The Impact of Investigator Heterogeneity in Clinical Trials on Detecting Treatment Differences

Abstract

In the usual multi-investigator trial, patients are randomized to treatment groups according to a randomization schedule generated for each investigator. Appropriate analyses of data collected from such a trial block on investigator to account for inter-investigator differences. Such analyses are also consistent with the randomization scheme. In some multi-investigator trials, the primary investigator or center may have other investigators who also enroll and provide physician care to patients. If differences between investigators within centers are expected, separate randomization schedules should be generated for them as well. For various reasons, often this is not done, and investigators within centers enter patients according to the randomization schedule for the center. To be consistent with randomization, statistical analyses of data from such trials would block only on the center. The possible impact that randomization to treatment within each investigator versus randomization to treatment within each center can have on the statistical detection of treatment differences is considered. It is observed that homogeneity of investigators within centers can have a demonstrable impact on the statistical detection of treatment differences. Homogeneity of investigators within centers may be accomplished through greater standardization, training and more careful monitoring.