Exposure to antiepileptic drugs and the risk of hip fracture: A case-control study


Address correspondence to Ioannis Tsiropoulos, Department of Neurology, Odense University Hospital, Sdr Boulevard 29, 5000 Odense C, Denmark. E-mail: iotsi@dadlnet.dk


Purpose: To investigate whether the use of antiepileptic drugs (AEDs) increases the risk of hip fracture.

Methods: We performed a case-control study using data from the Funen County (population 2004: 475,000) hip fracture register. Cases (n = 7,557) were all patients admitted to county hospitals with a hip fracture during the period 1996–2004. Controls (n = 27,575) were frequency matched by age and gender. Information on use of AEDs, other drugs, and hospital contacts was available from local registers. Odds ratios (ORs) with 95% confidence intervals (CI) for hip fracture were estimated by unconditional logistic regression.

Results: Fracture risk was increased with ever use of any AED (OR: 1.31; 95% CI: 1.16–1.48). The risk was also increased with use of only enzyme inducing (OR: 1.31; 95% CI: 1.14–1.51), but not with use of only noninducing AEDs (OR: 1.03; 95% CI: 0.77–1.37). Current (OR: 1.92; 95% CI: 1.58–2.33) and recent use, as well as high daily (OR: 1.50; 95% CI: 1.24–1.82) and cumulative dose increased fracture risk, but long treatment duration or previous use did not. The risk was modified by the presence of an epilepsy diagnosis.

Conclusion: Use of AEDs modestly increases the risk of hip fracture. The risk increase is probably associated to a higher degree with a dose dependent effect on CNS with current and recent use, than with an effect on bone tissue.

Among known complications to treatment with antiepileptic drugs (AEDs), their detrimental effect on bone health has been a long-standing issue. Osteomalacia with use of AEDs was first recognized in institutionalized persons (Wright, 1965; Richens & Rowe, 1970) with a long history of severe epilepsy (Kruse, 1968). Later studies suggested that decreased bone mineral density (BMD) is not particular to special subgroups, but also occurs in otherwise ordinary AED users, both old (Stephen et al., 1999) and young (Verrotti et al., 2002), even after short observation periods. Reduction in BMD was initially thought to be caused by enhanced vitamin D catabolism through liver enzyme induction. However, AEDs which do not induce liver enzymes also affect BMD (Sato et al., 2001; Boluk et al., 2004), AEDs may directly affect bone tissue (Feldkamp et al., 2000; Sato et al., 2001) and osteoporosis can occur without vitamin D deficiency (Fitzpatrick, 2004). Several studies documented an increased fracture risk among AED users (Cummings et al., 1995; Van Staa et al., 2002; Vestergaard et al., 2004; Souverein et al., 2006), reporting current use (Cummings et al., 1995), duration of treatment (Souverein et al., 2006), liver enzyme induction, and number of AEDs (Vestergaard et al., 2004) as significant parameters for the risk increase. AEDs can also cause fractures by increasing the risk for falling (Ensrud et al., 2002). A recurrent problem in studies linking bone changes to AEDs is that they are performed on persons with epilepsy, still the main indication for AED use, and in itself a well-documented fracture cause (Vestergaard et al., 1999; Persson et al., 2002; Souverein et al., 2005), although the seizure related proportion of fractures has varied from 6% (Stephen et al., 1999) to 34% (Vestergaard et al., 1999) in different studies.

The aim of this study was to assess the hip fracture risk associated with AED use and determine parameters of AED use influencing this risk, particularly epilepsy as treatment indication.


Data sources

Data for the study were retrieved from three different sources. The Funen County Hip Fracture Register was started on January 1, 1996. Every resident with a hip fracture admitted to one of the orthopedic departments in the county (population 2004: 475,000) has been consecutively included. The register has been validated and completed through retrieval and review of the individual  patient records (Nymark et al., 2006b). Persons in the register are identified by a Central Person Register - CPR - number, a unique person identifier, shared with virtually all other health related registers in Denmark, thereby allowing record-linkage studies (Gaist et al., 1997; Hallas, 2001).

