Why Clinical Trial Modernization Fails Without Point-of-Care Data: Q&A with Richard Graham, PhD, TruTechnologies

National initiatives are accelerating, artificial intelligence (AI) is reshaping drug discovery, and regulatory pressure for real-time oversight is mounting, but none of it will move the needle, argues Richard Graham, if the industry continues to build faster roads to the same broken bridge.

To explore this further, Applied Clinical Trials recently spoke with Richard Graham, PhD, co-founder and chairman of the board at TruTechnologies, about the site-level execution failures he has watched repeat across his career, why poor data quality places a hard ceiling on what AI can deliver, and what genuine modernization of clinical trial infrastructure actually requires.

ACT: In what ways has clinical trial execution failed to modernize over the past six decades, and what has kept legacy problems in place for so long?

Graham: While no single issue explains why trials fall behind, there's one that I've seen plague the industry since I worked on the sponsor side, accountable for clinical trial execution that largely happened out of my control. Fragile manual processes at clinical sites have compromised even the most promising science, and there's been resistance to taking a good, hard look at these processes and putting the effort into future-proofing them.

Why? At best, modernization requires activation energy. Clinical operations teams are under enormous pressure to start and finish trials as quickly as possible, all while staying within limited budgets. Adopting a different approach requires them to step back from those immediate pressures, think through how the change would work in practice, and bring along company stakeholders.

At worst, though, parts of the clinical trial ecosystem may benefit financially when studies take longer to complete, creating little incentive to challenge the status quo.

But these manual processes take enormous effort to maintain, too. If we want to get life-changing drugs into the hands of patients, remain competitive on the global stage in drug development, and also protect bottom lines, we have to confront the operational issues that have been slowing trials down from the start.

The clinical trial industry is much like a 60-year-old home. It's badly in need of repair, and the homeowners are faced with two different routes: build new additions onto the existing structure, or tear it down and rebuild it from the ground up.

From my point of view, the clinical trial system is so outdated that there's no point in adding on new technology or policy if it doesn't address the foundational issues of execution. That is easier said than done when so many stakeholders—including sponsors, sites, regulators, patients, payers, CROs, central labs, investors, and boards—are optimizing for different outcomes, sometimes in direct opposition to one another. Everyone has a different idea of what the remodel should look like, turning modernization into a tug of war rather than a coordinated effort.

But the stakes are too high to not look at the problem holistically. For one, patients are badly in need of the drugs that are in development. And the delays that keep those drugs out of reach also carry a significant economic cost. According to a 2024 paper from Tufts, each additional day can mean approximately $500,000 in lost prescription drug sales. In effect, our inability to fix these bottlenecks is leaving an enormous amount of money on the table that could be reallotted to further research and development, not to mention driving up the costs for patients.

ACT: What do national-level initiatives like Operation TrialBlazer signal about the industry's willingness to finally address site-level execution gaps?

Graham: The HHS-wide Operation TrialBlazer initiative is in line with something I have been an advocate for during the majority of my career. If we want to stay competitive in drug research on the global stage, we have to make our country an inviting and attractive place to conduct clinical trials.

Alongside the FDA's Real-Time Clinical Trials Pilot Program announced a few months ago, you can feel the momentum picking up and the industry waking up to the need for modernization. TrialBlazer is a definite step in the right direction, and it shows that exact willingness you mentioned, but I do want to call attention to some important aspects of the problem that it doesn't necessarily address.

At a high level, TrialBlazer is about making early development in the United States easier to navigate, with a proposed expedited IND pilot, clearer Phase I expectations, and more direct FDA support for sponsors preparing first-in-human studies. This is critical, as sponsors don't choose trial locations based upon patriotism. They make these decisions in line with which country has clear requirements and predictable processes, as well as where reliable human evidence can be generated efficiently.

Faster IND pathways, real-time review, and AI-enabled decision support could all make the country a more welcoming location to run trials. However, none of this will truly move the needle if we don't fix how trial data is collected, monitored, and made trustworthy as a clinical trial is executed day-to-day.

Data can't continue to be captured late, cleaned up after the fact, or looked at only when problems have already occurred. If any newly introduced policies and technologies are built on top of current practices, we've just created a faster road to the same exact broken bridge.

ACT: What are the most consequential analog, human-prone errors happening at the site level today, and how do they propagate through a trial?

