Positioningyour trial

In deciding whether a clinical trial of a therapy is warranted, it is important to know what has been done before, in particular what phase of development the treatment has reached. Systematic searching for relevant systematic reviews and trials is therefore an essential pre-study requirement. The first step should be to find out whether a systematic review or meta-analysis has already been done. The Cochrane library (see Chapter 11) is a good starting point as it includes Cochrane Collaboration systematic reviews in the Cochrane Database of Systematic Reviews (CDSR) and also abstracts of systematic reviews done outside of the collaboration in the DARE database. Literature searches may reveal additional systematic reviews. If no systematic review has been done, a systematic search for relevant trials should be carried out (see Section 11.5.4) to identify both

Fig. 3.2 Trial decision tree.

* Search for or/and conduct systematic review to identify trial.

Fig. 3.2 Trial decision tree.

* Search for or/and conduct systematic review to identify trial.

published and unpublished trials. This information is essential in justifying a new trial as it would be unethical to embark on a new trial if there is already sufficient information from completed trials to answer the question posed, and possibly impractical if similar trials are already ongoing.

Fig. 3.2 gives an outline of the questions to ask in determining whether a trial is needed, and if so, the type. Points to consider in answering the questions are described below.

Has a safe dose been established in clinical use?

If a new therapy has yet to be used in humans, then a phase I dose-finding study of some form is almost always needed; such a stage may even be needed if a treatment is in clinical use, but is being evaluated as part of a new combination therapy.

Is there evidence of activity?

If a therapy has been subject to some clinical evaluation to determine an appropriate dose and schedule, then it maybe appropriate next to conduct a phase II study, particularly if none have yet been performed. If phase II studies have been performed, but have failed to demonstrate a useful level of activity, then options are to abandon the treatment, or conduct another phase II study. Careful thought is needed before taking either path, and certainly it is worth considering if there is any potential to enhance activity before doing so.

Has the treatment been used in a comparative trial?

Chapter 11 describes search strategies for identifying randomized trials in the published literature, and sources of information which may identify unpublished trials. Searching for non-randomized comparative studies requires a more ad hoc approach but is nonetheless important; such studies may help build a case for or against a randomized trial.

Is the case for a phase III trial sufficiently strong?

If systematic searching has failed to identify any comparative studies involving the new therapy, then it should not automatically be concluded that a phase III trial should be performed. Given the limitations of phase II trials with respect to identifying active (as opposed to inactive) treatments, it will generally be the case that even after one or even more have been completed, uncertainty will remain about the potential benefits of a treatment. Simply repeating a phase II trial in a similarly small group of patients may bring limited data to help this process, but may well be necessary if existing trials have been flawed in some way - perhaps toxicity recording was inadequate, or the group were not registered prospectively. Other reasons for repeating a phase II study might include the need to look at different patient populations or the performance or practicalities of a therapy under different health care systems. If the treatment you wish to study has been modified in some way compared with the previous studies, then it will often be a clinical judgement as to whether the degree of modification is sufficient to warrant a further phase II trial, or whether the existing data are sufficient.

Has the trial(s) answered the questions of interest?

If other phase III trials have been done or are ongoing, the question is, is another phase III trial required? Repetition and replication are important principles of most scientific disciplines. Essentially, clinical research is no different, but the particular ethical requirements surrounding clinical rather than laboratory or theoretical research demand that the case for repetition should be carefully made. There is often a good case for repetition; as was mentioned in the context of repeating phase II trials, there maybe a need to repeat what is essentially the same trial in a different - perhaps broader - group of patients, or in a different health care system where other aspects of patient care may differ. More often, a decision is taken to conduct a further randomized trial because it is felt that existing trial data are in some way inadequate. There are several reasons why this conclusion may be reached, each of which is discussed in more detail in the following sections.

♦ Previous phase III trials were not randomized. As we have explained earlier, a 'phase III' trial is comparative, but not necessarily randomized. Non-randomized comparisons can suffer from serious bias and problems in interpretation; their limitations are discussed in Section 3.4.