Odense Pharmacoepidemiological Database (OPED) is a population based prescription database derived from reimbursement data. It contains data from 1990 with complete coverage of the County of Funen since November 1992. Only subsidized prescriptions are covered, thus excluding over the counter drugs (laxatives, high-dose aspirin, acetaminophen, antihistamines) and some nonreimbursed prescription drugs such as minor tranquilizers, oral contraceptives, and partly H-2-receptor antagonists and ibuprofen. Drugs dispensed by the county hospital pharmacies for inpatient use are not registered either. Inpatient AED use in 2004 amounted to 4.2% of the total use, according to official sources (Danish Medicines Agency, 2007).

Each prescription record contains the same unique person identifier (CPR number), age, sex, dispensing date, a pharmacy code, a prescriber code, and a full account of the dispensed product including the Anatomical Therapeutic Chemical (ATC) classification code, brand name, quantity, formulation, number of defined daily doses (DDD) purchased and price. OPED contains also a residence history of all county inhabitants with information on immigration to and emigration from the county, as well as date of death.

The Funen County Patient Administrative System (FPAS) is a comprehensive electronic register of admissions and outpatient contacts to all county hospitals since 1973. Besides demographic information and the same unique person identifier, records contain information on dates of admission and discharge, outpatient contacts and procedures performed, as well as up to 20 diagnoses coded according to the International Classification of Diseases (ICD), 8th revision for 1973 to 1993 and 10th revision since 1994.

Case and control definition

We performed a case-control study identifying cases by their first presence in the Funen County Hip Fracture Register from January 1, 1996 to October 31, 2004, which constituted the study period. Hip fractures were defined by ICD-10 diagnoses: S72.0 (femoral neck), S72.1 (intertrochanteric), and S72.2 (subtrochanteric—5 cm below trochanter minor). The date of admission for the first hip fracture registered was the index date.

Control subjects were randomly selected from the demographic section of OPED among county residents, frequency matched to cases by age and sex. Four controls were selected for each case. Control subjects were assigned a random date within the study period as their index date. Subjects who were not registered as county residents on the index date and during the preceding 6 months were excluded. Cases were eligible as control subjects before their case-defining event. Our control sampling followed the incidence density sampling principle (Rothman, 2002) that provides odds ratios (ORs) as unbiased estimates of the incidence rate ratios in the source population. The control: case ratio was lower among the oldest, as we had to exclude some subjects, when updated migration and death data became available (Table 1).

Table 1.   Characteristics of the study population by outcome (hip fracture)
 (n = 7,557)(n = 27,575)
Characteristicn (%)n (%)
Age, years, mean (SD)79.3 (11.2)77.5 (11.3)
 0–49152 (2.0)608 (2.2)
 50–59328 (4.3)1,307 (4.7)
 60–69692 (9.2)2,745 (10)
 70–792,000 (26.5)7,861 (28.5)
 80–893,272 (43.3)12,343 (44.8)
 ≥ 901,113 (14.7)2,711 (9.8)
Female gender5,427 (71.8)19,346 (70.2)
Antiepileptic drugs (AEDs)
 No use6,973 (92.3)26,471 (96.0)
 Use of any AED584 (7.7)1,104 (4.0)
 Use of liver enzyme inducing AEDs507 (6.7)922 (3.3)
 Use of noninducing AEDs160 (2.1)293 (1.0)
 AED monotherapy429 (5.6)897 (3.2)
 AED sequential monotherapy81 (1.1)137 (0.5)
 AED polytherapy74 (1.0)70 (0.3)
Comorbid conditions
 Epilepsy229 (3.0)260 (0.9)
 Affective disorder562 (7.4)903 (3.3)
 COPD591 (7.8)1,133 (4.1)
 Coronary HD522 (6.9)1,697 (6.2)
 Dementia108 (1.4)148 (0.5)
 Diabetes629 (8.3)1,407 (5.1)
 Hyperthyreoidism249 (3.3)589 (2.1)
 Obesity112 (1.5)436 (1.6)
 Osteoarthrosis730 (9.7)3,143 (11.4)
 Osteoporosis605 (8.0)865 (3.1)
 Parkinson's disease236 (3.1)220 (0.8)
 Polyneuropathy114 (1.5)185 (0.7)
 Previous nonosteoporotic fracture1,428 (18.9)3,069 (11.1)
 Previous osteoporotic fracture risk2,062 (27.3)3,492 (12.7)
 Stroke256 (3.4)569 (2.1)
Comedication (ever use of)
 Antiparkinsonian drugs200 (2.6)262 (1.0)
 Antipsychotics1,682 (22.3)3,440 (12.5)
 Bisphosphonates322 (4.3)627 (2.3)
 Corticosteroids2,356 (31.2)7,858 (28.5)
 Estrogens/SERMS1,383 (18.3)5,505 (20.0)
 Opioids3,209 (42.5)8,254 (29.9)
 Sedatives/anxiolytics1,070 (14.2)2,493 (9.0)
 SSRI2,102 (27.8)4,005 (14.5)
 Statins184 (2.4)959 (3.5)
 Tricyclic antidepressants1,037 (13.7)2,323 (8.4)

