MDM: NNT and NNH

Before you decide about treatment, you should ask about NNT and NNH. These are numbers that can help you understand how likely the treatment is to help (or harm) you. They are easy to understand - though they can seem confusing to calculate. 

NNT refers to the Number Needed to Treat, and is an estimate of the number of patients who will need to be treated in order to benefit one patient. The smaller the number the better: if the NNT is 1, then every patient who is treated will benefit, but if the NNT is 100, then only 1 out of 100 treated patients will benefit. 

The NNT is calculated by comparing the results of a treatment to non-treatment or a placebo control and is always rounded up to the next whole number. For example, imagine that MagiCreme is compared to no treatment for wrinkles in a study where 100 wrinkled patients get MagiCreme and 200 wrinkled patients get no treatment with the following results:

With this information, the NNT can be calculated with this formula: NNT = 1/(treatment_event_ratio - control_event_ratio) = 1/(.16 - .15) = 1/.01 =100

For our imaginary MagiCreme, we see that 16 out of every 100 cream users will see an improvement, but only 1 out of every 100 wrinkled people will see an improvement that was caused by the cream.

Of course, we also need to think about side effects. The NNH is also calculated by comparing the results of a treatment to non-treatment or a placebo control and is always rounded up to the next whole number. For example, the same MagiCreme study showed the following rates of itchy rash:

With this information, the NNH can be calculated with this formula: NNH = 1/(treatment_event_ratio - control_event_ratio) = 1/(.22 - .03) = 1/.19 = 5

For our imaginary MagiCreme, 6 out of 200 wrinkled people who ignore their wrinkles will get an itchy rash while 22 out of every 100 people who use the cream will have a bothersome side effect, so 19/100 cream users get a rash from the cream and the NNH is ~5 (remember, NNT uses only whole numbers).

Looking at these two numbers we can see that MagiCreme is not worth buying, because we have to treat 100 people to benefit one, but we only have to treat 5 people to harm one. About 20 times as many people are harmed as helped.

Asking your clinician for NNT and NNH information can help you understand the likelihood of benefits and harms from treatments, helping you make a good decision and avoid surprises.

You can often find numbers for NNT and NNH by Googling the treatment and NNT and NNH. NNT data is often easier to find, and you may have to calculate the NNH yourself. There are easy to use and free online calculators here and here.

Here are some NNT examples for common treatments:

• Antibiotics for otitis media (ear infection) in children: NNT = 7
• Oral sumatriptan (Imitrex) for migraine: NNT = 2.6
• Antibiotics x 5-7 days for a dog bite to prevent an infection: NNT 16
• Hypertension treatment for 1 year in patients over 65 to prevent a heart attack or stroke during that year: NNT = 18
• Fungal toenail infection (onychomycosis) treated with terbinafine for 24 weeks): NNT (to be clear of fungus at 24 weeks) = 2.7
• Statin for 4 years to prevent heart attack or stroke during those 4 years in patient with no previous heart disease (primary prevention): NNT = 35
• Statin for 3 years to prevent a second heart attack during those 3 years in a patient with a previous heart attack (secondary prevention): NNT = 11

(You may have noticed that NNT of 1 doesn’t appear. Almost all our treatments – even our BEST treatments – help only some of the people we treat. Patients and their clinicians tend to forget this – and then are surprised, frustrated and angry when benefits that were expected do not appear.)

Here are some links to useful online resources for more information:

• http://www.nntonline.net/
• http://www.thennt.com/the-nnt-explained/   (you can ‘personalize’ their calculations by entering your own information)
• http://www.medicine.ox.ac.uk/bandolier/painres/download/whatis/nnt.pdf
• http://www.cebm.net/index.aspx?o=1044
• http://www.medicine.ox.ac.uk/bandolier/band50/b50-8.html

If you know of a source that is useful for listing or calculating NNT or NNH information, please post it in the Comments below (registration required).