Something in the Pattern

Part of Pharmacovigilance and Drug Safety (PV) Training: Signal Detection and Risk.

About 2 minutes. Free, in your language, no account needed.

What a signal is

A signal is information suggesting a new potentially causal association, or a new aspect of a known association, that is judged to warrant further investigation.

Read that carefully. It is not a confirmed harm. It is a reason to look properly.

Where signals come from

Disproportionality in the case data. An event occurring more often with your product, relative to all other products in the database, than would be expected. This is the quantitative side, and it is a screening tool rather than an answer.

A small number of striking cases. Three well documented cases of a rare condition with a plausible mechanism and a positive dechallenge can be a stronger signal than a hundred cases of something common.

A single case, occasionally. Certain events are so rarely spontaneous that one well documented case is a signal in itself. Aplastic anemia, Stevens-Johnson syndrome, an unusual liver injury.

The literature, regulators and partners. Signals arrive from outside as well.

Why numbers alone are not enough

Disproportionality measures reporting, not occurrence. Reporting rises after media coverage, after a label change, after a competitor's problem, and in the first years of a product's life. A statistic that moves for any of those reasons looks the same as one moving because something is happening to patients.

This is why a quantitative flag is reviewed by somebody who knows the product and the therapeutic area, and not acted on by itself.

The case processor's contribution

Signals are found in the accumulated data, which is made of individual cases. Consistent coding, recorded outcomes, honest dechallenge fields and narratives that carry the sequence are what make a signal visible.

Sloppy individual case work does not produce visibly bad cases. It produces a signal that nobody finds.