Exposure definition

Exposure status was established from OPED information. Information on exposure was thus available for a period longer than the study period, i.e., since 1992, and partly for the period 1990–1992 as well. Subjects were considered exposed to AEDs if they had presented at least one AED prescription at any time before the index date. We defined AEDs as substances with ATC-code N03A or N05BA09 and used the 2004 version of DDD (WHO Collaborating Centre for Drug Statistics Methodology, 2004).

We considered subjects presenting only one AED prescription as exposed in order not to overlook the potential risk associated with acute toxicity. We defined three treatment patterns: polytherapy as concomitant use of two or more specific AEDs for at least 3 months during the study period; sequential monotherapy as use of two or more specific AEDs with an overlap of less than 3 months and monotherapy as use of only one specific AED during the entire study period. The short treatment overlap in the case of sequential monotherapy was allowed in order to adjust for variations in daily dose of specific AEDs. To study the temporal relationship between exposure and outcome, and before performing the analysis, we classified subjects into three groups depending on the interval between the date of the latest prescription before the index date and the index date: current with an interval of up to 3 months, recent with an interval between 3 and 6 months, and previous when the interval was longer than 6 months. The duration of treatment with AEDs was defined as the period between the dates of the first and the last prescription. The cumulative AED dose was the total of DDD dispensed. The daily dose was calculated by dividing the cumulative dose by the treatment duration after subtraction of the dose dispensed with the last prescription. For users presenting a single prescription, the treatment duration was set to the number of days equaling the number of DDD dispensed, and the dose was accordingly defined as one DDD daily. With regard to induction of liver enzymes we classified AEDs into two groups, enzyme inducing (EI) (ethotoin, carbamazepine, oxcarbazepine, phenobarbital, phenytoin, primidone, and topiramate) and not enzyme inducing (NEI) (clobazam, clonazepam, ethosuximide, gabapentin, lamotrigine, levetiracetam, valproic acid, and vigabatrine) AEDs (Anderson, 2004). Concerning use of other drugs, the same parameters as described for AEDs with regard to duration of treatment, dose, and temporal relationship to outcome were estimated and incorporated in the dataset.


All analyses were adjusted for the effect of age and gender. Based on knowledge about risk factors for osteoporosis and fractures, we considered a number of additional potential confounders. Information on coronary heart disease, chronic obstructive pulmonary disease (COPD), convulsions, dementia, diabetes, epilepsy, hyperthyroidism, affective disorder, stroke, osteoporosis, previous osteoporotic (proximal humerus, vertebral, wrist, femoral neck) fractures, and nonosteoporotic fractures (all other  localizations) was retrieved from FPAS for the period 1980–2004. Information on use of antidepressants, antipsychotics, antiparkinsonian drugs, beta-agonists, bisphosphonates, corticosteroids, nonsteroidal antiinflammatory drugs (NSAIDs), opioids and sedatives/anxiolytics was retrieved from OPED. Known protective factors as obesity, osteoarthrosis, use of beta-blockers, estrogens, statins, and thiazides were also included in the analysis. To reduce the number of independent variables and define covariates more precisely according to published evidence, whenever possible, composite variables were used in the analysis, e.g., use of thiazides was included in the analysis only in users above 55 years of age, when treatment duration exceeded 1 year and treatment was current or discontinued within 3 months before the index date (Schoofs et al., 2003).

