Increasing accrual rates

May allow a larger sample size in order to:

♦ Increase power to detect the same target difference

♦ Permit smaller target differences to be detected reliably changes to one or more of the 'input' factors; these may result from a lack of good data to design the trial originally, or emerging data from other sources. Secondly, substantially higher accrual rates which make larger trials feasible.

As discussed elsewhere in this chapter, all the possible changes to the 'input factors' have the potential to change the necessary sample size. It is therefore useful to assess periodically how well the original assumptions made with respect to these factors are holding up. Importantly, all can be assessed without revealing comparative outcome data. In general, there are few problems with adjusting sample sizes provided the decision is independent of the current trial results. Decisions which are conditional on, or even marginally influenced by, the current data can be problematic. Thus, although it might seem appropriate to seek the advice of the trial's independent data monitoring and ethics committee (DMEC, see Section 8.10), if one exists, it will generally be better for the decision to be made by an independent group. For example, a DMEC may be reluctant to recommend a major increase to a trial's sample size (with the practical and funding implications that may have) if they are aware that the results are tending to suggest early closure is likely. The DMEC must however be made aware of any changes to the trial design and must, at their next meeting, perform their usual functions and make their usual recommendations with reference to the amended design.

A slightly different issue is raised when accrual rates are better than anticipated, and make larger trials feasible. As discussed in Section 5.5, compromises are often necessary in determining the size of trial to be conducted, and it is not unusual for a trial to be launched which has lower than ideal power for the size of difference which is likely and/or clinically important. If the sample size is increased under these circumstances, it is important that it is done with a specific target in mind, not simply because 'more patients are always better.' Thus for example, it is simple to calculate how many additional patients would be required to provide 90 per cent power rather than 80 per cent power to detect the same treatment difference, and to set a revised sample size in the light of this. Equally, it is possible to recalculate the sample size based on wishing to detect a smaller difference than originally specified. However, in this case it is important that such a change is discussed, and that the decision is made because a smaller difference is genuinely of clinical interest, not simply because it is 'do-able.' It is much more straightforward to make this kind of decision if the original trial protocol details the range of treatment differences which are clinically relevant, even if for practical reasons the trial is planned with a sample size which is less than that required to detect the smallest of these. Again, it is important that decisions about changing sample size in these circumstances are made independently of the accumulating data, but that the DMEC are informed and can advise on the continuation ofthe trial in the light ofthese changes at their subsequent meetings.

0 0

Post a comment