AI agents to match clinical trials in oncology
AI agents can replace the weeks or months of outdated, manual screening and matching through automation. By analyzing the patient’s medical history and biomarkers, preferences, location, and other important details, the agent can identify the highest-fit medical research.
For providers and clinicians, official sites and sponsors, this means up to 40% less manual effort per candidate; for patients – the chance of accessing promising therapies.
In this short article, we highlight how adopting AI agents can change how patients are finding clinical research. By checking a patient’s profile against the complex eligibility criteria, continuously scanning medical research, and handling preliminary communication, AI agents can act as tireless virtual advocates.
Less effort, faster results – without doubt, a win-win for patients seeking help and institutions running studies.
The challenges of finding clinical trials that fit
Clinical research is what connects patients to cutting-edge experimental treatments yet not publicly accessible. The process of matching the patient to relevant medical research, however, stays quite tedious and expensive: it involves the analysis of long-term medical history and needs, sifting through eligibility criteria, and more.
No wonder that over 8/10 experiments are struggling with recruitment, which limits the patient in accessing potentially life-saving emerging therapies.
How do you connect a patient – an individual with unique medical history, preferences, location, and status, often limited in terms of traveling or otherwise – to a clinical research with strict inclusion/exclusion criteria? The process as it is today still relies on searching medical research in databases and contacting numerous sites, which leads to overload for clinicians and frustration for patients.
In oncology, where progress is critical for patients dealing with advanced disease, every week spent searching can impact future outcomes.
The solution: a personalized clinical trial AI agent
We suggest a combination of advanced AI algorithms, NLP capabilities, and specialized clinical knowledge – Abto Software’s AI agent POC design to analyze and match the studies to patients mostly automatically. Recent studies highlight that AI algorithms can automate this process with high-level, near-human accuracy, thereby bringing the patient much closer to best-fitting medical research.
Let’s dive into that.
AI agents in cancer clinical trials: key features
Initial patient profile registration
The interactive profile enrichment is designed to ensure all relevant inclusion/exclusion factors are captured:
- Initial patient profile registration is standard and includes the provision of personal medical history, cancer type/stage, prior treatments, demographics, biomarkers, and other relevant information
- The process might include the upload of electronic health records or answering a questionnaire
- The agent will use NLP capabilities to parse unstructured notes and extract clinical details
- The agent might also additionally query the user for any missing details to complete the profile:
- “Have they been through prior immunotherapy?”
- “Is the EGFR mutation still present?”
Intelligent search and retrieval
With the profile ready, the agent will start to search the databases and registers for relevant clinical research. This baseline, first-pass filtering is designed to browse wide selections and remove the clearly unfit candidates.
Rather than standard search, the agent will use an intelligent retrieval strategy to ensure thorough outcomes. For example, the agent might generate unique keywords or use embeddings considering the patient-specific medical history and condition.
In fact, this intelligent retrieval strategy can recall more than 9/10 experiments while scanning only a single-digit percentage, thereby ensuring higher efficiency.
An automated eligibility screening and matching
The next-stage eligibility screening and matching will provide a list of best-fitting clinical experiments:
- If the profile lacks any information for a certain criterion, the agent may ask follow-up questions
- Each candidate is marked either eligible/ineligible or as one that needs clarification
- The agent can also provide summaries to explain which criteria are met for enrollment
- By evaluating eligibility non-stop, the agent will filter out everything that’s irrelevant, sparing patients from pursuing dead ends (this means fewer calls/emails to sites, which end as disqualifications)
To note: the common eligibility criteria are often quite lengthy and written in hard-to-understand clinical terms. But using NLP capabilities, each criterion (cancer type/stage, laboratory thresholds, and more) is interpreted and checked for compliance
Ranked recommendations
The agent will present the user with an annotated list of the best-fit candidates, all ranked by how they match. The factors that impact the ranking are criteria being met, the phase, location proximity, and urgency of start.
