SEARCH

SEARCH BY CITATION

Keywords:

  • Epilepsy;
  • Fracture;
  • Risk;
  • Antiepileptic drugs;
  • Carbamazepine;
  • Valproate;
  • Phenobarbital

Abstract

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

Summary: Purpose: To assess fracture risk associated with different antiepileptic drugs (AEDs). An increased fracture risk has been reported in patients with epilepsy. Classical AEDs have been associated with decreased bone mineral density. The effects of newer AEDs are unknown.

Methods: We undertook a population-based pharmacoepidemiologic case–control study with any fracture as outcome and use of AEDs as exposure variables (124,655 fracture cases and 373,962 controls).

Results: All AEDs were associated with an increased fracture risk in an unadjusted analysis. After adjustment for prior fracture, use (ever) of corticosteroids, comorbidity, social variables, and diagnosis of epilepsy, carbamazepine [CBZ; odds ratio (OR), 1.18; 95% confidence interval (CI), 1.10–1.26], [and oxcarbazepine (OXC; 1.14, 1.03–1.26)], clonazepam (CZP; 1.27, 1.15–1.41), phenobarbital (PB; 1.79, 1.64–1.95), and valproate (VPA; 1.15, 1.05–1.26) were statistically significantly associated with risk of any fracture. Ethosuximide (0.75, 0.37–1.52), lamotrigine (1.04, 0.91–1.19), phenytoin (1.20, 1.00–1.43), primidone (1.18, 0.95–1.48), tiagabine (0.75, 0.40–1.41), topiramate (1.39, 0.99–1.96), and vigabatrin (0.93, 0.70–1.22) were not statistically significantly associated with fracture risk after adjustment for confounders. The relative increase was modest and in the same range for the significant and nonsignificant results. CBZ, PB, OXC, and VPA displayed a dose–response relation. Fracture risk was more increased by liver-inducing AEDs (OR, 1.38; 95% CI, 1.31–1.45) than by noninducing AEDs (1.19; 95% CI, 1.11–1.27).

Conclusions: A very limited increased fracture risk is present in users of CBZ, CZP, OXC, PB, and VPA. A limited significant increase cannot be excluded for the other AEDs because of the statistical power.

An increased fracture risk has been reported in patients with epilepsy (1,2), in particular, linked to tonic–clonic seizures (1). Furthermore, the use of antiepileptic drugs (AEDs) has been associated with increased bone turnover and occasionally with a mineralization defect, leading to decreased bone mineral density (BMD) and thus decreased bone strength (3) with increased fracture risk (4–6).

Some AEDs reduce intestinal calcium absorption (7) and can induce anticonvulsant osteopathy (7–11). Several AEDs [phenytoin (PHT), phenobarbital (PB), and primidone (PRM)] are inducers of microsomal enzymes and increase hepatic catabolism of various vitamin D metabolites, leading to hypovitaminosis D, secondary hyperparathyroidism, and osteomalacia (7,10). In addition, these AEDs may accelerate the catabolism of female sex steroids, which are vital to bone turnover and thus bone biomechanical competence (12). Furthermore, some of the AEDs may exert direct deleterious effects on bone mineral metabolism, independent of their effects on vitamin D and liver function (13). For example, PHT may have a direct adverse effect on bone cells (13), leading to a decrease in BMD (3,6). Evidence for the effect of newer AEDs on BMD and bone metabolism is sparse (5).

Studies comparing BMD in users of AEDs that increase bone turnover through vitamin D insufficiency and secondary hyperparathyroidism, and in those who do not, have differed in their findings. Stephen et al. (5) reported lower femoral neck BMD in patients taking liver-inducing AEDs [carbamazepine (CBZ), PB, PHT) than in those taking noninducing drugs [lamotrigine (LTG), topiramate (TPM), valproate (VPA), vigabatrin (VGB), gabapentin (GBP)], whereas no difference could be demonstrated in the spine. Kafali et al. (14) actually found a higher forearm BMD in children taking CBZ (liver-inducing AED) than in those taking VPA (noninducing AED). This was supported by Sheth et al. (4), who reported that VPA reduced spinal BMD, whereas CBZ did not. Akin et al. (15) found normal BMD levels in children taking both CBZ and VPA (valproic acid/sodium valproate). Duration of treatment is of significance to the observed decrease in BMD (3).

We conducted a population-based pharmacoepidemiologic case–control study to assess the fracture risk associated with different drugs used to control epilepsy.

SUBJECTS AND METHODS

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

Denmark offers good opportunities for studying fracture risk associated with drug use because all fractures are registered in a central register and because all prescriptions from pharmacies also are registered in a central database.

The National Hospital Discharge Register was established in 1977 and records all discharges from the hospitals for administrative purposes. It has nationwide coverage and an almost 100% completeness of registrations (16). From 1977 to 1994, only patients treated on an inpatient basis were registered, but from 1995 onward, all outpatient contacts also were registered (16). On discharge, the physician codes the reason for the contact by using the International Classification of Diseases (ICD) system. From 1977 to 1993, the ICD8 was used, and from 1994, the ICD10 system. Dates of admission and discharge are registered along with codes for any surgical procedures performed. In general, the validity of registrations is high (17), especially for fractures, for which a precision of 97% has been reported (18). The health system is tax financed and free of charge for the inhabitants. All inhabitants have a unique registration code (the central person number, to some degree, similar to the American social security number), which allows registration on an individual basis.

