The Number Needed to Treat (NNT) is a simple and effective way to communicate the potential benefit or harm of a treatment. While it is a statistical concept it is also quite intuitive and understandable by both clinicians and their patients. It signifies the likelihood that a patient will be helped, harmed, or unaffected by a treatment. NNT is growing in popularity and is often reported in RCTs and systematic reviews on therapy.
Let’s take as an example a 2016 Cochrane systematic review of acupuncture for frequent tension type headaches. The review looked at clinical trials of at least 6 acupuncture treatments compared to sham acupuncture or to routine conventional care. The specified efficacy outcomes were at least 50% reduction of headache frequency at 3-4 months after randomization in the trial. Harm endpoints were any adverse events during treatment or dropping out from the trial due to an adverse effect.
In Summary, for those who received acupuncture:
Benefits in NNT | Harms in NNT |
---|---|
1 in 3 had 50% reduction of headache frequency (compared with routine care) = NNT 3 1 in 11 had 50% reduction of headache frequency (compared to sham acupuncture) = NNT 11 | Any harm, such as triggering headache, dizziness, or other pain 1 in 416 participants withdrew because of adverse events in the acupuncture group when compared with routine care 1 in 20 participants experienced any adverse event from acupuncture treatment compared with sham acupuncture |
Interpreting the NNT requires knowing the clinical context including:
A perfect NNT would be 1. That means that for very patient treated one got better in the study who would not have otherwise without that particular intervention. The larger the number, the fewer people will be helped. As a general rule of thumb, an NNT of 5 or under for treating a symptomatic condition is usually considered to be acceptable and, in some cases, even NNTs below 10.
The NNT may be cited in a report of a clinical trial or a systematic review. If not, the NNT has been calculated and reported for a number of therapies, diagnosis and risk assessments at https://www.thennt.com/.
Special thanks to Ron LeFebvre, DC and his article Understanding Treatment Results series at https://www.uws.edu/wp-content/uploads/2013/10/Number_Needed_to_Treat.pdf