Option 2 Randomize patients with respect to both questions at the same time

Here all patients who consent to both randomizations are randomized 'up front' to their chemotherapy regimen and its duration, and so the choice of whether or not to be randomized with respect to duration does not depend on their experience of the initial chemotherapy regimen nor its effectiveness. However, since the question of stopping or continuing is relevant only to those who achieve a response or stable disease (and continuation of the same regimen in the face of progressive disease would not be considered appropriate), it is apparent from the start that a number of patients will not follow their allocated treatment. For example, approximately 50 per cent of patients in all the chemotherapy arms were expected to have progressive disease at three months. Therefore, half of those randomized to continue would in fact have progressive disease, and would not continue. Here then, the major effect could be on the stop versus continue randomization - analysis by intent to treat will mean comparing two groups in which the proportions stopping and continuing may be much more similar than you would like. It would then be necessary to assess the appropriate sample size, which will certainly need to be larger than under option 1 to account for the 'muddying' (the calculation follows the same lines as that described under Zelen randomization in Section 4.3 under assumptions of varying levels of non-compliance). Depending on the proportion of drop-outs, it maybe that the sample size required for the stop versus continue question becomes larger than that for the primary question. On the plus side, it might well be argued that patients randomized upfront and knowing their overall treatment plan may have greater levels of compliance with stopping or continuing (where their response makes this question relevant) than patients given the option of a later randomization.

0 0

Post a comment