Clinical Trial Enrollment is Hard

Clinical trial enrollment is one of the most significant bottlenecks in advancing medical research. Poor participant recruitment is the leading reason for premature discontinuation of randomized clinical trials, particularly investigator-initiated ones. Only 3-5% of candidates get matched today.

Why does this happen? Here are a few critical challenges:

1. Complex Eligibility Criteria. Eligibility requirements are often dispersed across multiple systems and buried in unstructured data. This fragmentation makes it difficult for researchers to assess patient eligibility efficiently.
2. Geographic and Demographic Constraints. Finding a diverse set of suitable patients in a specific location is very challenging. If eligibility criteria are overly intricate or the data is poorly structured, identifying a sufficient number of eligible patients becomes nearly impossible.
3. Limited Awareness Among Physicians and Patients. Physicians are deeply involved in facilitating matching for trials. With so many trials to consider and so many other competing priorities, it can be hard to remember trials that aren't popular or personally interesting to the physician.

‍

Tools like TrialGPT Show Promise for Coarse-Grained Trials

A November 2024 study demonstrated that TrialGPT, an AI-powered patient-to-trial matching framework, achieved 87.3% accuracy for clinical trial matching, nearing expert-level performance. This reduced screening time for trials by an impressive 42.6%.

This work is currently designed for coarse-grained clinical trials, or trials with criteria available in a patient summary or structured database. However, it leaves a gap for trials with fine-grained enrollment.

We define Fine-Grained Enrollment as follows.

Fine-grained Trial Enrollment requires data that is only available in unstructured records like medical notes, and performs logical and/or temporal functions on that data to determine eligibility.

The effort to apply to clinical trial enrollment is just getting started, but while TrialGPT excels for trials with coarse-grained criteria, it is clearly less suited for studies requiring detailed or temporal matching criteria.

‍

We Need Tools for both Coarse-Grained and Fine-Grained Enrollment

Clinical Trials have a wide range of goals, and a wide range of designs to meet those goals.

Trials with coarse-grained criteria benefit from higher potential enrollment and easier data parsing. However, there are certain types of hypotheses that require more detail or must exclude patients based on information that is in the medical notes.

Not all fine-grained criteria are available in a structured database - they may be detailed in medical notes, or require computation across a few pieces of data (for example the amount of time between two events in a patient’s chart).

To illustrate this point, here are two simplified trials based on a real breast cancer treatment use case.

Trial 1: Coarse-Grained
Trial Phase: Early Exploratory / Phase I
Objective: Understand the Safety and Tolerability of a new chemotherapy drug.
Criteria: Patient has breast cancer and is under the age of 70

Trial 2: Fine-Grained
Trial phase: Phase II
Objective: Understand the drug's efficacy for the specific genetic marker, knowing that it has higher risks for patients who have taken a certain list of drugs recently.
Criteria: Patient has breast cancer with a specific genetic marker, is under the age of 70, has not taken any of a list of drugs in the last year, and is at least 12 months from their last surgery.

It is unlikely that all of the criteria in Trial #2 can be found in a structured database -- we need to parse the clinical notes to determine eligibility.

In addition, many trials benefit from a combination approach: coarse-grained recruitment, followed by a fine-grained prescreening step. For example, tumor next-generation sequencing reports often generate trial recommendations for patients based on their high level information. When humans prescreening was added after these trial recommendations, the physician burden and level of false positives was significantly reduced.

A closely-related set of trials might use coarse-grained criteria, fine-grained criteria, or even a mix. However, determining eligibility based on the contents of medical notes adds significant implementation difficulty.

Fine-Grained Trial Enrollment has More Specialized Requirements

In order to support clinical trials with fine-grained criteria, you need additional capabilities:

1. Advanced Data Processing. Tools must handle large amounts of complex and unstructured data, including longitudinal clinical notes.
2. Scalability and Accuracy. Patient notes can be very long per patient and add a huge amount of data to enrollment criteria. This makes it critical to process an even larger amount of data with high accuracy and while remaining cost-effective.
3. Parse Complex Eligibility Criteria. Fine-grained trials typically have multiple layers of logic in each criterion. A system must be able to parse these and apply them to data in a contextual manner, mimicking a human enrollment specialist.

These additional requirements add difficulty to trials with fine-grained enrollment criteria. They also mean that tools designed for coarse-grained criteria may not apply without significant modification.

‍

Brim is an AI-Powered Solution for Unprecedented Clinical Trial Enrollment

Brim Analytics offers a cutting-edge solution that revolutionizes clinical trial enrollment, so you can find more eligible patients, faster. Developed by a team with deep experience in extracted structured data, and engineered to leverage powerful foundational AI models, Brim delivers an unparalleled combination of accuracy and efficiency.

Here’s how Brim stands out:

1. Extract More Patient Details: Brim capitalizes on the scalability and cost efficiency of modern large language models to dig deep into your data – including complex medical notes – to identify eligible patients overlooked by other methods.
2. AI to Accelerate Enrollment. Brim automates patient matching with lightning fast AI-powered matching combined with contextual understanding and layered reasoning.
3. Humans-in-the-Loop for Accuracy. Brim unites the power of AI with expert oversight, enabling the highest level of accuracy and patient safety. 
4. Easily Parse Complex Criteria. Easily implement even the most intricate inclusion/exclusion criteria with Brim’s intuitive interface and powerful “Dependent Variables" feature, which easily deconstructs complex trial criteria into layered, contextual variables.
5. Boost Efficiency. Brim’s intuitive interfaces are purpose-built for researchers, clinicians, and other key users, enabling seamless workflows and better outcomes.‍

With Brim, you can:

• Significantly expand your candidate pool
• Reduce screening time and costs
• Accelerate research and bring life-saving treatments to market faster

Ready to unlock the potential of your clinical trials? 

Contact us for a demo here and in the meantime, you can see how easy it is to create complex trial criteria in Brim below.