Adaptive Design: In Theory and Practice
by Lisa Zamosky
April 2007
Adaptive trial design has the potential to reduce costs, save time and enable greater patient safety … but only if thoroughly planned and properly executed.
Adaptive design refers to a multi-stage trial in which accumulated data is analyzed at pre-determined intervals to decide how to modify aspects of a study—for either safety or efficacy concerns—while the trial is still in progress. This periodic analysis allows sponsors to learn more about the safety and potential benefits of new medicines earlier than they would in some of the more traditional study designs. Adaptive design (in common with group sequential design) also enables sponsors to stop a study for futility when pre-specified conditions under the protocol are met. Additionally, adaptive design can often help identify efficacious drugs and determine effective dosing sooner than with traditional trial designs. For these reasons, adaptive trial design has become a topic of growing interest to many within the drug development community.
Although adaptive clinical trial design is full of potential when carried out well, biopharmaceutical executives making large investments in adaptive trials should engage in thorough prospective planning and remain aware of the complications and pitfalls that can occur with such programs.
Correcting Misperceptions
Some industry experts have expressed concern about widespread misunderstanding of the purpose and scope of adaptive designs. “I think that the concept of adaptive designs may be confusing to many people who are not familiar with all the details and who therefore might view ‘adaptive designs’ as only one special type of design, when there is in fact a whole range of different types of adaptive designs” says Christian Sonesson, Ph.D., senior statistician, regional lead for increased capability of biostatistics, and global lead for the decision analysis team at AstraZeneca, in Mölndal, Sweden.
Many misperceptions of what adaptive design can and cannot accomplish exist within the industry, says Robert O’Neill, Ph.D., director of the office of biostatistics in the Center for Drug Evaluation and Research at the U.S. Food and Drug Administration (FDA), in Rockville, Md., USA. “There is a lot of confusion about what the term even means,” he says. In 2004, the FDA announced its Critical Path Initiative, a program intended to modernize the process through which drugs, biological products and medical devices are moved from the discovery phase through to marketable, medical product. To that end, the FDA announced in July 2006 that it is in the early stages of developing guidelines for adaptive trial design.
“The guidelines would try to define in a relatively clear way what we all mean by adaptive design,” O’Neill says. “We want to discourage non-planned adaptations, which is an issue among those who have heard third-hand about adaptive design. We have recognized that there may be some opportunity with the use of this design. We also recognize that there may be an overselling of opportunities as to where it may be useful. We’re concerned about a lot of false starts and false hopes, particularly among executives who are not close enough to the action to know that this is not the key to their salvation,” O’Neill says. “We’re all interested in trying to put a measured approach to this area.”
Look Before You Leap
When set up in the best possible way, adaptive designs often would provide great advantages, according to Jonathan R. Smith, Ph.D., vice president of the strategic biostatistics unit at Quintiles Transnational, in Durham, N.C., USA. When adaptive design is appropriate, in some cases it can be set up so that, on average, time and money can be saved, Smith says. “In other cases it can be set up to increase the information that a trial provides.”
Appropriateness is the key. Understanding when not to use adaptive design is as important as knowing how to use adaptive design. Adaptive designs are particularly useful in trials aimed at identifying appropriate doses, especially when little is known about the drug, Sonesson says. In this type of trial, different doses of the investigational drug are compared to one or two controls. “In a traditional design we would generally allocate the same number of patients to each dose,” Sonesson says. “In an adaptive trial, on the other hand, we could learn about the dose-response relationship for the experimental drug during the trial, both with respect to effect and safety. Then we have the possibility to stop arms for which the effect is insufficient or safety inadequate.”
In such a scenario, new patients entering the study can be randomized only to the doses showing greatest promise. Not only can this save time, it protects patients from being randomized to treatment arms with drug doses that are ineffective or present safety issues. However, in cases in which the effectiveness and safety of a drug dose can be predicted with a high degree of certainty, the value of adaptation is low, Sonesson says.
In the future, a high proportion of Phase II trials are likely to be adaptive and will enable sponsors to gain more information about the optimal dose and regimen from a single Phase II study without any increase in the number of patients, Smith says. “This, in turn, will increase the chance of success in Phase III, because such a Phase II adaptive trial makes it more likely that the most appropriate dose and regimen will have been selected for Phase III,” he says.
There are ways to set up adaptive trials in any phase so that, where advantageous, an early stop due to a demonstration of efficacy—or an early stop for futility when it has been determined that the trial has a very low chance of being positive—can be built in. This can save time and money, Smith says. “In early phase trials, it is possible to incorporate very many different types of adaptation within the same trial,” he says. However, in Phase III confirmatory trials, there are more limitations because all adaptations need to be pre-specified in the protocol, and if too much adaptation is incorporated, then the trial would not be viewed as “confirmatory” by the regulators. “We always advise our customers that it can’t be done in every trial,” Smith says.
Blurring the Lines
Adaptive designs have been used for quite some time in drug trials, particularly in the areas of oncology and HIV/AIDS. What is new, however, is combining trial phases, including Phases I and II, as well as Phases II and III. In the case of a Phase II/III seamless study, investigators are able to achieve the objectives of a learning, dose-finding trial (Phase II) with a confirmatory trial (Phase III) within the same, continuous study. This allows trials to be defined by the information obtained along the way, rather than by distinct phases.
