ASRS 2026: AI screening platform increases retina clinical trial randomization

Manual prescreening for retina clinical trials is labor-intensive, often taking 30 minutes to an hour per patient per trial, according to Louie Cai, MD. Cai and colleagues developed an AI platform designed to screen patient charts and imaging in real time to check documentation against trial eligibility criteria. Data presented at the American Society of Retina Specialists 2026 annual meeting (ASRS 2026) evaluated the platform's effect on screening and randomization efficiency in a prospective study.

Study design

Researchers divided a single retina practice into 2 groups of 11 physicians each: one group used the AI platform, and the other did not. For the first 3 months, neither group used the platform, establishing a baseline period. Over the following 3 months, only 1 group used the tool.

Key findings

Physicians using the platform saw a 37.5% relative increase in trial randomization compared with their own baseline, along with roughly a 13% decrease in the screen failure rate. Cai reported that use of the model saved approximately 4,000 hours of human labor that would otherwise have gone to manual chart review.

Cai noted that adoption hinges on 2 factors: usability and accuracy. "I think the number one thing has to be usability and user friendliness," Cai said, adding that a platform must also maintain a high bar for accuracy, since repeated errors erode clinician trust in its results.

Clinical implications

According to Cai, clinicians typically spend 5 to 10 minutes with a patient and do not have time to walk through every trial criterion, which can lead to referring patients who are not ideal candidates. This delays care: patients may later learn they do not qualify and must return to standard-of-care treatment, a process that can take weeks.

In current practice without AI support, clinical research coordinators typically preview the next day's appointment list and manually check each patient's chart for trial eligibility. Cai's data show that 87% of patients seen in a retina practice do not qualify for any clinical trial, and the remaining 13% who might qualify require roughly 30 additional minutes of manual review to confirm eligibility.

Looking ahead

Cai said broader adoption of AI-based screening could shorten trial enrollment timelines, potentially reducing a typical year-long enrollment period to 6 months, accelerating drug development overall. Cai would like to see multi-site validation to determine whether the results generalize beyond a single practice, along with further study of the platform's effect on overall drug development timelines.

Reference:

Cai L. Presented at: American Society of Retina Specialists 44th Annual Scientific Meeting (ASRS 2026); July 15–18, 2026; Montréal, Québec, Canada.