Exploring methods for comparing the real-world effectiveness of treatments for osteoporosis: adjusted direct comparisons versus using patients as their own control

SUMMARY: Using Swedish and Dutch registry data for women initiating bisphosphonates, we evaluated two methods of comparing the real-world effectiveness of osteoporosis treatments that attempt to adjust for differences in patient baseline characteristics. Each method has advantages and disadvantages; both are potential complements to clinical trial analyses. PURPOSE: We evaluated methods of comparing the real-world effectiveness of osteoporosis treatments that attempt to adjust for both observed and unobserved confounding. METHODS: Swedish and Dutch registry data for women initiating zoledronate or oral bisphosphonates (OBPs; alendronate/risedronate) were used; the primary outcome was fracture. In adjusted direct comparisons (ADCs), regression and matching techniques were used to account for baseline differences in known risk factors for fracture (e.g., age, previous fracture, comorbidities). In an own-control analysis (OCA), for each treatment, fracture incidence in the first 90 days following treatment initiation (the baseline risk period) was compared with fracture incidence in the 1-year period starting 91 days after treatment initiation (the treatment exposure period). RESULTS: In total, 1196 and 149 women initiating zoledronate and 14,764 and 25,058 initiating OBPs were eligible in the Swedish and Dutch registries, respectively. Owing to the small Dutch zoledronate sample, only the Swedish data were used to compare fracture incidences between treatment groups. ADCs showed a numerically higher fracture incidence in the zoledronate than in the OBPs group (hazard ratio 1.09–1.21; not statistically significant, p > 0.05). For both treatment groups, OCA showed a higher fracture incidence in the baseline risk period than in the treatment exposure period, indicating a treatment effect. OCA showed a similar or greater effect in the zoledronate group compared with the OBPs group. CONCLUSIONS: ADC and OCA each possesses advantages and disadvantages. Combining both methods may provide an estimate of real-world treatment efficacy that could potentially complement clinical trial findings. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (10.1007/s11657-017-0375-7) contains supplementary material, which is available to authorized users.