Number needed to treat: how many for one to benefit

Here is one of the most clarifying questions you can ask about any treatment: how many people have to take it for one person to be helped? That number is the number needed to treat, or NNT, and it turns a slippery percentage into something you can picture. If a drug has an NNT of 10, then ten people take it for one to benefit. The other nine get nothing, apart from the cost and any side effects.

The NNT is useful because it resists the spin that percentages invite. Cuts your risk in half sounds like a sure thing. But if it means one in a hundred instead of two in a hundred, then a hundred people take the drug for one to be spared, an NNT of 100. Same drug, and suddenly you can see how little most people get. The NNT quietly folds in the starting risk that a bare percentage hides.

NNTs vary enormously, and a feel for the range is worth having. Some treatments are close to miraculous: an antibiotic for a serious infection can have an NNT near 1, meaning almost everyone treated benefits. Others are far higher than people assume. A daily pill taken for years to prevent one heart attack in a low-risk person can have an NNT in the hundreds. Neither number is bad by itself. A high NNT can still be worth it for a cheap, safe drug against a dreadful outcome. The point is that the number lets you weigh the deal honestly.

There is a companion worth knowing too: the number needed to harm, how many people you treat before one is hurt by a side effect. A treatment that helps one in ten but harms one in eight is a very different proposition from what the headline percentage suggests. So when you read that a drug works, look for the NNT, and if you can find it, the number needed to harm. How many for one to benefit, and how many before one is harmed. Those two numbers tell you more than any percentage.