Osteoporosis therapies: evidence from health-care databases and observational population studies

Abstract

Osteoporosis is a well-recognized disease with severe consequences if left untreated. Randomized controlled trials are the most rigorous method for determining the efficacy and safety of therapies. Nevertheless, randomized controlled trials underrepresent the real-world patient population and are costly in both time and money. Modern technology has enabled researchers to use information gathered from large health-care or medical-claims databases to assess the practical utilization of available therapies in appropriate patients. Observational database studies lack randomization but, if carefully designed and successfully completed, can provide valuable information that complements results obtained from randomized controlled trials and extends our knowledge to real-world clinical patients. Randomized controlled trials comparing fracture outcomes among osteoporosis therapies are difficult to perform. In this regard, large observational database studies could be useful in identifying clinically important differences among therapeutic options. Database studies can also provide important information with regard to osteoporosis prevalence, health economics, and compliance and persistence with treatment. This article describes the strengths and limitations of both randomized controlled trials and observational database studies, discusses considerations for observational study design, and reviews a wealth of information generated by database studies in the field of osteoporosis.

Fig. 1

Nonvertebral fractures in pooled randomized controlled trials (RCTs) and observational database studies. This comparison of previously published data on risk reduction in nonvertebral fractures includes outcomes in RCTs and observational data for two bisphosphonates. Observational data are based on Watts et al. [48]. RCT data are based on Black et al. [49] for alendronate and Harrington et al. [50] for risedronate

Fig. 2

Cumulative nonvertebral fracture incidence at 12-months. REAL data are based on Silverman et al. [52]. REALITY data are based on Curtis et al. [60]. VIBE data are based on Harris et al. [61]