Drugs purchased at pharmacies also are registered under the central person number to allow reimbursement (The National Pharmacological Database run by the Danish Medicines Agency; http://www.dkma.dk). Any drugs bought are registered with Anatomical Therapeutical Classification (ATC) code, dosage sold, and date of sale for the period January 1, 1996, to December 31, 2000.

Furthermore, data are available from the tax authorities and the National Bureau of Statistics on income in 1999 (the tax authorities), social status in 1999 (National Bureau of Statistics), working status in 1999 (National Bureau of Statistics), and contacts with general practitioners and practicing specialists (The National Health Organisation Register) for the period 1996 to 2000.

By combining these data, it is possible to assess whether any drugs are associated with an increased fracture risk.

Design

The study was designed as a case–control study, with any fracture as outcome and use of AEDs as exposure variable. Thus by using The National Hospital Discharge Register, we identified all subjects who had sustained a fracture from January 1, 2000, to December 31, 2000 (cases; n = 124,655). By using the Civil Registration System, which has electronic records on all changes in vital status, including change of address and date of death for the entire Danish population since 1968, we randomly selected three controls for each case, matched by gender and year of birth. The controls were selected by using the incidence-density sampling technique (i.e., the controls had to be alive and at risk for fracture diagnosis at the time the corresponding case was diagnosed). In total, 373,962 controls were included in the study.

Outcome variable

The outcome variable was the occurrence of any fracture in the year 2000 (ICD10 codes: S02.0-S02.9, S07.0-S07.9, S12.0-S12.9, S22.0-S22.9, S32.0-S32.8, S42.0-S42.9, S52.0-S52.9, S62.0-S62.9, S72.0-S72.9, S82.0-S82.9, S92.0-S92.9).

Exposure variables

The drug-exposure variable was use of AEDs with ATC codes: N03AA02 (PB), N03AA03 (PRM), N03AB02 (PHT), N03AB05 (fosphenytoin; FOS), N03AD01 (ethosuximide; ESM), N03AE01 (clonazepam; CZP), N03AF01 (CBZ), N03AF02 (oxcarbazepine; OXC), N03AG01 (VPA), N03AG04 (VGB), N03AG06 (tiagabine; TGB), N03AX09 (LTG), N03AX11 (TPM), N03AX12 (GBP), N03AX14 (levetiracetam; LEV), and N05BA09 (clobazam; CLB) from 1996 to 2000. A separate analysis was performed for AEDs considered to be liver-inducing (CBZ, OXC, PB, PHT, and PRM) and non-liver-inducing (CZP, ESM, LTG, TGB, TPM, VPA, and VGB). For all drugs, information was available on the total amount consumed (defined daily dosages, DDDs) between 1996 and 2000. The exposure to AEDs was assessed before fracture occurrence. The dose–response relation was studied by dividing the total amount ingested into approximate tertiles of DDDs. The duration of drug was closely linked to the total amount of drug ingested, and only results for the total amount of drug used are presented. The aim of the study was to determine the combined effect of use of AEDs, so a separate analysis was performed for the number of AEDs used to analyze the combined effect of multiple drug use.

Adjustment was made for the presence of epilepsy (ICD8: 34509, 34511, 34510, 34518, 34519, 34529, 34530, 34531, 34538, 34539, 34599, ICD10: G400, G401, G402, G403, G404, G405, G406, G407, G408, G409, G410, G411, G412, G418, G419) or its absence in the period from 1977 to 2000.

A separate analysis was performed with separation between different epilepsy types:

  • 1
    Generalized epilepsy [ICD8: 34510 (grand mal), 34518 (other generalized), 34519 (generalized), 34529 (status), ICD10: G403 (generalized), G404 (other generalized), G406 (grand mal), G410, G411, G412, G418, G419 (status)];
  • 2
    Other specified types of epilepsy [ICD8: 34509 (petit mal), 34511 (special), 34530, 34538, 34539 (focal), 34531 (complex), ICD10: G400, DG401 (focal), G402 (complex), G405 (special), G407 (petit mal)], and
  • 3
    Nonspecified epilepsy [ICD8: 34599, ICD10: G408, G409 (all nonspecified)] from the point of view that patients with generalized types of epilepsy would have more pronounced seizures (tonic–clonic seizures, etc.) and thus perhaps be more prone to fractures. This grouping into type of epilepsy was performed to maintain sufficiently large groups to retain statistical power.

Adjustment for comorbidity was done by using the Charlson index, which is a validated index of 19 items of comorbid conditions (acute myocardial infarction, cancer, liver disease, kidney disease, chronic obstructive pulmonary disease, etc.) (19). The data for the Charlson index were retrieved from the National Hospital Discharge Register for the period 1977 to 2000.

Use of corticosteroids, in the period 1996 to 2000, income, marital status, contacts with general practitioners (GPs) or specialists, working status, bed days, as well as prior fractures in the period 1977 to 2000 were included as confounders.

Statistical analyses

Mean and standard deviation were used as descriptive statistics. Crude odds ratios (ORs) were calculated and 95% confidence intervals approximated by using the method of Miettinen (20). A conditional logistic regression analysis was used to assess the association between any fracture and the exposure variable. Tests for trend were performed by the rank-sum method. A separate analysis was performed for the total number of different AEDs used during the 5-year period spanned by the prescription database [i.e., never used, one drug (i.e., if PHT or CBZ was used), two drugs (i.e., if LTG was replaced by CBZ or if a combination of two drugs was used)].