Confounders in the final model were selected by bivariate regression analysis. Potential confounders which changed the OR by more than 5%, when added to a model that contained only age and gender, were all included in the final model. Such confounders were current use of opioids, antiparkinsonian drugs, serotonine reuptake inhibitors (SSRIs), antipsychotics, ever use of tricyclic antidepressants and sedatives/anxiolytics, diagnosis of affective disorder, previous osteoporotic, and nonosteoporotic fracture. On theoretical grounds, we additionally included use of steroids, bisphosphonates, thiazides, statins, estrogens, and diagnoses of obesity and osteoarthrosis. Covariates included in the model are listed in Table 2.

Table 2.   Association between AED use and risk of hip fracture, adjusted for single
confounding factors
  1. List of covariates included in the adjustment model (OR from regression analysis including ever use of AEDs, age, gender, and the specific covariate).

Ever use of AEDs (adjusted only for2.03 (1.83–2.25)
 age and gender) 
Affective disorder1.91 (1.72–2.13)
Current use of antiparkinsonian drugs1.85 (1.65–2.07)
Current use of antipsychotics1.91 (1.72–2.12)
Current use of bisphosphonates2.02 (1.82–2.25)
Current use of estrogens/SERMS2.04 (1.83–2.26)
Current use of opioids1.81 (1.63–2.01)
Current use of SSRIs1.89 (1.70–2.10)
Current use of statins2.03 (1.83–2.26)
Ever use of sedatives/anxiolytics1.92 (1.73–2.14)
Ever use of TCAs1.84 (1.66–2.05)
Obesity2.03 (1.83–2.25)
Osteoarthrosis2.05 (1.85–2.28)
Previous nonosteoporotic fracture1.92 (1.73–2.13)
Previous osteoporotic fracture1.88 (1.70–2.10)
Use of steroids2.00 (1.80–2.22)
Use of thiazides2.03 (1.83–2.25)

Data analysis

We performed unconditional logistic regression to calculate ORs and 95% confidence intervals (CI) of the risk of hip fracture. Adjustment for the effect of age and gender was performed as part of the standard analysis and additional adjustment was done for the selected covariates. Never use of AEDs was standard reference group in all  regression analyses. Analyses were performed for AED use in general and the subgroups of inducing and noninducing AEDs. Analyses were also performed for specific AEDs, which were used by more than five cases. Separate analyses were performed to estimate the effect of current, recent, and previous AED use, the effect of AED use in mono- and polytherapy, and the effect of treatment duration, cumulative and daily dose. We treated epilepsy diagnosis as an effect modifier, rather than as a confounder and performed separate analyses for the stratum of subjects with, and respectively without, an epilepsy diagnosis. We did not adjust for diagnosis of epilepsy in the other analyses. Unless otherwise stated, reported ORs were adjusted for all covariates included in the final model for regression analysis. All analyses were repeated on a dataset, where data on AED use in subjects presenting a single AED prescription were disregarded. Data analysis was performed using Intercooled Stata 9.2 for Windows. Born 20 July 2007 (Stata Corp., College Station, TX, U.S.A.) Relevant permissions were obtained from the Danish Data Protection Agency and the registers involved. Approval from an Ethics Committee was not required.


We identified 7,557 cases and 27,575 controls. The characteristics of the study population are shown in Table 1. Ever use of AEDs was nearly twice as frequent among cases (7.7%) than among controls (4%). Epilepsy and other comorbid conditions or comedication with drugs that could affect fracture risk were overrepresented among cases, although use of steroids was comparable between the two groups.

As displayed in Table 3, the risk for hip fracture was increased with ever use of any AED (OR: 1.31; 95% CI: 1.16–1.48). Use of EI and NEI AEDs, not excluding concomitant use of the opposite group, showed a comparable risk increase with both subgroups, somewhat higher in the EI, than in the NEI group. Ever use of only EI AEDs was also associated with fracture risk increase, whereas ever use of only NEI AEDs (OR: 1.03; 95% CI: 0.77–1.37) was not. The temporal pattern of use significantly modified risk, which was highest with current use, lower with recent use, and not increased with previous use. This was underscored by a risk increase with current use of only NEI AEDs as well.

