|Allocation concealment||Ensure treatment assignment cannot be predicted in advance of patient entry.|
|Double-blinding||Patients, investigators, and those evaluating outcome remain unaware of assigned treatment after randomization.|
|Single-blinding||Investigators (and possibly patients) are aware of the assigned treatment, whereas those evaluating outcomes remain blinded.|
Superior to an unblinded design, but introduces greater opportunities for bias than a double-blind design.
|Methods of randomization|
|Simple randomization||Like coin tossing, with no connection between allocations.|
May lead to treatment imbalance in numbers or key patient factors.
|Random permuted blocks||Treatment numbers are equal after each block of patients.|
Order of treatments within each block is random.
Block sizes may vary to avoid predictability if trial not double-blinded.
|Stratification||Aims to ensure balance for key patient factors across treatment groups.|
Each combination of factors (e.g., center and sex) has its own random permuted blocks.
Must avoid overstratification (e.g., 4 binary factors = 16 strata), which may introduce imbalances.
|Minimization||A dynamic approach.|
Each treatment allocation is done to achieve the best balance across several patient factors.
|Unequal randomization||Can allocate more patients on new treatment (e.g., 2:1 ratio).|
Increases knowledge of new treatment and may enhance investigator/patient enthusiasm.
Requires more patients.