Risk of hip fractures associated with benzodiazepines: Applying common protocol to a multi-database nested case-control study. The protect project

Background: The association between benzodiazepines (BZD) and hip fractures has been estimated in several observational studies although diverse methodologies and definitions have hampered comparability. Objectives: To evaluate the discrepancies in the risk estimates of hip/femur fractures associated with BDZs across different databases and to assess the impact of different matching strategies. Methods: A case control study nested in a cohort of BZD users, examining their association with the risk of hip/ femur fracture between 2001 and 2009, was performed within 3 databases, the BIFAP (Spain), the CPRD (UK) and the Mondriaan (Netherlands) database. A risk set sampling matching was performed using two strategies: 1) controls matched by age (up to ± 2 years), sex and time in the cohort (up to ± 6 months) and 2) controls selected with the smallest Manhattan distance according same matching factors. Co-morbidity and co-medication adjusted OR and (95% confidence intervals) were estimated for current use (up to 30 days after last supply) vs. past (>60 days after current use) using conditional logistic regression models. Sensitivity analysis was performed in CPRD including matching by general practice (GP). Results: Adjusted ORs (matching 1) for current use were 1.14 (1.03-1.27) in BIFAP; 1.32 (1.22-1.42) in CPRD, and 1.34 (0.63-2.82) in Mondriaan. Matching 2 resulted in ORs of 1.09 (1.03-1.27), 1.29 (1.17-1.42) and 1.28 (0.60-2.71) in BIFAP, CPRD and Mondriaan respectively. In CPRD, adding GP-practice as a matching factor to matching strategy 1 increased the OR to 1.46 (1.35-1.59). Conclusions: By applying a common protocol, the estimated risk of hip/femur fractures associated to BZD was consistent between studies. The different matching strategies did not influence the risk estimates substantially, however the inclusion of GP-practice as matching factor should be carefully considered in further studies.