Clinical Researcher—August 2025 (Volume 39, Issue 4)

PEER REVIEWED

Avery M. Davenport, MPH, CCRC

 

Healthcare has always been slow to adapt to change. When other industries appear to be on time or at least close to catching up to the latest trend in the market, rest assured, healthcare is about a decade behind. However, one area where healthcare cannot afford to lag behind the times is the use of digital health—specifically, artificial intelligence (AI). Whether it is using AI to read an X-ray or to help dictate a physician’s progress note on a patient during follow up, the conversation cannot be ignored. Those who fail to enter the discussion will be left behind.

The slow adaptation to using AI in clinical research is understandable. With strict regulatory bodies, an “at no risk” approach, and worries about safety, compliance, and being sued, the fears surrounding AI are clear. Integration of AI into clinical trials is imperative, but it faces some cultural opposition. The good thing is that resistance to change can…well…change.

The Only Constant is Change

The lack of the kinds of changes necessary to adopting a more forward-thinking approach to utilizing AI will suppress the progress of clinical research. Clinical trials must change as technologies change. AI offers optimal potential to navigate the change we see within the industry and within healthcare in general.

The Cost

Biological technology—like oncology medications and medical technologies such as surgical devices—makes up most of clinical research. These two product categories are what comes to mind for most people when we talk about clinical trials, and they represent huge businesses. It will cost about $3 billion, on average, to get a novel drug completely through development, compared to lesser but still considerable costs for devices (about $2 million to $10 million for Class II devices and $10 million to $50 million for Class III devices.{1} It is no secret a lot of money is poured into trials and innovation. Using AI can help tamp down rising expenses by allowing complex processes to be more time- and cost-efficient.

Data Overload

With massive amounts of data being generated in healthcare—from protected health information (PHI) in electronic medical records (EMRs) to pain scores being recorded in healthcare system–designed phone applications—clinical trials only generate more data while adding to the complexity of information to which clinicians and researchers have access. Utilizing AI in clinical trials can help possibly navigate adverse events or even optimize drug dosing.

Woes of Eligibility

The most promising area for applying AI in clinical research is patient recruitment. Staring at a screen while manually pre-screening patients takes forever. Honestly, we are at the time now where we should be spending more time with patients and enrolling more of them, not spending 20% to 30% of our day or week checking if patients meet eligibility criteria. AI has the potential to match those patients, based on their EMRs, directly to the inclusion and exclusion criteria, as laid out in the protocol.

Optimizing Your Trial Design

Everyone loves adaptability and—better yet—adapting in real time. Who wants to wait to adapt? Wait too long, and it is time to adapt again. Unfortunately, trial designs have historically been very “templated”; you build a design, perform the study, copy that design, perform another study, and repeat. That sort of copy and paste design could lead to an ineffective arm or subgroup being included into a clinical trial that maybe should not have been included. Driving AI into trial designs could potentially correct this during the design itself. It has the potential to catch these archaic designs during a pilot and may suggest alternative cohorts to include in the larger funded study. The end result could possibly be cutting more costs and reducing timelines.

Never-Ending Regulatory and Compliance Pains

Even the folks at the U.S. Food and Drug Administration (FDA) are begging for the use of AI. Okay…maybe not begging, but they are open to AI-driven approaches in clinical research. With high costs, lengthy timelines, and patient recruitment difficulties, it seems the FDA is shifting toward this sort of recommendation. As recently as 2023, the FDA had issued guidance outlining possible frameworks for establishing AI and machine learning (ML) as part of standard practice for drug and device development. The recognition that falling behind in using this technology reveals a fear that it could mean non-compliance or missed opportunities for faster approvals, not to mention patient suffering.{2}

The Competitive Edge is the Sharpest

As the saying goes, “If you aren’t first, you’re last.” The same is true with adapting technology. Over time, it is going to be easier to predict that late adopters of AI are going to lose market share and damage their reputations for innovation leadership in their field. A lot of people like to sit back and watch how something new works before investing in it, but how long will companies sit back and watch competitors streamline trials more effectively, reduce more costs, and accelerate into the market more quickly?