♦ Previous phase III trials did not compare the new treatment with the existing 'gold standard'. The appropriate choice of control arm is a major issue in trial design and, while the choice may sometimes appear obvious, it may not always be so and in fact it maybe in some people's interest to compare a new treatment with a sub-standard control. This issue is discussed further in Chapter 4.

♦ Previous phase III trials were underpowered for clinically important differences. As discussed earlier, it is rare that a new treatment will pass through just one trial of each phase before becoming accepted. This is particularly true of phase III. In some situations, small or moderate treatment effects can be very important clinically, but may require trials of several thousand patients if such effects are to be detected reliably. It will often be the case that a smaller trial is performed first. This might be as a planned feasibility study for a subsequent larger trial, but perhaps more commonly smaller trials are done either because those conducting them had limited resources, or because they believed the impact of the new treatment would be substantial. Thus, although otherwise designed to assess the impact of a new treatment in clinical practice, existing trials may simply not have been large enough to produce reliable results. Chapter 5 discusses ways of determining how large a trial should be. In this situation, however, it might be appropriate to ask whether a formal systematic review and meta-analysis could answer the question without needing another trial.

♦ Data on potentially important outcome measures are lacking. A more difficult situation is perhaps where previous randomized trials have shown evidence of potentially useful efficacy over other treatments with respect to one outcome measure, but data on other relevant outcomes were not collected. Randomizing further patients in this situation may be difficult ethically, but equally may be important; first one needs to consider how extreme the differences between treatments must be with respect to other outcome measures - perhaps toxicity or quality of life - before they would overwhelm the benefit shown with respect to the primary outcome measure. If the differences would have to be substantial, it is possible that a carefully designed nonrandomized study may provide sufficient data. If small or modest differences in the secondary outcome measures could be important, then a further randomized trial collecting such data may be appropriate.

♦ The trial population was highly selective. Perhaps a treatment has shown efficacy in a particular type of patient, or perhaps in a particular setting, and the question arises as to whether the results can be extrapolated to other patients or settings. Sometimes, treatments pass through two broad stages of phase III trials in a planned manner; an early one on a modest scale designed with the aim of giving a new therapy the best possible chance of showing improved efficacy levels and a later, often larger scale trial, aimed at assessing impact in routine clinical practice. These two broad types of trials have been termed 'explanatory' and 'pragmatic' [12], and their different aims can have a major impact on many aspects of trial design, conduct and analysis. These issues are discussed further in Chapter 4, but one characteristic concerns patient selection.

For the explanatory trial, the patients who are considered most likely to demonstrate a difference in response to the alternative treatments are chosen; for the pragmatic trial one would include any patient who might be considered for the experimental treatment outside of the trial. For example, one might choose higher riskpatients (who have most potential to benefit) for the early trial but be interested more generally in the impact on lower risk patients should the treatment prove effective. There are no hard and fast rules for determining when a trial result can be extrapolated and when it cannot. More often than not, the differences will be of degree and not direction, and thus it will be a question of asking by how much would the treatment effect have to differ in a broader population before it was no longer considered worthwhile.

Could a systematic review or meta-analysis answer the question?

A systematic search of the literature alone may reveal a considerable quantity of relevant data pertaining to the treatment of interest, but the usual situation is that the results are to some degree at least, apparently conflicting. After all, if there were unanimity, you probably wouldn't even be considering the trial (unless you were very out of touch with the field). In this situation, it may often be that a systematic review and formal combination of the data from all relevant trials is appropriate. Chapter 11 describes the motivation and methods for systematic reviews and meta-analyses, and the pros and cons of using only published data to do this. If the systematic review indicates that the new therapy is already of proven benefit, or the evidence against a clinically worthwhile benefit is sufficiently strong, then a further trial is unlikely to be needed. If it demonstrates that a potentially useful effect maybe obtainable, but perhaps the number of patients and events in previous trials is insufficient to be conclusive, then a new trial may be warranted and the findings of the systematic review may be helpful in designing the trial. We describe in Box 3.5 two such situations in which a meta-analysis was used to guide decisions about future trials.

0 0

Post a comment