Narratives are derived from multiple data sources including the clinical database and case report forms. Additionally, safety data base (ARGUS etc.) or CIOMS forms are used for further information on the AEs.
Crucially, output from those sources should be easily readable, and should be provided in a table format which allows filtering and sorting to facilitate retrieval of relevant information. If the medical writer has to scroll through unsorted and unfiltered heaps of data, the writing process is prolonged and not as efficient as it could be. Excel might be a good data format because the writer can customize the view for an optimal overview.
While using different data sources (e.g., CIOMS forms vs. clinical data base), discrepancies can occur, and there should be some person at the sponsor’s end, who takes the responsibility for the resolution and description of irreconcilable discrepancies.
If possible, narrative writing should not start before the data have already been cleaned and reconciled completely: it is much more efficient to write narratives on the basis of clean and reconciled data. However, if timelines are short and if numbers of narratives are high (e.g. > 150), it might be prudent to start the writing process already with unclean data. Obviously, if written based on pre-data-base lock data, narratives have to be updated or modified upon data-base lock.