Objectives: To investigate the association between anticonvulsant use and fracture risk among older patients, including those with bipolar disorder (BD), an indicated condition for treatment with this class of medications.
Design: A prospective cohort study with 4.5 years of follow-up analyzed using both Cox proportional hazards modeling and propensity score techniques.
Setting: National sample selected from Veterans Health Administration records.
Participants: A representative sample of 67,387 Veterans Administration patients aged 50 years and older selected in fiscal year 2002 and followed until 2006, including 29,029 with a diagnosis of BD identified from the Veterans Administration National Psychosis Registry. Pharmacy records identified 19,635 patients who had ever used anticonvulsant medications.
Measurements: Incident fracture at any site and incident hip fracture indicated by administrative data.
Results: There were 4,367 fractures over the 4.5-year study period. Approximately two-thirds of patients with BD were prescribed anticonvulsants, and diagnosis of BD was associated with 20% increased risk of fracture independent of anticonvulsant use. Cumulative incidence of fracture was higher among anticonvulsant users relative to never users (35.7 per 1,000 versus 14.2 per 1,000 person-years). In fully adjusted models, anticonvulsant use was associated with over twofold greater risk of fracture (hazard ratio: 2.42, 95% confidence interval: 2.23-2.633). Current use was associated with the greater risk of fracture relative to former use. Duration of anticonvulsant use was significantly associated with increased fracture risk in a graded, nonlinear manner.
Conclusions: Use of anticonvulsants is associated with increased risk of fracture among older patients with BD and among those without serious mental illness.