From one member of the GCP Com. >>
The start and stop dates on which the event meets the seriousness criteria would be part of the narrative (if relevant) but the duration of the adverse event in itself, the contemporaneousness with study drug intake, and the evaluation of concurrent events are of greater relevance in the clinical assessment.
With respect to the point about 'underlying adverse event', we need to remember that all SAEs are fundamentally AEs. Clinical trial protocols will specify the study duration for which AEs need to be followed up after the last clinic visit - this is often until resolution or stabilization. The same rules apply to SAEs as the SAE is not cut off from the AE. It is the AE - it lasts for as long as the AE does.
With respect to the point about the clinical database, the start and stop dates of the AEs in the clinical database and the AEs in the safety database are often reconciled as part of the process of SAE Reconciliation at the end of the study.
As per ICH E2A, the duration of the reaction is the significant data point. Hence, the stop date (and time) of the reaction (event) are required.
Some guidance is also found in ICH E2B (R3) about how to report start and end dates of related events :
'If the events are thought to be related (i.e., if event1 is a sign or symptom of event2), it would be clinically reasonable to use the earliest start date or latest end date, as relevant, for both events. However, a sender should not infer dates unless there is a clear clinical rationale and this rationale should be stated in the case narrative'.