Adaptive designs are "a contradiction to the confirmatory nature" of Phase III clinical trials, EMEA's Committee for Medicinal Products for Human Use says in a reflection paper.

As such, modifications to the design for a Phase III trial or a late-stage Phase II trial that is part of a confirmatory package are "rarely acceptable" and must be justified in the study protocol, CHMP concludes. Their number should be limited, it adds.

Adaptive design guidance under development at FDA also would require trial protocols to describe potential changes. "The way we're defining adaptive designs for regulatory purposes, all the potential modifications need to be thought of and pre-specified in advance," Office of Biostatistics Director Robert O'Neill said in an interview with "The Pink Sheet."

He agreed that the number of design changes should not be unlimited. "What is reasonable, I don't know. But certainly after you talk about more than maybe three modifications, you're probably talking about getting into an exploratory study more than a confirmatory study."

An adaptive design "should not be the everyday trial," O'Neill said. There is a greater risk to this design "because much more upfront planning has to occur relative to a traditional design. ... It's not a matter of whether you shouldn't do it. It's a matter of whether the trial is going to give you the information that you need."

Interim analysis allowed by an adaptive design can provide data to refine design assumptions or objectives, he noted. It provides the "ability to interate on what might have been a very poor or very imprecise guess earlier on because you didn't know enough about the disease or the drug at that point in time."

FDA remains on track to release the draft adaptive design guidance by the fall of 2008, according to O'Neill (¹ Also see "Late-Stage Trials Pose Challenge As FDA Molds Adaptive Design Guidance" - Pink Sheet, 16 July, 2007., p. 18).

The EMEA reflection paper, which discusses methodological issues arising from adaptive design of confirmatory trials, "reflects on regulatory thinking about obstacles to interpretation" of adaptive trial results, Spiros Vamvakas, Acting Deputy Head of EMEA's Scientific Advice and Orphan Drugs Sector told "The Pink Sheet." These obstacles are particularly important in late-stage drug development, he said.

Although not official guidance, the reflection paper, to some extent, clarifies the circumstances in which adaptive trials can be used and how to justify their application, Vamvakas pointed out. Regulators already are seeing proposals that "more carefully address the critical points."

Thus far, EMEA has not approved a drug based on a late-stage adaptive trial, he said. Adaptive trials have been part of successful applications to individual countries, but these have made "only limited use of the potential flexibility of these designs," so practical experience in applying this design is still limited.

It is best to use adaptive designs in difficult experimental situations, according to the reflection paper. Such situations exist if placebo response is difficult to predict, small populations or orphan diseases limit the amount of evidence that can be collected through other trial designs, and there are ethical constraints on experimentation.

Justification For Design Modifications

To justify adaptive designs, EMEA says, a sponsor needs to outline the limitations in knowledge that are driving the potential design modification, explain why an anticipated change can safely be incorporated into the study and not endanger the interpretability of trial results, and discuss why the outstanding issues should not be answered through more exploratory studies.

EMEA does not anticipate adaptive designs that allow changes in primary endpoints that are selected to describe a clinically relevant treatment effect. "It is difficult to imagine any situation where the perception of what constitutes a relevant clinical benefit should change based on interim results."

In a confirmatory setting, the agency says, "the mere rejection of a global hypothesis combining results from different endpoints will not be sufficient as proof of efficacy."

EMEA's concerns about adaptive trials include whether analysis of their data can produce a true picture of efficacy.

At a minimum, the agency calls for using statistical methods that provide full control for a potential false positive. In addition to presenting the P-value, the methods also must estimate the size of the treatment effect and provide confidence intervals with pre-specified coverage probability.

If there are different scales for measuring the effect of treatment on a certain endpoint, the preferred scale is the one readily interpretable for clinicians.

With interim analyses, the various trial stages can have markedly different estimated treatment effects, EMEA points out. So "studies with adaptive designs need at least the same careful investigation of heterogeniety and justification to combine the results of different stages as is usually required for the combination of individual trials in a meta-analysis."

Depending on the design modification, demonstration of a convincing treatment effect may require more than the rejection of a global null hypothesis across all stages of the trial.

As with the design modifications, EMEA says analytical methods and issues should be addressed in the planning stage.

Because it is difficult to demonstrate that unblinded interim results have not been released, interim analyses "always introduce the possibility of damaging the integrity of a trial," the reflection paper says.

To minimize this problem, EMEA says "interim data should be provided by an independent statistician to an independent decision-making committee ... sponsor involvement is discouraged."

A late-stage Phase II trial could be combined with a Phase III trial when the dose is already established and the two studies would be performed in the same patient populations with the same endpoints, EMEA says.

But, it suggests, "Sponsors should consider carefully the value of independent replication of results and the value of waiting to confirm the detailed design of the confirmatory study until all important exploratory data are available."

When two Phase III trials are the usual basis for drug licensing, EMEA suggests that a combined trial usually is not acceptable, but two independent combination trials could be. For indications where a single confirmatory trial might be acceptable, a Phase II/III trial, without other supporting evidence, would be insufficient.

In the case of orphan diseases, a Phase II/III trial may be a more efficient method of obtaining information from a limited number of patients and permit more extensive dose-response studies, EMEA says.

The reflection paper and issues surrounding adaptive designs will be discussed at a Dec. 14 workshop in London sponsored by EMEA and the European Federation of Pharmaceutical Industries and Associations.

- Cathy Dombrowski (c.dombrowski@elsevier.com)

EMEA Looking At Limited Use Of Adaptive Design In Late-Stage Trials

More from Archive

More from Pink Sheet