Table 3.   Association between temporal profile of AED use and risk of hip fracture, reference:
never AED use
 Numbers exposed (%)a 
AED useCasesControlsOR (95 % CI)Adj. OR (95% CI)
  1. aPresented as % of all cases, respectively controls.

  2. OR, odds ratios adjusted for age and gender; Adj. OR, odds ratios adjusted for age, gender, current use of antiparkinsonian drugs, antipsychotics, bisphosphonates, estrogens/SERMS, opioids, SSRIs, statins, ever use of sedatives/anxiolytics and TCAs, use of steroids and thiazides, diagnoses of affective disorder, obesity, osteoarthrosis, previous osteoporotic, and nonosteoporotic fracture; current, interval from date of latest prescription to index date up to 3 months; recent, interval between 3 and 6 months; previous, interval longer than 6 months; EI/NEI AEDs, AEDs inducing/not inducing liver enzymes; only EI/NEI AEDs, analysis restricted to subjects using only AEDs inducing/not inducing liver enzymes.

 Ever use584 (7.7)1104 (4.0)2.03 (1.83–2.25)1.31 (1.16–1.48)
 Current use303 (4.0)389 (1.4)3.06 (2.62–3.56)1.92 (1.58–2.33)
 Recent use49 (0.7)82 (0.3)2.30 (1.61–3.28)1.84 (1.26–2.68)
 Previous use232 (3.0)633 (2.3)1.38 (1.19–1.61)0.95 (0.80–1.12)
 Ever use507 (6.7)922 (3.3)2.11 (1.88–2.36)1.36 (1.20–1.55)
 Current use244 (3.2)291 (1.1)3.30 (2.78–3.93)2.10 (1.66–2.65)
 Recent use44 (0.6)65 (0.2)2.59 (1.76–3.80)2.12 (1.41–3.17)
 Previous use219 (2.9)566 (2.1)1.46 (1.25–1.71)1.03 (0.86–1.22)
 Ever use160 (2.12)293 (1.06)2.14 (1.76–2.60)1.27 (1.02–1.58)
 Current use88 (1.2)107 (0.4)3.25 (2.45–4.32)1.99 (1.45–2.73)
 Recent use10 (0.1)18 (0.1)2.25 (1.04–4.88)1.52 (0.67–3.44)
 Previous use62 (0.8)168 (0.6)1.43 (1.06–1.91)0.81 (0.58–1.12)
Only EI AEDs
 Ever use424 (5.6)811 (2.9)1.99 (1.77–2.25)1.31 (1.14–1.51)
 Current use96 (1.3)94 (0.3)2.15 (1.61–2.87)1.32 (0.90–1.94)
 Recent use22 (0.3)23 (0.1)1.96 (1.09–3.53)1.89 (1.02–3.51)
 Previous use138 (1.8)277 (1.0)1.05 (0.85–1.30)0.93 (0.74–1.16)
 Ever use77 (1.0)182 (0.7)1.62 (1.24–2.12)1.03 (0.77–1.37)
 Current use53 (0.7)59 (0.2)1.95 (1.34–2.84)1.62 (1.08–2.43)
 Recent use4 (0.1)13 (0.1)0.63 (0.20–1.93)0.56 (0.18–1.78)
 Previous use41 (0.5)95 (0.3)0.93 (0.64–1.34)0.79 (0.53–1.17)

Fracture risk was increased with increasing daily dose (OR: 1.5; 95% CI: 1.24–1.82), the OR denoting risk change by one DDD per day. The risk also increased, but to a minor extent, with increasing cumulative dose (OR: 1.05; 95% CI: 1.03–1.08), the OR denoting risk change by 365 DDD. Treatment duration, estimated in calendar years, did not increase fracture risk (OR: 1.02; 95% CI: 0.99–1.05). Risk estimates were similar with use of EI AEDs, whereas use of NEI AEDs did not increase the risk. Results from this analysis in relation to current, recent and previous use are shown in Table 4.

