COMMENTARIES

http://dx.doi.org/10.5546/aap.2013.374

Adaptive designs in clinical research

In research, it is essential to carefully plan all the steps to be taken. Likewise, it is important to consider and define, in advance and in full, all the key elements of the trial. Such planning is based on facts assumed as real or certain, therefore a considerable part of the success of the trial lies on the accuracy of this initial assumption.

Unfortunately, not everything goes as originally planned. Many times, some of the considered endpoints do not behave as expected and, regardless of everything else working out smoothly, the study fails or is seriously compromised.

In an attempt to limit the impact of these circumstances on trials, the so called “adaptive designs" became popular in the past recent years. However, it is worth clarifying that such designs were initially envisaged as a variation to shorten the trial period in the development of a new drug.¹

What are adaptive designs? Also known as “flexible designs," they are defined as studies that include a prospectively planned opportunity for modification of one or more specified aspects of the trial design based on the analysis of data from subjects in the same study (interim analysis).² In other words, it should be established that if a certain issue occurs in a trial, instead of conducting procedure “A," as initially planned, “B" should be the approach used. It is relevant to note that any change to be introduced will adjust to clearly pre-specified rules. Although such designs offer certain flexibility, it is a limited flexibility because any change must have been considered in the initial project.

What are adaptive designs for? This type of design allows to modify the original plan and thus avoid failure of the study to meet its goals. This includes the possibility of modifying the sample size, study duration, treatment group allocation, number of treatment arms, or study endpoints.³

When are adaptive designs used? Adaptive designs are most commonly used in the initial phases of drug development in exploratory studies (for example, studies evaluating toxic doses). In more advanced phases of pharmacological research, adaptive designs can be used in combined phase studies, for adaptive randomization, group sequential designs, sample size re-estimation, or a combination of these.⁴ (Figure 1)

Figure 1. Summary of different types of adaptive designs for clinical trials. (From Kairalla JA, et al., 20124. Reproduced under a Creative Commons Attribution License)

Combined phase studies have become very popular. It is very common to combine Phases I and II so that the study includes the initial drug research in a small group of human beings and the subsequent safety study in a larger population, if safety interim data obtained in the first phase thus allows it.

This can be a covariate adaptive randomization (where the likelihood of being allocated to an arm varies in order to minimize the impact of an asymmetrical distribution of potential confounding factors or an unequal “covariation") or an outcome adaptive randomization (where the likelihood of being allocated to an arm increases if the preceding subject response was favorable). In a group sequential design the decision to continue or to stop a trial is based on the results obtained in the previous arm. Sample size re-estimation is used when actual significant frequencies are verified as different from those used to estimate the initial sample size, and it should be especially considered when demographic distribution data are not well known.

What are the limitations of adaptive designs? In spite of their increasing popularity, adaptive designs are not welcome by all. It is worth noting that seeking a higher flexibility than that established by the protocol is a risk that also jeopardizes research reliability. The concern is that, after applying several “adaptations,“ the population included in the analysis will be remarkably different from that originally targeted and, as a result, there will be an inadequate control of type I error (to mistakenly attribute effectiveness to an actually ineffective treatment).⁵ This is a source of concern even for the United States national regulatory agency, the Food and Drug Administration.² In addition, planning an adaptive design is a hurdle and poses its own statistical, operational, logistic, and regulatory issues.⁶

However, when adaptive designs are judiciously used and within their particular scopes, they can certainly be an alternative to be considered in very specific circumstances.

María Fabiana Ossorio, MD Fernando Ferrero, MD
Research and Teaching Hospital General de Niños Pedro de Elizalde, Buenos Aires, Argentina

References

1. Schmidt C. Adaptive design may hasten clinical trials. J Natl Cancer Inst 2007; 99(2):108-9.         [ Links ]

2. U.S. Food and Drug Administration: Draft Guidance for Industry: adaptive design clinical trials for drugs and bio-logics. Accessed on 05/07/2012. Available at: http://www.fda.gov/downloads/DrugsGuidanceComplianceRegulatoryInformation/Guidances/UCM201790.pdf.

3. Chow SC, Corey R. Benefts, challenges and obstacles of adaptive clinical trial designs. Orphanet J Rare Dis 2011; 6:79.         [ Links ]

4. Kairalla JA, Coffey CS, Thomann MA, Muller KE. Adap-tive trial designs: a review of barriers and opportunities. Trials 2012; 13:145.         [ Links ]

5. Chow SC, Chang M. Adaptive design methods in clinical trials - a review. Orphanet J Rare Dis 2008; 3:11.         [ Links ]

6. Brahmachari B, Bhatt A. Adaptive design - an innovative tool in drug development. Indian J Med Res 2011; 133:243-5.         [ Links ]