Action Leads to Reshaping

The AI healthcare market is undergoing predictable, exponential growth due to its ability to address critical inefficiencies in healthcare, specifically in clinical trials. The market is predicted to reach $187 billion to $674 billion by 2030–2034, with AI in clinical research dominating due to rising investments in drug discovery and the need for faster, more accurate trial outcomes.{3} The industry is aware of the issues we are facing—data organization/analysis, patient recruitment, trial design, drug/device discovery acceleration, compliance/regulatory guidelines, market competition, and cost…cost…cost….

Delaying integration risks widening the gap between those inventing and those on the sidelines. Competitors will leverage AI for faster, cheaper trials. Moreover, patients—especially those with urgent needs—bear the cost of delays, as prolonged trials postpone access to therapies. We want personalized medicine to better our patients. We have a way to do it. Embracing AI now aligns with FDA encouragement and market trends, ensuring trials are more efficient and patient centered.{1}

Follow the “Ethical” Brick Road

Now what? The challenges are many and easy to discuss, but the solutions may be more reclusive. One of the easiest ways to start getting buy in would be to address the PHI issue in healthcare—more specifically, keeping patient data in a secure and private environment. Companies like Unlearn.AI and Medable are already implementing solutions for clinical trials, following FDA urgency in this area.{4} There are several approaches that should be considered and hopefully can serve as models moving forward:

• AI models can be trained on specific hospital servers without transferring PHI. This is possible for local hospital systems and systems that are more geographically spread. For example, NVIDIA’s Federated Learning Application Runtime Environment (FLARE) platform enables collaborative learning for clinical trials, preserving privacy while leveraging diverse datasets. This would be useful for major EMR systems like EPIC or Athena when shared work needs to be done without sharing PHI and risking exposure of actual datasets.{5}
• Speaking of EMRs…data sharing needs to be normalized across all EMRs. There must be communication on some level. Standardization of fast healthcare interoperability resources (FHIR) would make communication seamless. Systems like HL7 FHIR would help support AI applications in clinical research while reducing nearly all privacy risks.{6}
• Blockchains are another solution but may be poorly understood. When a decentralized technology is used to record transactions between multiple computers, you essentially have created a blockchain. The catch is, the movement of data from one “block” to another cannot be altered…unless the previous “block” has been altered. It is intense data security and relies on complex algorithms. Companies like BurstIQ are using this technology to manage trial data.{7}

References

1. Johns Hopkins Bloomberg School of Public Health. 2018. Cost of Clinical Trials for New Drug FDA Approval are Fraction of Total Tab. https://publichealth.jhu.edu/2018/cost-of-clinical-trials-for-new-drug-FDA-approval-are-fraction-of-total-tab
2. U.S. Food and Drug Administration. 2023. Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence/Machine Learning (AI/ML)-Enabled Device Software Functions. https://www.fda.gov/media/167069/download
3. MarketsandMarkets. 2024. Artificial Intelligence in Clinical Trials Market to Hit US$2.74 Billion by 2030 with 12.4% CAGR. https://www.prnewswire.com/news-releases/artificial-intelligence-in-clinical-trials-market-to-hit-us-2-74-billion-by-2030-with-12-4-cagr–marketsandmarkets-302466093.html
4. World Health Organization. 2021. Ethics and Governance of Artificial Intelligence for Health. https://www.who.int/publications/i/item/9789240029200
5. Pew Research Center. 2023. How Americans View Data Privacy. https://www.pewresearch.org/internet/2023/10/18/how-americans-view-data-privacy/
6. U.S. Food and Drug Administration. 2023. Using Artificial Intelligence and Machine Learning in the Development of Drug and Biological Products. https://www.fda.gov/media/167973/download
7. IBM Security. 2024. Cost of a Data Breach Report 2024. https://www.ibm.com/reports/data-breach

Avery M. Davenport, MPH, CCRC, (adavenport@mypainsolution.com) is Director of Clinical Research at Commonwealth Pain & Spine and Chair and Founder of the ACRP Kentucky Chapter in Louisville, Ky.