Administrative Analyses for Funding Decisions in Adaptive Clinical Trials
Navigating Barriers to Innovation in Adaptive Trial Design
Introducing innovative trial designs into clinical research has historically encountered complex barriers. Evidence from the Adapt-IT project, where Berry Consultants and peers integrated adaptive methods into neurological emergency trials sponsored by the NIH and FDA, highlighted that statisticians were key obstacles to innovation in the mid-2000s. Though this is less the case now, adaptation to new statistical methods remains challenging when conventional standards predominate.
Today's principal barrier is seldom a lack of technical solutions but can be an absence of agreement about trial objectives and definitions of success. It can be challenging to create a more efficient design when the goals are not well defined. Stakeholders frequently resort to the “I’ll know it when I see it” fallback—leaving “success” undefined and precluding analytical rigor. This lack of specificity prevents effective simulation and objective comparison between different designs. A common example of this is the ability to capture what would be needed early in a trial to trigger continuing the trial – engaging full funding to complete the development. Berry Consultants tries to solve the I’ll know it when I see it – by showing many examples of possible trial results. By employing modeling, simulation, and hands-on examples, allowing clear communication, can bring quantitative clarity. By presenting scenarios, simulated datasets, example outcomes, or graphical data summaries—such as Kaplan-Meier survival plots or scatterplots of observed effects—stakeholders are compelled to articulate what outcome crosses the threshold for success and, by extension, further investment.
Seamless Design in Practice: Statistical and Operational Impact
In an example development program for a rare disease, the Phase II trial could specify enrolling 30 patients on a control and on each of two doses, followed by a Phase III of 90 on the selected dose versus 90 control, often with a year of operational “white space” between them while data are analyzed, protocols created, and regulatory interactions proceed.
A seamless Phase II/III adaptive designs could be set up by combining the dose-finding and confirmatory trials. After enrolling 30/30/30 patients and selecting a preferred dose, teams immediately transition to randomizing 60/60 more patients with a primary analysis combining each stage, creating a combined 90 versus 90 sample size for the primary analysis. Alpha adjustment is necessary to account for dose selection; established statistical methods provide this correction to control Type I error. What distinguishes this approach is increased efficiency: the seamless design can provide higher statistical power, despite 210 total patients compared to 270 for the separate trial approach. The operational benefit is equally clear—median time from initiation to final readout is greatly reduced – saving the time to enroll 60 less patients and avoiding the white.
Yet resistance remains, especially among smaller biotech sponsors. Even with scientific superiority, seamless designs are often rejected because these sponsors lack resources to commit beyond an initial cohort; they must obtain subsequent funding before proceeding. This is a practical, not scientific, barrier—a funding problem encountered across the sector. There is a deemed a need to expose the results of the phase II trial to obtain funding for the phase III trial, hence forcing the larger trial and timelines.
Administrative (Financial) Analyses: The Objective Funding Solution
A solution to this funding bottleneck centers on administrative (financial) analyses. These are checkpoint assessments, not interim analyses in the traditional (operational) sense. The difference is critical: administrative analyses are designed only to inform potential investors or funders about trial status related to their own objective criteria, never to alter the design of the trial.
Operationally, these analyses are conducted by an independent statistical group—sometimes an external CRO—tasked with evaluating only whether predefined criteria have been reached. The results, usually in the form of a “checkbox” (yes/no) per criterion, are shared only with C-suite figures for the sponsor and the investing party. This rigorous firewalling from operational and clinical teams maintains trial blindness, eliminating operational bias concerns and preserving trial credibility.
Care is required in protocol terminology. Administrative analyses should not be labeled as “futility analyses.” Such mislabeling risks confusion with efficacy/futility analyses intended to affect trial continuation. Accurate distinction protects both regulatory posture and scientific integrity.
Defining Criteria: Predictive Probability and Modeling
The most common statistical metric for these administrative analyses is a Bayesian predictive probability of trial success. Predictive probability quantifies the chance, given the current data and all its uncertainty, that the final analysis will meet significance. This is a dynamic measure, updating in real time—much like win probability metrics in televised sports, which continuously adjust as the score and time remaining evolve. This contrasts with traditional statistical “power,” which remains constant for a given design and effect size.
For funding partners—particularly venture funds, private equity, or large pharmaceutical collaborators—predictive probability is an immediately relatable, risk-driven metric. They routinely use such dynamic forecasts in their own fields; seeing a predictive probability of success at an interim point enables invest-or-wait decisions, quantifying risk and potential reward.
When stakeholders are slow to quantify objective criteria, simulation becomes a valuable communication the bridge. Simulation allows the design team to present realistic data sets, effect matrices, and visualizations, directly asking: “If the results looked like this, would you fund the next phase?” Eventually, a pattern emerges, and quantitative decision thresholds can be written into a funding trigger.
Terminology can be problematic as the same concept can be interpreted differently to different teams. Terms like “target product profile” (TPP) can cause confusion; some stakeholders consider it the effect size for powering a study, others view it as the minimum necessary for commercial competitiveness or funding readiness. Appropriate quantification of funding rules, without these confusing concepts are critical.
Conclusion
Seamless adaptive trial designs deliver quantifiable gains in power, efficiency, and speed to answer, but practical funding constraints often stand in the way. Objective administrative (financial) analyses—using clearly modeled, Bayesian-driven criteria such as predictive probability—enable funders and sponsors to define explicit, actionable go/no-go funding triggers outside traditional interim analyses. Through simulation, visualization, and meticulous information governance, these analyses make innovative trials viable for organizations constrained by uncertainty or resource gaps. This empirical approach, rooted in clarity and analytic rigor, can be central to modernizing clinical development.