Three complementary ways to examine study data
We reviewed the deidentified chemotherapy-only control arm of EXPAND, a completed phase III study in advanced gastric or gastroesophageal-junction cancer, using data made available through Project Data Sphere.
436 participants · 48 datasets · 237,784 rows
Programmed checks
Twelve discrepancy types across measurement integrity, tumor and disease assessment, eligibility, treatment exposure and record chronology, with counts, severity and explanations.
AI checks
Clinical-context review connected laboratory episodes with adverse-event histories and resolved four of five initial missed-dose concerns after examining the wider treatment history.
Pattern review
Participant groups showed recurring differences in ECG assessment timing. The case study compares those schedules with the supplied protocol and explains why they warrant review.
Extend the depth of review as questions emerge
This focused investigation illustrates how study materials and data can drive additional review questions, with traceable evidence to help clinical, data management and biometrics teams evaluate the results. The paper describes the scope, findings and limitations in three pages.