Table 4.   Association between AED use and risk of hip fracture in relation to dose and treatment duration, reference: never AED use
AED useAdjusted OR (95 % CI)
Daily dose (DDD)Cumulative doseTreatment duration (years)
  1. Daily dose, average daily dose calculated by dividing the cumulative dose by treatment duration, after subtraction of the dose dispensed with the last prescription; cumulative dose, total of defined daily doses dispensed, estimated as annual load = total dispensed quantity in DDD/365.24; treatment duration, period between the dates of first and last prescription, estimated in calendar years; adjudsted OR, odds ratios adjusted for age, gender, current use of antiparkinsonian drugs, antipsychotics, bisphosphonates, estrogens/SERMS, opioids, SSRIs, statins, ever use of sedatives/anxiolytics and TCAs, use of steroids and thiazides, diagnoses of affective disorder, obesity, osteoarthrosis, previous osteoporotic, and nonosteoporotic fracture.

 Ever use1.50 (1.24–1.82)1.05 (1.03–1.08)1.02 (0.99–1.05)
 Current use1.77 (1.32–2.36)1.03 (1.00–1.05)0.99 (0.95–1.03)
 Recent use1.94 (0.89–4.23)1.11 (1.00–1.23)0.98 (0.89–1.08)
 Previous use1.42 (1.08–1.86)1.05 (0.95–1.15)0.97 (0.90–1.04)
 Ever use1.51 (1.22–1.87)1.06 (1.03–1.09)1.03 (1.00–1.07)
 Current use1.61 (1.14–2.27)1.03 (0.99–1.06)1.00 (0.96–1.05)
 Recent use2.33 (0.96–5.69)1.12 (1.00–1.25)0.97 (0.88–1.08)
 Previous use1.47 (1.10–1.97)1.06 (0.97–1.15)0.97 (0.90–1.04)
 Ever use1.03 (0.68–1.55)1.06 (0.98–1.14)1.00 (0.93–1.08)
 Current use1.67 (0.81–3.43)1.03 (0.96–1.11)0.97 (0.88–1.06)
 Recent use0.20 (0.02–2.32)0.38 (0.05–3.10)0.85 (0.65–1.12)
 Previous use0.83 (0.43–1.62)0.86 (0.54–1.36)0.88 (0.67–1.14)
Only EI AEDs
 Ever use1.48 (1.17–1.87)1.05 (1.02–1.08)1.03 (0.99–1.06)
 Current use1.54 (1.06–2.24)1.02 (0.98–1.06)0.99 (0.94–1.04)
 Recent use1.85 (0.77–4.45)1.14 (1.01–1.29)1.01 (0.90–1.14)
 Previous use1.58 (1.13–2.22)1.05 (0.96–1.16)0.97 (0.90–1.05)
 Ever use1.03 (0.50–2.12)0.80 (0.48–1.33)0.95 (0.81–1.11)
 Current use1.96 (0.57–6.75)0.60 (0.32–1.12)0.88 (0.72–1.07)
 Recent use 3.03 (0.14–64.72)0.04 (0.00–3.56)0.59 (0.24–1.44)
 Previous use1.15 (0.41–3.28)0.73 (0.15–3.49)0.91 (0.59–1.39)

In analysis of AED use in strata defined by the presence, respectively absence of, an epilepsy diagnosis, we did not find a significant risk increase in ever use, neither with a concomitant epilepsy diagnosis (OR: 1.29; 95% CI: 0.77–2.18) nor without (OR: 1.08; 95% CI: 0.94–1.24). Fracture risk was however increased with recent use (OR: 3.02; 95% CI: 1.10–8.30) in subjects with an epilepsy diagnosis, and with current use (OR: 1.70; 95% CI: 1.33–2.17) in subjects without an epilepsy diagnosis (Table 5).

Table 5.   Association between AED use, epilepsy and risk of hip fracture, reference: never AED use
AED useNumbers exposed (%)a  
CasesControlsOR (95 % CI)Adjusted OR (95% CI)
  1. aPresented as % of all cases, respectively controls.

