Lights, Camera, Adaption

Adaptive is no longer the future—it's happening now. Companies such as Pfizer, Merck, Novartis, and Wyeth are going full steam ahead, not just doing designs, but implementing them. Several have appointed a VP-level person to drive the adoption of these methods.

Ranganath Nayak

The FDA and EMEA are also much more receptive than they were a couple of years ago. FDA has appointed a senior person, Dr. Sue-Jane Wang, to lead the review of adaptive design submissions. Adaptive conferences and Web seminars are proliferating. Suddenly, every consultant or vendor to the biopharm industry is an adaptive expert.

However, the picture is a shade different from what people had projected two years ago. Then, much of the focus was on adaptive Phase III trials as well as on seamless Phase II/III trials. Those are, indeed, being proposed, and some are under way. But the real action is in Phase II dose-finding trials and even in Phase I trials for safety. There are several reasons:

• The FDA's concern about the potential for bias is much less with these trials than with Phase III registration trials.

• Adaptive trials are most helpful when you do not know enough at the beginning of a trial to design it with confidence. This is more likely to be true in Phases I and II than in Phase III.

• Getting the dose right through well-designed Phase I and Phase II trials is the best way to maximize success in Phase III, which then leads to a higher rate of NDAs.

Figure 1

Issues and concerns

Do we know how to deal with the statistical issues? The answer is an unequivocal "yes," as long as you work with experts in adaptive design methods.

Can we "sell" adaptive to the regulators? Our experience is that you can, as long as you can show that there is a really good reason for using an adaptive design and thoroughly evaluate several options, including a nonadaptive one, and then submit the details of your statistical work as part of your protocol submission. When it comes to Phase I and II adaptive trials, regulatory acceptance is much less of a hurdle than in Phase III studies.

Are there safety issues to be considered? Safety issues can never be dismissed, but it is worth noting that in most adaptive designs, the number of patients on doses that go to market is almost never reduced, and is often increased. That means that the amount of relevant safety information almost never diminishes.

Will sponsors accept an independent decision-making body? It is probably hard to accept that decisions with major economic consequences will be made by someone outside the sponsor company, but it may be necessary to do so to satisfy regulators that the possibility that the trial will be biased is minimized. The FDA's position on this is evolving, and it may accept a "Chinese Wall" approach in certain instances, where a statistician from some other part of the sponsor's organization sees the unblinded data and recommends what to do next. This is again less of an issue for Phase I and Phase II adaptive trials.

Will site investigators and IRBs accept the design? It's hard to generalize, but if you discuss this with them early and figure out what their concerns are, the answer is probably "yes."

Will patient consent be easy to get if they know the trial is adaptive? Yes. You just have to explain again why this will not hurt their interests and may help. There is a question, though, whether patients will wait to enroll in a trial knowing that they are more likely to get the right dose if they wait.

Can we get the data needed for the adaptive decision in a timely manner? That depends on two things. One, can the patient end point be observed reasonably quickly, and two, are there adequate EDC and data management systems in place so that data for the adaptive decision are available? If it takes a long time to observe the end point, then adaptive trials will only work if a validated surrogate end point is available. Phase II trials of modest size can be run without sophisticated EDC and data management, but that will add to the cost. For Phase III, EDC and first-class data management are essential.

Is the adaptive change to a trial going to be easy to implement at all the sites? That is a multifaceted question. One facet is: Can we drop doses easily? The answer is "yes," if you have centralized randomization. You just assign a probability of zero to the dose in the randomization system, and the dose is no longer allocated. Another facet is: Can we enrich subpopulations? The answer is "maybe," depending on whether the sites have the capacity to do this or not. Careful selection of the sites will be necessary. A third facet is: What will the investigator need to be paid if the number of patients to be recruited is indeterminate until the adaptive change is made? Again, there is no general answer. You will need to negotiate the solution and possibly see some of the economic savings from adaptive trials evaporate. When this happens, it is important to remind yourself that the primary motivation for doing adaptive trials is not to save money, but to heighten the chances of eventual market success and to get there faster. The economic value of that far outweighs any cost savings or increases.

How do we budget for an adaptive trial? Good question! Imagine that the trial design team says, "The number of patients will be between 500 and 750, with an expected value of 600." What number do you budget for? Practice will evolve here, and as experience is gained, statistical tools will be developed to guide budgeting. Quite likely, sponsors will need to maintain a central pool of funds into which savings from adaptive trials flow and from which monies go to adaptive trials that incur added cost beyond the original estimate.

Can we manage the medical supply logistics? The concern here is overages and stockouts. Overages have generally been considered unimportant, but managing supplies for an adaptive trial in the conventional manner could lead to overages of as much as 500%. At this time, the only way to keep the probability of a stockout low without having massive overages is to simulate the supply system ahead of time and use the simulations to optimize the system of manufacturing campaigns, inventories, re-order points, re-order quantities, and shipping methods. Good simulation tools are now available, but their ability to optimize is still low.

In the future, it is likely that supply optimization tools will be developed, and this will be a boon to the industry's ability to systematically conduct complex adaptive trials with ease. It is also likely that some of the vendors who supply clinical trials will create technological innovations in manufacturing, packaging, and labeling.

What are some best practices?

• Always consider the adaptive alter-native, whether in Phase IV, III, II or I.

• In particular, use adaptive methods to get the dose right in Phase II trials.

• Consider the more classical group sequential design in any Phase III trial.

• Compare the costs of the adaptive and conventional alternatives, but also compare the probability of success.

• If you choose to do adaptive, do a thorough analysis of the statistics and be clear about how the adaptive change is going to be made, and by whom.

• Make this thorough analysis a part of your protocol submission to the FDA.

• Do thorough planning. Get a cross-functional team together that includes representatives of medical supply, manufacturing, packaging and labeling, biostatistics, project management, and the therapeutic areas, and charge them with doing the planning together. Do not leave the logistical and project management issues as an afterthought. With adaptive designs, the design is not complete until you have figured out how to supply the trial.

• In particular, simulate the trial ahead of time: accruals, randomization, response, and medical supply.

It's time to act

One thing is clear: If you're planning a clinical study without considering an adaptive design as one of your options, you are not getting good advice. It's

not good enough anymore to talk about the difficulties as though they were reasons for not doing adaptive trials. It's time to figure out how to overcome those difficulties. It's time to move, to learn by doing. Those biopharmaceutical companies that do this right will be the winners because their return on investment in drug development will outstrip the industry.

Ranganath Nayak, PhD, is chief executive officer at Cytel Inc, 675 Massachusetts Avenue, Cambridge, MA 02139, email: info@cytel.com