Clinical trials are the cornerstone of advancing cancer research and care, where connecting the right trial to the right patient can significantly improve treatment outcomes. But for healthcare teams, matching patients to suitable trials is a complex task—one that requires carefully sifting through medical records and comparing them against the strict criteria of hundreds of studies. This process is not only time-consuming but also prone to human error, leading to missed opportunities for patients to access potentially lifesaving cancer treatments.
Investigators at the MCW Cancer Center and the Clinical and Translational Science Institute of Southeast Wisconsin (CTSI) have found a way to streamline this process and ensure that no patient is overlooked for trial eligibility: by harnessing the power of artificial intelligence (AI). Their new pilot study automates patient-trial matching by using a large language model (LLM), a Triomics platform called OncoLLM, to analyze vast amounts of data in a fraction of the time it would take a human, and with greater precision. As the first medical institution in the nation to adopt the AI-driven platform, MCW is not only setting new standards for clinical trial equity and inclusion, but also is redefining the benchmarks for how innovative technology can be used to enhance patient care.
“Early results of this study are very promising. Since implementing the OncoLLM platform in July, we’ve been able to evaluate all the patients in the Disease Oriented Teams (DOTs) that are part of our pilot project. This means that every patient who has an upcoming appointment will be rigorously assessed for eligibility for open clinical trials at the MCW Cancer Center,” said co-principal investigator (PI) of the study Anai Kothari, MD, MS, Assistant Professor of Surgery and Co-Director of the Center’s Geospatial, Epidemiology and Outcomes Shared Resource.
“Clinical trials are an essential part of optimizing cancer care but only when patients have fair opportunities to participate in them. Using a generative AI-powered model like OncoLLM can help us ensure that every patient is being assessed comprehensively for the trials we have open. This could potentially be transformative for the clinical research process and ultimately, improving overall outcomes,” said Dr. Kothari.
Collaboration and a Robust Infrastructure Propel MCW to the Forefront of Technological Innovation
Dr. Kothari explained that the team’s vision has long been: “How do we best take all the information that’s not well structured—the clinician notes, lab results, imaging, and so much more—and turn it into real knowledge?” OncoLLM was developed using a training model similar to OpenAI’s ChatGPT but with a focus on clinical data. It uses advanced natural language processing techniques to interpret and extract meaningful information from the text, then translates the data into a structured format that can be analyzed systematically.
Dr. Kothari and co-PI of the study Brad Taylor, MBA, FAMIA, Chief Research Informatics Officer in the CTSI, recently co-authored a research paper showing that OncoLLM not only outperforms existing AI models like GPT-3.5 but also matches the accuracy of qualified medical professionals, potentially leading to more efficient and accurate patient-trial matching.
“While the impact of this work is significant, we know there’s often concern about privacy when it comes to unstructured clinical documentation. Instead of using ChatGPT or some other commercial source, we train OncoLLM in a secure, privacy-preserving manner. This method ensures that the LLM is isolated from the internet, in a HIPAA-secure private platform,” said Taylor.
The pilot study is currently testing OncoLLM in the gastrointestinal, genitourinary, breast, and thoracic DOTs. In light of the positive early results, the team intends to implement the platform across all of the Cancer Center’s DOTs and over time, will allow investigators across the institution to access it. Dr. Kothari emphasized that the reason MCW is able to lead this work is because of the state-of-the-art infrastructure already in place.
“Many people might not realize that effectively using clinical data requires it to be meticulously organized and structured to unlock its full potential. Thanks to Brad Taylor’s efforts in the CTSI and Integrated Cancer Data Resource, we’ve been able to achieve this and provide researchers with the tools they need to harness data in a way that makes an impact. The Triomics program is a prime example of what’s possible when such collaboration occurs and is going to help propel us to the forefront,” he said.
Beyond Matching: The Role of Patient Navigation
While OncoLLM excels in identifying suitable clinical trials for patients, the journey does not end with the matching process. The MCW Cancer Center recognizes that enrolling in a clinical trial can be a complicated and daunting process, which is why it recently launched a new Patient Navigation Program. At the forefront of this initiative is the inaugural navigator, Adrena Luckett, who helps guide patients through logistical challenges such as understanding the trial requirements, coordinating appointments, managing transportation, and addressing financial concerns. Dr. Kothari said that if the pilot study is successful in matching more patients to trials, this program will play a critical role in ensuring patients can enroll and participate.
“We’re fortunate to have such a strong infrastructure to take a patient and help them enroll and deal with the structural barriers. This holistic approach ensures that patients not only have access to trials but also the resources needed to participate fully, thereby maximizing the potential benefits of their treatment,” he said.