Often sample sizes for randomized clinical trials (RCT) are too small because prior estimates of design parameters are inaccurate. Therefore, a computerized approach to increasing sample size during a RCT without increasing d-level has important commercial application for organizations that perform RCTS. This proposal describes a new class of designs for an RCT. The new design designates a portion of the main clinical trial as the internal pilot (IP) study. The sample size may be recalculated based on data from the IP. The final analysis includes data collected during the IP. The proposed project will develop prototype software programs for IP designs. Phase I has three aims: (a) To investigate operating characteristics of a binomial test for a IP trial. The binomial represents distributions with nonindependent sample mean and variance. Numerical integration, checked for accuracy by simulation, will calculate size and power for several types of IP's. (b) To determine by numerical integration for normal distributions the optimal proportion of the planned sample to include in the IP phase. (c) To write a demonstration computer program that designs an IP-RCT using a variety of inputs (e.g., parameters to be recalculated and maximum allowable sample size). Applications to cardiovascular disease will be stressed.
| Zucker, D M; Wittes, J T; Schabenberger, O et al. (1999) Internal pilot studies II: comparison of various procedures. Stat Med 18:3493-509 |
| Wittes, J; Schabenberger, O; Zucker, D et al. (1999) Internal pilot studies I: type I error rate of the naive t-test. Stat Med 18:3481-91 |
