
AI in Clinical Trials – Fujitsu Japan Limited [1] (Fujitsu Japan), together with the National Cancer Center Hospital East [2] (NCCHE) and U.S.-based Paradigm Health, Inc. [3] (Paradigm), conducted joint research into the use of a Large Language Model (LLM) and Paradigm’s clinical trial matching platform, “Paradigm Platform” [4], to support the identification of potential candidates for clinical trials of new drugs. The research confirmed the effectiveness of the approach in significantly improving the efficiency of the candidate identification process.
Between June and the end of October 2026, the partners conducted a verification study using data from 22,334 patients in the Department of Respiratory Medicine, simulating a clinical trial targeting patients with EGFR gene [5] mutations. The process involved using an LLM to structure patient data from electronic medical records and linking this data to the “Paradigm Platform” for matching. As a result, it was confirmed that AI could identify 57 priority candidates along with the clinical evidence for these selections.
Building on the results of this research, Fujitsu Japan will proceed with the commercialization of a service to support the entire process from candidate screening to clinical trial enrollment, aiming for a launch by the end of fiscal year 2027. Through this service, the company aims to promote participation in clinical trials for new drugs and contribute to expanding treatment opportunities for patients.
Background
The issue of “drug loss,” where new treatments available overseas are not developed or approved in Japan, is a pressing social challenge. To address this, there is an urgent need to promote clinical trials for new drugs. In recent years, personalized medicine based on genetic mutations and biomarkers [6] has advanced, particularly in oncology, making the criteria for clinical trial participation increasingly complex. Consequently, medical professionals must comprehensively review information dispersed across electronic medical records, treatment histories, test results, and genomic data. This places a significant burden on medical staff, not only affects the preparation and progress of clinical trials, but also serves as a factor in patients missing opportunities to consider new treatment options. Furthermore, differences in medical data formats and the prevalence of unstructured data in Japanese electronic medical records create obstacles for global data utilization.
Joint Research Overview
In this joint research, which began in June 2026, Fujitsu Japan, the NCCHE, and Paradigm utilized LLMs to structure unstructured data —such as patient medical records, treatment histories, and genomic information— necessary for clinical trial eligibility assessment [7]. This was then used to verify the application of the “Paradigm Platform,” which has a proven track record in the United States.
Specifically, in a verification process simulating a clinical trial for patients with EGFR gene mutations, Fujitsu Japan performed data processing, including anonymization and structuring using LLMs, on electronic medical record information for 3,027 lung cancer patients out of 22,334 respiratory patients at the NCCHE, and linked this data to the “Paradigm Platform.” Paradigm evaluated and matched the data of 540 patients among these 3,027 with EGFR gene mutations, based on five criteria used to determine clinical trial eligibility, successfully identifying 57 priority candidates. Subsequent review by physicians of the clinical records presented by the AI as evidence confirmed that 51 of these individuals were promising candidates.
Previously, clinical trial candidate identification relied on medical professionals—such as physicians and clinical research coordinators—to identify candidates and manually review individual medical records based on their own experience and knowledge. The results of this study confirmed that by utilizing AI and the “Paradigm Platform,” medical institutions can comprehensively determine the eligibility of target patients for clinical trials based on standardized criteria, without relying on individual expertise. Furthermore, by identifying candidates through automated overnight processing, the system reduced the number of patient records requiring manual review by approximately 89%, demonstrating the potential to alleviate the burden on medical staff and expedite clinical trial opportunities for patients.

Figure : Joint Research Overview
Note: These results represent the candidates identified during this joint research and do not guarantee their actual participation in clinical trials. Final eligibility is determined by physicians based on trial requirements and the patient’s consent.
Future Plans
Based on the results of this joint research, Fujitsu Japan aims to commercialize a service that supports the end-to-end process from clinical trial candidate identification to patient enrollment, with a target launch by the end of fiscal year 2027. Going forward, the partners will promote the adoption of this solution at medical institutions across Japan to help expand their capacity to conduct clinical trials. By providing more patients with opportunities to participate in clinical trials, Fujitsu Japan will contribute to faster clinical trial initiation and patient registration, ultimately facilitating broader access to new therapies.
Fujitsu Japan promotes the “Uvance” [8] business model, committed to achieving business transformation and creating social value through AI. The company will continue to use “Uvance” to streamline and advance clinical trials, aiming to eliminate drug loss and provide higher quality healthcare to every patient.
For more information visit: global.fujitsu.
















