randomization
Meaning of randomization: Deciding treatment by chance, which remains the only method that balances the factors nobody involved thought to measure.
Definition of randomization
Randomization is the assignment of participants in a study to treatment groups by chance alone, using a generated sequence rather than any judgment about who should receive what. Its purpose is to make the groups comparable at the start, so that differences appearing at the end can be credited to the treatments. It is the defining feature of a randomized controlled trial.
Why does chance beat judgment?
Because judgment carries preferences that shape the outcome. Left free to choose, a clinician tends to offer a promising new treatment to those most likely to do well, or sometimes to those with nothing left to lose, and either tilt creates a difference at the end that owes nothing to the treatment itself. Chance has no such preference. It also handles what nobody measured, and that is the deeper argument: age and disease stage can be balanced deliberately, but factors nobody has thought of, or cannot measure, are balanced only by leaving the allocation to chance. No statistical adjustment made afterward substitutes for it.
Is everyone given an equal chance?
Not necessarily, and this is a common misunderstanding. Random does not mean equal. Trials frequently allocate two participants to the treatment group for each one to the comparison group, so that more experience accumulates with the new option or so that fewer people receive a dummy treatment. Allocation remains random because nobody can predict which group the next person will fall into. What has to be absent is influence over the assignment, not inequality in the size of the groups.
How is it actually done?
| Method | How it works |
|---|---|
| Simple | Each allocation drawn independently, much like a coin toss |
| Block | Allocations arranged in small blocks, keeping the groups even in size |
| Stratified | Separate sequences within categories such as age or disease stage |
| Cluster | Whole clinics or communities allocated instead of individuals |
| Minimization | Each allocation weighted to balance the characteristics already recruited |
One safeguard sits alongside it and is often confused with it. Allocation concealment keeps the next assignment hidden from whoever is enrolling participants, since a sequence that can be foreseen can be gamed, and that is a separate matter from blinding, which hides the assignment after it has been made. Setting a randomized clinical trial beside a nonrandomized clinical trial shows what the step buys, because the second can never rule out that the groups differed before anything was given.
Also known as
Used in a sentence
Randomization was stratified by disease stage at entry.
Good to know
Randomization does not require equal group sizes, since unequal allocation ratios are still random, and it cannot guarantee balanced groups in a small trial where chance alone may leave them uneven.