Antidepressantuse and the risk of hip fracture: A self-controlled case series approach in two primary care databases

Background: The use of antidepressants(AD,selective serotonin reuptake inhibitors,SSRI, or tricyclic antidepressants,TCA) has been associated with hip fractures(HF) in observational studies.However, it has been suggested that such results may be confounded by unmeasured patient characteristics. Objectives: To assess the effect of AD use on the risk of HF using self-controlled case series design. Methods: A self-controlled case series study was conducted in two primary care databases among patients with a first HF and a prescription for AD at any time during observation.Data were extracted from the UK THIN and the Dutch Mondriaan GP databases of the period 2001-2009.The incidence rate ratio(IRR) of HF for periods of AD use versus no AD use was estimated using conditional Poisson regression. Results: There were 6,632 and 136 HF patients for analysis in THIN and Mondriaan, respectively.After adjustment for age, an increased risk of HF was observed during the 30 day period after AD initiation: IRR 1.57 (95% CI, 1.39-1.78) in THIN and 3.22 (1.51-6.84) in Mondriaan.The increased risk was also observed during the next six months of AD use: IRR 1.52(1.39-1.65) in THIN and 2.76 (1.69-4.50) in Mondriaan.In the period after six months, the risk remained higher in THIN (IRR: 1.47; 1.31-1.65) but not significant in Mondriaan (IRR 1.94; 0.84-4.47). Furthermore, an increased risk of HF was observed during the 30 day period prior to AD initiation (IRR 1.22; 1.06-1.41) in THIN and (IRR 2.51; 1.00-6.33) in Mondriaan.In both THIN and Mondriaan, when cases were censored at the event times (HF) there appeared to be a substantial bias in IRR in all the periods: IRR during the 30 day period after AD initiation: 3.07 (2.66-3.54) in THIN and 12.34 (2.13-71.54) in Mondriaan.The risk of HF was higher in SSRI users than TCA users (4.39; 2.42-7.99 vs. 1.39; 0.58-2.41) in Mondriaan but was similar in THIN (1.49; 1.36- 1.63 vs. 1.49; 1.33-1.66). Conclusions: The incidence rate of hip fractures was higher in the periods both immediately after the start of AD use and during the first six months of AD use compared to periods of no use, although the magnitude of the risk varied between databases.