Each listing comes with a short, plain-language summary of the picked candidate and note on why it’s relevant.
This readability is important to make jargon-heavy descriptions more patient-friendly.
AI algorithms can generate easy-to-understand summaries without losing key details and instantly make them much clearer for those who don’t have specialized clinical backgrounds.
Further support: taking action
The agent can help to take the next steps required, for example, by providing contact information or guidance. An agent can draft an email to reach the coordinators on behalf of clinicians and patients, which includes important details (for example, “a patient with metastatic lung cancer, EGFR exon 19 deletion, ECOG 1, is a good match for NCTXXXXX.”)
This kind of interaction can streamline the process of recruitment and enrollment, thus saving valuable time. This ensures that when the clinician or patient will speak with investigators, both parties will have the details on hand, thereby making those conversations more efficient and productive.
Continuous monitoring and updates
With the profile registered and enriched, the agent will continue to act on its behalf until the status is changed. It will periodically check for opportunities or updates that might be relevant.
If a new experiment closely matching the patient’s profile appears, the agent will alert the clinician or patient. And conversely, if the patient’s condition should change (for example, if a new mutation is found), the agent can re-run the search to look for better fitting experiments.
This monitoring can turn the quick, one-time search into an ongoing scouting, thereby giving a patient a chance to catch opportunities immediately.
AI agents for cancer clinical trials: the application
Let’s envision a hypothetical patient scenario:
Step 1. Getting started
Emily, a 55-year-old patient with mature lung cancer, went through standard protocols but with small success. Her therapist has suggested alternative strategies.
Emily registered on the AI platform and provided her essential medical history & details.
Step 2. Profile enrichment
The agent will process provided information and ask follow-up questions to close the gaps, for example:
- “What is your current ECOG status?”
- “Do you have any other significant health issues (autoimmune diseases, heart conditions, or other)?”
The agent will update profile details (Emily’s able to care for herself and has well-controlled hypertension). Moving further, it notes that some lab values are missing and pulls them from provided records by itself.
Step 3. Finding candidates
The agent will search the research that targets Emily’s condition – lung cancer with an ALK or similar mutation. After narrowing down from endless existing clinical experiments, it’ll find 15 that are in late-phase development and located within a reasonable distance or offer travel support.
This filtering is performed in seconds, whereas traditional manual search might have taken weeks or months. The agent will identify a bunch of different clinical experiments – some testing ALK inhibitors, others studying immunotherapy combinations.
Step 4. Eligibility screening
The agent will go through each of the 15 options:
- For one (Phase II ALK inhibitor medical experiment), it verifies that Emily meets all key inclusion criteria: ALK-positive status, prior treatment with first-line ALK inhibitor, and others
- This option is flagged as eligible and a strong match
- For another (immunotherapy combination), it catches a key exclusion criterion: no history of previous autoimmune diseases
- The agent will confirm Emily’s answer and mark it eligible as well
- A third accepts patients with no brain metastases
- The agent can find no mention of those, so she presumably qualifies
- Just to make sure, it asks Emily whether she had any known brain metastases
- She confirms she hasn’t, so this one is also marked as eligible
- Some get filtered out, and, after the screening, the agent will present 5 options with potential
Step 5. Final results
At the very top is the Phase II ALK inhibitor medical experiment: “High match: You’re meeting 10/10 criteria. The study is for ALK+ patients who had prior therapies, which matches your history. Location: 50 miles away.” Another highlights: “Moderate match: You meet most criteria. It requires no prior immunotherapy sessions – you have had chemo, so this is okay. Location: 200 miles away (travel support is available).”
For each of those, the agent will provide concise summaries, for example: “#NCTXXXX studies a new targeted pill to see if it works when earlier treatments have failed. The goal is to shrink tumors who have ALK mutations.” This helps Emily understand the options without needing to parse medical jargon.