Age- (40 years or younger, 41 to 60 years, and older than 60 years) and gender-stratified analyses were performed.

A separate analysis of the fracture risk associated with use of different AEDs was performed in the patients with epilepsy.

Analyses were performed by using STATA 8.1 (STATA Corp., College Station, TX, U.S.A.) and SPSS 10.1.0 (SPSS Inc., Chicago, IL, U.S.A.), both in the UNIX version.

RESULTS

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

Table 1 shows baseline characteristics of the cases and the controls. They were well matched for age and gender. Cases tended to have a higher frequency of comorbidity (fewer with a Charlson index of 0) and a higher number of comorbid conditions than did the controls. The cases were more often retired, were more likely to be divorced or unmarried, and had a lower income than the controls. The cases had a much higher frequency of prior fractures than the controls and more often had used bisphosphonates and selective estrogen-receptor modulators (SERMs; e.g., raloxifene).

Table 1. Characteristics of patients and controls: any fracture
VariableCases (n = 124,655)Controls (n = 373,962)p Value
  1. GP, general practitioner; SERM, selective estrogen-receptor modulators (raloxifene); HRT, hormone replacement therapy (estrogen compounds with or without progestogens): women only.

  2. aA composite index of 19 comorbid conditions (see text).

Age (yr)43.44 ± 27.3943.44 ± 27.39-
Gender
 Men60,107 (48.2%)180,321 (48,2%) 
 Women64,548 (51.8%)193,641 (51.8%)-
Annual income (DKR)161,036 ± 138,789172,322 ± 193,704<0.01
Marital status
 Widowed18,365 (14.8%) 52,550 (14.2%) 
 Divorced10,423 (8.4%) 23,239 (6.3%) 
 Married35,859 (28.9%)123,719 (33.3%) 
 Unmarried59,335 (47.8%)171,349 (46.2%) 
 Other   90 (0.1%)   264 (0.1%)<0.01
Occupational status
 Independent3,374 (3.3%)11,816 (3.9%) 
 Assisting wife  209 (0.2%)   951 (0.3%) 
 Working37,797 (36.9%)124,984 (40.8%) 
 Retired40,201 (39.3%)109,447 (35.7%) 
 Other20,752 (20.3%) 59,278 (19.3%)<0.01
Charlson indexa
 097,256 (78.0%)314,099 (84.0%) 
 1–219,634 (16.8%) 47,745 (12.8%) 
 3–45,450 (4.4%) 9,132 (2.4%) 
 ≥52,315 (1.9%) 2,986 (0.8%)<0.01
Previous fracture41,315 (33.1%) 56,200 (15.0%)<0.01
Number of bed days in hospital in 1999 9.7 ± 39.7 4.2 ± 20.3<0.01
Number of contacts with GP or specialists in 199923.9 ± 43.318.1 ± 31.4<0.01
Use of antiosteoporosis drugs
 Any antiresorptive drug12,900 (10.3%)28,101 (7.5%)<0.01
 Bisphosphonates5,679 (4.6%) 3,887 (1.0%)<0.01
 HRT  440 (6.3%)   244 (6.6%)<0.01
 SERMs7,789 (0.4%)24,623 (0.1%)<0.01
 Ever use of any corticosteroid67,695 (54.3%)189,636 (50.7%)<0.01

Table 2 shows the distribution of epilepsy diagnoses and use of AEDs. None of the participants had used GBP during the study period. The cases were more likely to have a diagnosis of epilepsy than were the controls and also more often had used AEDs. More subjects had used AEDs (5.7 and 2.9%) than were diagnosed with epilepsy (2.5 and 1.3%, respectively). Similar trends were found when analyzing hip, Colles', and spine fractures.

Table 2. Epilepsy and use of antiepileptic drugs
VariableAny fractureHipColles'Spine
Cases (n = 124,655)Controls (n = 373,962)Cases (n = 10,530)Controls (n = 31,535)Cases (n = 20,035)Controls (n = 60,030)Cases (n = 3,364)Controls (n = 10,079)
  1. ap < 0.05 for comparison of cases and controls.

