This an area that I work in, so I guess I'll offer my one or two cents. I will say up front that I'm not a biostatistician (and I'm glad that I'm not one because they are a weird bunch). Statistics are mostly up over my head, but I know a tiny bit about it.

If you bring up the number needed to treat (NNT) for any given drug with your physician, they will likely explain that summarizing all the trial data through any one single number might be problematic and have limitations (and they would technically be correct). Or they might just shrug their shoulders and scratch their head because they don't really know. There are even some crazy doctors out there that would love to put statins in the water supply. And this is why it is good for all of us patients to have a bit of knowledge to be able to protect ourselves from harm.

Data calculations on large clinical drug trials are very complex. First, the stats people will do hypothesis testing on the data. Maybe you've even heard of terms like p-value or null hypothesis or t-test. They are just trying to answer some simple questions - Does the drug work? Is treatment A better than treatment B? Is there a dose response? Are treatments A and B clinically equivalent?

At this point they know to what extent there is a treatment effect. But there are other ways to measure treatment benefit and harm, so they will calculate odds ratios (odds of a patient suffering one or more serious adverse events), relative risk, relative benefits, and NNT. There are advantages and disadvantages to each of these calculations, which is beyond my understanding.

Most people would find the NNTs for statin drugs shocking. The data for statins really only points to helping middle-aged men who have already experienced a heart attack (secondary prevention). The other group that might want to consider a statin are those with familial hypercholesterima who have a genetic defect with their LDL receptor. So if people fall into those two categories, by all means, take a statin. But if you are outside of those categories, definitely consider things like number needed to treat or absolute reduction in risk.

Many studies like to put their numbers in terms of relative reduction of risk and not absolute reduction of risk because it makes them look more impressive. I find this to be very annoying and dishonest. For example, if you take a drug and it reduces the risk of having a disease from 2% to 1%, well then the absolute reduction in risk = 1%; this number is not very impressive sounding. But if were to use relative reduction in risk, then the number would be 50% reduction in disease risk. This sounds way more impressive. So always check the fine print on any study - did they use absolute risk or relative reduction in risk.

I found this neat little website that gives the NNT by medical specialty and organ system. I haven't totally vetted it out, but lists sources for all the numbers that it gives.

http://www.thennt.com/home-nnt/

Sorry I rambled, but I enjoy this topic. Have a good, pain-free weekend everyone!