A Comparison of a Blinded Procedure to Estimate Treatment Difference and Standard Deviation with an Unblinded Procedure in Clinical Trials

• Activity Number: 121
• Type: Contributed
• Date/Time: Monday, August 9, 2004 : 12:00 PM to 1:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Farid Kianifard*+ and Michael Chen
• Companies: Novartis Pharmaceuticals and Novartis Pharmaceuticals
• Keywords: drug development ; folded normal distribution ; sample size re-estimation

Abstract

In drug development, one often uses the estimated treatment difference and standard deviation of a normally distributed response variable from previous clinical trials to determine the sample size for future trials. Chen and Kianifard introduced a procedure, based on properties of the folded normal distribution, to estimate these parameters without unblinding for an ongoing trial, thereby avoiding bias due to unblinding and potentially shortening drug development time. One may decide to abandon plans for future trials if estimated treatment difference is much smaller than the minimal clinically meaningful treatment difference, or choose to reduce the sample size for future trials if the difference is much larger than expected. In another scenario, most of the planned number of patients may have completed the trial, but patient enrollment is somewhat short of target. This approach can be used to estimate treatment difference and standard deviation for available patients, with the possibility of stopping enrollment. In a simulation study, we compare the power of this procedure to an unblinded procedure for sample size re-estimation in an ongoing clinical trial.