On this page
Number needed to harm (NNH) is the expected number of patients who must be exposed to a treatment before one experiences an adverse effect that would not have occurred in a control group. It translates adverse event rates into a single, tangible number for clinical decision-making.
- NNH is the reciprocal of absolute risk difference: divide 1 by the percentage point difference in adverse event rates between treatment and control groups.
- A treatment with NNT of 10 and NNH of 200 has a favorable profile because benefits occur far more frequently than harms.
- NNH applies only to the specific adverse event studied and the populations in the trial; it does not generalize automatically to other groups or longer time frames.
- Treatment effect heterogeneity means individuals within a trial population experience different risk from the same treatment, so published NNH averages may not represent every patient.
- NNH and relative risk reductions can tell opposite stories: a large relative reduction can coexist with a high NNH if baseline risk is low.
Clinical trials and observational studies report adverse events as percentages or relative risks. A treatment might cause nausea in some percentage of patients, or double the risk of infection compared to placebo. These figures are mathematically correct but psychologically opaque. A clinician reading that a drug doubles infection risk cannot immediately tell whether that means one patient in a hundred will be harmed or one in ten.
Number needed to harm (NNH) exists to solve that problem. It translates adverse event rates into a single number: how many patients must receive the treatment before one experiences a harm the control group would not. That number makes risk tangible and comparable across treatments. This article explains what NNH measures, how it is calculated, why it matters, and the common mistakes that arise when interpreting it.
What number needed to harm actually measures
Number needed to harm (NNH) is the expected number of patients who must be exposed to a treatment or risk factor for one additional adverse event to occur versus a control condition. It answers a straightforward clinical question: if I give this drug to my patients, how many must I treat before one of them experiences this specific adverse effect that would not have occurred under the alternative treatment?
NNH serves the same interpretive role for harms that number needed to treat (NNT) serves for benefits. Where NNT counts how many patients must receive an intervention to prevent one bad outcome, NNH counts how many must receive it to cause one additional bad outcome. Both measure absolute rather than relative risk, which makes them directly comparable to each other and to the baseline risk in a given population.
The measure applies only to the specific adverse event being counted and only within the populations studied. A treatment might have an NNH of one value for a common side effect but a much lower NNH for a serious adverse reaction in a subgroup at higher baseline risk. NNH is not a property of the treatment alone; it emerges from the comparison between two groups under specific conditions.
How to calculate number needed to harm
NNH is derived from the difference in adverse event rates between the treatment group and the control group. The formula is simple, but understanding what goes into it clarifies what the result means.
The NNH formula and its inputs
The absolute risk difference is the foundation of NNH. It is the percentage of patients in the treatment group who experienced an adverse event minus the percentage in the control group who experienced the same event. Once you have that difference, NNH is its reciprocal: 1 divided by the absolute risk difference.
The formula appears simple because it is. The subtlety lies in what you feed into it. Both the treatment adverse event rate and the control adverse event rate must come from the same study or comparison, measured the same way, in populations as similar as possible. If the control group is healthier, younger, or selected for lower baseline risk, the adverse event rate will be artificially low, and the NNH will be artificially high, making the treatment appear safer than it truly is in a more typical population.
Worked example with straightforward numbers
Suppose a randomized trial compares a new arthritis medication to placebo. In the treatment group, 16 out of 100 patients develop a rash. In the placebo group, 6 out of 100 develop a rash. The absolute risk difference is 16 percentage points minus 6 percentage points, which equals 10 percentage points.
If you treat 10 patients with this medication, you expect one additional case of rash compared to treating them with placebo. This same treatment will appear throughout the following sections so that you can see how NNH relates to benefit measures and how clinicians use it in practice.
Why NNH matters in clinical decisions
NNH forces explicit calculation of the absolute frequency of harm. A treatment that doubles the risk of an adverse event sounds dangerous; if that event occurs in a small percentage of untreated patients and the treatment increases it by a small percentage, the NNH may still be high, and the treatment may be worthwhile if the benefit is large. Conversely, a treatment that increases a common side effect by a small relative percentage can have a low NNH if the baseline event rate is high.
Clinicians use NNH to decide whether a treatment is worth offering to a population. If a medication prevents serious outcomes but causes adverse effects in a proportion higher than it prevents those outcomes, that unfavorable risk profile should be apparent in the NNT and NNH comparison. Without explicit calculation, relative risk figures can obscure this balance. NNH makes the tradeoff visible and quantitative, which is essential when patients and clinicians are discussing whether the benefit justifies the risk.
