Limitations of conventional phase II trials

The decision is not, however, limited to these three options. The fact that so many of the treatments which reach phase III testing fail to demonstrate clinically important differences suggests some problems with the conventional wisdom of phase II studies. The value of tumour response as a predictor of long-term efficacy might be questioned in some cancer sites, particularly when the phase II study involves patients with advanced disease - possibly heavily pre-treated - but the treatment is intended ultimately for use in the adjuvant setting. Perhaps, though, the key problem is that patients in phase II studies are typically highly selected, and the impact of such selection is perhaps not always well understood. How would standard therapy have fared in exactly the same set of patients? The typical phase II design makes an implicit assumption that this is known and invariable, but this may well be questionable.

Alternative designs have been developed to attempt to tackle this latter point, for example, designs which formally incorporate historical data to quantify the size and variability of the minimum acceptable response rate have been proposed - these are described byThall [7].

Randomized phase II trials

While there probably remains a place for the conventional, non-randomized, phase II trial early in the development of a treatment, there has been a welcome extension of this phase of testing to include some early, small, randomized studies. The description 'randomized phase II trial' is widely misused, often applied to what is simply a small randomized phase III trial, or a larger randomized phase III trial in which response or an alternative 'early' endpoint is the primary outcome measure but where, in both cases, the aim is to compare the outcome in the treatment groups, and this has been used to determine the study size. As shown in Box 3.1 and discussed below, the characteristic feature of the phase III trial is that it is comparative. A true randomized phase II trial is NOT primarily intended to be comparative, and it is certainly not powered, statistically, to be comparative. Rather, it should be seen as a combination of two or more parallel phase II trials in which conventional phase II sample size selection and stopping rules, such as those described above, are applied to one or more of the individual arms [8]. Typically, one would design a standard single arm phase II trial of a new therapy, but would then add an additional arm by randomizing patients equally between the standard therapy and the new therapy until the required sample size for the new therapy is achieved. The response rate on the standard therapy is not used to determine whether or not to stop the phase II trial early, but provides a check on the extent to which patient selection alone accounts for the results in the experimental group. This is then used, rather informally, to aid decisions about whether a new treatment is sufficiently promising to be taken forward into a phase III trial. An example of a trial in which such an approach was taken is given in Box 5.3. This example illustrates another of the benefits of a randomized phase II design in that, if the experimental treatment shows sufficient activity to be taken into a phase III trial, patients from the randomized phase II can contribute to the overall sample size.

Randomized phase II trials can be particularly useful when good data on response to standard therapy are lacking, perhaps because the disease is rare; where there is limited knowledge of prognostic factors and their potential impact on outcome, or where several potential new therapies require testing. In the latter case, the aim is to eliminate inactive treatments, not compare arms, and the Simon design seems to lend itself well to this situation as the determinant of sample size.

With randomized phase II trials, as indeed for any small randomized trials, it is important to use a method of treatment allocation which ensures balance for the most important prognostic factors; these are described in Chapter 4.

0 0

Post a comment