Step 6. Getting qualified
The agent can help with the next steps as well:
- By clicking “Contact site for me”, Emily triggers the agent to send an email to contact the coordinator
- Her oncologist will receive a report that can be added to the medical file
In only a week, Emily hears back from the site and is then scheduled for an official screening for the top option. The groundwork has spared Emily from multiple calls and uncertainty, thereby accelerating her path to finding a potentially successful treatment.
Clinical trial matching agents: market readiness
- The global AI in clinical trials market size is expected to be worth around $22.89 billion by 2034
- In 2024, North America has led the sector by reaching 31.5% share
While these figures reflect AI adoption in this market segment more broadly, they also suggest that AI agents – an approach that’s still relatively fresh – are likely to enjoy comparable momentum as their adoption widens.
As projected by Research and Markets in their even more bullish report:
- The global AI in clinical trials market volume might see growth from $6.57 billion in 2024 to striking $38.7 billion by 2029
- The growth in the historic period is attributed to rapid data accumulation and evolving AI algorithms
While these projections cover AI as a whole, they suggest a similar growth trajectory for the AI agent.
Clinical trial matching agent: the benefits and impact
For providers & clinicians
The agent takes over initial screening, which commonly takes away from personal doctor-patient interactions. That means, it augments the process of decision-making rather than replacing it: the filtering gets automated, the decisions and discussions with patients are left to experts.
The result: less frustration, better outcomes.
For patients
The agent quickly analyzes viable options, thus saving several weeks (or months) of tedious manual research. What’s more, it minimizes false leads, so patients don’t waste their time on options they don’t qualify for – good candidates are found much faster, a benefit when fighting aggressive diseases.
In an NIH study, the doctors that used AI spent 40% less work hours on screening without losing any accuracy.
Trial sites & sponsors
With agents, the sites will see the candidates more likely to qualify, which means a more successful enrollment. This helps to finish the research on schedule and bring new therapies to market way sooner.
The pre-screening directly addresses the mentioned 80% matching failure rates, thus boosting the outcomes.
Broader impact
The agent can learn over time and improve the algorithm, for example, to highlight why patients are excluded. It can even identify unmet needs when patients consistently lack potential matches and notify their therapists.
More important, if getting widely available, the agent can expand healthcare access.
How we can help
Clinical trial AI agents can transform a stressful, resource-draining process into automated, fruitful assistance. That’s what AI should be used for today.
Talk to our team to dive deeper into our vision.
Our expertise:
- AI for digital physiotherapy
- AI solutions engineering services
- Robotic process automation services
- Hyperautomation services
Our services:
FAQ
Clinical research in oncology has complex eligibility criteria that require sifting through enormous databases. Manual work it requires is oftentimes quite labor-intense, comes with human errors, and might face barriers including bottlenecks in recruitment, administrative overload, and other common issues.
That causes two core business problems:
- The time that’s spent on experiments
- And cost of experimenting
That’s where AI holds strong promise.
Clinical research in oncology can generate voluminous records and involve quite many complex workflows. Manually handling all those slows down the recruitment – and automating those processes might deliver numerous benefits.
In brief, we mean:
- Freeing coordinators from everyday routine tasks
- Automating workflows from screening to matching and supporting the patient
And, naturally, AI has sufficient benefits for patients who seek efficient treatment.
Medical research often involves repetitive tasks, from reviewing the records to managing the documentation.
That’s where AI agents can shine:
- Screen records and datasets against strict eligibility criteria
- Extract details from documents
- Assist with patient-trial matching and recommendations
- Support the coordination workflows between everyone who’s involved
This reduces administrative burden for teams and allows to focus on care and quality rather than manual tasks.
Patient recruitment is one of the biggest challenges, in particular in oncology with stringent eligibility criteria.
Another thing AI agents can cover:
- Scan records and databases
- Match profiles
- Prioritize candidates
- Alerting coordinators, and more
This reduces recruitment delays and facilitates the rates.