Literature Screening in Practice: Tools & Workflow

Everything else in this module has focused on statistical signal detection — PRR, ROR, EBGM, IC — run against spontaneous report databases. Literature screening is the other major signal source, and it runs on an entirely different rhythm: not statistical thresholds, but systematic, recurring human review of published scientific literature, required under GVP Module VI at a minimum weekly cadence.

In practice, this means a defined search strategy run against databases like Embase and Medline (sometimes both, since neither indexes everything the other does), covering the product’s active substances and known safety topics. Results get deduplicated — the same publication indexed in both databases, or citing an already-known case — and then screened, often by two reviewers working independently before a result is confirmed as reportable or dismissed, precisely the kind of dual-check quality control this course has emphasized throughout.

AI has changed the volume problem more than it’s changed the judgment problem here. Tools like Clarivate’s Drug Safety Triager and platforms such as Clinevo, Compier, or QUOSA can pre-screen thousands of abstracts and flag likely candidates far faster than manual review alone, which matters given how much published literature exists across even a modest product portfolio. But the AI output is a triage layer, not a decision layer — every flagged article still needs a human reviewer to confirm it actually describes a valid case (the same four minimum criteria from Module 1 apply) or a genuine signal candidate, exactly the human-in-the-loop requirement covered in Module 13.

The stakes are real: a case reportable from literature that’s missed by a weak or inconsistent screening process is treated by inspectors the same way a missed spontaneous report would be — as a gap in the PV system, not a lesser category of miss. That’s why literature screening, despite running on a calendar rather than reacting to an inbound report, gets audited with the same rigor as intake triage.