  2. OR, odds ratios adjusted for age and gender; adjusted OR, odds ratios adjusted for age, gender, current use of antiparkinsonian drugs, antipsychotics, bisphosphonates, estrogens/SERMS, opioids, SSRIs, statins, ever use of sedatives/anxiolytics and TCAs, use of steroids and thiazides, diagnoses of affective disorder, obesity, osteoarthrosis, previous osteoporotic, and nonosteoporotic fracture.

 Subjects with an epilepsy diagnosis
Never use 40 (0.5) 60 (0.2)1 Reference1 Reference
Ever use189 (2.5)200 (0,7)1.25 (0.77–2.03)1.29 (0.77–2.18)
Current use140 (1.9)148 (0.5)1.23 (0.74–2.04)1.14 (0.64–2.01)
Recent use 20 (0.3) 17 (0.1)1.89 (0.80–4.48)3.02 (1.10–8.30)
Previous use 29 (0.4) 35 (0.1)1.15 (0.57–2.31)1.36 (0.62–2.95)
 Subjects without an epilepsy diagnosis
Never use6,933 (91.7)26,411(95.8) 1 Reference1 Reference
Ever use395 (5.2)904 (3.3)1.66 (1.47–1.88)1.08 (0.94–1.24)
Current use163 (2.2)241 (0.9)2.63 (2.15–3.22)1.70 (1.33–2.17)
Recent use 29 (0.4) 65 (0.2)1.72 (1.11–2.66)1.31 (0.82–2.08)
Previous use203 (2.7)598 (2.2)1.28 (1.09–1.50)0.85 (0.72–1.02)

Ever use of AEDs was associated with lower risk in monotherapy (OR: 1.21; 95% CI: 1.06–1.39) and in sequential monotherapy (OR: 1.39; 95% CI: 1.01–1.91), than in polytherapy (OR: 2.54; 95% CI: 1.66–3.90). We identified use of up to seven different AEDs by a single user in the study population. Bivariate analysis of AED ever use and number of AEDs used, with adjustment for age and gender, showed a substantial risk increase for each additional AED (OR: 1. 64; 95% CI: 1.53–1.77).

In monotherapy with specific AEDs, only current (OR: 2.26; 95% CI: 1.58–3.24) and recent (OR: 3.04; 95% CI: 1.49–6.19) use of phenobarbital, as well as current use of oxcarbazepine (OR: 2.70; 95% CI: 1.57–4.64), was associated with a statistically significant risk increase. Dose and treatment duration analysis revealed only an association to the daily dose with current use (OR: 3. 62; 95% CI: 1.03–12.65) of carbamazepine, but no effect with use of the other specific AEDs (data not presented). Because of the small number of users, we did not perform analysis in strata defined by an epilepsy diagnosis. In polytherapy with specific AEDs, there was no increase in fracture risk with the use of clonazepam and gabapentin. Fracture risk was increased with current use of all other AEDs, recent use of carbamazepine and phenobarbital, and previous use of phenobarbital (data not presented).

Analysis in a dataset where data on AED use for single AED prescription users were omitted showed almost identical results. The trends reported above regarding liver enzyme inducing AEDs, current use, and dose association were reproduced without changes (results not presented).


Our results suggest that AED treatment, after adjustment for confounders, is associated with a modest increase in risk for hip fractures. The risk increase was associated with use of liver enzyme inducing AEDs, daily dose, and to a lesser extent cumulative dose, but not with treatment duration. Polytherapy was associated with a higher risk than monotherapy. Fracture risk was significantly modified by stratification on diagnosis of epilepsy. In monotherapy with specific AEDs, current and recent use of phenobarbital, and current use of oxcarbazepine increased fracture risk.

A number of limitations in our study should be mentioned. Information on a number of known risk factors, e.g., smoking, was not available. The use of proxy markers as a COPD diagnosis and use of inhaled beta-agonists can partially compensate for the lack of precise information on this specific risk factor. The distinction between current and recent use is arbitrary, and thus our results must be interpreted with caution.

