Minimization: A Flexible Randomization Method

Randomization is the key element of comparative clinical trials and has thankfully evolved substantially from the early days of randomization lists and sealed envelopes at the sites for emergency code breaks. A commonly used method is stratified permuted block randomization, but covariate-adaptive randomization (also known as minimization) can be better in many situations.

In this webinar, the minimization method will first be explained, including demonstration of some of the advantages of minimization over permuted blocks, discussion of the implementation of the method and closing with some challenges related to minimization.

Minimization was originally suggested by Taves (1974) and by Pocock and Simon (1975) so the method has been used for several decades. Its popularity has increased as web-based Randomization and Trial Supply Management (RTSM) systems have been developed and have become more sophisticated. Minimization is often used in multi-center studies when two objectives are simultaneously sought: to maintain a good balance between all treatment groups with respect to important prognostic factors and to maintain the unpredictability of the next treatment assignment.

Minimization attempts to achieve optimum treatment balance between several factors simultaneously rather than within the separate strata defined cross-classification of these factors. The major advantage of minimization, as compared to a stratified treatment allocation, is that good treatment balance can be achieved across a large number of “minimization” factors whereas for stratified permuted block, the number of stratification factors is limited. In this webinar, simulations that help identify which methods give the best results in terms of the size and power of the test, as well as the precision of the estimation, will be presented.

There is discussion among the statistical community regarding which methods of randomization and analyses are preferred. This webinar will discuss some challenges that might come from using minimization. For example, some regulatory agencies may not be comfortable and/or they will require the use of re-randomization tests. In this presentation, it will be demonstrated how such tests are easy to implement and usually do not differ from tests based on asymptotic theory.

Speakers

Marc Buyse, ScD, Chief Scientific Officer, International Drug Development Institute (IDDI)

Marc Buyse holds a ScD in biostatistics from the Harvard School of Public Health (Boston, MA). He is the founder of the International Drug Development Institute (IDDI) and of CluePoints, two biostatistical service organizations based in the US and Europe. Buyse is interested in clinical trial design, meta-analysis, validation of biomarkers and surrogate endpoints, statistical methods in oncology, statistical detection of errors and fraud, statistical monitoring of clinical trials and medical data sharing (http://publicationslist.org/marc.buyse).

Message Presenter

Linda Danielson, MS, Chief Operating Officer, International Drug Development Institute (IDDI)

Linda Danielson has been Chief Operating Officer at the International Drug Development Institute (IDDI) since 2011. Danielson obtained her MS degree in Biostatistics in 1990 from the University of Wisconsin, Madison. She has over 25 years of experience in the Pharmaceutical Industry and in the Clinical Trial environment. Before coming to IDDI, Danielson was the Head of Biostatistics at UCB Pharma in Belgium. Danielson has had a special interest in the relationship between Statistics and Randomization since helping select a randomization vendor at UCB in the late 90s.

Message Presenter