________________________________________________________________________________________________________
Instead of leaving feasibility in the early stages of clinical trial planning, a longitudinal, continuous feasibility method will ensure any unplanned changes to the protocol are dealt with swiftly and thoroughly.
Aaron Mackey at Lokavant
Space research, like clinical research, is highly risky. A lot can go very wrong, very quickly. When NASA plans to launch a spaceship, engineers carefully create a flight plan with a calculated, multi-step trajectory: the initial rocket thrust required to achieve exit velocity; a clever slingshot around the moon to gain some additional acceleration; a reverse thrust burn to slow the approach; and so on.
But during the voyage, NASA engineers also know there will be many smaller details to monitor, and certainly some necessary adjustments to make small engine burns, course corrections, vehicle altitude, or orientation adjustments. NASA anticipates these midflight modifications, even without knowing exactly what they will entail or what specific challenges might necessitate them. NASA labels these 'anticipatory work' and breaks them down into two groups: long-term and real-time.1
This anticipatory planning is also reflected in the overall budget -- for instance, the amount of reserve fuel kept aboard the ship -- and the expectations of everyone involved. Prepping for the unknown, closely monitoring in flight progress and executing minor 'just in time' adjustments are essential for NASA to be successful. NASA's preemptive engineering is also the key component to clinical trial success.
Houston, we have a problem
Initial feasibility analysis for a clinical trial begins at the start of portfolio planning, often referred to as 'strategic feasibility'. Typically, this is a preplanning, thumb-in-the-wind exercise, or a gut check: 'Could we enroll this trial within six months, given our budget? No? Then how about in five years? Definitely. Okay then, what Is a realistic target window for completed enrollment narrower than six months and five years? At what cost? With a regulatory plan that supports our market strategy?'
Then the questions get more complex as the team moves from strategic to 'operational feasibility':
lnitial goals and constraints around total costs, enrollment time, site selection, and collected endpoints will all impact enrollment expectations and inherent probabilities for success.
They are also different levers that study teams can pull to adjust during the initial planning phase -- often leaving one aspect 'better' at the expense of the others.
To increase the probability of success, you may need to reduce the number of patients, plan for a bigger recruitment budget to achieve a shorter timeline or instead extend the timeline to allow an existing budget to do its work. If this is the most important trial in a sponsor's portfolio, the plan may require a probability of enrollment success of 95% or higher for it to be considered feasible, so what adjustments will be necessary to hit that goal? Study teams must work the available levers and find an optimal combination of adjustments, given the various goals and constraints. With the right balance of conditions, feasibility is agreed upon and a trial moves forward.
Do it early & often - iterative feasibility analysis
Traditional feasibility study processes are standard fare for any clinical study today. The problem: it’s too often a one and done exercise, a point in time in which all these factors are considered statically, without the 'What if?' planning and foresight that NASA lives and dies by. Once an initial feasibility analysis and operational planning are complete, feasibility is rarely revisited to a meaningfuI degree while a trial is ‘in flight'.
We're in a time where that must change. According to a 2022 study by the Tufts Center for the Study of Drug Development, the prevalence of phase 1 to 4 clinical trial protocols with at least one significant protocol amendment -- many designed to improve enrollment -- has climbed from 57% to 76% since 2015.2 Plus, the average number or amendments per protocol increased 60% from 2.1 to 3.3.2 Redoing a feasibility analysis at an appropriate point mid study -- either one or several times, depending on the protocol amendments and challenges -- could be the critical step in keeping a trial on track, and avoiding expensive and time-consuming protocol amendments.
Right now, when there are small course corrections needed (ie, opening new sites; launching a new patient outreach campaign; engaging a centralized patient registry), they're often identified too late, or without clear direction on how to handle them in a way that will keep a trial from failing to meet its goal, as 80-90% of trials do.3
There are incredible amounts of additional cost at each step, so by the time a clinical trial is in the clinic, it's at its most expensive, and time-consuming. Effectuating change early on and identifying risks and operational issues in a particular plan early -- notably, in the development of the protocols stage can not only cut costs, but also accelerate the trial. Otherwise, when a study team encounters an issue, such as an amendment, the trial may get delayed anywhere from six months to a year.
