html xmlns=”http://www.w3.org/1999/xhtml” xmlns:mml=”http://www.w3.org/1998/Math/MathML” xmlns:epub=”http://www.idpf.org/2007/ops”>
Einstein is often thought to have said that “God does not play dice with the universe.” Without entering either in the scientific or theological issues of this statement, there is at least one place where you should “play dice”: when making treatment allocations in an interventional study. This process, called “randomization,” is used to minimize the chance that the groups in the study differ on any important prognostic factors, and thus to avoid prognostic bias. Randomization implies that a chance process under the direction of the investigator is used to determine which of the two or more specific interventions a participant receives. This does not mean, however, that the investigator controls which treatment is assigned to a specific participant. In a randomized interventional study, participants are enrolled into a clinical study and randomized after giving informed consent and completing the screening process, which ensures that the participant meets the enrollment criteria for the study. In this chapter we discuss what randomization is and why it is essential, while the next chapter discusses how randomization should be done.
20.1 What Is Randomization?
Randomization involves the random allocation of something to the participants. We are deliberately using the awkward word “something” here, because the something might be whether the participant receives an active intervention or is studied without an intervention, or which of several interventions the participant receives, or it might be the order in which different interventions are given. For convenience throughout this book, when we talk about an intervention, we may mean any of these somethings.
Randomization implies that
(a) there is a random process allocating the intervention;
(b) the process is under the direction of the investigator; and
(c) no one knows what the intervention will be until the participant has completed all the screening and is enrolled into the study.
All three are essential. The process being random means that there is no pattern – that is, that no one can predict in advance what intervention the process will assign – so knowing the history of the process (that is, all the assignments until now) does not help you predict what the treatment assignment will be for the next participant. A computer algorithm that simulates rolling a die or tossing a coin (but not a person doing that!) is used to determine the intervention allocation. The result is generally called a “randomization table” or “randomization schedule.” The second item on the list means that you cause a randomization schedule to be created and applied when participants are enrolled in the study. But importantly, the third item means that neither you nor anyone else involved with participant screening or care knows what intervention a participant will receive before they have completed the screening process and are enrolled into the study and randomized. Although you direct the process, you do not control the assignment to a specific participant. This is important, since generally inclusion and exclusion criteria for a study allow participants to be excluded if they are “unsuitable.” If you know which treatment the next participant will receive, you may decide, consciously or unconsciously, that a participant is unsuitable for the study because of the treatment the participant would receive if enrolled. See Chapter 21 for how randomization can be implemented. Most randomized studies also include blinding (Chapter 22), if possible, so that the investigator does not know what treatment a participant receives until after the study is completed.
An investigator wants to assess whether a new drug is better than the standard of care. The investigator decides that certain participants will receive the new drug and certain participants will not, so that the investigator will have a comparison group. Perhaps the investigator believes that the new drug will be better for some patients while the standard of care is better for others. This is not a randomized study, since the investigator assigns the treatment. Even though the investigator may try to make the groups as similar as possible, this procedure is subject to potential prognostic bias, either conscious or unconscious, leading to systematic differences between the two groups, which could confound any effect of the intervention.
Instead of assigning certain participants in the study to receive the drug and others not to receive the drug, the investigator decides to use the participant’s birthday to decide which treatment the participant will get. Although this is not predictable and may or may not be known before the participant is accepted into the study, this process is not under the investigator’s direction, so this is not randomization.
Instead of using birthdays to determine which participants receive the drug and which do not, the investigator decides to actually toss a coin. Tossing a coin seems random, and since the investigator is doing it, it is under the investigator’s direction. However, appealing as this is, a coin can be biased, and there are people who can toss a coin to come up heads (or tails) in a nonrandom sequence.
Instead of tossing a coin, the investigator uses a computer program to prepare a randomization table (that is, a list of the treatment assignment for each participant) to determine which participants receive the drug and which do not. As the investigator has prepared the list of assignments, however, the results are known to the investigator before a participant is enrolled, so this is not a randomized study.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
