New simulation tool expands the range of eligible trials.
By Deb Borfitz
November 16, 2010 | The first upgrade to a one-of-a-kind simulation tool co-developed by Tessella and Berry Consultants significantly enlarges the proportion of clinical studies that can benefit from simulation of the design. With FACTS 2, 85% of all clinical trials can be productively simulated, says Scott Berry, the consultancy’s president and senior statistical scientist. That’s up from about one-third of trials with the original FACTS (Fixed Adaptive Clinical Trial Simulator), introduced last year, which focused exclusively on single-endpoint Bayesian dose finding and dose escalation trials.
Berry and Tom Parke, head of clinical trial design at Tessella, launched FACTS 2 at last month’s ADAPT 2010 conference in Arlington, Va. Berry Consulting (College Station, Texas) is a statistical consulting group specializing in adaptive clinical trial design and Tessella (Oxford, UK) is a global information technology and consulting services company with clients in the life sciences industry.
After collaborating on trial optimization projects for more than a decade, the firms have created a simulation package that can be used to study a trial design’s operating characteristics at the planning stage. The software allows the impact of different dose response profiles, trial accrual rates, and subject drop rates on the likely trial outcome to be explored.
The user interface for FACTS 2 is much like a conventional PC application, explains Parke, with the user-specified parameters handed to a “statistical powerhouse” that does all the mathematical processing. The program can also be distributed on the cloud to run simulations in parallel, he adds. Summary results are presented in the form of easy-to-read graphs and tables.
Trial design simulation is intended to predict how a study is apt to play out if executed under a variety of potential scenarios and assess the impact of various design features on the study objectives, allowing for the best use of resources and better odds of success. Trial sponsors up to now have been wary of trial design simulation, says Berry, despite positive mentions in Food and Drug Administration draft guidance on adaptive clinical trials issued earlier this year. The vast majority of companies have been designing clinical trials using a classical statistical framework developed over 50 years ago.
Four major pharmaceutical companies are currently using FACTS and two of them are at an “advanced stage,” says Berry. This includes the firm partnering in the tool’s development, which may soon be simulating up to 90% of its drug development portfolio.
New features introduced by FACTS 2 could expand the practice of trial simulation considerably, particularly for oncology and central nervous system studies. FACTS can now simulate dose-finding trials with multiple endpoints, such as studies simultaneously driven by efficacy and toxicity goals. The efficacy value of a drug can be scaled by a safety factor that reduces the utility of a dose if side effects are above a chosen threshold, explains Berry. FACTS 2 will make clinical teams and sponsors “think hard about why they’re doing a trial.” Like the existing FACTS 1 dose-finding simulator, longitudinal modeling of the relationship between subjects’ early responses and their final outcome can be used to maximize the information from subjects that have dropped out or have not yet completed.
Time-to-event trials based on a time duration measurement—e.g. time-to-death and time-to-progression (oncology), time-to-failure (medical devices) and time-to-recovery (post-surgery)—can also be accommodated with FACTS 2. Relative to a quantitative response variable, time to event endpoints are now “fairly common,” Berry says. The new feature makes the potential benefits of early stopping, dropping ineffective arms, and adaptive randomization easy to explore.
Rather than a collection of pre-canned trial designs, FACTS offers sponsors a “list of [pre-programmed] ingredients” they can combine to suit virtually any study circumstance,” says Parke. With the dual endpoint “utility function,” biostatisticians can create entirely novel trial designs, including those combining early and long-term endpoints or a pair of efficacy endpoints. Phase I “combination design” trials that vary the dose of two drugs given in combination remains a candidate for inclusion in a future release of FACTS.
The world of clinical trial design is expanding with the upsurge in personalized medicine, so FACTS will ultimately assist with the design of studies focused on molecular-level efficacy, says Berry. Forthcoming upgrades will also expand the “multiple endpoint engine” to four or five endpoints and simulate designs for trials of a single drug in multiple diseases or tumor types. The latter feature could be especially useful in speeding the development of therapies for orphan diseases. •