myTomorrows launches an AI supported clinical trial search tool

Today we launch TrialSearch AI, a physician-focused, pre-screening clinical trial search tool, that keeps patients at the heart. By leveraging AI technology, it allows physicians to identify potentially eligible clinical trials or expanded access programs (EAPs) for their patients, within minutes.

Physicians will look at a patient’s entire medical profile when making a clinical decision. When it comes to searching for suitable clinical trials, they are required to take a patient’s complex history and manually interpret it against multiple different trial eligibility criteria, which can be a lengthy exercise for treating physicians.

TrialSearch AI will help a physician by providing quicker insight in available pre-approval options. By cross-referencing a patient medical summary with multiple clinical trial eligibility criteria sourced from public registries, the tool presents the physician with a list of potentially available clinical trials, and where appropriate, EAPs. It will prompt the physician to do manual checks and ask them to confirm whether a pre-approval option is suitable for the patient.

In doing so, the tool saves precious time for physicians. The physician can still make an informed decision, based on transparent and personalized results, and is then able to use the myTomorrows end-to-end platform to make patient referrals directly.

TrialSearch AI aims to transform the search for clinical trials for patients with unmet medical needs; all whilst keeping physicians firmly at the center of the process.

Physicians can sign up to test the beta here.

Why are we doing this and what are the benefits?

myTomorrows is a forward-looking, ambitious organization. We use a combination of in-house industry expertise alongside a reliable tech platform with the aim of connecting all stakeholders within clinical trials and EAPs.

With this approach, myTomorrows assists patients and physicians with searches for possible clinical trials and EAPs worldwide and aims to give patients and physicians an unbiased, inclusive, and comprehensive breakdown of pre-approval options where patients could be eligible. We do this by leveraging the myTomorrow’s search which is curated for people looking for investigational pre-approval options and sourced from public registries.

To make sure we continue to deliver a valuable service, we continuously look at how we can make things more efficient, and where we can utilize new tech solutions that can lead to improved user experiences while keeping the patient and physician at the center.

We believe that by harnessing the power of AI, we can take one step closer to helping more patients, discover and access treatments. With TrialSearch AI, we aim to improve the speed and accuracy of physicians finding clinical trials whilst simultaneously reducing their admin time. Helping patients and physicians discover and access potential pre-approval options may also support the wider BioPharma industry with clinical trial enrollment.

What is AI and how does it work?

Around the world, we see more and more companies adopting AI into their business models, to deliver improved services to customers – but what does AI mean and how does it work?

Artificial intelligence (AI) refers to machine or computer system processes that are designed to simulate human intelligence. The goal of AI activities is often to speed up a process by using computer coding to collect, interpret, and reason data, in a short amount of time.

TrialSearch AI will use a type of AI called Large Language Models (LLM), specifically the GPT series of models licensed by Azure OpenAI.

An LLM works to read, summarize, translate text, and reason data, based on a set of instructions.

In the case of TrialSearch AI, the LLM will take the patient’s medical summary provided by the physician, and search against all available clinical trials listed in public registries. The results shown to the physician will still require manual verification to ensure the physician remains at the center of determining whether a trial would be suitable for a patient.

TrialSearch AI does not depend on information sourced from the LLM, but instead utilizes its capabilities to quickly verify the medical profile against all listed eligibility criteria in public registries. The information comes from public clinical trial registries and the medical summary provided by the physician.

To take a deeper dive into how AI is applied within TrialSearch AI, check out our recent scientific research paper.

How will TrialSearch AI work in action?

Upon accessing TrialSearch AI, physicians will be asked to provide basic patient medical information such as condition, age, country of residence and gender. Next, to make the process as efficient as possible, the physician will be able to type the medical summary into a free-text box. This means they can save time not having to answer lengthy questionnaires, and only need to provide the relevant summary information required. All patient information submitted should be non-identifiable.

A medical summary could include:

• histology characteristics of the condition
• previous treatments
• relevant lab results for example blood pressure, renal function, or other condition-led test results,
• condition biomarkers,
• other health conditions which may be pertinent
• notes on the overall health of a patient including, for example, an ECOG score.

Upon entering the patient information and clicking submit, TrialSearch AI will go to work.

It will identify potentially eligible clinical trials and relevant pre-approval options by cross-checking the trial’s eligibility criteria against the information in publicly available registers, clinicaltrial.gov, and the European Clinical Trials Register.

Any not met criteria must be reviewed by the physician before the final TrialSearch AI-assisted overview is produced. This critical checking point ensures that we harness both the physician’s invaluable expertise and the power of AI.

TrialSearch AI makes no medical recommendations, and eligibility final checks are always made by the physician.

What does the research tell us?

Alongside the launch of TrialSearch AI, members of the myTomorrows team together with a representative from Eindhoven University of Technology, have co-authored and published a scientific paper entitled ‘Improving patient pre-screening for clinical trials: assisting physicians with Large Language Models’. It investigates, in the context of LLM clinical validation testing, the use of GPT-3.5 to determine eligibility for clinical trials based on the summarized description of a patient’s medical profile.

The paper takes an insightful look at the performance of TrialSearch AI in 10 synthetically created patient profiles covering 10 different diseases. The research found that using GPT-3.5 paired with a physician-in-the-loop could reduce the number of eligibility criteria to be checked by 90%, while correctly identifying 100% of potentially eligible trials and correctly classifying approximately 3 out of 4 ineligible trials.

These findings underpin both the validity of TrialSearch AI and the anticipated value it offers to healthcare providers and ultimately patient care.

Who can try it?

We are currently in the beta testing phase. This means we are giving a group of Healthcare Professionals the opportunity to use and try out TrialSearch AI, ahead of a general wider release.

We will work with these early adopters to collect feedback so we can proactively improve the user experience for all. Because physicians will be the tool’s primary end user, during the beta testing, we will give priority access to physicians to test the tool and supply feedback.

Sign-up for the waitlist here.

For those whose access is approved, they will be notified via email to set up a TrialSearch AI account. A member of the myTomorrows team will be in touch, supporting on how to start using the tool.

We are confident that by simplifying the pre-screening experience, TrialSearch AI can deliver real-time benefits for Healthcare Professionals and their patients. Together we can help more patients discover and access treatments.