Randomization

A good principle in any randomized trial is to carry out randomization as near as possible to the time at which the treatments diverge - the shorter the time between treatment being allocated and being administered, the less room for factors interfering with the goal of giving treatment as allocated. For example, in rapidly progressing cancers such as malignant glioma, a delay of a few weeks between randomization and starting treatment might mean a not insubstantial proportion of patients become too ill for radical treatment. The only exception to this rule, and one that would apply more often to the pragmatic trial, is that there maybe situations where treatment policy would be decided at an earlier stage. For example, in an MRC/EORTC trial in malignant glioma, patients received conventional radiotherapy alone, or followed by a stereotactic radiotherapy boost. As the boost was to take place after conventional radiotherapy, it would have been preferable to carry out randomization on completion of conventional radiotherapy. However, with stereotactic facilities available at only a limited number of centres, many patients had to be referred to other centres if allocated the stereotactic boost. Without advance warning of a potential patient, the need for review of scans and for securing a place on the waiting list for treatment could have led to a substantial delay between the completion of conventional radiotherapy and the start of the boost course. Therefore, for practical reasons, randomization took place immediately before the conventional radiotherapy course was due to start, and the sample size allowed for possible drop-outs.

In general, there may be aspects of the treatments that are common to all arms which would be modified to some extent if it was known that a specific additional treatment would or would not follow. If this is accepted as appropriate in routine clinical practice, then the public health trial would need to consider accommodating this by randomizing before the common treatments commence.

0 0

Post a comment