Research you can
look into.
Explore whether giles® could support the time-intensive data extraction stage of a systematic literature review, while producing findings that align with a manually extracted reference dataset.
R-S-S (Regulatory Scientific and Health Solutions) provides regulatory and technical expertise to the biopharmaceutical industry. Its teams conduct systematic literature reviews to support evidence synthesis and Health Technology Assessment. The collaboration evaluated how giles® could help with the repetitive work of identifying and extracting study outcomes.
Atopic dermatitis · Published randomised controlled trials
How the evaluation worked.
- 01
Build the reference dataset
The team manually extracted key outcomes using its standard SLR process. The evaluation focused on atopic dermatitis publications, including EASI and IGA outcomes across treatment arms and timepoints.
- 02
Refine the extraction approach
The same publications were uploaded to giles®. The prompt was refined through 20 iterations before a stable format was applied across articles and treatment arms.
- 03
Compare the results
The evaluation covered 13 published articles and 1,166 data points. An average of 6.23 questions per paper was used to retrieve the required information.
Strong agreement with the manual review.
For the EASI outcomes assessed, giles® achieved approximately 94% average extraction accuracy compared with the manual SLR. The mean correlation coefficient was 0.98, with a reported range of 0.93–0.99.
“By integrating giles® into our SLR workflow, we reduced the manual effort by up to 85%.”
Dr. Iftekhar Khan · R-S-S case study
Read the results in context.
These results describe this atopic dermatitis evaluation and its refined extraction workflow. They are not a guarantee of performance across other documents, questions or therapeutic areas. The case study also reports an R-S-S estimate of up to 85% less manual effort.
Explore the full publication