Bisphosphonate use and improved implant survival: A nationwide cohort study

Background: Osteolysis and aseptic loosening are the most common causes of revision arthroplasty worldwide. Bisphosphonates might improve implant survival through their anti-osteoclast effects. We aimed to study the association between bisphosphonate use and implant survival. Methods: A retrospective cohort study was conducted within the Danish nationwide registries (5.5 million residents). We identified patients aged>=40 years undergoing total joint replacement (TJR) during the study period (1998-2007) using ICD10 codes. Patients with inflammatory arthritides, Paget disease of bone, hip fracture and use of DMARDs were excluded. Each participant was followed up until end of study, date of emigration, revision surgery, or patient's death, whichever came first. Participants were classified as bisphosphonate users (BPU) if they had been on treatment for at least 6 months. A time-varying exposure was used to avoid immortal-time bias. Up to six BP non-users (BPNU) undergoing arthroplasty were matched to each BPU using propensity scores. Stratified Cox regression was used to model implant survival according to bisphosphonate use. Further, we studied the association between duration of use, adherence (medication possession ratio=MPR), and timing of therapy initiation (pre-op vs post-op) and implant survival. Results: 80,342/95,392 (84.2%) subjects were eligible. We identified 1,950 (2.4%) BPU, and 1,911 (98.0%) of them were matched to 10,755 BPNU. In total, 226/12,666 (1.78%) of the participants (22/1,911 BPU and 204/10,755 matched BPNU) underwent revision surgery during study follow-up (median 1.11 years, inter-quartile range 0.43-2.29). Cox regression models showed reduced revision risk in BPU (HR 0.59, 95% CI 0.37, 0.94). This protective effect was highest in patients with longest duration of treatment and highest adherence (Table 1). Conclusions: BPU are at 40% reduced risk of revision compared with matched BPNU. These results are similar to previous findings using similar retrospective data from the UK [Prieto-Alhambra D et al. BMJ 2011]. Confirmation of causality in randomized controlled experiments is required. (Table presented).