Key Message: How to reduce the risk of selection bias in RCTs

Probably the simplest and most effective method to prevent selection bias in randomized trials is simple randomization.

Simple randomization works by assigning each patient to one of the treatment groups with a 50% probability; this probability is the same for every patient, regardless of previous allocations. For example, consider a trial where 40 of the first 60 patients were assigned to the intervention, when the 61st participant is enrolled, the probability of being assigned to either treatment group is still 50%, regardless of the imbalance in previous allocations. Because the probability is always the same, recruiters will not be able to guess with any accuracy which treatment the patient will be assigned to; therefore, selection bias cannot occur in this scenario.

In practice, simple randomization is infrequently used, possibly because investigators, for different reasons, prefer randomization methods which provide balance in the number of patients assigned to each treatment group. However, provided the overall sample size is not too small, this lack of balance has only a very small impact on power, and should not be used as a reason to avoid simple randomization. Therefore, many experts recommend the use of simple randomization more frequently in practice.

Referecnces

  1. Hewitt CE, Torgerson DJ: Is restricted randomisation necessary? BMJ 2006, 332(7556):1506-8.
  2. Kahan BC, Morris TP: Reporting and analysis of trials using stratified randomisation in leading medical journals: review and reanalysis. BMJ. 2012, 345:e5840.
  3. Kahan BC, Rehal S, Cro S. Risk of selection bias in randomised trials. Trials 2015, 16:405
  4. Rosenberger WF, Lachin JM: Randomization in clinical trials. Wiley, Chichester; 2005.
  5. Schulz KF, Grimes DA: Unequal group sizes in randomised trials: guarding against guessing.Lancet 2002, 359(9310):966-70.
Advertisements