Title: The Advantages Of Minimization In The Prevention Of Selection Bias And In The Design And Analysis Of Clinical Trials
- Date/Time: May 23, 2014
12:30 pm - 2:00 pm - Speaker: Donald R. Taves, M.D., M.P.H., Ph.D.
- Location: Room 5E32/34, 9609 Medical Center Drive, Rockville MD
Photo ID and sign-in required.
Take the red line to the Shady Grove metro stop. You can take the NIH shuttle to the 9609 Medical Center Drive Building. - Sponsor: Public Health and Biostatistics Section, WSS and the NCI
Abstract:
The possibility that minimization can prevent selection bias has not been recognized. It exists because of the advances in detection of selection bias made by Vance Berger. He uses the probability of the next assignment going to a particular treatment group; he labels it the reverse propensity score (RPS). To cause selection bias an investigator must use the high RPSs for selected subjects. Finding a correlation between the RPS and the outcome suggests that selection bias has occurred. The RPS used as a variate in minimization will automatically reduce the chances of selection bias. In addition if the individual investigators are included as variates the possibility of selection bias can be further decreased and the perpetrator identified. The decrease will occur because of interference with each of the three necessary conditions for selection bias as outlined by Berger. The use of the RPS and investigators as variates in randomized block designs is not possible because they are limited to too few variates. Thus minimization has a unique and important advantage over randomized block designs in preventing selection bias. The claim that minimization improves the design and analysis of clinical trials has received little attention. The design is improved by the greater attention paid to a large number of variates, giving better clinical guidance and follow up studies. The analysis is improved because the balancing of secondary variates without taking them into account in the primary analysis results in a steeper power curve and better repeatability of clinical trials.
