home > ebr > autumn 2010 > changing with the times
European Biopharmaceutical Review

Changing with the Times

Shein-Chung Chow at Duke University School of Medicine explains the different types of adaptive trial designs and takes a general look at their challenges and advantages

In early 2000, when the US FDA recognised that increased spending on biomedical research does not necessarily reflect an increase of the success rate of pharmaceutical development, the organisation kicked off the Critical Path Initiative to assist sponsors not only to diagnose the scientific challenges underlying medical product pipeline problems, but also to identify solutions to bridge the gap between the quick pace of new biomedical discoveries and the slower pace at which those discoveries are currently being developed into therapies. The use of advancing innovative trial designs is identified as one of the top opportunities for streamlining clinical trials, not only to increase the probability of success, but also to expedite the biopharmaceutical development process. The potential use of adaptive design methods in clinical trials has been discussed in detail within the pharmaceutical and biotech industries and regulatory agencies since then.

Adaptive clinical trial designs have become very popular in clinical research due to their flexibility and efficiency for identifying any potential (preferably the best or optimal) clinical benefit of the test treatment under investigation. The use of adaptive methods for modifying trial and/or statistical procedures of on-going clinical trials based on accrued data has been practiced for years in clinical research. Their attractiveness to clinical scientists is due to three main factors. First, they reflect medical practice in the real world. Second, adaptive designs are ethical with respect to both efficacy and safety (toxicity) of the test treatment under investigation. Third, they are not only flexible, but also efficient in the early phase of clinical development.

In practice, while one enjoys the flexibility of the adaptive trial design, the quality, integrity and validity of the trial may be at a greater risk. From a regulatory perspective, it is always a concern whether the p-value or confidence interval regarding the treatment effect under an adaptive trial design is reliable or correct. In addition, the misuse or abuse of adaptive design methods in a clinical trial may lead to a totally different trial which is unable to address the scientific and medical questions that the project was originally intended to answer. This has raised awareness of the impact, obstacles, and challenges of a fair and unbiased assessment of the treatment effects of the test treatment under investigation when utilising adaptive trial designs in clinical research.


In clinical trials, it is not uncommon to modify the trial and statistical procedures while still conducting them, based on a review of interim data. The purpose is not only to efficiently identify clinical benefits of the test treatment under investigation, but also to increase the probability of success of clinical development. In this article, an adaptation is referred to as a modification or a change made to the trial or statistical methods during the conduct of a clinical trial. By definition, adaptations that are commonly employed in clinical trials can be classified into the categories of prospective, concurrent (or ad hoc) or retrospective adaptation. Prospective adaptations include adaptive randomisation, stopping a trial early due to safety, futility or efficacy at the interim analysis, dropping inferior treatment groups, sample size re-estimation, and so on. Concurrent adaptations are usually referred to as any ad hoc modifications or changes made as the trial continues. Concurrent adaptations include modifications in eligibility criteria, evaluability criteria, dose/regimen and treatment duration, changes in hypotheses and/or study endpoints, and so on. Retrospective adaptations are usually referred to as modifications or changes made to the statistical analysis plan prior to database lock or unblinding of treatment codes. In practice, prospective, ad hoc and retrospective adaptations are implemented by study protocol, protocol amendments, and the statistical analysis plan with the regulatory reviewer’s consensus.

Although the concept of adaptive design can be traced back to the 1970s when adaptive randomisation and a class of designs for sequential clinical trials were introduced, there was no universal definition of adaptive design until recently (1). For example, Chow et al refer to an adaptive design as a design that allows adaptations to a trial and its statistical procedures after its initiation, without undermining the validity and integrity of the trial (2). In a recent publication, discussing features of ‘by design’ adaptations only (rather than ad hoc adaptations), the Pharmaceutical Research Manufacturer Association (PhRMA) Working Group on Adaptive Design refers to an adaptive design as a clinical trial design that uses accumulating data to decide how to modify aspects of the study as it continues, without undermining the validity and integrity of the trial (3). In recent draft guidance, the US FDA defines an adaptive design as a study that includes a prospectively planned opportunity for modification of one or more specified aspects of the study design and hypotheses based on an analysis of data (usually interim data) from subjects in the study (4). An adaptive design is also known as a ‘flexible design’ by the European Medicines Agency (EMA) (5,6).


Based on the adaptations employed, the most commonly considered adaptive designs in clinical trials include:

  • An adaptive randomisation design
  • A group sequential design
  • A sample size re-estimation design
  • A drop-the-loser (or pick-thewinner) design
  • An adaptive dose finding design
  • A biomarker-adaptive design
  • An adaptive treatment-switching design
  • A adaptive-hypothesis design
  • An adaptive seamless (such as a Phase I/II or Phase II/III) trial design
  • A multiple adaptive design

Adaptive Randomisation Design

An adaptive randomisation design allows modification of randomisation schedules based on varied and/or unequal probabilities of treatment assignment in order to increase the probability of success.

