the biologically active form of the molecule (17). This explains the maintenance of 1,25-OHD levels seen in AED-treated patients (18,19). The chronic elevation of PTH required to maintain 1,25-OHD levels—referred to as secondary hyperparathyroidism—causes an increase in bone turnover, which leads to long-term loss of bone mass (14,20). The increase in bone turnover seen in our CBZ subjects is consistent with the findings of several other studies (5,8,9), with one exception (16). Considerable debate exists about whether this increase in bone turnover is really due to decreases in 25-OHD, however. One recent study (5) found increased BAP in patients taking PHT without a statistically significant alteration of 25-OHD levels; nonetheless, PHT patients had a mean 25-OHD level 33% less than that of LTG-treated patients (20 vs. 30 ng/ml), suggesting a true, clinically meaningful difference that their study was not adequately powered to detect (similar to their findings in CBZ, as discussed earlier). Two other studies found marked increases in bone turnover after CBZ treatment with no change at all in 25-OHD levels (8,9). Our own study yielded mixed results in this regard, with some correlation between 25-OHD and PTH being apparent in only one of two analyses. In our patients, the observed alterations in 25-OHD, PTH, BAP, and OCLN were in the directions predicted by the aforementioned hypothesis (decreased vitamin D, increased PTH, increased BAP and OCLN). It is possible that detecting a statistically significant correlation between vitamin D and the other biomarkers would require a larger number of patients for confirmation, given the many biologic variables affecting PTH secretion. It is also possible that the two are not closely correlated in this population, and that, as has been suggested, the drugs produce increased bone turnover through a different mechanism (5,8,9). In addition, it is worth noting that the bone-resorptive marker NTX was not elevated in our drug-treatment groups, as would be expected from the previous hypothesis. This may be because we were unable to obtain NTX samples consistently in the morning, as has been done in other similar studies, as evidence exists that NTX levels are highest then and diminish over the course of the day (21). More investigation is clearly needed to elucidate the relations between these measures.

Our study was designed to elucidate differential effects of CBZ and OXC on the study variables. The two groups yielded very similar values, indicating that their effects on bone and 25-OHD metabolism are very similar. This is corroborated by the fact that when CBZ patients were crossed over to OXC and retested after 6 weeks, the values did not change at all. Although it is conceivable that this latter finding was due to insufficient washout of any CBZ effects, we believe this is highly unlikely. Both enzyme induction from CBZ use and deinduction from its discontinuation occur quite rapidly, with a half-life of ∼4 days (12). Thus each process should be effectively complete after 3 weeks, and the 6-week period between measurements in the present study should have provided a more than adequate “cushion” to ensure that all CBZ effect was gone. Furthermore, our analyses demonstrated absolutely no correlation between duration of time taking the drug and any of the study variables. Because many of these patients had been taking their AED for only 8–12 weeks, this strongly suggests that the observed changes in bone and 25-OHD metabolism occur soon after initiation of drug treatment and, in light of the aforementioned pharmacokinetic time-course data, are consistent with the hypothesized mechanism for the changes (i.e., CYP450 induction).

The finding that OXC had effects nearly identical to those of CBZ was unexpected. OXC is clearly not a broad-spectrum inducer of CYP450 enzymes (22), but it does have some enzyme-induction properties; for example, it is known to induce CYP450 isoenzymes 3A4 and 3A5, leading to a small number of significant drug–drug interactions (23). The results of our study suggest that OXC may induce P450C1 and P450C24, the enzymes responsible for the metabolism of 25-OHD (17). The correlation between OXC dose and two of the bone-turnover markers also is intriguing in light of preliminary evidence suggesting that OXC may have some dose-dependent CYP450 induction properties (24). An alternative hypothesis is that OXC might exert a direct effect on osteoblast proliferation in a manner similar to that described with CBZ in one investigation (11). Directed experiments are required to examine these propositions.

Perhaps the greatest limitation of our study is that our data cannot exclude the possibility that the observed alterations in bone and 25-OHD metabolism stem from the underlying disease state rather than from the medications. Although this would be a reasonable interpretation of our data alone, the results of other studies make this hypothesis unlikely. Pack et al. (5) found numerous differences in serum calcium and bone-turnover markers among epilepsy patients taking different AEDs. Gough et al. (15), with even greater numbers, found significant decreases in 25-OHD and increases in alkaline phosphatase in epilepsy patients taking CBZ, PHT, or phenobarbital relative to controls, with no changes in valproate-treated patients. Both of these studies suggest that, whether or not epilepsy itself affects bone status, the specific AED treatment contributes to altered bone metabolism.

We excluded patients taking bisphosphonates or corticosteroids from participating, but we did not exclude patients taking estrogen preparations (either alone or in an oral contraceptive compound), although estrogen is known to increase bone turnover. This does not explain our findings, however, because only one of the drug-treated patients was taking an estrogen-containing compound. In a similar vein, bone turnover is known to be substantially increased in menopause, but no postmenopausal women were in our drug-treated cohort. Two drug-treated women had had hysterectomies, however, so that their ovarian hormone status was unknown. When these two patients plus the one estrogen-using epilepsy patient were removed from the analysis, the association between drug treatment and increased BAP remained significant (p < 0.05).

Another limitation of our study is the restricted statistical power to detect differences among the individual groups. This limitation was overcome by combining the CBZ- and OXC-treated groups, which some might view as a proverbial “mixing of apples and oranges.” We would argue that the marked structural similarity between the two drugs and the marked similarity of the results obtained in the two groups (Table 2) indicate that the drugs clearly have highly similar effects on the study variables, thereby justifying this statistical procedure. Nonetheless, further studies should be done with larger samples from each individual AED-treated group to validate our findings.

In summary, our data suggest that epilepsy patients treated with CBZ or OXC experience reduction in 25-OHD levels, along with elevations in other biomarkers that are consistent with increased bone turnover. These alterations may well predispose patients to bone loss over time. It is possible that OXC has adverse effects on bone metabolism at higher doses, but not at lower doses. More study is clearly needed to verify these findings, to elucidate the relation between 25-OHD changes and bone turnover among individuals taking AEDs, and to determine whether these alterations truly produce clinically meaningful bone loss. Pending such studies, it may be prudent to encourage 25-OHD replacement (≥800 IU/day) in all CBZ- or OXC-treated patients (25,26), because such replacement is inexpensive, safe, free of side effects, and beneficial even for normal healthy individuals.


Acknowledgment:  This study was sponsored by a grant from Novartis (CTRI476BUS38, S.M.).