PROMETRIKA’s Biostatics and Statistical Programming teams continually seek new and innovative ways to address the many functions of data analysis in clinical research. Our statistical programmers have had numerous successes in researching and developing new programs and methods that reduce time and steps in analyses.
Our recent attendance at PharmaSUG 2025 in San Diego, CA was an insightful and thought-provoking experience. This year’s conference featured a wide range of sessions, with artificial intelligence taking center stage. Presenters shared innovative approaches, both at the individual and company level, for integrating AI into clinical research.
Our ePoster session, “Automating Recurring Data Reconciliation for Serious Adverse Events Using SAS,” touched on one of the larger themes of the conference, automation.
The e-Poster highlighted the process of reconciling serious adverse event (SAE) data from pharmacovigilance vendor files against the electronic data capture (EDC) files. At regular intervals during a project, our clinical safety team receives the external vendor file with SAE data that must be reconciled with the adverse event (AE) data from our EDC system to ensure consistency across systems.
Previously, this process was manual and tedious, requiring side-by-side comparisons, discrepancy tracking, and cumulative comment logging; typically, all done and maintained by one user in Excel. This process took several hours per month and left room for human error.
We noted that this process is well-suited to automation, and proceeded to create a useful program. The SAS program created, saereconiliation.sas, automates the SAE reconciliation process in two major phases. Firstly, it reconciles the current SAE vendor file and the current AE EDC data. The program performs a series of eight cascading merges between the two datasets using composite keys (e.g., Subject + Term + Start/End Dates), tiering the exactness of the criteria as they fail to account for data inconsistencies. It then compares each matched record across all fields and tracks any differences. A new variable, “Discrepant Fields,” specifies which fields differ or states “100% Match” when both sources are the same. Secondly, the tool reads the previous month’s reconciliation report to track changes over time. It flags records that are: new, resolved, missing, or newly discrepant. In addition, all comments from past runs are preserved and carried forward, creating a transparent audit trail. The final output is a newly generated Excel report, automatically color-coded based on the flagged records of interest mentioned.
The session for our e-Poster was well attended and sparked thoughtful conversations. Attendees engaged with questions around the necessity of the application, other potential use cases, and ideas for future enhancements. Since implementing the program, our safety team has reduced reconciliation time, while improving consistency, traceability, and scalability. This program has many more use cases, as the architecture is adaptable to any setting where two datasets (with or without a foreign key) need to be reconciled and tracked over time.
We enjoyed the challenges and discoveries of developing this program. We would like to acknowledge the contributions of our colleague, Valerie Jurasek, Senior Drug Safety and Pharmacovigilance Specialist, and mention our appreciation for PROMETRIKA’s continuing support of employee innovation.
We left the conference energized by the exchange of ideas and inspired to keep pushing the boundaries of automation and innovation in clinical research.
