What would you investigate when a safety signal appears in spontaneous reports but not in clinical trials?

Why

Trials and spontaneous reports answer different questions. A trial of 3,000 people for one year has little chance of showing an event that occurs in 1 in 10,000 users. Trials often exclude older patients, people with kidney or liver disease, pregnant women and those on many other drugs, so an interaction may only appear after approval.

Long latency events may not occur within trial follow-up. So absence in trials is weak evidence of absence. The signal still needs scrutiny.

Review cases one by one for time to onset, dechallenge and rechallenge, alternative causes and quality of information. Check whether publicity or a regulatory communication stimulated reporting, whether duplicates inflate counts, and whether the event is part of the treated disease. GVP Module IX describes the EU steps of validation, confirmation, analysis and prioritisation, and assessment, and signals that meet the definition of an emerging safety issue must be notified quickly.

Strengthening evidence may need a pharmacoepidemiological study, because spontaneous reports cannot give incidence.