Artificial Intelligence (AI) can start processing a protocol amendment as it is being drafted, especially identifying potential impediments to hitting clinical trial milestones.
One significant impediment is inclusion and exclusion criteria for trial participation; Al can run models based on historical real-world data to determine if it’s feasible to recruit 500 patients based on the protocol's strict participation criteria across these 13 countries.
Further, AI analysis early on can evaluate the effect of the protocol on the capability of the site to run the trial. Will there be compliance issues and deviations? How much of a patient and site burden does the protocol create -- for instance, what's the impact of its requirements for different clinician assessments and endpoint measurements at the site on the patient -- and is it realistic to expect patients won't drop out? With the advent, and expansion of, large language models and modern versions of Al, it is much easier to process these potential, seemingly endless combinations of inputs-outputs, and predict probability of success with greater accuracy much sooner in the process.
Consider this common scenario: A year and a half into a trial, six months from the expected close, management wants to understand the trial's status and if it's on track. 'Not good,' says the operations team -- enrollment numbers are off and we're going to run six months over. Had a feasibility revisit been triggered sooner, perhaps more budget could have been allotted, the time frame extended and additional sites opened to help meet the goal. Like our NASA analogy, engines could have been burned, and some course corrections made, to reach the target successfully.
This is not to say the operational experts behind trials aren't vigilant in their monitoring of recruitment progress and data review during trials. However, they may not be using the natural milestones in the life cycle of a trial to regularly re-examine the plan with an updated forecast, such as when a study expands into a new country, or when that country is granted regulatory approval sooner or later than originally expected. They also may not be responding quickly enough to red flags, or have the proper tools to know what the red flags really are.
Prescriptive and causal Al for continuous, iterative feasibility
Current AI-fueled technology platforms don't understand red flags correctly. For example, some sites recruit more participants in a month than projected, others less. The next month could be different. Many predictive tools inadvertently identify decreases in rates -- spiky behaviour that's part of the natural ebb and flow of recruitment -- as red flags. Many of the tools on the market also don't properly account for realities like site exhaustion, when there are simply no more trial participants left to recruit in each area. Most are, essentially, just scenario forecast planners -- plug and chug calculators. Staff enter the numbers for what's expected in a trial, click a button, and get an answer about what that scenario equates to in terms of cost, time, and probability of success.
To conduct continuous feasibility analysis in a way that matters, operations teams need better solutions than just basic forecasting tools with the potential for false alarms. They need to see the projected particulars of what would happen in alternate scenarios, combined with informed guidance on what to do when advance tooling detects things are beginning to veer off course, or when a new factor emerges that wasn't there when feasibility was first examined. Causal Al, a prescriptive form of AI, provides recommendations based on a combination of historical real-world data and current, fluctuating trial data in real-time so teams can take remedial action before trials veer off course. What if a competitor suddenly enters the picture after a trial is underway? What if an opportunity arises to end a trial six months earlier? What would that look like? People don't want to go to the moon anymore -- they want to go to Mars. How would the plan change? These kinds of insights can lead operations teams to think of scenarios they might not have otherwise considered -- ones that could mean the difference between success and failure because they'll be getting more of the 'why'. They also provide reliable, data-driven guidance to junior study staff; a key advantage, especially amidst current industry-wide staffing challenges.
Most importantly, it's time to start looking at feasibility analysis as an ongoing requirement, an iterative process rather than just a box to check once at the beginning of a trial. And just like NASA, the advanced preparation and planned mid-journey checks and adjustments will rocket clinical trials to success.
References:
Aaron Mackey PhD, senior vice president of Al and Data Science at Lokavant, has held various tenure track faculty research positions in academia, including the Center for Public Health Genomics at the University of Virginia and the University of Pennsylvania Genomics Institute, both US. Additionally. Mackey has held various scientific and leadership roles at GlaxoSmithKline, Hermoshear Therapeutics, Covance/Labcorp, Valo Health, and Roivant Sciences. Most recently, he served as vice president of Data & Al/Ml at Sonata Therapeutics. At Lokavant, Mackey leads the company’s Al and data science team and plays a critical role in the company’s product development.