Change is Constant
by Lisa Zamosky
April 2008
To fully optimize adaptive clinical trial design, the proper technological tools must be in place.
Photography by Raimund Koch/Riser/Getty Images
Adaptive trial designs are a relatively new trend in clinical trial methodology and have sparked a growing interest throughout the biopharmaceutical industry. An adaptive trial design allows the original hypothesis of a study, and therefore, the study’s course, to be modified as data are collected and analyzed. These trial designs offer drug companies an alternative to lengthy and exceedingly expensive traditional studies, which in recent years have failed to produce the volume of innovative treatments commensurate with their costs.
The U.S. Food and Drug Administration (FDA) has estimated the cost of bringing a new medicine to market in the range US$0.8 to $1.7 billion, a 55 percent increase over the last five years. At the same time, the costs to develop treatments are on the rise, and the number of treatments produced has declined. It’s no wonder the industry is busily searching for ways to alter traditional methods of developing and ultimately delivering new drugs to market. Adaptive trials, in theory, offer the promise of drug development at a significantly reduced cost because of increased efficiency with respect to their limited use of both time and patients.
An adaptive trial is pre-planned to be altered based on what is learned along the way. As a result, many who champion these designs believe they are ethically superior because adaptive studies more quickly steer patients away from treatments that interim analyses show aren’t working. They typically require the use of fewer patients and much less time than traditional trials. Adaptive trials also benefit companies looking to make wiser investments by helping them realize sooner which compounds have no chance of success.
For many drug companies, effectively implementing adaptive trial designs is dependent upon a number of factors, including a major shift in organizational mind-set and a streamlining of existing business processes. And in a setting where researchers’ abilities to make critical decisions quickly is dependent upon access to current and accurate data, technology is key to transforming the way the industry conducts clinical research.
To Tech or Not to Tech
With all the talk in the industry of electronic data collection (EDC) and interactive voice response systems (IVRS)—two primary technologies that are important to carry out adaptive trials—one might assume that without the will and deep pockets to invest in such technology, implementing trials using adaptive designs would be out of the question. Not so, says Ranganath Nayak, Ph.D., chief executive officer of Cambridge, Mass., USA-based Cytel Inc. Nayak points out that many companies conduct smaller Phase I and early Phase II studies in which proof-of-concept and safety are established almost entirely manually.
“When people say you must have EDC or IVRS, to an extent, they are raising the bar unnecessarily high,” Nayak says. “They are discouraging the experimentation with adaptive trials by making them feel that unless you invest in a lot of technology and pay a ton of money, you can’t do adaptive. It’s not true.”
However, technology does become imperative, Nayak acknowledges, during later, more confirmatory stages in which trials are making dose-finding adaptations, for example, that need to be implemented globally across multiple sites.
Real-Time Equals Real Progress
EDC systems, which collect patients’ clinical data directly from the investigative site electronically, enable researchers to look at data in real time, at least in principle. “EDC helps you make turnaround times more efficient and include more data in your interim reviews,” says Zoran Antonijevic, senior director, strategic development, biostatistics at Quintiles in Durham, N.C., USA. “It helps you decide what kind of adaptation you’re going to make and do analysis more quickly.”
Because adaptive trials require a continuous evaluation of data, with decisions being made quickly to adjust a number of study elements, EDC systems enhance the quality of the data and therefore, the decision-making processes associated with adaptive designs.
“When we say we want to adapt,” says Dr. Michael Krams, assistant vice president of adaptive designs at Wyeth in Collegeville, Pa., USA, “we want to adapt to information that is accruing. If information accrues more slowly, we can’t be as efficient in reacting to the latest level of understanding of the data that’s out there,” he says. “The more timely the capture, the better.”
According to Krams, EDC systems have the potential to offer a secondary gain to the overall set up of a trial as well. Sponsors have the ability to provide feedback to investigators about the speed they enter data at each of the study sites.
The Low-Tech Side of High-Tech
As with all technology, if the users are not on board, it cannot serve its highest purpose. For drug companies attempting to shift from retrospective to prospective decision-making through adaptive trial designs, operational and personnel issues inevitably come into play.
According to Krams, EDC may be the enabler, but without investigators actually plugging in the data in a timely fashion, the technology will not live up to its potential. “It sounds like a trivial point, but in a number of trials we are running, we are looking at whether we have 100 percent of all available data, but initially only a fraction of that rate is available—as low as 20 percent—due to delays in data entry and other reasons.”
With traditional clinical trials, investigators have much more time to enter clinical data. Feedback about missing data may not come back to a clinical team until the end of the trial, at which point they would begin chasing down that information. With adaptive trials, the earlier the data are available, the better. According to Krams, Wyeth is aiming to have all the data in the database no later than 48 hours after an observation was made. “The technological enabler needs to be coupled with the monitoring and investigator attitude to get the data in there,” Krams says. “It’s not good enough to just have the technique sitting there.”
“What we always have in the forefront of our mind,” says Tom Parke, head of clinical trial solutions with Tessella, in Oxford, England, “is in order for the adaptation to work, the people who are supposed to be sending us the data have got to actually do it. It’s worth making the system particularly convenient for them, so you [lessen] the obstacles of participating in the trial.”
In cases where an EDC system that allows the clinical team to extract real-time data as the trial progresses is not available, Parke suggests considering the most convenient way of capturing data, and that will of course depend upon the circumstances in which the measurements are being taken. “If the data are being collected in hospitals and being written down on some kind of form first and then faxed off, for example, it’s actually more convenient than using a computer. If you’re using patient-recorded outcomes, then let them use the form or SMS (short message service or text messaging) because that’s normally the most convenient for the patient to use,” Parke says.
