Wednesday, 6 April 2005

Poster Abstracts: Theoretical / Patient Care

Design Considerations for Head-to-Head Trials Comparing Osteoporosis Treatments

Paul D. Miller, MD and David Hosking, M.D.

Properly conducted clinical trials provide the highest level of evidence of treatment efficacy, and head-to-head trials are the ideal way to compare efficacy and tolerability of agents. Although fracture endpoints are the ideal measure of efficacy, the need for huge sample sizes and other complexities of trial design make it unlikely that they will be performed, and alternatives should be considered.

For head-to-head fracture trials, the sample size is influenced by the expected incidence of fractures and other factors, including the analysis objective (an equivalence or non-inferiority trial versus a superiority trial). Because both active treatments may reduce both the overall incidence of fracture and the difference between treatment groups, sample sizes will be considerably larger than in a placebo-controlled trial.

In a superiority trial, with a power of 80% and a fracture rate of 1.2% among participants in one group, 235,000 participants would be required to show a difference of 10% in efficacy; 18,080 would be required to show a difference of 30%. Even if the fracture rate in one group were 15%, 14,700 participants would be needed to demonstrate a difference. In an equivalence trial, sample sizes in the 30,000 to 40,000 range are required. Comparing fracture incidence in smaller studies is misleading and fundamentally meaningless because the power to detect a real difference is negligible. Non-significant differences are almost certainly due to chance. In this regard, meta-analysis may be valuable, since the procedure pools all available data to derive the most precise estimate of effect of an agent.

Head-to-head trials could examine endpoints that are well-validated surrogates for fracture events, such as BMD and bone turnover. These trials can be considerably smaller and more quickly completed than fracture trials, and a large number of studies have documented the strong, graded association between fracture incidence and these surrogates.

Disclosure Information:

Faculty Member's Name: Paul D. Miller, MD
Grants/Research Support: Merck, Proctor and Gamble


See more of Poster Abstracts: Theoretical / Patient Care
See more of The Sixth International Symposium on Osteoporosis: Current Status and Future Directions