Web-based open source application for the randomization process in clinical trials: RANDI2

TLDR

Randomization is essential for multi‑arm clinical trials to ensure comparable groups and minimize random error, which is best reduced by large sample sizes. The application implements randomization using four mathematical concepts—complete, restricted, covariate‑adapted, and response‑adapted—driven by a pseudo‑random number generator such as a linear congruential generator.

Abstract

The randomization process is pivotal forclinical trials with more than one treatmentarm. This is because the same boundary con-ditions have to be used whenever new thera-pies are being compared with standard treat-ments or other experimental treatments.Kendall [3] describes “Chance” as a randomerror appearing to cause an association be-tween an intervention and an outcome. Themost important design strategy to minimizesuch random error is to have a large samplesize. Randomization is based on four mathe-maticalconcepts: (i) completerandomizationwith probability p = 1/n for the choice of onetreatmentout of n with identically distributedBernoulli random variables, (ii) restrictedrandomization with a variance-covariancematrix unequal 1/4 I (unique matrix), (iii)covariate adapted, and (iv) response adaptedrandomization. The random process itself isdrivenbyapseudo-randomnumbergeneratorusing, e.g. the linear congruential generatorX