IVRS: Making Adaptation Possible
IVRS performs a variety of patient and clinical trial material-related functions. Patient randomization and dosing, as well as trial material forecasting, are all functions where researchers rely on IVRS when conducting adaptive trials. Algorithms are incorporated into IVR systems that allow adaptations to be automatically implemented, thereby protecting the integrity of the trial by eliminating the need to unblind study personnel. If an interim review indicates, for example, that one treatment arm of a study should be dropped, an IVRS would have the ability to implement this adaptation and adjust enrollment targets and drug supply needs for the other arms automatically.
“IVRS is required in particular if we want to do an adaptive dose-ranging study,” Krams says. “If we want to have a highly dynamic reallocation to different doses where the proportions of patients being allocated to different treatment arms changes from week to week, that would be very difficult to achieve if we didn’t have IVRS.”
According to Parke, you can divide adaptive trials into two distinct camps, defining them primarily by their statistical approach. One kind—frequentist methodology—incorporates just a few interims during the trial where only a small number of adaptations are made at each interim. In such cases, a review board will meet to look at the current accumulating data, and based on pre-defined decision rules, make a decision on the conduct of the study.
The second type of statistical approach for analyzing data from adaptive trials is Bayesian methodology, in which adaptations are made on a continuous basis, for example, every randomization or every week. Because this involves a more automated approach, one where algorithms have been programmed into a “black box” working with the IVRS, reviewers may meet less frequently and primarily to look at safety data.
“In this second case,” says Parke, “it’s absolutely vital that the statistical model for the adaptation is coded up in software and you use it to run the trials. You don’t want to have a manual intervention every time you have to make an adaptation. You don’t want to have to unblind too many people.”
Executing patient randomization and dosing in adaptive trials is a critical function of IVRS, but researchers must rely on these systems in conjunction with EDC for the real-time capture of safety and adverse event, lab data and much more that IVRS alone cannot provide.
According to Nayak, the integration of various technologies, including EDC, IVRS and others used during adaptive trials, such as e-diaries, is not as great a challenge as it might seem. “At the moment it’s happening because the sponsor says ‘I’m going to work with this company on the IVRS and this company on the EDC and you do the adaptive stuff. I want you guys to [all] work together for the next three months to figure out the technology integration and make sure it will be ready at the time we launch the trial.’”
According to Parke, like many companies, Tessella has run a number of trials where multiple parties have been involved, and they’ve been expected to integrate the data coming from various sources. “Actually, that integration is one of the easier parts of getting the system up and running.”
New Challenges for Supply Management
One of the critical functions of IVRS is its use in supply management forecasting. Adaptive trials present a particular challenge for supply personnel, given that the number of treatment arms and study patients, as well as dosing levels, will all be changing depending upon what the accruing data indicate as the trial progresses. The pressure on the supply management team in adaptive trials, as a result of these dynamic conditions, can be extreme.
“You can see that by almost continuously making adjustments to what doses may be assigned to patients at the next stage, you run into a potential problem of either not ordering enough medication for each potential scenario or you can have a serious overage and the study can become very expensive,” Antonijevic says. “This is where optimizing drug supply is very important.”
According to Nayak, statistical simulation provides companies an opportunity to plot out more accurately the amount of medicine that will be required. During a clinical trial testing biological medicines, for example, where the cost per patient can be as high as US$10,000, “you can’t say ‘we’ll just order five times as much as needed,’ which is what can happen in a dose-finding adaptive trial if you use conventional approaches to supply planning,” says Nayak, who adds that Cytel is being asked to do simulations for supply management with increased regularity.
Data Management: What is Clean Enough?
With adaptive trial designs, discussions of interim analysis usually lead to questions about the level of required data cleanliness. There seems to be a general consensus among the experts that more data is preferable to relying exclusively on clean data.
“Making adaptations where you’re adjusting the randomization between different dose arms, it is much better to get your data in quickly than to have it scrupulously cleaned,” Parke says. Once the data are entered, he says, they can be replaced with clean data as the trial progresses. This is especially effective in continuous adaptation trials, rather than those that have only a specified number of interim analyses. “So you’re growing all your clean data, and it’s only the stuff you’ve collected recently that’s not that clean,” Parke says.
Krams agrees. “Of course cleaner data are always better than not so clean data, but no data is much worse than having data that is not entirely clean.” Krams says his approach is to always use all data in the learn trial, with the data cleaned as quickly as possible. “If there initially have been unclean data, we’ll substitute that with the clean as soon as possible and build safety margins so we don’t let the system adapt to anything early on when it is very sensitive to potentially misleading input.”
The Way Forward
As adaptive trial designs continue to take hold throughout the industry, their success will no doubt rest on thoughtful planning and further integration of clinical technologies. “From my 35 years of consulting,” Nayak says, “it’s a mistake to think that technology alone is going to be the answer to one’s prayers. You really need to work on both the technology and the organizational side. Streamlining a lot of the organizational processes is absolutely critical to making the technology work.”
Send us Your Comments
Web Exclusives
- An Unbalanced Burden
-
A Q&A with Dr. Kathleen Squires, director of the infectious disease department at Jefferson Medical College in Philadelphia, illuminates the gender differences in the diagnosis and treatment of HIV.
- Turning the Corner
-
After decades of neglect, leading to a rapid increase in incidence, tuberculosis started gaining attention from industrialized nations over the past 10 years. With this new focus, TB’s incidence has started to decrease—but barely.