Prerandomizations: Pros and cons.



Adam Taube
Adam.Taube@statistik.uu.se
Statistiska institutionen, Uppsala Universitet


Abstract

In a conventional randomized trial the potential participants first get the question: Do you want to participate in the trial? If the answer is yes, the patient is then randomized to one of two treatment modalities and participates in the trial. An alternative strategy was proposed in Zelen (1979) for the case when the trial is concerned with a comparison between a new treatment (A) and a conventional standard treatment (B): The patients are first randomized and then those who were randomized to A get the question: Do you want to have the new treatment A or do you prefer the standard treatment B? (If the answer is no, they get the standard treatment B but the results are analysed according to the intention to treat principle). Those who have been randomized to B are not asked about their preferences since they get the same treatment as they would have got anyhow, had they not been included in the trial. Dr Zelen later generalized his approach to a symmetrical design where patients are first randomized to A or B and then asked whether they want to accept the treatment modality they have been offered or if they would prefer the other one. In Zelen (1992) a general theory is given about when the conventional strategy or prerandomization is to be prefered.

In the seminar, the prerandomization strategies will be presented and the pros and cons of prerandomization as compared with the conventional approach will be discussed.

References

Zelen M. (1979): A new design for randomized clinical trials. N Eng J Med 300:1242-1245.
Zelen M. (1992): Randomized consent designs for clinical trials. An update. Statistics in Medicine. 9:645-656.