Surveillance

Surveillance — also called post-marketing surveillance, adverse-event surveillance, or pharmacovigilance — is the systematic collection of harm reports generated during real-world use of a substance, after it has entered routine practice rather than controlled trial conditions.

In the evidence hierarchy, surveillance sits at the far end of the research pipeline. Trials generate data by design, on selected participants, over fixed windows. Surveillance generates data by accumulation: clinicians, patients, or manufacturers submit reports of events that occur in practice, and regulators gather and analyse the stream.

Because it runs at the scale of routine use — often across millions of people — it can surface effects too rare to appear in any trial cohort. That scale is also the source of its central limitation: the signal depends on someone recognising an adverse event, connecting it to a substance, and filing a report.

What this design can establish

A well-run surveillance system can establish that a particular adverse event has occurred in people taking a particular substance — and, when reports accumulate, that it occurs more often than chance alone would predict. This is a safety signal: not proof of causation, but a credible prompt for further investigation.

Surveillance can also document that an effect appears across populations, settings, and use patterns that no single trial covered — the real-world diversity that eligibility criteria deliberately exclude. Where reporting completeness is reasonably well understood, it can support frequency estimates for serious, recognisable events.

What it cannot

Surveillance cannot establish causation. A report records that a patient experienced an event while taking a substance; it does not establish that the substance caused the event. Confounding — other drugs, underlying conditions, timing coincidences — is never controlled for, and a rising signal warrants investigation, not automatic attribution.

It also cannot establish how often an effect occurs with any precision. Under-reporting is endemic: most adverse events, especially mild or delayed ones, are never submitted. An absence of reports is therefore not evidence that an effect is absent — it may mean no one recognised it, no one filed the form, or the effect has no obvious signature.

For substances that are not prescribed, not licensed, or not formally monitored — which describes most compounds catalogued here — surveillance data simply does not exist. An empty record means an absent watch, not an absent risk.