Usage Examples
- Signal detection on that substance runs quarterly, not six-monthly, because it is on the additional monitoring list.
- The disproportionality score flagged the combination, but signal detection is not finished until someone reads the cases.
- We detected it in EudraVigilance on the 3rd, validated it on the 9th, and notified the Agency the same day.
What is Signal Detection?
Signal detection is the pharmacovigilance activity of searching safety data for evidence that a medicine may cause a new adverse reaction, or that a known reaction has changed, before that evidence is confirmed.
Signal detection exists because a marketing authorisation is granted on a safety picture built from a few thousand trial subjects studied for a limited period. Reactions that occur once in 10,000 patients, emerge only after years of exposure, or affect populations the protocol excluded cannot be seen before approval. Signal detection is the mechanism that finds them once the product reaches real-world use.
Signal detection covers looking for and identifying signals using data from any source: individual case safety reports, EudraVigilance and other spontaneous reporting databases, published literature, active surveillance, and studies. Signal detection stops at identification. Validating the evidence, confirming the signal, prioritising it, assessing causality, and deciding regulatory action are separate activities within signal management, and a detected signal is not yet a risk.
Signal detection in practice runs on a documented periodicity set per active substance rather than ad hoc. Teams combine review of individual case safety reports with statistical measures such as disproportionate reporting, apply clinical judgement to the drug-event combinations that surface, and record the method used, the frequency chosen, and the justification for that frequency in their written procedures.
Not to be confused with
- Signal validation
- Signal detection identifies a candidate association; signal validation then evaluates whether the available documentation contains sufficient evidence of a new potentially causal association to justify further analysis. A detected signal that fails validation is a non-validated signal, and no notification obligation follows.
- Pharmacovigilance
- Pharmacovigilance is the whole discipline of monitoring medicine safety. Signal detection is one activity inside its signal management process, which in the EU also spans validation, confirmation, analysis and prioritisation, assessment, and recommendation for action.
- Adverse event reporting
- Adverse event reporting is a case-level duty with a per-case clock, such as the 15 calendar days for serious and unexpected experiences under 21 CFR 314.80(c)(1)(i). Signal detection is population-level: it looks across accumulated cases for a pattern. Meeting every reporting deadline proves nothing about detection.
- Emerging safety issue
- An emerging safety issue is defined by urgency rather than novelty and carries its own written notification to the Agency and affected Member States no later than 3 working days. Signal detection carries no such clock; its output feeds validation first.
Signal detection obligations differ by jurisdiction; these are the operative anchors for a marketing authorisation holder with EU and US products.
What you must do
- 1Monitor EudraVigilance continuously at a frequency proportionate to the identified risk, at minimum every 6 months per active substance, and more frequently for substances in products on the additional monitoring listEU GVP Module IX Sec IX.C.3.2
- 2Document the monitoring frequency chosen for each active substance and the justification for it, and update both as the risk profile changesEU GVP Module IX Sec IX.C.3.2
- 3Validate any new signal detected in EudraVigilance and inform the Agency and national competent authorities forthwith; where you conclude the authorities must analyse it further, file a standalone signal notification no later than 30 days after completing your assessmentEU GVP Module IX Sec IX.C.4
- 4Notify an emerging safety issue in writing to the Agency and to the competent authorities of each Member State where the product is authorised no later than 3 working days after establishing that it meets the definitionEU GVP Module IX Sec IX.C.2
- 5Promptly review all adverse drug experience information received from any source, foreign or domestic, for every approved US application21 CFR 314.80(b)
- 6Report each serious and unexpected adverse drug experience as soon as possible and no later than 15 calendar days from initial receipt21 CFR 314.80(c)(1)(i)
Common mistakes
Letting the algorithm be the detection process
Disproportionality output is a screening aid, not a signal. ICH E2E states that data mining techniques should always be used in conjunction with, and not in place of, analyses of single case reports, and that the method does not quantify the magnitude of risk. A programme that escalates only what crosses a threshold will miss signals visible in a handful of well-documented cases and will drown a small safety team in statistical noise.
Reading a quiet database as a safe product
ICH E2E states plainly that the absence of a signal does not mean that a problem does not exist. Spontaneous reporting rates move with time since launch, media attention, and regulatory activity, so a flat month reflects reporting behaviour as much as product safety. Treating "no signal detected" as a positive finding is how under-reporting becomes a false assurance in a PSUR.
Running one monitoring frequency across the whole portfolio
EU GVP Module IX requires the frequency to be proportionate to the identified risk, with more frequent monitoring for substances on the additional monitoring list, and requires the chosen frequency and its justification to be documented. A single default cadence applied to a 15-year-old generic and a newly launched biologic alike is an inspection finding waiting to happen, and it delays the notification clock on the products where delay costs most.
When This Matters
- Signal detection on that substance runs quarterly, not six-monthly, because it is on the additional monitoring list.
- The disproportionality score flagged the combination, but signal detection is not finished until someone reads the cases.
- We detected it in EudraVigilance on the 3rd, validated it on the 9th, and notified the Agency the same day.
Frequently Asked Questions
Signal detection identifies a candidate association; signal validation then tests whether the documentation behind it contains sufficient evidence of a new potentially causal association, or a new aspect of a known one, to justify further analysis. Detection produces candidates. Validation decides which of them survive as validated signals.
