Most drugs don’t do anything significantly good or bad for most people who take them

Monday, October 25th, 2021

A statistic like relative risk reduction — which is far and away the most common one you’re getting — is not the statistic that you need in order to make an informed decision, David Epstein explains:

NNT is an abbreviation for “number needed to treat.” In other words: How many patients must be treated with the drug in order for a single patient to get the desired benefit?

When you read about drugs in the news — or even in most medical journals — you will almost never be explicitly given the NNT (which I will explain in more detail below). Instead, you’ll get relative risk reduction, a metric that a Michigan State med school dean once told me “is just another way of lying.”


Here’s a fictional example:

You read that a new drug reduces your chance of dying from Ryantastic syndrome by 40 percent. Here’s what that means in practice: if 10 in 100,000 people normally die from Ryantastic syndrome, and everyone takes the new drug, only 6 in 100,000 people will die from Ryantastic syndrome. Now let’s think about it from an NNT perspective.

For 100,000 patients who took the new drug, four deaths by Ryantastic syndrome were avoided, or one per 25,000 patients who took the drug. So the NNT is 25,000; that is, 25,000 patients must take the drug in order for one death-by-Ryantastic to be avoided. Ideally, you also want to know the NNH, or “number needed to harm.”

Let’s say that 1 in 1,000 patients who take the new drug suffer a particular grievous side effect. In that case, the NNH is 1,000, while the NNT is 25,000. Suddenly, the decision seems a lot more complicated than if you’re just told the drug will lower your chance of dying from Ryantastic syndrome by 40 percent.

Now let’s move to the real world: aspirin. Nearly five years ago, the NNT and NNH of aspirin caught my eye, so I included them in an article about medical evidence:

For elderly women who take it daily for a year to prevent a first heart attack, aspirin has an estimated NNT of 872 and an NNH of 436. That means if 1,000 elderly women take aspirin daily for a decade, 11 of them will avoid a heart attack; meanwhile, twice that many will suffer a major gastrointestinal bleeding event that would not have occurred if they hadn’t been taking aspirin.


Once I started looking at NNT and NNH data instead of relative risk, one of my main takeaways was that most drugs don’t do anything significantly good or bad for most people who take them.


  1. McChuck says:

    So, in other words, only take the drugs you need to treat the problems you have.


  2. Borepatch says:

    McChuck, only take the drugs that are more likely to make you better than to make you worse. ;-)

  3. McChuck says:


    I’ve got a bit of experience with that. I react oddly to many medications. I’ve had permanent effects from medications I took for two or three weeks, years ago. I’m a walking six-sigma event. (This is not a compliment. Do not recommend.)

    On a more serious note, what the study shows is that many medicines can be harmful to the genpop who does not need them. They need to be evaluated for the positive vs/ negative consequences for the small percentages of people who actually need them for a specific purpose. For example, if daily aspirin reduces your chance of dying from a heart attack by 50%, but only increases your chances of abdominal bleeding by 10%, it’s still a win. Especially if there are easy tells, like spontaneous bruising. (I’ll keep on taking mine, because I have ridiculously thick blood, and aspirin is the only painkiller that still works for me.)

    In other words, none of this is new. Bayes theorem should always be applied. And that’s before we get into the enormous percentage of fraudulent medical studies and reports.

  4. Goober says:

    It seems pretty simple, but there’s more to it than even this.

    For instance, take a theoretical drug NNT is 5, NNH is 2. This theoretical drug prevents second heart attacks in patients likely to have one. The harmful side effect is GI bleeding.

    It is completely arguable and reasonable to say that having a bleeding ulcer is a way better thing than having a heart attack, and so even if the NNH is one, meaning ever person that takes it gets a bleed, at least they have a chance of not getting a heart attack.

    I’ve been on meds that caused me to have GI bleed. It wasn’t fun, and I even became anemic during the time it took to figure out that I had a GI bleed. But the med kept me from throwing any more clots so I didn’t have another stroke, so… yeah. Give me a GI bleed any day.

Leave a Reply