Pharma forecasting: The use adaptive designs in clinical research
How adaptive design can reduce development times for drug and provide earlier long-term safety data
By Lisa Comarella, Luca Girardello, and Stefano Vezzoli
Over the past few years, we have seen an exponential increase in the cost of research and development of new drugs, with the average R&D expenditure for a new drug doubling in the last decade.
This increase in costs, combined with a slower approval process, has resulted in a drastic reduction in the percentage of new drugs that are successfully brought to market.
According to pharmaceutical companies, the two most problematic aspects for clinical studies are excessive study lengths and the high probability of failure.
Often, in cases of failed studies, the conviction of those involved is that if some aspects were managed differently, the program could have been saved and the potential of the therapies demonstrated.
Consequently, the question pharmaceutical companies continue to face is: How can we make a study more efficient while maintaining its validity and integrity?
Sponsors, clinical researchers, and biostatisticians are becoming more interested in designs with greater flexibility than standard ones and in procedures that anticipate ‘go/no-go’ decisions.
In light of these objectives, it is of the utmost importance to make modifications to a study while it is in process on the basis of new information derived from accumulated data.
Adaptive designs are able to cope with exactly these issues.
Adaptive design: The basics
Adaptive design consists of many different possible approaches; some simple and relatively common, others more sophisticated and—in certain respects—controversial.
All of these designs, however, have in common the use of collected data to modify several aspects of a study already in progress—without compromising its validity or integrity.
It should be emphasized that modifications are not a remedy for inadequate planning, but must be pre-defined and justified in the study protocol.
The possible modifications include changes in sample size, in the criteria for inclusion/exclusion, in the doses or treatment regiment, in the study endpoints, in the elimination or addition of treatment groups, or in the early closure of the study for efficacy/futility.
Statistical methodologies permit the application of these modifications to a study in progress, while keeping under control the probability of error associated with multiple hypothesis testing.
The European Regulatory Authority (EMEA) has addressed adaptive designs in the 2007 document Reflection Paper on Methodological Issues in Confirmatory Clinical Trials Planned with an Adaptive Design.
Depending on the type of changes foreseen, adaptive designs can be classified into group sequential design, designs with numerous adjustments (N-adjustable design), adaptive seamless Phase II/III design, adaptive randomization design, and others.
While it is possible to re-estimate the number of patients needed for a study based on observed data in interim analysis, it is less widely known that design can allow objectives to be achieved in one study that would normally require the scheduling of two distinct studies in phase IIb and III.
An adaptive study in phase II/III combines two sequential and separate studies into one study and allows the use of collected information in the first stage to adapt the design in the second phase.
The advantages of this design are a reduction in the overall development time for a drug, fewer patients required, and the early availability of long-term safety data.
Case study
As an example of the application of designs in phase II/III, let’s consider a recent study.
In the first stage, corresponding to phase II, there are three treatment arms included: two different experimental treatment doses and standard treatment.
In the second stage, corresponding to phase III, only two treatment arms are continued: the experimental treatment dose, selected based on observed results in the first stage, and standard treatment.
The selection of dosage depends on the observed success rates in the treatment groups at the end of phase II, according to the decision rule defined.
In case of a lack of efficacy of both doses of the experimental treatment, it is possible to stop the study.
It is clear in this example that adaptive designs have flexibility.
Based on our experience, a pharmaceutical company considering adaptive designs should
- Budget more time for the planning of an adaptive design as compared to a standard design
- Interact with regulatory authorities in the planning phase, especially for phase II/III studies
- Define the decision rules for interim analyses and provide statistical justification to support the study design
- Use simulations to calculate the power of the sample size and the probability of success in the study
- Evaluate whether to stop patient recruitment during the interim analyses
- Consider the use of electronic data capture to manage data efficiently and rapidly for the interim analyses
- Use independent statisticians to perform the interim analyses and independent committees to review the results of the analyses
- Schedule frequent monitoring visits to provide as much data as possible for the interim analyses
- Inform in advance those responsible for supplying the drug of the type of design, and take into consideration, for randomization, the possibility of an interactive centralized system.
Adaptive designs have the potential to change the way in which we conduct clinical research, with the understanding that the flexibility in comparison to traditional designs requires a greater effort in the initial planning stage of the study.
Lisa Comarella is a senior statistician and head of methodological biostatistics at international contract research organization CROS NT in Verona, Italy. Luca Girardello has been a statistician at CROS NT since 2007 and Stefano Vezzoli has been a senior statistician at CROS NT since 2008.
For more on forecasting, join the sector’s other key players at Forecasting Excellence USA on October 4-6 Boston.
For an overview of eyeforpharma’s forecasting coverage, see ‘Highlights from eyeforpharma’s Forecasting coverage’.