Skip to Main Content

Gut Check looks at health claims made by studies, newsmakers, or conventional wisdom. We ask: Should you believe this?

The Claim:

Drugs and other treatments are likely to help patients get better. 

Tell me more:

Although we think if a doctor prescribes a treatment it’s going to help, a lot of therapies benefit very few patients. The “number needed to treat” captures this: It’s how many people must get some intervention for a bad thing to be averted in one person. For instance, the NNT for antibiotics curing conjunctivitis within 10 days is about 12: A dozen people with that eye infection need to take the drugs for one person who would otherwise remain infected to become cured. The other 11 either would have gotten better on their own or aren’t helped by the drugs.


Calculating the number needed to treat is simple. Here at Gut Check, when we describe how much a drug or procedure reduces people’s risk of some bad thing, like a stroke or cancer, we try to use absolute numbers rather than relative ones. Take post-surgical compression stockings, which are meant to prevent the serious blood clots called deep-vein thrombosis. A study of their benefit found that without the stockings, 27 percent of patients recovering from surgery developed deep-vein thrombosis, but with them 13 percent did. That’s a relative risk reduction of 52 percent, but an absolute risk reduction of 14 percentage points (27 minus 13). To calculate the NNT, you divide that 14 into 100, which gives you about 7. Seven surgical patients need to don compression stockings in order to save one from dangerous blood clots.

Compression stockings are cheap and essentially risk-free, so doctors prescribe them even though most patients won’t benefit. But when a procedure is risky, the number needed to treat “helps the clinician weigh the benefits of an intervention for preventing bad outcomes against its associated risks and costs,” said biostatistician George Tomlinson of the University of Toronto, an expert on NNTs. In a 2006 study, he and colleagues concluded that “an NNT of 5 or less was probably associated with a meaningful health benefit,” while “an NNT of 15 or more was quite certain to be associated with at most a small net health benefit.”

Yet interventions with NNTs above 15 are common, as a quick tour of a leading NNT database shows.


Statins, which have become synonymous with “heart-attack-and-stroke-preventing,” have an NNT of 60 for heart attack and 268 for stroke: That’s how many healthy people have to take statins for five years for those respective outcomes to be prevented. In people with heart disease already, the number is smaller: Just 39 must take statins for five years for one non-fatal heart attack to be prevented, while 83 have to do so for one life to be saved. If 125 people with high blood pressure take drugs for five years to lower it, the meds will prevent a fatal stroke or heart attack in only one.

The NNT for aspirin to prevent cardiovascular calamities is even higher. A whopping 1,667 healthy people need to take aspirin every day for a year to prevent one stroke or heart attack. But only 77 people who previously had a heart attack or stroke need to do so for one heart attack to be prevented; it’s 200 for one stroke to be prevented.

The statin and aspirin examples underline that the NNT is different in different populations, said Dr. H. Gilbert Welch, a professor of medicine at the Dartmouth Institute for Health Policy & Clinical Practice. “People at higher risk of an adverse outcome tend to benefit more [from an intervention], so the NNT is always lower” than in lower-risk people. For instance, the NNT for preventing hip fractures with the bone-strengthening drugs called bisphosphonates is 100 in post-menopausal women with previous broken bones, but essentially infinite in those without previous fractures.


It may come as a shock that drugs shown in rigorous studies to be “effective” actually help so few people — where help means keeping a bad thing from happening to someone who would have otherwise suffered it. But there are sound medical reasons for that.

For one thing, “treatment is never perfect,” Welch said: “Effective” drugs may help a lot of people, but “a lot” is rarely everyone. Sometimes you get a stroke despite taking statins religiously.

Just as important, “not everyone is destined to have something bad happen to them” after they develop some condition or suffer an accident, Welch said. Most surgical patients will never develop blood clots even without compression stockings. The vast majority of people with a broken bone that pierces the skin will not develop infections even without antibiotics: Antibiotics for that situation have an NNT of 16. Most people with sinusitis will also get better without antibiotics, which has an NNT of 18 for sinusitis. “People can heal on their own,” Welch said.

To be sure, NNTs are only as good as the studies underlying them. But when they’re based on numerous studies, it’s unlikely that additional research would turn a very high NNT into a low one or vice versa. Undergoing a treatment with an NNT of, say, 50 is a bet that you’re the one who will benefit and not the 49 who won’t. Yet few patients have ever heard of NNTs, and they’re widely ignored by physicians.

The verdict:

Because even “effective” treatments are imperfect, because conditions often get better on their own, and because risk factors only sometimes lead to anything bad, many commonly used treatments need to be given to dozens and even scores of patients for a single person to benefit.

  • When I follow the link to look at the NNT for statins for healthy figures, I see different figures. You say there is a NNT of 104 for heart attack and 154 for stroke. However, the page you link to has an NNT of 60 for heart attack and 268 for stroke. Your figures for people with heart disease match the page you link to.

Comments are closed.