Aims: In this study we aimed to investigate how glucocorticoids (GCs) affect fracture risk in patients with rheumatoid arthritis (RA) regarding the strength of dose, duration of use and recency of exposure, using new methods of analysing pharmacoepidemiological data. Methods: We used data from the clinical practice research datalink (CPRD) an electronic medical records dataset from the United Kingdom of over 10million patients. Patients were identified as suitable for inclusion in the study if they were adults, had a diagnosis of RA and had at least one year âup to standardâ follow up. Patients were classified as exposed or unexposed to GCs upon entry and exposed patients were matched to up to two unexposed patients. Cox proportional hazard models were used to analyse common risk attribution models used in previous literature, for example categorical indicators of dose and recency of use. The weighted cumulative exposure (WCE) was also used to incorporate the effects of strength of dose, duration of use and recency of exposure in one model. The weight function generated by the WCE model, reflecting the importance of recency of treatment, was applied to different clinically plausible exposure patterns to assess the risk of fracture. Results: The study sample comprised 16,507 patients of which 8,357 were exposed to oral GCs. Average follow up was four years with 70% of the population female. The GC exposed patients were older, had higher body mass index and more comorbidities. Risk of fracture increased with current daily dose such that the hazard ratio (HR) was 1.44 (1.17, 1.77) for 5 to 9.9mg prednisolone equivalent (PEQ) and 3.02 (1.77, 5.15) for 15 to 19.9mg PEQ. In addition, risk of fracture increased with cumulative dose, from a HR of 1.22 (1.03, 1.44) (0.1 to 0.9g) to 1.83 (1.35, 2.48) (7.5 to 10g). Cumulative dose is a combination of current dose and duration of use; hence it was unclear about the exact effects of dose and duration. Unexpectedly the risk of fracture increased during the first month after discontinuation of GCs HR 1.66 (1.27, 2.16) from 1.43 (1.21, 1.68) for current users before decreasing to non-significant levels for later time windows post discontinuation. The weighted cumulative exposure model showed a positive weight for the first year after GC treatment, suggesting doses taken in the past year increase the risk of fracture: the weight function was highest for doses taken most recently with the magnitude of weight declining throughout the first year. Risk of fracture increased with duration of use: HR (95%CI) 1.09 (1.13, 1.05) to 1.25 (1.16, 1.35) to 1.42 (1.30, 1.55) for 5mg/day for 1, 3 and 6 months respectively. We demonstrated that the dosing route to a given cumulative dose matters. For example, the cumulative dose of 900mg had a HR of 1.42 (1.30, 1.55) from 5mg/day for 6 months compared to 1.70 (1.37, 2.09) for 30mg/day for 1 month. After GC discontinuation, risk of fracture returned to baseline between 6 months and 1 year. HR 1.06 (0.97, 1.15) (5mg for 6 months stopped 6 months ago) and 0.99 (0.92, 1.06) (30mg for 1 months stopped 1 year ago). Conclusions: Using novel methodology to explore the issue between GC use and fracture risk we have demonstrated that the risk of fracture appears to be affected by the strength of dose, duration of use and recency of exposure to oral GCs.