“In a Phase II/III trial, the idea is to eliminate the traditional concepts of different phases in clinical development and make a transition into a continuous flow from ‘learning’ to ‘confirming,’” Sonesson says. “The first part of the trial is used to select the appropriate [drug] doses to use in the second and confirmatory part of the trial.”
A Phase II/III trial accelerates drug-to-market speed by removing the lead time between phases. This is the main benefit of combining the two phases, Sonesson says. “Typically, this could be one to two years of development time,” he says. “Part of the time savings comes from having just one clinical study protocol.”
Although adaptive trial methods can lead to significant advantages, careful planning and operational management are critical to success. “If [an adaptive trial] tells investigators or the sponsor too much, or you inadvertently unblind someone or change the nature of the population, it compromises the integrity,” Smith says.
Having particular kinds of knowledge in advance is critical to the adaptive trial’s set up, according to Smith. For example, when beginning a seamless Phase IIB/III trial, the set of doses and regimen to be included must be identified up front, because this type of design does not allow investigators to add new arms—either dosage or regimen—given the confirmatory nature of such a trial.
It’s All in the Setup
With all the talk of time and cost savings associated with adaptive designs, it’s easy to overlook the extensive up-front planning required in order to carry out a well-planned trial, particularly in the budget planning stage. The investment that trial sponsors need to make early in the development of a Phase II/III trial places new demands on the business with respect to how resources are managed.
During a seamless trial, additional prospective work is conducted because all adaptation must be pre-specified and all adaptive designs need to be carefully evaluated in advance. According to Smith, “this evaluation will generally look at factors such as false positive rate … expected minimum and maximum number of patients; and expected minimum and maximum duration of the trial; as well as demonstrate gains over traditional trial designs.” This assessment often will lead to the adaptive design being refined to more closely meet the sponsor’s needs, and all of this needs to be carried out before the trial starts, he says.
In evaluating seamless Phase IIB/III trials, review boards use pre-defined criteria to make determinations about which doses are to be carried into the next phase or whether the trial should be stopped after the first stage. “This requires that you specify ‘go/no-go’ criteria for the second stage as well as the dose selection criteria prospectively,” Sonesson says. “This is important in order to preserve the blindness to sponsors and thus protect the integrity of the trial.”
The Way Forward
In order to be effective, adaptive trial designs must be considered within a larger context. “To obtain maximal gains from adaptive designs, it is important to consider the needs of the whole clinical development program,” Smith says. “Very different types of adaptive designs will be needed in the learn stage compared to those that will be needed in the confirm stage.” Additionally, any gains that adaptive design would bring to a clinical development program need to be carefully assessed and quantified, Smith says.
Since there is a constant need to optimize clinical programs based on available information regarding effect, safety and the medical need in society, a broader focus beyond adaptive design is needed, Sonesson says. “In a clinical development program, there are numerous decisions that must be made under conditions of uncertainty,” he says. “I think we should strive for structured and quantitative thinking from the first time the drug enters a human body to the launch of a new drug.”
Adaptive design possibly could help to increase the success rate of some clinical trials, O’Neill says. “Whether or not that actually occurs remains to be seen.”
Web Extra: Reducing White Space
Seamless Phase II/III adaptive trial designs present logistical and statistical challenges that necessitate substantial up-front planning and judicious implementation. When adaptive design is applied appropriately to a drug development program, the potential savings in cost and time are undeniable.
Adaptive seamless designs ideally reduce the time to plan and implement the next clinical phase, often referred to as the “white space” between two phases of a trial, says Frank Bretz, Ph.D., senior expert statistical methodologist at Novartis Pharma AG, in Basel, Switzerland. Seamless designs save costs by combining two studies, and they also can increase the value of the information obtained during the trial. Seamless Phase II/III adaptive trial designs may be of great value when developing targeted therapies for oncology, Bretz says.
“Developing a therapy targeting one or more biochemical pathways suggests that a specific sub-population of patients is more likely to achieve response to treatment,” Bretz says. A more efficient approach than traditional trial designs is a seamless Phase II/III adaptive design to combine the two study objectives—the selection of either the full or a pre-specified sub-population, and proof of efficacy—into a single two-stage study, he says.
“All stakeholders are interested in an efficient drug development process, such that an effective new treatment can be made available as early as possible to the patients,” Bretz says. “It is our opinion that carefully planned and conducted studies based on adaptive or adaptive seamless designs are an important tool, which can help to fulfill these requirements.”
Adaptive designs alone, however, will not solve the pharmaceutical industry’s problems of increasing inefficiency and cost in the drug development process, Bretz says. “While adaptive designs can contribute to the solution and in many instances are substantially more efficient than standard clinical trials, they will not solve the problem on its own. A lot of work remains to be done beyond adaptive designs.”
Send us Your Comments
Web Exclusives
- Statistical Advantage
-
Gary Shorter, director of biostatistics for Quintiles, talks to Envisage editor Adam Istas about the growing importance of incorporating biostatistical analysis into all stages of drug development.
- Measure for Measure
-
A proposed Health Care Comparative Effectiveness Research Institute in the United States may force the biopharmaceutical industry to emphasize patient outcomes and quality of life issues when developing a new medicinal product.