Osteoporosis Medication Use and Fracture Risk in the Women’s Health Initiative
Evidence is inconclusive about whether long-term (>3-5 years) bisphosphonate therapy reduces or increases fracture risk. This dissertation examined the association of long-term bisphosphonate use with fracture, using short-term use as the referent group, among older women and postmenopausal women with breast cancer. Additionally, this dissertation, which used self-reported medication data, included a validity study of self-reported medication use.
Participants were Women’s Health Initiative participants who completed a mailed medication inventory. Fracture outcomes were ascertained by self-report on an annual study form; a subset was confirmed with medical records. The analyses examined hip, wrist/forearm, clinical vertebral, and any clinical fracture among older women (n=5,137) and any clinical fracture among women with breast cancer (n=902). The association between bisphosphonate use and fracture was estimated using multivariate Cox proportional hazards models that compared 3-5, 6-9, and 10 or more years of use with 2 years among older women and compared 4-7 and 8 or more years of use with 2-3 years among women with breast cancer. Self-reported medication use was compared with pharmacy records for four chronically-used classes of medications (statins, calcium channel blockers, beta blockers, and bisphosphonates) among 223 participants using sensitivity, specificity, and positive predictive value for current medication use and kappa statistic for duration of use (<2, 2, 3, 4, >5 years).
Among older women, 10 or more years of bisphosphonate use was associated with increased risk of any clinical fracture compared with 2 years of use (HR: 1.29 [95% CI: 1.06-1.56]). Among women with breast cancer, 8 or more years of bisphosphonate use was associated with increased risk of fracture compared with 2-3 years of use (HR: 1.65 [95% CI: 1.01-2.58]). Compared with pharmacy records, sensitivity, specificity, positive predictive value, and kappa statistic were near perfect for all medication classes, except for bisphosphonates, which had sensitivity and positive predictive value of 80%.
The longest duration of bisphosphonate use was associated with increased risk of any clinical fracture compared with short-term use among older women and women with breast cancer. The medication inventory was a highly accurate source of self-reported current medication use for chronically-used medications.