Clinical trials are required for any new treatment to reach the market and help patients. However, the clinical research industry has a huge problem: it’s not very good at finding the patients needed to participate in the required studies.
How bad is this problem? It’s pretty bad: It is estimated that 80 % of trials are delayed due to challenges enrolling participants, and 15-20% of trials report not being able to enroll participants at all, meaning these studies cannot take place. One barrier to enrolling participants is that patients have to read through lengthy, complex and jargony documents in order to look for trials and in order to consent to be part of the study if they identify one that is suitable. Physicians, who have the expertise to understand trials and advise patients, lack the time to consider them for every patient that could benefit as the information is lengthy and wieldy to review.
Recruitment delays are costly, ultimately driving up drug prices and resulting in treatment innovations taking longer to reach patients who need them. Technology, specifically AI, can be our greatest asset in democratizing trial information, increasing access to clinical trials, making trials more representative, reducing trial costs and accelerating the pace of innovation in medicine.
Trial Enrollment Issues
To understand why this is a problem let’s start by understanding clinical trials and their role in medicine. Clinical trials are the critical step in generating evidence for an investigational treatment to get regulatory approval. Bluntly put, clinical trials are when new medicines are tested on people for the first time and these people are closely monitored to see whether 1) the medicine is safe 2) it works and 3) it works better than what is the standard treatment today. To answer these questions, clinical trials are run on participants that are closely observed for months or years.
The first step of a clinical trial is to recruit and enroll participants. This process has 3 main issues today:
- Knowledge & Awareness Challenges: Not many people outside of medical professionals even know about trials as an option with 50% of adults reporting they have never heard of the concept and less than 25% of patients reporting hearing about trials as an option from their physicians. From the perspective of medical professionals, 80% report that they believe patients benefit from clinical trials but doctors do not have the time to look into them for their patients as tools available to do so are cumbersome. In essence, many people do not even know to look for trials and the majority do not have the time and/or knowledge to do so.
- Representation Issues: To really answer the question, “is this medicine safe and efficacious for all”, the trial also needs to enroll representative populations. Unfortunately, this is not standard practice today. For example, African Americans make up 13% of the US population, but only 5% of clinical trial participants. If we also consider that in many indications the burden of the disease is greater in this population, and outcomes tend to be poorer, it is a major oversight not to collect data from trials for patients who need it the most.
Without this representation we introduce populations excluded from trials to unnecessary risk. One of the examples of the risk that can be introduced is Plavix, an anti-platelet drug. Plavix was approved but was not tested extensively on a representative and diverse population. After it was approved and in the market, it was eventually discovered that 50% of Asian and 75% of Pacific Islander patients’ do not have a functional enzyme needed to activate the drug.
Unequal Access- Healthcare already has an accessibility and affordability problem in many markets. The current state of clinical trial enrollment is yet another example of health inequity. Currently access to trials is unequal and restricted to those with access to doctors at research hospitals or who have the right network to help them find a trial. “VIP” trial search concierges charge thousands of dollars for bespoke services to do this. The fact that patients need privilege to be able to find a trial results in unequal access to the latest advances in medicine, which can be life-saving treatments.
To summarize, currently clinical trials face great challenges with recruiting participants with their current approach, meaning they are delayed, that we lack representative populations on trials, and that we reserve trials to those who are privileged or lucky enough to find one.
AI Democratizing Clinical Trials
So how can AI help? First, let’s walk through the common ways of finding a clinical trial today. If you as a patient or medical professional are interested in finding a trial, you will likely start with clinicaltrials.gov, a public website that is a database of clinical trials.
Here you will be searching for your medical condition and then reading through titles and unstructured eligibility criteria. Eligibility criteria is a series of text that specifies what participant characteristics are required to be able to participate in the trial (i.e. inclusion criteria) and what characteristics would exclude someone from participating in a trial (i.e. exclusion criteria). This eligibility criteria needs to be read through to be able to understand what trials might be an option for you personally.
Eligibility criteria is clinical information text represented in short-hand in a large, unstructured text field. Currently there are no standards in how this data is represented, meaning as you read this information you also have to traverse the variability in how this clinical information can be represented for different studies. Additionally, because trials investigate new concepts, treatments and tests, the topic being described may not have standards yet (i.e. cell therapies, new treatments that do not yet have a molecule name).
Eligibility Criteria example for a trial:
While this task is daunting for patients who lack the knowledge as well as medical professionals who lack the time to do a comprehensive search and comparison of all relevant studies, it is an excellent and appropriate task for AI to tackle. Natural Language processing, or NLP, is defined as “a field of Artificial Intelligence that gives the machines the ability to read, understand and derive meaning from human languages.” Essentially, NLP is a tool used to help computers action information from written human language. NLP is well suited to assist in the task of reading through thousands and thousands of clinical trials and finding appropriate trials for an individual patient, and offering personalized trial recommendations. We can use NLP in the following ways to assist users with trial information:
- Leverage NLP to extract eligibility criteria into phrases
- Use named entity recognition (NER) to identify key entities needed for evaluating clinical trials, such as the medical condition, stage of the disease, previous treatments, etc.
- Harmonize the information into a standard using ontologies
- Use parts of speech tagging to identify how to apply criteria so that patients can be matched with suitable studies
AI for Good in Clinical Research:
It is imperative that we take care in how this AI is built so as to not perpetuate the inherent bias in our current medical system but act to remove it. This means that trial matching logic needs to be transparent and strictly guarded against current treatment bias found in real world examples. Eligibility criteria needs to continuously be reviewed and challenged as to ensure fair access to trials for all. This technology needs to be actively monitored to ensure we not only build, but also continue to sustain Good AI.
And so, by applying AI to unlock clinical trial information we can run personalized trial searches for any and every patient, which is our mission at Ancora.ai. This allows for patients to be seamlessly matched to trials. The goal is to integrate this technology into standard clinical workflows, so trials can be programmatically and systematically considered from the point of diagnosis with no effort from the patient or physician. With this, all patients who could benefit can access trials and thus we can:
1) improve representation on trials
2) remove bias from trial enrollment
3) accelerate trial enrollment
This technology could help save lives by giving more people access to trials and potentially life-saving treatment, and could mean clinical trial timelines and costs are greatly reduced, giving a huge boost to how we innovate in medicine. Most importantly, this is only one of many applications of how AI can democratize healthcare so that every person has equal access to care and can live a healthy life.
ABOUT THE AUTHOR
Danielle Ralic is a technologist interested in the application of technology to revolutionize healthcare in an effort to democratize access and improve the quality of care. Danielle is the CEO, CTO, and Co-founder of Ancora.ai, an award-winning, AI-powered platform that supports cancer research with patient-first technology.