AN INDEPENDENT PUBLIC HEALTH READING ROOM
Places. People.
Possibilities.

A Number Needed to Treat Needs an Outcome and a Time Period

Paper collage of a notebook, tablet, magnifying glass and source card.
AI-created conceptual illustration of a fictional setting; not a record of people or outcomes.

A number needed to treat can look like a simple score for comparing treatments. Without the outcome, comparison and follow-up period, however, the number has lost much of its meaning.

The Cochrane Handbook explains NNT in relation to absolute differences in event risks. Keep the relevant population and time horizon attached when reading it; it is not a universal property of a medicine.

Consider an invented arithmetic example, not treatment evidence. Over a stated period, an undesirable event occurs in 10 of 100 people in one group and 5 of 100 in the other. The absolute difference is 0.05, and its reciprocal is 20. That calculation describes one additional avoided event per 20 people on average under those example conditions.

It does not identify which person benefits, promise an outcome for the twentieth person, or include every possible harm. It also does not tell you whether the underlying comparison was reliable.

Before reusing a published NNT, record the event definition, alternative treatment or control, follow-up duration and uncertainty. If two reports use different outcomes or periods, resist placing their numbers in a simple best-to-worst list.

For a public summary, keep the underlying risks alongside the NNT where available. Readers can then see the scale of the comparison. Decisions about an individual’s care require the broader evidence and discussion with an appropriate health professional.

Image: an editorial illustration, not a documentary photograph.