Epilepsy3,178 (2.5%) 4,947 (1.3%)a 377 (3.6%)  345 (1.1%)a419 (2.1%)  762 (1.3%)a100 (3.0%)129 (1.3%)a
Any AED7,091 (5.7%)10,974 (2.9%)a1,081 (10.3%)1,578 (5.0%)a910 (4.5%)1,706 (2.8%)a332 (9.9%)410 (4.1%)a
Carbamazepine1,804 (1.4%) 2,893 (0.8%)a 287 (2.7%)  438 (1.4%)a233 (1.2%)  475 (0.8%)a 67 (2.0%)109 (1.1%)a
Clonazepam  835 (0.7%) 1,135 (0.3%)a 116 (1.1%)  120 (0.4%)a103 (0.5%)  163 (0.3%)a 34 (1.0%) 32 (0.3%)a
Ethosuximide    22 (0.02%)    45 (0.01%) 2 (0%) 1 (0%)0      8 (0%) 1 (0%)0     
Phenobarbital1,401 (1.1%) 1,230 (0.3%)a 211 (2.0%)  185 (0.6%)a175 (0.9%)  174 (0.3%)a 61 (1.8%) 48 (0.5%)a
Phenytoin  278 (0.2%)   338 (0.1%)a  61 (0.6%)   60 (0.2%)a 41 (0.2%)   62 (0.1%)a  8 (0.2%)  9 (0.1%)a
Lamotrigine  600 (0.5%)   844 (0.2%)a  61 (0.6%)   60 (0.2%)a 63 (0.3%)  128 (0.2%)a 24 (0.7%) 13 (0.1%)a
Oxcarbazepine1,006 (0.8%) 1,455 (0.4%)a 132 (1.3%)  132 (0.4%)a119 (0.6%)  202 (0.3%)a 43 (1.3%) 52 (0.5%)a
Primidone  164 (0.1%)   236 (0.1%)a  32 (0.3%)   46 (0.1%)a 23 (0.1%)  44 (0.1%) 11 (0.3%) 5 (0%)a
Tiagabine    25 (0.02%)      34 (0.01%)a 4 (0%)  1 (0%)a 3 (0%)10 (0%) 1 (0%) 0 (0%) 
Topiramate  121 (0.1%)    121 (0.03%)a   8 (0.1%)  4 (0%)a 6 (0%)26 (0%)  5 (0.1%) 3 (0%)a
Valproate1,154 (0.9%) 1,804 (0.5%)a 166 (1.6%)  224 (0.7%)a146 (0.7%)  265 (0.4%)a 45 (1.3%) 66 (0.7%)a
Vigabatrin  151 (0.1%)    203 (0.05%)a 5 (0%) 4 (0%) 17 (0.1%)  39 (0.1%) 10 (0.3%) 5 (0%)a

Table 3 shows the OR associated with any fracture in cases compared with controls both unadjusted (crude), and after adjustment for a number of potential confounders. A prior fracture and ever use of corticosteroids both were associated with an increased risk of any fracture, and the estimates changed little after adjustment. Comorbidity (Charlson index) also was associated with an increased risk of any fracture, and the relative risk increased with the number of comorbid conditions. A diagnosis of epilepsy was associated with an increased risk of any fracture, but the risk estimate was somewhat attenuated after stratified analysis. It did not change the results to adjust for type of epilepsy (generalized, other specified, and nonspecified).

Table 3. Crude and adjusted odds ratios for any fracture in patients who used antiepileptic drugs
VariableCrude OR (95% CI)Adjusted OR
  1. Logistic regression analysis; values expressed as odds ratio (OR) and 95% confidence interval (CI). All are ever use vs. never use of actual drug.

  2. GP, general practitioner.

Carbamazepine1.88 (1.78–2.00)1.18 (1.10–1.26)
Clonazepam2.16 (2.03–2.42)1.27 (1.15–1.41)
Ethosuximide1.47 (0.88–2.43)0.75 (0.37–1.52)
Phenobarbital3.44 (3.20–3.70)1.79 (1.64–1.95)
Phenytoin2.47 (2.12–2.88)1.20 (1.00–1.43)
Lamotrigine2.14 (1.93–2.37)1.04 (0.91–1.19)
Oxcarbazepine2.09 (1.93–2.26)1.14 (1.03–1.26)
Primidone2.09 (1.72–2.54)1.18 (0.95–1.48)
Tiagabine2.21 (1.33–3.65)0.75 (0.40–1.41)
Topiramate3.00 (2.36–3.82)1.39 (0.99–1.96)
Valproate1.93 (1.79–0.07)1.15 (1.05–1.26)
Vigabatrin2.23 (1.82–0.74)0.93 (0.70–1.22)
Corticosteroid1.16 (1.14–1.17)1.03 (1.01–1.05)
Epilepsy vs. no epilepsy1.95 (1.87–2.04)1.21 (1.14–1.28)
Charlson index
 1–21.33 (1.30–1.35)1.16 (1.14–1.19)
 3–41.93 (1.86–1.99)1.44 (1.39–1.50)
 ≥52.50 (2.37–2.64)1.80 (1.70–1.91)
Previous fracture2.80 (2.76–2.84)2.58 (2.53–2.63)
Number of bed days in 1999
 1–31.73 (1.70–1.76)1.52 (1.49–1.56)
 3–91.93 (1.89–1.97)1.55 (1.52–1.59)
 ≥102.51 (2.46–2.56)1.87 (1.83–1.92)
Number of contacts with GP or specialist in 1999
 1–71.22 (1.19–1.25)1.12 (1.08–1.16)
 8–181.34 (1.31–1.38)1.19 (1.14–1.23)
 ≥191.73 (1.69–1.781.33 (1.29–1.39)
Working or not0.83 (0.82–0.84)0.84 (0.82–0.87)
Income in 1999 (≥150,000 vs. <150,000 DKK)0.87 (0.85–0.88)0.95 (0.93–0.97)
Living with another person vs. living alone0.81 (0.80–0.82)0.81 (0.80–0.83)

All AEDs were associated with an increased fracture risk in the unadjusted analysis. However, after stratified analysis, the risk estimates were attenuated. CBZ (and OXC), CZP, PB, and VPA were all associated with an increased fracture risk after adjustment, the highest relative risk remaining for PB. The relative risks were modest.