However, an adaptive randomisation design may not be feasible for a large trial or a trial with a relatively long treatment duration, because the randomisation of a given subject depends on the response of the previous subject and will therefore take a much longer time to complete. Furthermore, the randomisation schedule may not even be available prior to the conduct of the study.

Group Sequential Design

A group sequential design allows a trial to be prematurely stopped due to safety, futility, efficacy or a combination of these, with options of additional adaptations based on the results of interim analysis. Group sequential designs with planned interim analyses are very popular in clinical trials. However, it should be noted that the standard methods for group sequential design may not be appropriate (for example, it may not be able to control the overall Type I error rate) when there is a shift in target patient population due to additional adaptations or protocol amendments.

Sample Size Re-estimation Design

A sample size re-estimation design allows for sample size re-estimation based on the observed data at interim. Sample size reestimation is often performed in a blinded fashion based on the criteria of treatment effect size, conditional power, and/or reproducibility probability.

Drop-the-Losers Design

A drop-the-losers design drops the inferior treatment groups. This design is useful for Phase II clinical development, especially when there are uncertainties regarding the dose levels (7). The selection criteria and decision rules play important roles for drop-the-losers designs. Some clinical scientists prefer the term ‘pick-the-winners design’.

Adaptive Dose Finding Design

An adaptive dose finding design (such as escalation) is often used in early phase clinical development to identify the maximum tolerable dose (MTD), as this is often considered to be the optimal dose for next phase clinical trials. For adaptive dose finding designs, the method of continual re-assessment method (CRM) in conjunction with the Bayesian approach is commonly considered (8,9).

Biomarker-Adaptive Design

A biomarker-adaptive design allows adaptations based on the response of biomarkers such as genomic markers. A biomarker-adaptive design can be used to:

  • Select the right patient population (such as the enrichment process for selection of a better target patient population)
  • Identify the natural course of the disease
  • Detect the disease early
  • Help in developing personalised medicine (10)

Adaptive Treatment-Switching Design

An adaptive treatment-switching design allows the switch of a patient’s treatment from an initial assignment to an alternative treatment if there is evidence of a lack of efficacy or safety of the initial treatment (11). For example, in cancer clinical trials, estimation of survival is a challenge when treatmentswitching has occurred in some patients due to disease progression.

Adaptive-Hypothesis Design

An adaptive-hypothesis design allows modifications or changes in the hypothesis based on interim analysis results (12). Adaptive-hypothesis designs are often considered before database lock and/or prior to data unblinding. Some examples include the switch from a superiority hypothesis to a noninferiority hypothesis and the switch between the primary study endpoint and the secondary endpoints.

Adaptive Seamless Trial Design

An adaptive seamless trial design refers to a programme that addresses within a single trial objectives that are normally achieved through separate clinical development trials (13). Common adaptive seamless trial designs include a Phase I/II adaptive seamless in early clinical development and a Phase II/III adaptive seamless design in late phase clinical development. For example, a Phase II/III adaptive seamless design consists of a learning or exploratory stage (Phase IIb) and a confirmatory stage (Phase III). The validity and efficiency of Phase II/III adaptive seamless designs have been challenged by several authors (14).

Multiple Adaptive Design

A multiple adaptive design is any combination of the above adaptive designs. In practice, since statistical inference for a multiple-adaptation design is often difficult, it is suggested that a clinical trial simulation be conducted to evaluate the performance of the resultant multiple adaptive design at the planning stage.


In practice, the flexibility of adaptive trial designs offers the investigator the opportunity to correct wrong assumptions, select the most promising option early, make use of emerging external information to the trial, and react earlier to surprises (positive or negative). In addition, adaptive design methods provide the investigator with a second chance to re-design the trial after seeing data from the trial itself at interim (or externally). Consequently, adaptive clinical trial designs may help in speeding up the development process.

One of the obstacles when implementing adaptive trial designs in clinical trials is that too much flexibility may induce operational biases. Commonly seen operational biases include these due to patient selection, method of evaluation, early withdrawal, modification of treatment and protocol amendments. Thus, it is a concern that the overall Type I error rate may not be preserved at a pre-specified level of significance. In addition, p-values may not be correct and the corresponding confidence intervals for the treatment effect may not be reliable. Moreover, major adaptations (changes or modifications) may result in a totally different trial that is unable to address the medical questions that original study intended to answer. 

The FDA classifies adaptive designs as either well understood designs or less well understood designs, depending on whether the adaptations are blinded or unblinded (4). In practice, however, most adaptive designs considered are less well understood designs. As a result, one of the major challenges is the development of appropriate statistical methods under the less well understood designs for valid statistical inference of the test treatment under investigation.


