Number Needed to Treat: Weighing Benefit Against Harm

Hey, I'm Medi — and today we're tackling one of the most useful numbers in all of medicine: the Number Needed to Treat, or NNT. Here's the core idea. Suppose a clinical trial tests a new blood pressure drug. In the control group, 10 out of 100 patients have a heart attack over five years — that's a 10% risk. In the treatment group, only 6 out of 100 have a heart attack — a 6% risk. The drug reduced the absolute risk by 4 percentage points (10% − 6% = 4%). That difference is called the Absolute Risk Reduction, or ARR. NNT is simply the inverse of the ARR: NNT = 1 ÷ ARR. So here, NNT = 1 ÷ 0.04 = 25. That means you need to treat 25 patients for one extra person to avoid a heart attack who otherwise would have had one. Why does this matter? Because NNT frames treatment as a trade-off. A drug with NNT = 5 is very powerful — treat 5 people, help 1. A drug with NNT = 200 is much weaker — treat 200 people to help 1. But every one of those 200 patients also faces the drug's side effects. That's where NNH — Number Needed to Harm — comes in. NNH is calculated the same way, but from the harm event rate. If 2 out of 100 people on the drug develop a serious side effect versus 1 out of 100 on placebo, ARR for harm = 1%, so NNH = 100. An NNT of 25 and NNH of 100 means for every 4 patients helped, roughly 1 is seriously harmed — a real trade-off that you and your patient must weigh together. Key rules to remember: NNT is reported as a whole number rounded up to the nearest integer — because you cannot treat a fraction of a patient — but the underlying formula (1 ÷ ARR) yields a real number. A lower NNT is better. NNT only applies to the specific outcome, population, and time window studied.