Reading the Evidence
Number Needed to Treat and the Headline It Deflates
A treatment that halves your risk sounds transformative until you ask how many people have to take it for one to benefit. That number is calculable and almost never quoted.
The options around number needed to treat as a way of reading results are set out side by side below, with the conditions that genuinely favour one over the other.
The difference in one place
- The figure counts how many must be treated for one additional person to benefit.
- It falls when the baseline risk is high and rises when the risk is low.
- The equivalent number for harm exists and is rarely presented alongside it.
What the number expresses
The number needed to treat counts how many people must receive a treatment for one additional good outcome to occur. It is calculated directly from the difference in outcome rates between the treated and untreated groups. A number of ten means nine of those people took the treatment without gaining the benefit being counted.
That framing is uncomfortable and it is a more faithful description of what a trial found than a percentage reduction. It exists because relative reductions, discussed elsewhere on this site, systematically make small differences sound large.
Why the same treatment gives different numbers
The figure depends on how likely the outcome was in the first place, which varies enormously between populations. A treatment halving a high risk produces a small number, and the same treatment halving a tiny risk produces a large one. This is why the same medicine can be excellent for people at high risk and barely worthwhile for people at low risk.
Where the claim is carefully worded, it also explains why guidelines recommend treatments by risk category rather than for everyone. Quoting a number without stating the population it came from is close to meaningless.
The number nobody prints
The number needed to harm counts how many people must be treated for one additional bad outcome to occur. Presenting benefit and harm figures side by side is the only way to see the actual trade being offered.
Harm figures are usually less precisely estimated, because trials are sized for benefit rather than for side effects. Rare serious harms often emerge only after a treatment is widely used, which is what surveillance systems exist for. A discussion of benefit without any corresponding discussion of harm is an incomplete account of the same trial.
Applying it to supplements
Very few supplement trials report this figure, partly because their outcomes are surrogates rather than events. Where a supplement trial has found a benefit on a real outcome, the numbers have often been large.
Follow the mechanism and a large number is not the same as no benefit, and it changes how you weigh the cost, the effort and any risk. For products with an unproven benefit and any real risk, the calculation cannot even be started.
Asking how many people would need to take this for one to benefit reframes almost every wellness claim usefully.
Where it applies to behaviour too
The same arithmetic applies to screening, where many people are tested for each one who benefits. Screening also carries harms in false positives, unnecessary procedures and diagnosis of conditions that would never have caused problems. Good screening programmes publish these figures, and the honest ones present benefit and harm together.
It is why reasonable people can decline a screening test after reading the numbers, which is a legitimate decision. These are decisions to make with a clinician who knows your risk, rather than from a percentage in a headline.
Nothing here is a diagnosis or a treatment recommendation; a doctor or pharmacist is the person to ask about your own situation.
How to use it as a reader
When you see a percentage reduction, look for the underlying rates in both groups, which are usually in the paper. The difference between those rates, inverted, gives you the number, and the arithmetic is genuinely simple. If the article does not give you the rates, that omission is the most informative thing in it.
Compare against how much the treatment costs, in money, effort and side effects, over the period involved. This one calculation dissolves more misleading health headlines than any other single technique.
Side by side
| Consideration | What it means in practice |
|---|---|
| What the number expresses | The figure counts how many must be treated for one additional person to benefit. |
| Why the same treatment gives different numbers | It falls when the baseline risk is high and rises when the risk is low. |
| The number nobody prints | The equivalent number for harm exists and is rarely presented alongside it. |
The takeaway
Halves your risk tells you nothing until you know the risk. Find the two rates, take the difference, and invert it.
An extraordinary mechanism needs better evidence than testimonials, and usually has less.
Questions readers ask
Is a large number always bad?
No. For a cheap, safe treatment preventing something serious, a large number can still be worthwhile at a population level. It becomes a value judgement once you can see the actual figure.
Why is it rarely reported?
Relative reductions produce larger and more quotable numbers, and press releases favour them. The underlying rates are usually available in the paper if you look.