In recent years, many discussions are directed at the flexibility, efficiency, validity and integrity of adaptive trial designs in clinical research. To maintain the validity and integrity of an adaptive design with complicated adaptations, it is strongly suggested that an independent data monitoring committee (IDMC) should be established. The role and responsibility of an IDMC for a clinical trial utilising adaptive design should be clearly defined. IDMCs usually convey very limited information to the sponsors about treatment effects, procedural conventions and statistical methods with recommendations to maintain the validity and integrity of the study.

In summary, from the clinical point of view, adaptive design methods reflect real clinical practice in clinical development, and the flexibility they offer is useful in early clinical development. From a statistical point of view, the use of adaptive methods in clinical trials makes current Good Statistics Practice (GSP) even more complicated and the validity of the use of adaptive methods is not well established or fully understood. In practice, regulatory agencies may not realise that the adaptive design methods for review and approval of regulatory submissions have been employed for years without any scientific basis. However, it should be noted that the draft guidance on adaptive design clinical trials issued by the FDA is currently being distributed for comments, and is to be viewed as a warning document, rather than specific guidance for the use of adaptive clinical trial designs.


  1. Wei LJ, The adaptive biased-coin design for sequential experiments, Annal of Statistics 9: pp92-100, 1978
  2. Chow SC, Chang M and Pong A, Statistical consideration of adaptive methods in clinical development, Journal of Biopharmaceutical Statistics 15: pp575-591, 2005
  3. Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams M and Pinheiro J, Adaptive design in clinical drug development – an executive summary of the PhRMA Working Group (with discussions), Journal of Biopharmaceutical Statistics 16: pp275-283, 2006
  4. FDA Draft Guidance for Industry – Adaptive Design Clinical Trials for Drugs and Biologics, February 2010
  5. European Medicines Agency (EMA), Point to Consider on Methodological Issues in Confirmatory Clinical Trials with Flexible Design and Analysis Plan, CPMP/EWP/2459/02, 2002
  6. European Medicines Agency (EMA), Reflection paper on Methodological Issues in Confirmatory Clinical Trials with Flexible Design and Analysis Plan, CPMP/EWP/2459/02, 2006
  7. Bauer P and Kieser M, Combining different phases in development of medical treatments within a single trial, Statistics in Medicine 18: pp1,833-1,848, 1999
  8. O’Quigley J, Pepe M and Fisher L, Continual reassessment method: A practical design for Phase I clinical trial in cancer, Biometrics 46: pp33-48, 1990
  9. Chang M and Chow SC, A hybrid Bayesian adaptive design for dose response trials, Journal of Biopharmaceutical Statistics 15: pp667-691, 2005
  10. Chang M, Adaptive Design Theory and Implementation Using SAS and R, New York, 2007
  11. Branson M and Whitehead W, Estimating a treatment effect in survival studies in which patients switch treatment, Statistics in Medicine 21: pp2,449-2,463, 2002
  12. Hommel G, Adaptive modifications of hypotheses after an interim analysis, Biometrical Journal 43: pp581-589, 2001
  13. Maca J, Bhattacharya S, Dragalin V, Gallo P and Krams M, Adaptive seamless Phase II/III designs – background, operational aspects, and examples, Drug Information Journal 40: pp463-474, 2006
  14. Tsiatis AA and Mehta C, On the inefficiency of the adaptive design for monitoring clinical trials, Biometrika 90: pp367-378, 2003

Read full article from PDF >>

Rate this article You must be a member of the site to make a vote.  
Average rating:

There are no comments in regards to this article.

Shein-Chung Chow is a Professor at the Department of Biostatistics and Bioinformatics at Duke University School of Medicine, Durham, North Carolina, US. He is currently the Editor-in-Chief of the Journal of Biopharmaceutical Statistics and the Biostatistics Book Series of Chapman and Hall/CRC Press. He is a Fellow of the American Statistical Association, and has authored and co-authored over 200 methodology papers and 18 books, which include Adaptive Design Methods in Clinical Trials and Handbook of Adaptive Designs in Pharmaceutical and Clinical Development. Shein-Chung received his PhD in Statistics from the University of Wisconsin, US.
Shein-Chung Chow
Print this page
Send to a friend
Privacy statement
News and Press Releases

Landmark Partnership Aims to Improve Scotland’s Health

A ground-breaking collaboration between the NHS, academia and industry partners has been launched to co-create NHS Scotland transformation through large scale programmes to improve the health of the population and expand clinical research.
More info >>

White Papers

Crossing the Blood Brain Barrier (BBB): Which Route to Take?

Aptar Pharma

Aptar Pharma recently hosted an expert panel/roundtable on the various approaches to treating central nervous system (CNS) diseases. This international scientific forum was held in Paris, France on December 5th, 2014. The roundtable was organized to explore and exchange views on the challenges, approaches and opportunities that exist with regard to effectively crossing the blood brain barrier (BBB) and successfully treating a host of CNS diseases.
More info >>




©2000-2011 Samedan Ltd.
Add to favourites

Print this page

Send to a friend
Privacy statement