ESM, LTG, PHT, PRM, TGB, TPM, and VGB were not statistically significantly associated with fracture risk after adjustment. However, the power was low, and a small excess fracture risk may be present.

Age and gender stratification did not change the estimates. The results for the different AEDs were similar if the results were restricted to patients with epilepsy.

Table 4 shows the fracture risk associated with use of different AEDs stratified by skeletal site. An increased fracture risk was present across skeletal sites for PB.

Table 4. Fracture risk associated with use of antiepileptic drugs at various skeletal sites
VariableHipColles'Spine
  1. Logistic regression (adjusted OR and 95% CI). For all drugs, information on the total amount consumed (defined daily dosages; DDDs) between 1995 and 2000 were available.

  2. GP, general practitioner; OR, odds ratio; CI, confidence interval.

Carbamazepine1.33 (1.13–1.58)1.05 (0.88–1.25)0.91 (0.631.30)
Clonazepam1.91 (1.43–2.56)1.38 (1.06–1.80)1.71 (0.99–2.96
Ethosuximide1.25 (0.06–23.5)--
Phenobarbital1.69 (1.34–2.15)1.93 (1.53–2.43)2.03 (1.31–3.15)
Phenytoin1.41 (0.94–2.15)1.10 (0.71–1.70)1.34 (0.47–3.83)
Lamotrigine1.30 (0.85–2.00)0.81 (0.55–1.21)2.47 (1.13–5.39)
Oxcarbazepine1.48 (1.11–1.97)1.31 (0.99–1.74)1.13 (0.69–1.84)
Primidone1.10 (0.66–1.85)1.06 (0.62–1.81)3.82 (1.16–12.6)
Tiagabine4.01 (0.35–45.5)0.39 (0.09–1.71)-
Topiramate0.90 (0.20–4.03)1.12 (0.35–3.55)0.40 (0.06–2.78)
Valproate1.06 (0.84–1.34)1.28 (1.00–1.65)1.07 (0.68–1.70)
Vigabatrin0.52 (0.09–3.19)1.13 (0.53–2.40)2.49 (0.68–9.19)
Corticosteroid0.89 (0.84–0.93)1.01 (0.97–1.06)1.11 (1.01–1.21)
Epilepsy vs. no epilepsy1.57 (1.30–1.90)1.25 (1.05–1.49)1.02 (0.73–1.43)
Charlson index
 1–21.51 (1.43–1.60)1.03 (0.97–1.09)1.37 (1.22–1.53)
 3–42.10 (1.93–2.28)1.24 (1.12–1.38)1.84 (1.53–2.20)
 ≥52.79 (2.47–3.15)1.33 (1.13–1.57)2.83 (2.13–3.75)
Previous fracture2.25 (2.14–2.37)2.03 (1.93–2.13)2.50 (2.27–2.76)
Number of bed days in 1999
 1–31.33 (1.22–1.45)1.32 (1.23–1.41)1.52 (1.33–1.75)
 3–91.11 (1.03–1.20)1.28 (1.20–1.37)1.46 (1.28–1.67)
 ≥101.68 (1.58–1.78)1.30 (1.21–1.39)1.89 (1.67–2.14)
Number of contacts with GP or specialists in 1999
 1–70.93 (0.81–1.07)1.15 (1.04–1.26)1.24 (0.99–1.54)
 8–180.91 (0.80–1.04)1.14 (1.03–1.26)1.32 (1.06–1.65)
 ≥191.12 (0.98–1.28)1.17 (1.06–1.30)1.58 (1.26–1.97)
Working or not0.41 (0.35–0.48)0.84 (0.78–0.91)0.69 (0.59–0.81)
Income in 1999 (≥150,000 vs. <150,000 DKK)0.84 (0.83–0.94)0.95 (0.90–1.01)1.04 (0.93–1.17)
Living with another person vs. living alone0.78 (0.73–0.82)0.83 (0.79–0.87)0.88 (0.80–0.97)

In the hip, CBZ (and OXC), CZP, and PB were associated with an increased fracture risk, whereas only PB was associated with an increased fracture risk for Colles' fractures. In the spine, PB, LTG, and PRM were associated with an increased fracture risk.

No statistically significant increase in fracture risk could be demonstrated across skeletal sites for ESM, PHT, TGB, TPM, and VGB. However, the power was low, and a small excess fracture risk may be present.

A grouped analysis demonstrated an increased risk of any fracture in both inducing (OR, 1.38; 95% CI, 1.31–1.45) and noninducing AEDs (1.19; 95% CI, 1.11–1.27). The same was the case for hip (1.52; 1.33–1.73 and 1.37; 1.15–1.64), Colles' (1.33; 1.17–1.52, and 1.25; 1.05–1.48), and spine fractures (1.50; 1.17–1.92, and 1.48; 1.07–1.2.05). The OR associated with inducing AEDs was significantly higher for any fracture than that for noninducing AEDs (two-sided p value: 2p < 0.01), whereas this was not the case for any of the individual skeletal sites.

Table 5 shows the dose–response relation for risk of any fracture in users of the various AEDs. Most subjects had used a low number of defined dosages (<400 DDDs), and because the observation period was 5 years (approximately 1,800 days), most had used well below 1 DDD per day. CBZ, PB, OXC, and VPA displayed a dose–response relation, whereas a borderline significant relation was present for CZP. No dose–response relation could be demonstrated for ESM, PHT, LTG, PRM, TGB, TPM, and VGB. A significant trend toward more fractures was noted with increasing numbers of AEDs used (OR, 1.20; 95% CI, 1.17–1.24 per AED used). It did not change the results to analyze current and previous users separately.

