“Reality is frequently inaccurate” A case study examining the whens and whys of post-live database changes in a UK clinical trials unit *Douglas Adams

Introduction: Accurately estimating the costs of clinical trials is challenging. There is currently no reference class data to allow researchers to understand the potential costs associated with database change management in clinical trials. Methods: We used a case-based approach, summarising post-live changes in eleven clinical trial databases managed by Sheffield Clinical Trials Research Unit. We reviewed the database specifications for each trial and summarised the number of changes, change type, change category, and timing of changes. We pooled our experiences and made observations in relation to key themes. Results: Median total number of changes across the eleven trials was 71 (range 40–155) and median number of changes per study week was 0.48 (range 0.32–1.34). The most common change type was modification (median 39, range 20–90), followed by additions (median 32, range 18–55), then deletions (median 7, range 1–12). In our sample, changes were more common in the first half of the trial's lifespan, regardless of its overall duration. Trials which saw continuous changes seemed more likely to be external pilots or trials in areas where the trial team was either less experienced overall or within the particular therapeutic area. Conclusions: Researchers should plan trials with the expectation that clinical trial databases will require changes within the life of the trial, particularly in the early stages or with a less experienced trial team. More research is required to understand potential differences between clinical trial units and database types.