Testing Clinical Trials for Randomness

As it comes to well-balanced random sampling of representative experimental data, nature will be helpful to provide researchers with results that comply with the random property. It means that such data closely follow statistical frequency distributions. We continually make use of these statistical frequency distributions for analyzing the data and making predictions from them. However, we, virtually, never assess how close to the expected frequency distributions the data actually are. Unrandomness of the data may be one of the reasons for the lack of homogeneity in current research, and may jeopardize the scientific validity of research data.1–3 Statistical tests used for the analysis of clinical trials assume that the observations represent a sample drawn at random from a target population. It means that any member of the population is as likely to be selected for the sampled group as the other. An objective procedure is required to achieve randomization. When other criteria are used to permit investigators to influence the selection of subjects, one can no longer conclude that the observed effects are due to the treatment rather than biases introduced by the process of selection. Also, when the randomization assumption is not satisfied, the logic underlying the distributions of the test statistics used to estimate that the observed effects are due to chance rather than treatment effect fails, and the resulting p-values are meaningless. Important causes for unrandomness in clinical trials include extreme exclusion criteria4 and inappropriate data cleaning.1