How can you tell the effectiveness of a drug?

In 1988 epidemiologists came up with a way to determine how effective a drug really is. It’s called “Number Needed to Treat” (NNT). Simply put, NNT is how many patients would need to take a drug before one of them received the benefit the drug offers. If the NNT is 4, then 4 people would need to take the drug in order for one of them to get the benefit.

So if a drug were perfectly effective, then the NNT would be 1. It would require 1 person taking the drug for 1 person to get the benefit.

So far it seems simple enough.

##### Stealthy Statistics

The sneaky use of statistics happens when you see advertising about the drug. Whenever you see an ad on TV or in a newspaper or magazine, you’ll *never* hear about the NNT for the drug.

Instead you’ll typically hear about the ‘relative risk reduction.’ This is because relative risk reduction sounds a whole lot more impressive than the NNT.

For example, in men with high cholesterol, the relative chance of a heart attack is 31% lower if they’re taking the cholesterol-lowering drug Prevacol than if they’re not. Sounds impressive, doesn’t it? And if you didn’t know better you’d think that your chances of having a heart attack when taking this drug would be about 1/3 less.

In fact, the NNT for this drug is 50. That means only one out of 50 people who take it receive the benefit of lowering their chance of a heart attack. The other 49 don’t get this benefit, but do get exposed to the side effects of the drug.

##### How This Works

The reason for the radical difference between relative risk reduction and NNT is how the statistics are used. In the case of relative risk, they are comparing the percentage in each group who had heart attacks over the course of the study.

In men with high cholesterol the incidence of heart attacks was 7.5%, and in the group not taking the drug it was 5.3%. That’s an absolute difference of only 2.2% (not nearly so impressive sounding). But the relative difference between the two groups is, in fact, 31% (5.3% is 31% lower than 7.5%).

So how did they get the NNT of 50? Out of 100 men with high cholesterol 93 of them wouldn’t have heart attacks anyway. 6 would have heart attacks even while taking the drug and the remaining 2 would avoid the heart attack because of the drug. In other words, out of 100 people taking the drug, only two would get the promised benefit. That’s the same as 1 person in 50.

##### Other Examples

Not all drugs have such high NNTs. Here are the NNTs for some commonly prescribed drugs:

**Cortisone injections for painful, stiff shoulder: NNT=3****Amoxicillin for childhood ear infection: NNT=20****Proscar for enlarged prostate: NNT=18 (if taken for four years)****Aspirin to help avoid heart attacks: NNT=208**

In every case, **the NNT number should be as low as possible** when considering a prescription drug.

##### One Other Important Statistic

There is an equally important number to consider. It’s called the Number Needed to Harm (NNH). In this case the number reflects how many people would have to take the drug before one person was harmed by it. Unlike the NNT, **the NNH number should be as high as possible**.

Case in point is the recent problem with Vioxx. The NNT for Vioxx is 2.2, which is pretty good. For every 2.2 people taking it, 1 person got a significant improvement in pain reduction.

The problem came with the NNH for this drug. It was 42 for cardiovascular events (combined death, myocardial infarction, stroke, and heart failure) when someone was taking 400 mg twice daily. That meant that one in 42 patients had a risk of these symptoms at that dose.

With millions of people taking this drug, it’s not surprising that there were so many adverse cardiovascular problems.

##### How To Use These Numbers To Make Better Choices

If you’ve read this far I congratulate you. Statistics can be so boring at times. There are 2 payoffs for sticking with it:

- Next time you see a drug commercial anywhere, remember that they’re giving you the best possible spin on how well people do with the drug and take the information accordingly.
- If your doctor recommends a prescription drug, in addition to asking for information about the side effects, etc., also ask about the NNT and NNH for the drug.

It’s always a judgment call whether the risks outweigh the benefits for any treatment, but at least you’ll have the best information possible when making the decision.

## 2 replies on “Lies, Damn Lies and Drug Statistics”

great post

very useful for those who are using medication

which is next to everybody

thanks

U

[…] using amoxicillin to treat ear infections in […]