Why Choose an Adaptive Clinical Trial Design?
All trial sponsors want their trials to be as efficient as possible: to get the best data in the shortest time at the lowest cost. Adaptive designs are one tool in the clinical triallist’s toolkit that can sometimes help to achieve that aim. For example, an adaptive design may result in fewer patients than a non-adaptive design, thus clearly saving time and costs, as well as the ethical advantages of exposing fewer patients to experimental treatment.
But adaptive trials are not a magic bullet. While they may sometimes bring welcome efficiencies, sometimes they may bring only added complexity with little to show for it on the upside. They may even increase sample sizes in some circumstances. It is a mistake to think that all trials should be adaptive. Before choosing an adaptive design, it is important to be clear about the purpose of adaptation, and whether it is truly appropriate in the individual trial. Expert statistical advice is essential in deciding whether, and how, to adapt.
What adaptive trial designs are available?
There are a wide range of adaptive designs, ranging from simple and well-established designs to more complex and experimental ones. It is likely that simple, well-established designs can be used with little risk of regulatory objections provided that they are used appropriately, but more complex designs may require considerable up-front work and lengthy discussions with regulators before they are approved for use in a trial.
One of the simplest and best established adaptive design is a blinded sample-size re-estimation: re-evaluating the sample size estimation for the trial based on data obtained at an interim analysis, to ensure that the sample size estimation is based on data from the trial you are actually running, rather than data from the literature that may or may not be relevant to your trial. This is an excellent way of de-risking a trial, and I would recommend that this technique be used in most phase III trials in which it is feasible.
Another well established adaptive technique is group sequential designs: here, the trial may terminate after an interim analysis, either because the results are disappointing and there is little point in continuing the trial (stopping for futility), or because the results are excellent and there is already sufficient evidence that the trial is successful (stopping for efficacy). Stopping the trial early for either reason will not only save time and money, but also has ethical advantages: if the trial is stopped for futility, fewer patients are exposed to ineffective treatment, and if a placebo-controlled trial is stopped for efficacy, fewer patients will be given placebo when an effective treatment is available.
Many other adaptive features are possible: for example, a trial with multiple arms (for example different doses of the experimental drug) may drop some arms along the way, or changes may be made to the patient population recruited to the trial, the randomisation ratio, or even the primary endpoint.
It may be useful to run some simulations before the trial starts to determine the operating characteristics under a range of plausible assumptions, such as the probability of stopping the trial early, the expected total sample size, the power, and the type I error rate. For some of the more complex adaptive designs this may be essential before regulatory approval for the design will be given.
For complex designs involving simulations, the FDA recommend early discussion with regulators to avoid delays in the review process [1].
When to adapt?
Decisions in an adaptive design are made after an interim analysis of the study results, and the timing of such an analysis can be challenging. If the interim analysis is done too early, there may be insufficient data to make a good decision. On the other hand, if the analysis is done too late, it is possible that recruitment will already be complete by the time sufficient patients have been on study long enough to contribute data to the interim analysis, thus largely negating any benefits from being able to stop the trial early.
Indeed in some trials in which patients need to be followed up for a long time before they can contribute useful data, a meaningful interim analysis may simply not be possible, particularly if recruitment is fast. In such cases, an interim analysis based on a surrogate endpoint that can be measured at an earlier timepoint may be a reasonable strategy.
If the most appropriate timing of the interim analysis is not obvious, then again, simulations may be helpful in exploring the advantages and disadvantages of different options.
How to ensure statistical rigour in adaptive designs?
Having an adaptive design is absolutely not a licence to make things up as you go along. It is important that any potential adaptations and the rules for making them are carefully planned before the study starts.
There are often statistical penalties related to adaptive designs, and it is important to take account of these and make sure that the type I error rate (the probability of declaring the study to be successful by chance when in fact the study drug is ineffective) is properly controlled. In simple circumstances there may be straightforward calculations to ensure that this happens, but in more complex cases, simulations may be required to show that the type I error rate is controlled.
It is also important to ensure that, where an interim analysis requires knowledge of treatment assignments in a blinded study, the appropriate operational security measures are in place to ensure that no-one is unblinded other than the nominated unblinded team who run the interim analysis, and that any announcement of results is made carefully, giving regard to the minimum information necessary to ensure that any adaptations can be implemented.
Why partner with Ergomed?
Adaptive designs can bring great benefits, but they need to be done carefully and with expert statistical advice to help avoid expensive pitfalls. Ergomed’s biostatistics team are experienced in adaptive designs and can help ensure that your trial runs as efficiently as possible. We support sponsors in ensuring that their adaptive trials are both efficient and statistically rigorous, while also meeting global regulatory expectations.
References
- US Food and Drug Administration (2019). Adaptive Design Clinical Trials for Drugs and Biologics: Guidance for Industry.