Table 5. Dose-response relation for antiepileptic drugs, with any fracture as end point
Drug<50 DDDs50–400 DDDs>400 DDDsTest for trend
  1. Unadjusted odds ratios and 95% confidence intervals.

  2. DDD, sum of all ingested defined daily dosages of the drug in question. Test for trend was performed by rank sum.

Carbamazepine1.68 (1.53–1.84)1.81 (1.61–2.05)2.22 (2.01–2.44)<0.01
Clonazepam1.90 (1.68–2.14)2.28 (1.95–2.66)3.89 (3.10–4.89) 0.05
Ethosuximide0.50 (0.12–2.17)1.43 (0.70–2.93)2.70 (1.14–6.41) 0.98
Phenobarbital2.73 (2.34–3.19)3.64 (3.24–4.08)3.74 (3.33–4.19)<0.01
Phenytoin2.57 (1.61–4.12)3.25 (2.34–4.50)2.26 (1.87–2.72) 0.48
Lamotrigine1.90 (1.60–2.27)2.27 (1.93–2.68)2.32 (1.91–2.82) 0.17
Oxcarbazepine1.81 (1.53–2.14)2.14 (1.86–2.45)2.20 (1.95–2.47) 0.03
Primidone1.29 (0.92–1.80)2.14 (1.48–3.12)3.49 (2.53–4.82) 0.72
Tiagabine1.29 (0.50–3.34)3.21 (1.61–6.40)2.00 (0.58–6.91) 0.95
Topiramate3.15 (2.10–4.72)2.61 (1.83–3.72)3.95 (2.27–6.86) 0.73
Valproate1.94 (1.70–2.22)1.75 (1.55–1.96)2.17 (1.90–2.47) 0.02
Vigabatrin1.86 (1.15–3.01)2.31 (1.60–3.35)2.34 (1.76–3.12) 0.72

DISCUSSION

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

In this large-scale population-based study, we demonstrated a modestly increased fracture risk in patients with epilepsy and in users of several types of AEDs, after adjustment for confounders.

The major advantage of the study is that it is population based. Selection bias is thus limited, as all fracture cases are included. In the year 2000, when the cases were sampled, both inpatients and outpatients were included in the National Hospital Discharge Register (i.e., no cases were left out). Almost all fracture cases are treated in emergency departments and hospitals and are thus covered by the register. Information bias also is limited, as the precision of a fracture diagnosis is high in Denmark (18). All drugs bought at pharmacies were included, so only drugs given during hospital stay were not included (i.e., drugs given during an epileptic seizure that led to an admission to a hospital or an emergency department). However, the amount of drugs given this way is minimal and may not have influenced the results.

The major drawback is that the prescription database spanned only the past 5 years. A subject may thus have been taking an old AED for many years and only recently shifted to a newer AED, with only the newer AED being included in the prescription database because AEDs were introduced in different eras. The patients will thus have been exposed to the older AEDs (PHT, PB, etc.) for a longer period than to the newer AEDs (e.g., LTG). If an increased fracture risk were associated with the older AEDs, but not the newer ones, patients shifting from an old to a new AED before the start of the prescription database would have the fracture risk actually associated with the old drug confused with that of the new drug. This may thus have underestimated the risk associated with the use of older AEDs and overestimated the effect of newer AEDs. However, even the older AEDs were associated with only a modest increase in fracture risk, although the exposure time to these must be anticipated to be the longest. As we could not exclude patients free of AED exposure at the start of the observation period covered by the prescription database, this difference cannot be addressed, and patients with long exposure times may thus be compared with patients with short exposure times.

Another drawback is the fact that we did not separate between “active epilepsy” and “epilepsy in remission” (i.e., whether the subject in question had seizures during the observation period or just had a prior diagnosis of epilepsy with no current activity in the disease). This may have biased the results, as patients with active epilepsy are more likely to take drugs and to sustain fractures than are those without current seizures (e.g., VPA may be prescribed for generalized epilepsy with tonic–clonic seizures, but the tonic–clonic seizures may per se contribute to fracture risk, also called confounding by indication). Patients with more active epilepsy also may use higher doses of AEDs. However, the adjustment for numbers of contacts with general practitioners and bed days in hospitals may partly have adjusted for this, as patients with active epilepsy may be more likely to undergo hospitalization and have contacts with their general practitioners and specialists than are those with epilepsy in remission, although a complete separation is not possible.

The main aim of the study was to assess the fracture risk associated with use of AEDs. By including all subjects exposed to AEDs and not limiting the study to a cohort with a diagnosis of epilepsy, it was possible to address the issue of any potential detrimental effects of the drugs per se, and not just the effects in a group with an a priori increased fracture risk, as in patients with epilepsy. This allowed us to show the fracture risk associated with a diagnosis of epilepsy (OR, 1.21; 1.14–1.28 for any fracture) after adjustment for important confounders.

The fracture risk was present after adjustment for important confounders such as use of corticosteroids (21) and prior fracture (22). However, other important confounders such as a family fracture history (23) and smoking (24) could not be included because of the register-based nature of the study. The significance of this remains unclear.

