[Durgical](https://durgical.com/en/) › [DurgiPedia](https://durgical.com/en/durgipedia/) › [Epidemiology](https://durgical.com/en/epidemiology/) › [Measures of frequency and association](https://durgical.com/en/epidemiology/areas/measures/) › Number needed to harm

# Number needed to harm

- Measure

How many people must be treated, over a given time, for one to suffer an adverse effect caused by the treatment.

Updated Nov 5, 2026

## Key points

- NNH = 1 / absolute risk increase (risk of harm in the treated − in controls).
- A high NNH is good: the harm is rare.
- Compare it with the NNT from the same study and period.
- If the NNT is smaller than the NNH, more people benefit than are harmed, but the severity of each outcome matters.
- Always report it with the adverse effect and the period.

On this page

- The basics Key points
- What it is
- Formula and reading
- Worked example
- NNT and NNH from the example, per 1,000 treated for one year
- Markers
- How to read it
- Common pitfalls
- Memory aid

- Study Test yourself

- Connections Connections

## What it is

The number needed to harm (NNH) is the inverse of the absolute risk increase for an adverse effect: the risk of the event in the treated minus the risk in controls. It is the mirror image of the NNT. A high NNH means harm is uncommon. Putting NNT and NNH side by side, over the same period, shows how many people benefit for each one harmed.

## Formula and reading

- ARI (absolute risk increase) = EER − CER for the adverse effect; NNH = 1 / ARI.
- Events per 1,000 treated: benefits = 1,000 / NNT; harms = 1,000 / NNH.
- Reading: an NNH of 40 over one year = one extra harm for every 40 people treated for a year.

## Worked example

- Data: an anticoagulant is tested for one year. Major bleeding: 4% in the treated and 1.5% in controls.
- Absolute risk increase = 0.04 − 0.015 = 0.025.
- NNH = 1 / 0.025 = 40 over one year.
- The same trial lowers embolism from 9% to 4%: ARR = 0.05 and NNT = 1 / 0.05 = 20.
- Per 1,000 treated for a year: 1,000 / 20 = 50 embolisms prevented and 1,000 / 40 = 25 bleeds caused.
- Reading: for each extra major bleed, 2 embolisms are prevented; the decision depends on how serious each is.

## NNT and NNH from the example, per 1,000 treated for one year

| Measure | Value | Events per 1,000 treated |
| --- | --- | --- |
| NNT (embolism prevented) | 20 | 50 prevented |
| NNH (major bleed caused) | 40 | 25 caused |

## Markers

- NNH
- ARI

## How to read it

- Clinical practice guidelines weigh the NNT against the NNH together with the severity of each outcome and patient preferences.
- A mild adverse effect with a low NNH may be acceptable; a serious one with a high NNH may not.

## Common pitfalls

- Comparing a 5-year NNT with a 6-month NNH.
- Reading a low NNH as good news.
- Subtracting the risks in the wrong order and getting a negative number.

## Memory aid

Want the NNT small and the NNH large: few treated per benefit, many treated per harm.

## Test yourself

Think of the answer, then tap the question to see it.

## An adverse effect occurs in 7% of the treated and 2% of controls. What is the NNH?

Absolute increase = 0.05; NNH = 1 / 0.05 = 20.

## Is an NNH of 500 better or worse than one of 50?

Better: the harm is less frequent; 500 people must be treated to cause one extra adverse effect.

## What must be true to compare the NNT with the NNH?

They must come from the same population and the same follow-up period.

[Add to my cards (3)](https://durgical.app/estudiar/?add=/en/epidemiology/number-needed-to-harm/)

## Connections

## Related 4

- [Number needed to treat](https://durgical.com/en/epidemiology/number-needed-to-treat/)
- [Absolute risk reduction](https://durgical.com/en/epidemiology/absolute-risk-reduction/)
- [Attributable risk and attributable fraction](https://durgical.com/en/epidemiology/attributable-risk-and-attributable-fraction/)
- [Phases of clinical trials](https://durgical.com/en/epidemiology/phases-of-clinical-trials/)

[← All of epidemiology](https://durgical.com/en/epidemiology/)

Educational content on epidemiological and statistical methods; the examples use made-up numbers to teach the calculation. Not a clinical or legal guide: current Colombian rules are in the official sources.

Sources: [Chapter 15: Interpreting results and drawing conclusions](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-15) (Cochrane, 2024); [Chapter 6: Choosing effect measures and computing estimates of effect](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-06) (Cochrane, 2024).

## **Spotted a mistake?** We are always improving for the community. Tell us what is wrong and our team will review it. Report a mistake

Representative model, for educational purposes only.

---
Source: https://durgical.com/en/epidemiology/number-needed-to-harm/
License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Cite as: DurgiPedia (durgical.com), https://durgical.com/en/epidemiology/number-needed-to-harm/