A number of studies have explored the association of fracture risk and AED use. In a recent case-control study from an epilepsy cohort (Souverein et al., 2006), risk was not associated with use of liver enzyme inducing AEDs, but with current AED use, polytherapy, and duration of treatment. Current monotherapy with carbamazepine, valproate, phenobarbital, and phenytoin were also associated with an increased risk.

A previous Danish case–control study (Vestergaard et al., 2004) showed higher fracture risk with use of EI AEDs, increasing cumulative dose, number of AEDs used, and epilepsy. Current use did not affect fracture risk. Use of carbamazepine, oxcarbazepine, clonazepam, phenobarbital, phenytoin, and valproate were significantly associated with risk of any fracture.

Disparities between our results and those of the two studies may be accounted for through differences in methodology and the studied populations. The fact that a majority of the study population were elderly, and thus more prone to CNS related adverse effects of AEDs, may explain why the effect of current AED use was more pronounced in our study. We calculated daily dose from the cumulative dose and treatment duration. Thus it can be said to express the combined effect of both. Incidence estimates for hip fractures are higher in national register data than in the individually validated hip fracture register we used, probably because of misclassification of recurrent cases (Nymark et al., 2006a).

Despite solid evidence of an AED effect on bone metabolism, the causes of this association have yet to be clarified. Regarding the significance of hepatic enzyme induction, bone changes shown in the earliest studies (Kruse, 1968; Richens & Rowe, 1970) developed first after prolonged treatment with multiple AEDs and responded favorably to treatment with modest doses of vitamin D despite continued AED treatment.

Considering the direct AED effect on bone tissue, inhibition of osteoblast proliferation by phenytoin and carbamazepine in therapeutic doses has been reported (Feldkamp et al., 2000). However, lower phenytoin doses have been shown to promote osteoblast proliferation (Koyama et al., 2000). A net increase of bone resorption was found under valproate treatment (Sato et al., 2001), but recently valproate was shown to enhance osteoblast viability and differentiation in low concentrations (Schroeder & Westendorf, 2005). Despite evidence on BMD decrease under AED treatment, some studies failed to show such changes for carbamazepine (Sheth et al., 1995; Erbayat et al., 2000) and for valproic acid (Erbayat et al., 2000). Although the link between reduced bone density and fracture risk is well established (Chandler et al., 2000; Cauley et al., 2005), attention has been drawn to a possible overestimation of fracture risk depending on the method of estimation (Nordin et al., 2007).

A notable result in our study was that the current, as opposed to previous, AED use, carried the highest fracture risk. Absence of risk associated with past AED use and to treatment duration do not favor a long standing osteoporotic effect of AEDs as the explanation for the increased fracture risk.

Stratifying on epilepsy diagnosis modified the risk estimate for ever use, with a lower risk for subjects without epilepsy. In addition, we found a significantly higher risk only with recent use in the epilepsy stratum (OR: 3.02; 95% CI: 1.10–8.30), contrary to a significant risk increase only with current use in the nonepilepsy stratum (OR: 1.70; 95% CI: 1.33–2.17). Despite the weakness of the association, due to a limited number of subjects, one can speculate if higher risk with recent use can partly be attributed to a higher risk of seizure relapse, because of recent treatment discontinuation.

Our study confirms most findings from previous studies, particularly stressing the importance of current use. It also emphasizes the potential for modification of effect by the indication for AED use, an issue of increasing significance as the utilization spectrum of AEDs broadens.


The study was supported by a grant from Novartis Pharma AG, Switzerland, to the University of Southern Denmark.

Conflict of interest: We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. Dr. Tsiropoulos has received teaching fees and honoraria for participation in an advisory board from Janssen-Cilag Denmark A/S and teaching fees from UCB Pharma Nordic. Drs. Nymark, Lauritsen, and Gaist have nothing to disclose. Dr. Andersen has participated in studies receiving funding from AstraZeneca, Nycomed, and Lundbeck and has received teaching fees from the Danish Association of the Pharmaceutical Industry. Dr. Hallas has received study grants from Novartis and Nycomed and has received teaching fees from AstraZeneca and from the Danish Association of the Pharmaceutical Industry.