NCI-designated cancer center spotlights how AI accelerates clinical trial recruitment
Only 5% of Americans participate in clinical trials
Clinical trials are vital for advancing medicine and providing early access to cutting-edge treatments, particularly in cancer care. However, as the complexity of clinical research increases, and as targeting becomes more precise, it leads to fewer patient enrollments. The underutilization of clinical trials results in missed opportunities for patients to access life-saving therapies.
The power of AI in oncology clinical trials
Clinical researchers are now looking to technology like artificial intelligence (AI) and natural language processing (NLP) to precisely match patients with clinical trials, closing the gap between studies available and eligible patients. Instead of doing manual chart reviews that require hundreds of clicks through electronic medical records (EMRs) and sifting through physician notes, genomics reports, imaging results, labs, medications, family history, and allergies, this technology can instantly comb through millions of health records to identify patients that match the trial’s inclusion and exclusion criteria.
This is not a simple keyword search. Unfortunately, it is not that easy. Over 80% of the data in the EMR is in an unstructured format, meaning the free text of doctors’ notes and reports added to medical records. Searching complex, free-form clinical information is a perfect use case for AI and NLP because it can comprehend information in context. For example, if a patient note says there is “no indication of [a target condition],” that patient should not show up in the search results for the target condition. Similarly, “diagnosis of lung cancer” in a patient record means something different than “sister has lung cancer.” Keyword searches are also greatly limited by acronyms, where “ER” in a note can mean emergency room or estrogen receptor, among other things. NLP can understand these different contexts and provide busy researchers a short list of patients that match their search without missing important context.
Cancer center uses AI to accelerate recruitment
A National Cancer Institute-designated cancer center was recruiting patients for a small cell lung cancer trial. Small cell lung cancer accounts for 10 to 15% of all lung cancer patients, making this population harder to find. To identify eligible patients, researchers searched through standardized, structured diagnosis codes known as ICD-10 codes in their EMR system. However, the searchable codes did not distinguish between different types of lung cancer. The research team needed to manually read patient charts to look for phrases like “small cell” or “SCLC” or even “oat cell” cancer. This manual chart review was cumbersome, took hours, and often failed to identify the right patients.
To accelerate recruitment for the trial, the cancer center invested in Deep 6 AI’s software. Using AI, researchers were able to search across all patients for the concept of small cell lung cancer and return precise results from standard data and free-form text. Researchers added other eligibility criteria to their search, such as specific genetic mutations and a status of having received other treatments without a positive result. The AI immediately identified 116 patients as potential matches for this study. The researchers could then see how many of the matched patients had small cell lung cancer, and of those how many had the target genome, and of those how many previously received treatments that did not work. This reduced the burden on the cancer center’s staff and accelerated recruitment for the trial.
Still, AI and NLP algorithms on their own are not enough to advance clinical trials. The matching technology will never be 100% perfect even as data models improve over time. The researchers needed the ability to do a final review and approval of each patient match. The software showed them which data points and information in each health record created the match. This allowed the research team to quickly review and approve or decline the patient for enrollment.
The cancer center is planning to use the software for more than 150 clinical trials, with a goal of accelerating patient recruitment and reducing the burden on its research teams.
Next up: AI to accelerate feasibility assessments
The cancer center’s investment not only accelerates recruitment, but it also helps their research teams make faster decisions about clinical trial feasibility. Before starting a trial, they can use AI to mine their structured and unstructured EMR data to see how many patients match the trial’s inclusion and exclusion criteria. This allows them to focus on studies where they will be able to collect sufficient data for research, and then arms the study team with the tools to quickly identify and recruit those patients.
This kind of intelligence helps move the industry forward, getting treatments to patients faster. And it is key to transforming an industry where the next wave of breakthroughs will be even more personalized and targeted.
Could AI be the answer to accelerating your feasibility assessments and recruitment? Reach out to request a demo with the team at Deep 6 AI.