The perfect and the practicable

Although careful thought, and a good deal of'science,' must go into deliberations about sample size, it must be recognized that the final choice of sample size almost always involves compromize between what is ideal and what is practicable.

Therefore, it is important to try to anticipate the likely accrual rates and hence the time the trial will need to run if your ideal sample size is to be achieved.

Sources of information, in decreasing order of value include:

♦ rate ofpatient accrual in previous trials in the same disease,

♦ survey of potential participants,

♦ estimate of approximate number of eligible patients from national incidence rates.

It must be borne in mind that the latter 2 sources tend to over-estimate the number of patients available and should be considered extremely conservatively. While data on the number of patients with a given diagnosis seen annually might be reliable, it is often difficult to determine what proportion would satisfy the eligibility criteria, what proportion would consent to enter a study and what proportion might be lost through time pressures or other factors meaning they are never approached about the trial.

Estimating the rate of accrual is only one part of the equation. The total time to consider is that from the launch of the trial to the first publication of mature results. It is therefore important to consider the trial outcome measures, and to bear in mind that if the primary outcome measure is survival time, a period of follow-up will generally be required after entry of the last patient before the trial can be analysed definitively. As previously noted, this is because the important factor in survival studies is not the number of patients but the number of events.

Is this timeframe reasonable?

Some helpful points to consider are the following:

♦ How common is the condition? Remember that in the last forty years only a small proportion of cancer trials have accrued more than 5 per cent of the incident population (diseases treated in specialist centres are an exception).

♦ How long will enthusiasm for the trial continue - are other treatments 'in the pipeline'? If the trial is asking a fundamental question, or new treatments are at a very early stage of development, a long accrual time may not be a problem. If the answer is likely to be as important at the end of the trial as when you started, it is worth persevering.

What if the accrual period seems unreasonably long?

There are some options that should be considered before the trial is abandoned.

♦ Can the accrual rate be increased without altering the design of the study?

Consider whether the number of participating centres can be increased either nationally or internationally, if not by participating in the same, centrally organized, trial then perhaps through running parallel trials with a formal agreement to combine the data for the primary analysis.

♦ Can the number of potentially eligible patients be increased?

This implies considering whether any of the eligibility criteria (relating to both patients and perhaps clinicians) can be relaxed in such a way as to increase the number of patients who would potentially be suitable for inclusion in the trial without compromising its aims. Section 4.2.3 discusses the essential features of eligibility criteria.

♦ Can any of the 'inputs' to the sample size calculation be reconsidered?

This is perhaps the least attractive option. If the size of difference the trial is powered to detect is the result of careful assessment and broad discussion, then it is hard to see how this component could be changed. However, if the error rates had been chosen to be particularly stringent, there may be scope to relax these somewhat, accepting lower power for a more rapidly completed trial for example. For survival studies, remember that there are two ways to get a given number of events; a large trial with short follow-up or a smaller trial with longer follow-up. However, one must consider if, for this disease and these treatments, early results will necessarily reflect those which would emerge with longer follow-up.

If the answer to these 3 questions is no, should the trial be abandoned? If the only option appears to be a smaller trial than is ideal - is it worth doing? One can argue that any randomized evidence is better than haphazardly treating patients, but to embark on a trial which has no hope of providing useful information is wasteful of effort and resources, and unfair to patients who consent to be in a trial which they believe will truly help future patients if not themselves. As a general rule, however, the trial may be worth pursuing provided:

♦ It will contribute new information, or add substantially to existing data.

♦ The limitations of the study resulting from its size are made clear in both the protocol and publications.

♦ It is registered so that, even if it is never published, the trial can be identified and perhaps used in systematic reviews and meta-analyses.

0 0

Post a comment