Number needed to treat

It’s even worse than that. Suppose you treat 100 to help one. Then suppose that a different one out of the 100 suffers serious side effects. You’ve harmed one to help one.

I took a statin for “borderline” cholesterol. Otherwise, I had good health all my life: no family history of heart disease or high blood pressure. My dad is 82. His dad lived to age 91. The statin damaged my leg muscles and there’s not a thing the doctors or physical therapists could do to help me. It has been three years of just enough disability that I can’t do the job I loved anymore. I had to change my very nice PCP because I couldn’t even think of him anymore without getting angry. If I had known the NNT, I would never have taken the statin to begin with.

Though it is true that NNT is a very helpful number, and drug companies use relative risk (2/100 in placebo vs 1/100 would be a 50% reduction) compared to absolute risk (1% reduction in ex above), you have to be careful by saying “only 1/100 would benefit”. Sometimes NNT of 5 is unacceptable and NNT of 500 is acceptable. It depends on risk, benefits and costs. Chance of having a house fire is 1/200. You wouldn’t say that fire insurance is a waste because only 1/200 would actually have a benefit. For statins, a heart attack is a pretty bad thing, even if it doesn’t kill you. The most common side effect of a statin is muscle pain (myalgia) which is usually mild, usually goes away and occus in about 3% of patients. So, for 100 people taking statins, 1 would have a heart attack prevented and 3 would get myalgia. These are actually pretty good numbers, if you don’t take cost into account. So the real issue is are we willing to pay more than 1/4 million dollars (cost of 100 people on Lipitor for 3 years) to prevent one heart attack?
http://www.drmintz.com