Know your Numbers / How Statistics Can Be Misleading

April 2016

A recently published study in the New England Journal Medicine, The Hope-3, looked at the benefit of statin drugs to lower cardiovascular disease.  The study was funded in part by AstraZeneca, a company that makes statin drugs.  The idea is that perhaps healthy people should be given a low dose statin alone or would a mixture of a statin with a mild blood pressure medicine help to lower risk of heart attack or stroke.  Essentially, should ALL of us over 50 be on these drugs?  Turns out that the statin alone worked better than with a combination of a blood pressure medication.  

But that’s not what I found interesting.

What I found interesting is that the presentation of these numbers is a perfect example of how numbers are used to manipulate or confuse the truth, not about just medicines but EVERYTHING.

The study included 12,000 people; that’s a lot of people.  Good.  Some were given a statin alone. Some were given the combo pill, and some were given placebo.  Of the people who got placebo in the study, 4.8% had a heart attack or stroke.  The people who got the statin drug, 3.7% had a cardiovascular event.  The author noted that this represented “a statisically significant 24% reduction in risk."  (Actually the article came up with 24% but the math is 23% to my way of calculation.)

Here’s the math:  4.8 - 3.7 = 1.1

                            1.1 / 4.8 = .2292  or rounding, 23%

SOUNDS IMPRESSIVE!  Doesn’t it?  “A 24% reduction of risk.”

But the author is leaving out a very important qualifier; he should have use the term “relative risk.”  And to my mind, relative risk is often meaningless.

Let’s look at the numbers again:

4.8% of the people on placebo had an event.  This means that for every 100 people on placebo, not quite 5 individuals had a heart attack.  For every 100 people on the statin, 3.7 or about 4 people had an event. So 

5 - 4 = 1  So, for every 100 people on the drug, only 1 heart attack was prevented.


This is the difference between relative risk and absolute risk.  Relative risk is reduced by 23%; absolute risk was reduced by 1%.

By example, let’s say that there is a 1 in a million risk for a rare cancer. But! Hooray for medical science!  If one were to take a drug, that may cause liver problems or be assoicated with memory problems, or increase your risk for diabetes, (i.e. statins),  the risk becomes 1 in 2 million.  That represents a whopping 50% reduction in risk.  THAT would be the headline.  But is it wise to take a drug with known risks just to help with a 1 in a million chance?  Probably not.

I am not suggesting people stop taking their statin drugs, especially if they have already had a heart attack or if their cholesterol is astromically high (because their risk is much higher than the 4.8%.)  But I do call to question how risks and benefits are presented to the public.  Know your numbers.  I don’t think the author was trying to dupe the public, but it does demonstrate how statistics can be used to confuse the truth, from medicine to especially politics. 

Statistics play a role in my practice.  When one’s intra-ocular pressure (IOP) reaches 26, roughly 50% of people will develop optic nerve damage, which is the definition of glaucoma.  (There are more risk factors which should be considered however, increasing or decreasing risk; so this number of 26 is rather simplistic).  But at any rate, at a pressure of 26, I often offer treatment to the patient to lower their risk.  But I am careful to explain that if I put 10 people with a IOP of 26 on drugs, I am OVER TREATING 5 of them!  I then let them decide.  

Only when people are given the true meaning of numbers can good decisions be made.  

© Richard Randolph 2012