How NNH relates to number needed to treat
Number needed to treat (NNT) is the reciprocal of the absolute risk reduction for the benefit being studied. It tells you how many patients must receive an intervention to prevent one adverse outcome or achieve one positive outcome. NNH applies the same logic to harms.
Clinicians compare NNT with NNH to judge whether benefits outweigh harms. A treatment with an NNT of 10 and an NNH of 200 means that for every ten patients treated, one experiences the desired benefit, but you must treat 200 to cause one additional adverse effect. That is a favorable profile: benefits outweigh harms substantially. Conversely, a treatment with an NNT of 50 and an NNH of 20 causes more harm than benefit, because adverse effects occur in one out of every 20 patients while the desired outcome occurs in only one out of 50.
The comparison becomes more complex when a single treatment produces multiple benefits and multiple harms. An antidepressant might have an NNT of 9 for symptom improvement but also carry different NNH values for various adverse effects in different subgroups. Clinicians then weigh the severity and reversibility of each outcome, the patient's preferences, and the availability of alternatives.
Common misreadings and limitations of NNH
The most frequent error is treating NNH as if it applies to all patients equally. The measure is an average across the trial population. Treatment effect heterogeneity, the nonrandom, explainable variability in the direction and magnitude of treatment effects for individuals within a population, means that individuals within the population experience different risk from the same treatment. A drug might have an NNH of 50 for an adverse effect in the overall trial population, but in patients over 65 with prior conditions, the NNH might be much lower. A clinician must look beyond the published NNH to subgroup analyses and baseline risk profiles.
Another common mistake is confusing NNH with the incidence of adverse events in absolute terms. An NNH of 20 does not mean that percentage of patients will be harmed; it means that treating 20 patients will produce one additional adverse event compared to the control. If the control group has a certain adverse event rate and the treatment group has a higher rate, the NNH is the reciprocal of the difference, but a much larger percentage of treated patients still experience the event.
NNH is also specific to the time frame and population studied. A medication tested over 12 weeks may have a different NNH if followed for five years, as adverse effects accumulate or tolerance develops. Populations with different comorbidities, concurrent medications, or genetic backgrounds will have different NNH values for the same treatment. Applying NNH from one population to another without checking the similarities is a common source of misjudgment in clinical practice.
NNH in context with absolute risk reduction and relative risk
Absolute risk reduction (ARR) is the starting point for NNH. ARR is the difference between the control event rate and the experimental event rate. NNH is simply the reciprocal of ARR, so the two are inseparable: they contain the same information, expressed differently. Where ARR tells you the percentage point reduction in risk, NNH tells you the number of patients you must treat to achieve that reduction.
Relative risk reduction (RRR) and relative risk (RR) are different measures. They compare the treatment group's event rate to the control group's rate as a ratio rather than a difference. A treatment that reduces an event from a higher rate to a lower rate has one relative risk reduction as a percentage, but an absolute risk reduction measured in percentage points and a corresponding NNH. The same treatment can have a dramatic-sounding relative risk reduction and a much less impressive absolute number needed to treat, depending on the baseline risk. RRR and RR are useful for understanding mechanistic efficacy, but ARR and NNH are more useful for clinical decisions, because they show the actual frequency of benefit in the population being treated.
Do this automatically
Calculating number needed to harm requires only basic arithmetic, subtracting the adverse event rate in the control group from the treatment group rate, then taking the reciprocal, yet clinicians and researchers must perform these calculations across thousands of studies, trials, and safety reports to inform treatment decisions. Software can systematize this computation across large datasets, automatically extracting adverse event rates from trial data and generating NNH values for rapid clinical comparison. Healthcare research runs this kind of analysis end to end and cites every figure back to its original source, across as many subjects as you need, with Mimir attaching a citation to every figure. It will not predict a trial's outcome or advise on anyone's treatment. Those decisions stay with clinicians and regulators, and deliberately so. Every plan includes it, including the free one.
Start your first report and run it on something you already know well.
This article is general information about healthcare and life-sciences topics. It is not medical advice, diagnosis, or treatment guidance, and it is not a substitute for a qualified professional. Any treatment decision should be discussed with a clinician.
