The ‘number needed to treat’ (NNT) is a measure used in assessing the effectiveness of a particular intervention, The NNT is the average number of patients who require to be treated for one to benefit compared with a control in a clinical trial. It can be expressed as the reciprocal of the absolute risk reduction. The ideal NNT is 1, where everyone improves with treatment: the higher the NNT, the less effective is the treatment in terms of the outcome and timescale described in the trial. So if treatment with a medicine reduces the death rate over five years from 5% to 1% (a very effective treatment), the absolute risk reduction is 4% (5 minus 1), and the NNT is 100/4, i.e. 25.
To give an indication of the likely impact of the intervention in the next year, the Drug Efficacy tables that follow illustrate an “annualised NNT”. This assumes that the clinical effect is seen at an even rate over the period of the intervention. So, if the NNT for the medicine to prevent one death in 5 years is 25, the estimate for the annualised NNT, that is, to prevent one death in one year is then 125.
In other words, the number needed to treat with that medicine for one year to prevent one death is 25. The ideal NNT is 1 where everyone improves with treatment. The higher the NNT, the less effective is the treatment. There is always a need to consider:
- What is the outcome being avoided? Death is more significant than a vertebral fracture, but different outcomes will be more or less significant to individual patients.
- Over what period does the benefit accrue? Two drugs may have the same NNT to avoid one death, but the drug that achieves that over 6 months is more effective than the drug which takes 10 years. You can put NNTs on the same timescale by multiplying or dividing the NNT appropriately, but there is an assumption that benefit accrues consistently over time (a not unreasonable assumption, but one that is difficult to test).
- What are the TRUE costs of the drug? This will include monetary costs, but also costs associated with treatment burden, and harm/side effects. A medicine might save the life of one of the 25 people who take it, but if it led to all 25 suffering a debilitating side effect, its costs may outweigh its benefits.
NNTs are only estimates of average benefit, and it is rarely possible to know precisely what the likely benefit will be in a particular patient. Clinicians and patients should also be aware of a degree of ‘uncertainty’ in the number since it is usually not possible to calculate valid confidence intervals around NNTs.
'Number needed to harm’ (NNH) is a related measure which is the average number of people exposed to a medication for one person to suffer an adverse event. Again, a defined end point (e.g. GI bleeding or renal failure) requires to be specified and confounders may require correction of the raw data i.e. in very elderly patients the risk of particular side effects such as confusion and falls may be higher than on average . In discussion, the overall benefit – risk ratio (NNT / NNH) requires to be ‘weighed’ in the individual patient and may vary considerably in people with polypharmacy depending on absolute risk, life expectancy and vulnerability to adverse drug events.
Example:
“A meta-analysis of sedative hypnotics in older people with insomnia showed that the number needed to treat for improved sleep quality was 13 and the number needed to harm for any adverse event was 6. This indicated that an adverse event is more than twice as likely as enhanced quality of sleep. Glass, J. et al. Sedative hypnotics in older people with insomnia: a meta-analysis of risks and benefits.” BMJ 2005; 331: 1169
Polypharmacy guidance: realistic prescribing