Graham: This goes back to the fragile chain of manual steps I mentioned earlier. At many sites, protocol-required work is still documented on paper and entered into electronic systems later. That leaves sponsors with limited visibility into what is happening as the trial is executed and little ability to intervene when something goes wrong. A mislabeled collection tube, a swapped sample, or an incorrectly recorded timepoint may not be identified until the error has already moved downstream and become much harder, or even impossible, to correct. This is not a criticism of the hardworking individuals executing trials at the site level; they are doing the best they can with processes and tools that leave too much room for preventable error.

But during my time working on clinical trials, I saw firsthand how quickly these site-level errors could become much larger problems. While I was working on two investigational drugs in parallel, the FDA intervened during review of the applications. This is a drug developer's nightmare. Our data was called into question because the electronic records for the biological samples didn't match what had been reported in the clinical study reports.

When we traced the problem back to its root cause, all signs pointed to those manual processes that had created inconsistencies between the site records and the data included in the regulatory submissions, making the data untrustworthy. When all was said and done, we lost significant time tracing the discrepancies, the review process was delayed, and both approvals were put at risk.

That is how seemingly small site-level errors become serious trial-level problems. Across multiple sites and regions, what begins as an isolated issue can quickly become a much larger pattern that affects trial integrity. I've seen this time and time again with biological samples: issues emerging too late, after patients have already sacrificed so much of their time, blood, and hope to advance science and medicine. We have to do better.

ACT: How does poor data quality at the site level undermine the promise of AI in drug development?

Graham: My biggest concern is that we are asking AI to transform drug development while relying on data produced by a clinical trial execution system that remains error-prone. That places a fundamental limit on how much AI can do to accelerate the development of life-saving therapies.

There's no denying that AI is generating experimental drugs faster than ever at the discovery stage. But that progress doesn't address the clinical trial execution problems that can delay or undermine the evidence needed to advance those candidates. The fact of the matter is, no AI-discovered drug has reached FDA approval yet, and overall failure rates remain near 90%.

HHS says innovation has to be rooted in the US from the earliest stages. But top academic medical centers still send letters to technology vendors—one from a site running over 500 protocols—saying they'll only accept paper requisitions. You can't root innovation in sites that are declining it in writing.

A simple fact remains: more candidates moving faster into clinical trials doesn't fix the failure rate. It just means more volume hitting a clinical trial model still built on 60-year-old infrastructure. Speeding up discovery without fixing data quality and site-level execution just moves the bottleneck; it doesn't remove it.

AI can support faster, better decisions as clinical trial data are generated, but only if those data are captured accurately at the point of care and made available while there is still time to act. If we skip that foundational step, delayed, incomplete, or inaccurate data will continue to undermine trial execution, and AI will remain limited by the information it receives.

ACT: What does meaningful modernization of site-level execution actually require to get right?

Graham: Meaningful modernization is rooted in making sure the biological samples collected in trials, and the associated data, can be trusted. That is our exact mission at TruTechnologies and a piece of the puzzle we believe is crucial.

This means defining "real-time" data collection as data collected at the point-of-care, not hours or days later, which increases the probability for error or mismatched records. Our proprietary technology, the TruLab platform, is doing just that. Across more than 3,300 clinical sites in 60 countries, it has demonstrated up to a 97% reduction in source data errors, a 98% reduction in sample-related queries, an 85% reduction in sponsor time spent on sample tracking, and the potential to reduce time to market by up to three months.

Getting this right also requires recognizing the ethics behind poorly run trials. Bad data collection doesn't just hurt the bottom line for sponsors or cause strife for those working at the sites. It also, ultimately, affects the patients that sacrifice their time and take on risk to advance science and medicine.

One important issue missing from many modernization initiatives is patient transparency. If sponsors choose legacy, manual sample management processes despite the availability of validated technologies that substantially reduce preventable errors, Institutional Review Boards should consider whether that operational decision belongs in the informed consent process. Patients deserve to know when avoidable operational choices may increase the risk that the samples they provide will be lost, mislabeled, or otherwise rendered unusable.

Every stakeholder deserves a better way to understand what is happening in a trial while there is still time to act. Achieving that will require thoughtful policy aligned across a complex stakeholder ecosystem, but it will also demand that we fundamentally rethink how clinical trials are executed.