Number Needed to Treat: How Much a Treatment Really Helps

Your clinician suggests a daily pill to lower your risk of a future problem. You want to know: how likely is it to help me? A simple number can answer that question. It is called the number needed to treat. This guide explains what it is, how to work it out, and how to use it to weigh benefits against harms.

What number needed to treat means

The number needed to treat (NNT) is the number of people who need to receive a treatment for one extra person to benefit, compared with not treating or with another treatment. [1] [2]

For example, an NNT of 25 means that if 25 people take the treatment for the time studied, 1 more person avoids the bad outcome than if none of them took it. The other 24 do not get that benefit. Some of them would have been fine anyway, and some have the outcome despite treatment.

Researchers introduced NNT in the late 1980s because it is easier to use in practice than relative risk. [1]

How to work it out

NNT comes from the absolute risk reduction. This is the plain difference in risk between two groups. [2] [4]

The formula is simple: NNT equals 100 divided by the absolute risk reduction in percentage points.

A worked example

These numbers are made up to show the math.

• Without treatment, 4 in 100 people have a stroke over 5 years.
• With treatment, 3 in 100 people have a stroke over 5 years.
• The absolute risk reduction is 1 percentage point.
• The NNT is 100 divided by 1, which equals 100.

So 100 people need to take the treatment for 5 years to prevent 1 stroke.

The relative risk reduction in this example is 25 percent, because 1 is a quarter of 4. "Lowers stroke risk by 25 percent" sounds much larger than "100 people treated for 5 years to prevent 1 stroke". Both are true.

Starting risk changes the NNT

Now imagine a higher risk group, where 20 in 100 people have a stroke over 5 years. If the treatment still lowers risk by 25 percent, the risk falls to 15 in 100. The absolute reduction is 5 percentage points. The NNT is 100 divided by 5, which is 20.

The same treatment, with the same relative effect, has an NNT of 100 in the lower risk group and 20 in the higher risk group. This is why guidelines often recommend treatments for people at higher risk. [2]

Number needed to harm

The number needed to harm (NNH) works the same way, for side effects. It is the number of people who need the treatment for one extra person to have a certain harm. [2]

A larger NNH is better, because harm is rarer. For example, an NNH of 200 for a stomach bleed means 1 extra person in 200 has a bleed.

Comparing NNT and NNH helps you see the trade-off. If 50 people need treatment for 1 to benefit, and 1 in 200 has a serious harm, then for every 4 people who benefit, about 1 person is harmed. Whether that is acceptable depends on how serious each outcome is.

What doctors look at

When clinicians use NNT, they check several things. [2]

• The outcome: preventing death is very different from preventing a mild symptom.
• The time period: an NNT over 1 year is not the same as an NNT over 10 years.
• The comparison: is the treatment compared with placebo, no treatment, or another drug?
• The starting risk: does the study group look like you?
• The harms: what is the NNH for the most important side effects?
• The certainty of the evidence: how strong are the studies behind the numbers?

Research on risk communication shows that people understand numbers better when they see absolute risks, such as "1 in 100", rather than only relative risks. [3] NNT is one way to show absolute effect. Some people find "out of 100 people" easier to picture than NNT. Ask for whichever format works for you.

Limits of NNT

• NNT is an average for a group. It does not tell you whether you will be the one who benefits.
• NNT from short trials can be hard to apply to long-term use.
• Comparing NNTs across different studies can mislead if the outcomes, time periods, or groups differ.

How to use NNT in a decision

• Ask what outcome the treatment prevents, and over what time.
• Ask for the NNT, or the absolute risk with and without treatment.
• Ask about the most important harms and their NNH.
• Think about how much each outcome matters to you.
• Consider cost, effort, and how long you will take the treatment.

Questions to ask your clinician

• What is my risk of this problem over the next 5 or 10 years without treatment?
• How many people like me need to take this for one to benefit?
• What are the most common and the most serious side effects, and how often do they happen?
• How long would I need to take this treatment?
• Are there other options with a better balance of benefit and harm?
• How sure are we about these numbers?

Bottom line

The number needed to treat shows how many people must take a treatment for one extra person to benefit. It depends on your starting risk, the outcome, and the time period. Compare it with the number needed to harm. Together, these numbers help you see the real size of a benefit and decide what is right for you.

Primary AI can show you the guideline and study sources behind this topic in plain language, so you can bring better questions to your appointment.