The reasons that more subjects had used AEDs than were diagnosed with epilepsy may be that some subjects had their epilepsy diagnosed before 1977, when the Hospital Discharge Register was founded, and subsequently they had not been in contact with the hospitals or that some used the drugs for conditions other than epilepsy [CBZ for pain (e.g., neuralgia of the trigeminal nerve), PB for abstinence in alcohol abuse, psychiatric conditions]. Diabetes with complications was included in the Charlson index, and this would partly address the issue of use of AEDs for neuropathic pain. The use of PB for abstinence and other drugs for psychiatric conditions was partially corrected for by including number of contacts with GPs, work status, and number of bed days in hospital.

Some of the AEDs were used only by very few subjects, despite the large overall sample size (ESM, PRM, TGB, TPM, and VGB), and this made the results for these types of AEDs uncertain; small changes in relative fracture risk may have been overlooked, even though the sample size was large.

A very limited statistically significantly increased risk of any fracture was found for CBZ, CZP, OXC, PB, and VPA. A dose–response relation with fracture risk also was demonstrated for these types of drugs, making it probable that these drugs did indeed contribute to the increase in fracture risk. In the hip, an increased fracture risk was present for the same drug types, except for VPA. However, the increase in relative risk in those drugs found to be statistically associated with fracture risk was the same as that in those in which no significant association could be demonstrated. Because of sample size and power, small significant associations may have been overlooked.

The cumulative dose for all types of AEDs was very low (well below 1 DDD per day; i.e., many subjects took AEDs for a only limited time or shifted between several types of drugs). Despite this, an increase in fracture risk could be demonstrated. A trend toward more fractures was noted with increasing numbers of AEDs, probably linked to an increase in DDDs in those who had used many AEDs. This also reflects that those with active epilepsy may be more likely to have changed types of AEDs and therefore have used more than one AED, some of the risk thus being linked to disease activity.

The absence of an association between fracture risk and use of PHT is in contrast to one prior questionnaire-based cohort study performed on patients from a highly specialized neurology department (2), and a small study (n = 87) of patients exclusively treated with PHT (25). The patients in the previous studies may thus have had a more advanced degree of epilepsy, which may have been difficult to control by other than combinations of liver-inducing drugs. This may mean that a confounding by indication may be responsible for the association between PHT and fracture risk. In our study, the relative risk of any fracture associated with use of PHT was 1.20 (95% CI, 1.00–1.43; i.e., borderline statistically significant), whereas the relative risk of any fracture associated with CBZ was 1.18 [95% CI, 1.10–1.26; i.e., within the same range, but statistically significant because of the larger number of users (1,804 vs.278)]. Some of the difference in significance may thus be due to differences in statistical power.

Among children, one study (3) demonstrated a decrease in BMD of ∼0.9 Z-scores in the spine among users of PHT. By using the estimates by Marshall et al. (26) this converts to a relative risk of any fracture of 1.50.9= 1.4. Among adults, Kubota et al. (27) demonstrated a BMD deficit of 0.7 Z-scores in the lumbar spine and 0.8 Z-scores in the hip. These estimates equal relative risks of any fracture of 1.3 and 1.5 (26). In our study, the relative risk of any fracture associated with use of PHT was 1.20 (95% CI, 1.00–1.43; i.e., within the same range as those expected from BMD data). This observation is in accordance with other studies on different types of AEDs, in which the decrease in BMD associated with drug use has been modest (5).

The increase in fracture risk was seen in both liver-inducing AEDs (CBZ and PB) and in noninducers (VPA)—an observation also done in the grouped analysis—and absence of an increase was demonstrated in both inducers (PHT) and noninducers (LTG and TPM). It thus seems that the liver-inducing potential per se was not responsible for all the increase in fracture risk. There was a higher OR for any fracture for liver-inducing AEDs compared with noninducing AEDs, whereas this was not the case for any of the individual skeletal sites (hip, Colles', and spine fracture), but power was limited for some of the AEDs at the individual skeletal sites. It should be noted that the subdivision into liver-inducing and non–liver-inducing AEDs might pose problems, as not all AEDs can be classified with certainty (e.g., OXC).

The higher comorbidity in fracture cases probably was responsible for the observation that more were retired from the work force, the latter again explaining the lower income in the cases than in the controls.

The absence of GBP use in the cohort was probably due to the indications for use of GBP and the reimbursement policy in Denmark.

The higher percentage of cases than controls that used bisphosphonates and SERMs was due to confounding by indication (i.e., subjects with a prior fracture were more likely to receive treatment with these drugs than were those without).

In conclusion, a very limited increased fracture risk seemed present in users of CBZ, CZP, OXC, PB, and VPA after adjustment for important confounders. The increase in relative risk was in the same range for the significant and nonsignificant results. A limited significant increase cannot be excluded for the other AEDs because of the statistical power.

Acknowledgments

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

Acknowledgment:  We thank The National Bureau of Statistics (Statistics Denmark) for their help, without which this project would not have been possible. Financial support was provided by “Bagermester August H. Jensen og Hustrus Fond” and “Laura og Jens Veng Christensens Fond.”

