Responsiveness of endpoints in osteoporosis clinical trials--an update.

As an update of our earlier paper, published as part of the Outcome Measures in Rheumatology Clinical Trials (OMERACT 3) proceedings in 1996, we surveyed the types of outcomes incorporated in recent clinical trials. A literature search was conducted on MEDLINE and Current Contents, from January 1996 to March 1998, using the search strategy recommended by the Cochrane Collaboration for the identification of randomized controlled trials (RCT). Two independent reviewers selected trials according to inclusion criteria. The same reviewers extracted data on clinical and radiographic fractures, pain, quality of life, and bone mineral density (BMD). Seventy-four RCT conducted on bone loss in postmenopausal women were identified. Most trials incorporated biochemical markers and BMD as outcome measures. Fewer trials included vertebral fractures, pain, height, and quality of life. The responsiveness is presented in terms of the sample size needed per group to show a statistically significant difference. The most responsive outcomes were pain, BMD, and biochemical markers. The number needed to treat to prevent one vertebral fracture ranged from 13 to 54, depending on the intervention and population. Investigators should examine the characteristics of the patient population and the nature of the intervention in determining the sample size required to demonstrate a significant effect. The selection of endpoints should be based on their responsiveness, feasibility, and the importance of using standardized outcomes. Standardized outcomes greatly facilitate the synthesis of available information into systematic reviews by groups such as the Cochrane Collaboration.