NNT / NNH Calculator
Enter event rates (as percentages) for a control group and treatment group. This tool calculates absolute risk reduction, relative risk reduction, and the Number Needed to Treat (or Harm).
ARR = CER − EER
NNT = 1 / ARR (when ARR > 0)
NNH = 1 / |ARR| (when ARR < 0)
Educational use only. Always combine NNT with confidence intervals, baseline risk, and clinical judgment.
What is the Number Needed to Treat (NNT)?
The Number Needed to Treat tells you how many patients must receive a therapy for one additional patient to benefit compared with a control condition. It translates study results into something practical and easy to communicate.
If a medication lowers bad outcomes from 20% to 10%, the absolute reduction is 10 percentage points (0.10). The NNT is 1 / 0.10 = 10. That means treating ten people leads to one extra person benefiting over the studied time horizon.
How this calculator works
Step 1: Enter CER and EER
CER (Control Event Rate) is the event frequency in the control/placebo/usual-care group. EER (Experimental Event Rate) is the event frequency in the intervention group.
Step 2: Compute risk difference
The calculator finds Absolute Risk Reduction (ARR):
- If ARR is positive, treatment reduces events and NNT is reported.
- If ARR is negative, treatment increases events and NNH (Number Needed to Harm) is reported.
- If ARR is zero, there is no measurable difference and NNT is effectively infinite.
Step 3: Round for interpretation
NNT/NNH is conventionally rounded up to the nearest whole person, because you cannot treat a fraction of a patient.
Interpreting NNT in context
A lower NNT often indicates a stronger treatment effect, but lower is not always automatically “better.” Interpretation depends on:
- Severity of outcome: Preventing stroke is different from reducing mild discomfort.
- Treatment burden: Cost, side effects, route, and duration matter.
- Population baseline risk: NNT shifts across low-risk vs high-risk groups.
- Follow-up time: NNT over 1 year can differ from NNT over 5 years.
Example calculation
Suppose a trial reports:
- CER = 25%
- EER = 15%
ARR = 25% − 15% = 10% = 0.10. NNT = 1 / 0.10 = 10. Clinical reading: about 10 people need treatment to prevent one additional event during the trial period.
When to use NNH instead of NNT
If the treatment group has more adverse events than control (EER > CER), the intervention may be harmful for that endpoint. In this case, report Number Needed to Harm:
- NNH = 1 / absolute risk increase
- Smaller NNH indicates harms appear more frequently
Good decisions often compare both benefit and harm side-by-side (for example, NNT for preventing hospitalization vs NNH for serious adverse effects).
Common pitfalls
- Using relative risk reduction alone without absolute risks.
- Ignoring confidence intervals around ARR and NNT.
- Applying trial NNT values to very different patient populations.
- Forgetting the timeframe (30 days, 1 year, lifetime, etc.).
Practical takeaway
NNT is one of the most useful evidence-based medicine metrics because it turns abstract percentages into a concrete treatment impact estimate. Use it with outcome importance, patient values, adverse effects, and cost to make better decisions.