Almost every health headline reports a relative change: “twenty per cent lower risk”, “doubles the chance”. On its own that number is close to meaningless, because it says nothing about how likely the thing was to begin with. Twenty per cent off a one-in-two risk and twenty per cent off a one-in-a-million risk are the same headline and completely different news.
Put in the baseline and the reported percentage, and this returns the numbers that describe what actually happens to people.
In plain numbers
Out of 1,000 people, 20 would have had it. After a 20% reduction, 16 would. That is 4 people in 1,000 spared — 0.40 percentage points.
Why this matters. "Twenty per cent lower" sounds the same whether the risk was one in two or one in a million, and it is the only figure most reports carry. The absolute change is what a single reader experiences, and it is usually much smaller than the headline sounds. Both numbers are true; only one of them is informative on its own.
What this is not. This is arithmetic on numbers you typed. It knows nothing about the study, whether the finding is reliable, whether it applies to anyone like you, or what the treatment costs in side effects. A number needed to treat of 250 is excellent for a cheap, harmless pill and terrible for major surgery. Nothing here is medical advice — for that you need a clinician who can see your history.
Nothing you type here leaves your browser.
The three numbers
- Relative risk reduction — the headline. How much smaller the risk became, as a proportion of what it was. It is the largest of the three numbers, which is why it is the one that gets printed.
- Absolute risk reduction — the same change measured in people. Twenty per cent off a 2% risk is 0.4 percentage points: four people in a thousand. This is what a single reader experiences.
- Number needed to treat — how many people have to receive the treatment for one of them to get the benefit. It is simply one divided by the absolute reduction, and it is the number clinicians actually argue about.
Why a big NNT is not automatically bad
A number needed to treat of 250 sounds discouraging, and for major surgery it would be. For a cheap tablet with no meaningful side effects, given to a large population, it can be excellent public health — 250 people take something harmless and one avoids a serious illness. The figure only means something next to the cost, the harms and how bad the outcome is.
That is the honest limit of this tool. It converts one number into another. It cannot tell you whether the study was any good, whether it was large enough, whether the people in it resembled you, or what the treatment does to the people it does not help.
What to look for in a report
- Is the baseline risk given at all? If a report says a risk fell by a third and never says a third of what, the calculator cannot help and neither can the reader. That absence is itself the story.
- Risk of what, and over how long? A lifetime risk and a five-year risk are different quantities with the same units.
- Who was studied? A result in men over 65 is a result in men over 65.
- Was it people, or was it mice? A great deal of health reporting quietly loses this distinction between the paper and the headline.
None of this is medical advice, and this page cannot assess anyone’s risk. It is a way of reading a number. Decisions about your own health belong with a doctor or a pharmacist who can see your history.
Nothing you type here leaves the browser. Nothing is stored and nothing is sent to us.