Adaptive Design Clinical Trials for Drugs and Biologics

Name of Guidance
Adaptive Design Clinical Trials for Drugs and Biologics

Status of Guidance
Draft

When was the Guidance released?
February 2010

Which organization released the Guidance?
Center for Biologics Evaluation and Research (CBER) and Center for Drug Evaluation and Research (CDER)

Target Audience
Drug company sponsors who design and/or run clinical trials, and FDA staff who review clinical trial data

Laws and Regulations Referenced
21 CFR 314.126: Drug sponsors must conduct adequate and well-controlled studies to demonstrate the safety and efficacy of new drugs before they can be approved for marketing. This includes having a clear objective for the study, comparing the new drug to a valid control drug or treatment, choosing appropriate participants, assigning participants to a treatment or control group in an unbiased way (usually by random assignment), minimizing bias wherever possible, and using valid methods to observe and analyze response to treatment.

Definitions

• Adaptive design clinical study: A clinical trial that is structured with prospectively planned changes that usually go into effect after an interim analysis of data.
• Bias: A tendency to design a study in a way that favors a particular outcome, or a tendency to interpret data in a way that overstates the effect of a drug. May lead to Type I errors.
• Blinded analysis: A study of data collected during a clinical trial where the personnel doing the analysis does not know which participants received which drug or treatment. Important for avoiding bias in the study results.
• Endpoint: An event or outcome that marks an important change in a disease or condition due to treatment. For example, an endpoint in a clinical trial could be improvement or worsening of symptoms, or a blood test showing the drug has had some effect on the body. Most clinical trials have both primary and secondary endpoints.
• Interim analysis: A study of data collected during a clinical trial before the trial is complete.
• Primary endpoint: The outcome believed to be most useful and reliable in determining whether a drug is effective against a disease.
• Prospective: Planned in advance. In the Guidance this refers specifically to planning modifications in one or more aspects of a clinical trial design before the data is unblinded to personnel making the modifications.
• Protocol: The “recipe” for running a clinical trial. Includes descriptions of who is eligible/ineligible to participate in the trial, how participants must be assigned to receive drugs being tested, dose and delivery method of the drug, how participants should be monitored, how long participants should take part in the trial, etc.
• Secondary endpoint: An outcome believed to demonstrate the effect of a drug, but may not be as definitive as the primary endpoint.
• Type I error: A statistical term for believing a drug has an effect on a disease or condition when it actually does not.
• Type II error: A statistical term for believing a drug has no effect on a disease or condition when it actually does.
• Unblinded analysis: A study of data collected during a clinical trial in which the personnel doing the analysis knows which participants received which drugs or treatments. Considered to be an unreliable method of analysis because it introduces bias and increases the risk of committing Type I errors.

Background
Sponsors who run clinical trials must determine the design of each trial (including goals, protocols, and methods for collecting and analyzing data) before testing begins. By defining the design up front and adhering to it closely through the conclusion of the trial, the sponsor is more likely to get valid and meaningful results.

However, there can be benefits to adapting the design while a clinical trial is still in progress. For example, if an interim analysis shows convincing evidence that a drug is effective the sponsor may be able to reduce the duration of the trial and make the drug available to the public faster. Or the sponsor may be able to get more useful data on the effect of the drug by adjusting the dosages given to participants. In order to make trials more efficient or informative, sponsors should use an adaptive design that provides flexibility without detrimental effects on the results.

This Guidance outlines approaches to adaptive design clinical trials that can help preserve the integrity of trial data.



Summary
Clinical trials (also known as clinical studies) are necessary to determine if new drugs are safe and effective enough for public use. In addition to FDA regulations for adequate and well-controlled studies drug sponsors aim to follow principles for good statistical and clinical practices. Everything is planned carefully in advance. Once a trial starts, it is critical that the protocol and planned methods for gathering and analyzing data remain unchanged. Any revisions that are made along the way can introduce bias and compromise the outcome of the trial.

However, there are times when a clinical trial will benefit from having some flexibility in its design. For example, if a sponsor realizes 8 months into a 2-year trial that a drug is not effective, it would make sense to stop the trial early. Or if a sponsor is trying to establish that a drug is effective across several races but the participant pool for the trial is disproportionately composed of people from one race, it would make sense to change their recruiting or eligibility criteria so the participants are more racially diverse. Even if a sponsor has the best intentions when making changes, if not done carefully the trial results can be called into question. In cases like these sponsors need guidance on how to introduce flexibility without compromising quality. Trials that are planned to be flexible are known as “adaptive design clinical studies.”

The Guidance starts by describing general concerns associated with using adaptive design clinical trials. Specifically it addresses the increased risk of committing Type I statistical errors, added difficulty in interpreting positive study results, limitations introduced by adaptive design, and the need for more planning and coordination with FDA.

The Guidance then addresses both familiar and less familiar methods for adapting trial designs. The familiar methods are well established and therefore thought to be less risky. The less familiar methods, however, should be used with caution; the number of trials in which these forms of adaptation have been used is too low for CDER and CBER to determine whether the benefits outweigh the risks. Irrespective of which method a sponsor chooses, all changes must be planned prospectively to be considered valid.

Finally, the Guidance outlines statistical considerations for adaptive design clinical studies.


Rationale
Straying from a rigid, conventional clinical trial design can introduce bias, increase the risk of making errors (particularly Type 1), and cast doubt on the validity of the results. Adaptive designs allow sponsors to make preplanned revisions. Ultimately this can make clinical trials more efficient, successful, and informative. However, some methods of adaptation are more familiar than others. The Guidance is needed to help sponsors understand the different ways they can incorporate flexibility in their clinical trial designs.


Resulting Recommendations
These recommendations are meant for trials in which preplanned revisions occur before data is unblinded to the persons planning or making the revisions. Unplanned revisions made after unblinding to these persons raise concerns about study integrity and should be avoided.

Familiar approaches to adaptive design (well-established, relatively low-risk)

• Modify eligibility criteria to attract participants more quickly or round out the existing participant population with desired baseline characteristics
• Increase sample size to achieve desired study power
• Conduct an unblinded analysis if participants are experiencing negative dose-related issues (such as a serious side effect due to high doses) so they can discontinue treatment
• Discontinue a study if interim data shows the drug has little or no benefit
• Make limited changes to the statistical analytical plan (SAP) if study data makes the planned SAP invalid, but only if blinding has been meticulously maintained

Less familiar approaches to adaptive design (not well-studied, risk level is unclear)

• Discontinue treatment with dosages that are shown to be ineffective or unsafe after an unblinded interim analysis
• Adjust the method by which participants are assigned to receive treatment so that while the assignments are still random, more people are assigned to the treatment that has the best response data collected to date
• Increase the size of the participant pool if an unblinded interim analysis shows the effect of treatment is smaller than expected but still clinically relevant
• If baseline characteristics (eg, genetic or physiologic differences) are thought to cause different responses to the drug, either change the eligibility criteria so only participants with the preferred characteristic are enrolled or limit final analysis to include data from these participants only
• Change the order of the primary and secondary endpoints if they were not well understood when the trial was planned and the data collected in the trial warrants the change

Changes that are not considered adaptive design (should be avoided)

• Revisions that were not planned prospectively
• Revisions based on information from an external source

Impact
The Guidance will help drug sponsors run clinical trials more efficiently and effectively, increasing their chances of having a successful application to market a new drug or biologic product. Specifically, sponsors will be able to design trials that can be modified at preplanned stages with less risk of compromising the integrity of the data and any conclusions drawn from the data.