Previous: Number Needed to Treat: Treatments Don't Work Like You Think They Work
Next: Sunscreen Works, If You Use it Right



View count:76,142
Last sync:2023-11-05 09:00
Last week, we discussed the NNT, or number needed to treat. I'm sure it made a lot of you upset to realize that many therapies you've been sold as "awesome" were, in fact, somewhat incremental with respect to benefits. But another problem is that a lot of those therapies are anything but benign. They come not only with costs, but also with side effects or problems.

We can quantify harms, too. Watch and learn about NNH!

Almost all of the data for this came from the amazing website TheNNT ( You can go there to see those and more.

Additionally, Aaron's new book is out! Please consider buying a copy. He'd really appreciate it!

John Green -- Executive Producer
Stan Muller -- Director, Producer
Aaron Carroll -- Writer
Mark Olsen -- Graphics
Last week we discussed the NNT, or number needed to treat. I'm sure it made a lot of you upset to realize that many therapies you've been sold as "awesome", were in fact somewhat incremental with respect to benefits, but it's gonna get worse.

      A lot of those therapies are anything but benign: they come not only with costs, but also with side-effects or problems. We can quantify harms, too. That's the topic of this week's Healthcare Triage.

Intro Music and Graphics*

      Every time you take a drug, or engage in some sort of treatment, *pause* you take a risk. Exercise, for instance, can lead to some injuries. Drugs have side effects, too. They can cause rashes, diarrhea, or other issues. Surgeries can lead to infections, or even death.

      Sometimes those harms are calculated in terms of relative percentages; "your risk doubles!" and such. As with benefits, we should avoid relative risks. We care about absolute risk increases; those are reported in many trials, too.

      *Graphics come on*

      Let's make up an example. Let's say that twenty percent of people who take a certain antibiotic develop bad diarrhea. In the same study, ten percent of those who didn't take the drug develop diarrhea, too.

       If you go with a relative risk, then the drug doubled your chance of having diarrhea. If you go with an absolute risk, then the drug increased your risk of diarrhea from ten percent to twenty percent. So your absolute risk went up by twenty minus ten, or ten percent.

      Here's another example. Let's say that point one percent of people who don't have surgery for a certain illness get diarrhea. Let's say that point two percent who have the surgery develop diarrhea.

      If you go with relative risk, then surgery doubles your risk of diarrhea. If you go with an absolute risk, then surgery increased your risk of diarrhea from point one percent to point two percent, so your risk went from point two to point one, or point one percent.

      In relative risks these are similar, they both doubled the risk of diarrhea, but in absolute risks, they are way far apart. The drug increased your risk by ten percent, surgery increased it by point one percent.

      *switches back to view of Dr.Carroll*

     Just like we calculated numbers needed to treat, we can calculate numbers needed to harm.

      *Graphics come on*

     We can calculate the NNH by taking one-hundred and dividing it by the absolute harm increase. Let's start with the drug; it had an absolute risk increase of ten percent. Therefore, the NNH is one-hundred divided by ten, or ten. 

      *Switches back to view of Dr.Carroll*

      That means that for every ten people we give the drug to, one extra person will have diarrhea. Granted, more won't, but it's important to understand that one in ten people will have the bad outcome.

      *Graphics come on*

      With the surgery, the absolute risk increase was point one percent. Therefore, the NNH is one-hundred divided by point one, or one-thousand. That means that for every one thousand people who have surgery, one extra will have diarrhea.

      *Switches back to view of Dr.Carroll*

      That's much better than for the drug. Just as NNT is rarely discussed, so is NNH. But we should, because comparing the NNT and the NNH provide people with a simple and objective way to determine if therapy is worth it.

      Let's take some examples.

      *Graphics come on*

      Last week, I told you that the NNT for Aspirin to prevent a first heart attack or stroke was one-thousand, six-hundred, and sixty-seven. But the NNH is three-thousand, three-hundred, and thirty-three for a major bleeding event.

      *Switches back to view of Dr.Carroll*

      Granted, the chance of you having a benefit is greater than that of you having a harm, but the harms are real. Remember how I told you the NNT for a chest CT to prevent death in one year among high-risk smokers was two-hundred seventeen? I bet a lot of you thought that sounded okay, because it's, well, death. 

      But what about harms? It turns out that the NNH for a false-positive diagnosis was four. That means one in four people will have a positive scan that turns out not to be disease.

      The NNH for unnecessary surgery was thirty. The NNH for a surgical complication was one-hundred and sixty one.

     *Graphics come on*

      So one in two-hundred seventeen will see a benefit in death prevented, but one in one-hundred and sixty one will have a surgical complication, one in thirty will have unnecessary surgery, and one in four will have a lot of needless worry. Not as clear, now.

      *Switches back to view of Dr. Carroll*

      So lots and lots of people without know heart disease are put on statins. I recently discussed this in a piece in The New York Times. But statins have never been shown to prevent death in this population; period. The number needed to treat is, effectively, infinity.

      The NNT to prevent a heart attack in this population sixty. The NNT to prevent a stroke is two-hundred sixty-eight. Realize that this means that fifty-nine of sixty people who take the drug get no heart attack benefit, and two-hundred sixty-seven of two-hundred sixty-eight people who take it get no stroke benefit.

      And there are harms! The number needed to harm for developing diabetes is fifty! The number needed to harm for developing muscle damage was ten.

      Think about that: one in sixty may see a heart attack prevented, but one in fifty will get diabetes! You're more likely to get diabetes than to see the benefit. Worth it?

      The world is full of examples like this. I'm a pediatrician, and I see a lot of kids with ear infections. When parents want antibiotics, I like to talk in these terms.

      You could give ten-thousand kids antibiotics, and none would be less likely to see serious complications. None would have less pain in twenty-four hours. Antibiotics don't work there, and NNT is, again, effectively infinity.

      But yes, some kids will see less pain within a week if they take antibiotics. The NNT is sixteen.

      *Graphics come on*

      We have to treat sixteen kids with antibiotics for one to see that benefit. Fifteen of them will see no benefit at all, and one out of every nine kids given antibiotics, that's the NNH, is likely to develop diarrhea.

      *Switches back to Dr.Carroll*

      So by giving your child antibiotics, I'm more likely to give them diarrhea, than to reduce their pain in a week. Alternatively, you could give them pain meds, which don't have those side effects, and won't build up resistance in bacteria.

      Guess which option my patient's parents choose most of the time. The pain meds; thinking about benefits and harms in this way makes a difference.

      *Outro Music and Graphics*