REFERENCES

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES
  • 1
    Persson HBI, Alberts KA, Farahmand BY, et al. Risk of extremity fractures in adult outpatients with epilepsy. Epilepsia 2002;43: 76872.DOI: 10.1046/j.1528-1157.2002.15801.x
  • 2
    Vestergaard P, Tigaran S, Rejnmark L, et al. Fracture risk is increased in epilepsy. Acta Neurol Scand 1999;99: 26975.
  • 3
    Chung S, Ahn C. Effects of anti-epileptic drug therapy on bone mineral density in ambulatory epileptic children. Brain Dev 1994;16: 3825.
  • 4
    Sheth RD, Wesolowski CA, Jacob JC, et al. Effect of carbamazepine and valproate on bone mineral density. J Pediatr 1995;127: 25662.
  • 5
    Stephen LJ, McLellan AR, Harrison JH, et al. Bone density and antiepileptic drugs: a case-controlled study. Seizure 1999;8: 33942.DOI: 10.1053/seiz.1999.0301
  • 6
    Välimäki MJ, Tiihonen M, Laitinen K, et al. Bone mineral density measured by dual-energy X-ray absorptiometry and novel markers of bone formation and resorption in patients on antiepileptic drugs. J Bone Miner Res 1994;9: 6317.
  • 7
    Mosekilde L, Hansen HH, Christensen MS. Fractional intestinal calcium absorption in epileptics on anticonvulsant therapy: short term effects of 1,25- dihydroxycholecalciferol and 25-hydroxycholecalciferol. Acta Med Scand 1979;205: 4059.
  • 8
    Dymling JF, Lidgren L, Walloe A. Biochemical variables related to calcium metabolism in epileptics. Acta Med Scand 1979;205: 4014.
  • 9
    Nilsson OS, Lindholm TS, Elmstedt E, et al. Fracture incidence and bone disease in epileptics receiving long- term anticonvulsant drug treatment. Arch Orthop Trauma Surg 1986;105: 1469.
  • 10
    Mosekilde L, Melsen F. Dynamic differences in trabecular bone remodeling between patients after jejuno-ileal bypass for obesity and epileptic patients receiving anticonvulsant therapy. Metab Bone Dis Rel Res 1980;2: 7782.DOI: 10.1016/0221-8747(80)90001-6
  • 11
    Sotaniemi EA, Hakkarainen HK, Puranen JA, et al. Radiologic bone changes and hypocalcemia with anticonvulsant therapy in epilepsy. Ann Intern Med 1972;77: 38994.
  • 12
    Alexeeva L, Burkhardt P, Christiansen C, et al. Assessment of fracture risk and its application to screening for postmenopausal osteoporosis: report of a WHO Study Group. 1st ed. Geneva : WHO Technical Report Series 843, 1994.
  • 13
    Hahn TJ, Scharp CR, Richardson CA, et al. Interaction of diphenylhydantoin (phenytoin) and phenobarbital with hormonal medication of fetal rat bone resorption in vitro. J Clin Invest 1978;62: 40614.
  • 14
    Kafali G, Erselcan T, Tanzer F. Effect of antiepileptic drugs on bone mineral density in children between ages 6 and 12 years. Clin Pediatr (Phila) 1999;38: 938.
  • 15
    Akin R, Okutan V, Sarici U, et al. Evaluation of bone mineral density in children receiving antiepileptic drugs. Pediatr Neurol 1998;19: 12931.DOI: 10.1016/S0887-8994(98)00039-3
  • 16
    Andersen TF, Madsen M, Jørgensen J, et al. The Danish National Hospital Register. Dan Med Bull 1999;46: 2638.
  • 17
    Mosbech J, Jørgensen J, Madsen M, et al. The Danish National Patient Register: evaluation of data quality. Ugeskr Læger 1995;157: 37415.
  • 18
    Vestergaard P, Mosekilde L. Fracture risk in patients with celiac disease, Crohn's disease, and ulcerative colitis: a nation-wide follow-up study in 16,416 patients in Denmark. Am J Epidemiol 2002;156: 110.
  • 19
    Charlson ME, Pompei P, Ales KL, et al. A method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis 1987;40: 37383.
  • 20
    Miettinen OS. Theoretical epidemiology: principles of occurrence research in medicine. 1st ed. New York : Wiley & Sons, 1985.
  • 21
    Van Staa TP, Leufkens HGM, Abenhaim L, et al. Use of oral corticosteroids and risk of fractures. J Bone Miner Res 2000;15: 9931000.
  • 22
    Klotzbuecher CM, Ross PD, Landsman PB, et al. Patients with prior fractures have an increased risk of future fractures: a summary of the literature and statistical synthesis. J Bone Miner Res 2000;15: 72139.
  • 23
    Cummings SR, Nevitt MC, Browner WS, et al. Risk factors for hip fracture in white women: study of Osteoporotic Fractures Research Group. N Engl J Med 1995;332: 76773.
  • 24
    Vestergaard P, Mosekilde L. Fracture risk associated with smoking: a meta-analysis. J Intern Med 2003;254: 57283.
  • 25
    Lidgren L, Walloe A. Incidence of fracture in epileptics. Acta Orthop Scand 1977;48: 35661.
  • 26
    Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ 1996;312: 12549.
  • 27
    Kubota F, Kifune A, Shibata N, et al. Bone mineral density of epileptic patients on long-term antiepileptic drug therapy: a quantitative digital radiography study. Epilepsy Res 1999;33: 937.