Michael Guren wrote this on Jun 28, 2013
The other day I came across an article, "Novartis heart drug…gets 'breakthrough' status". I was intrigued. They had me at "breakthrough". So I pulled open the study to read the results.
(Note: The following has been greatly simplified for the purposes of this discussion.)
The highlight of the study was the fact that this new drug reduced the mortality (death) rate by 37% at 6 months; compared to placebo. That sounds incredible.
What I want us to consider is the following question. What does that mean? How should we think about a "37% reduction in mortality"? In order to make an informed decision about whether to prescribe, or take, a prescription drug it is important to understand what these types of results mean.
To answer this question, we have to dip our toes into the pool of statistics. Don't worry, we're not staying long.
Here are the raw study results from the breakthrough trial touting a 37% reduction in mortality (with decimals intentionally left off):
- Placebo: ~11% mortality rate
- Study Drug: ~7% mortality rate
Interesting. I don't see 37% anywhere. Hmmm. So where does 37% come from? The answer is, it's relative.
Let's explore (warning: high usage of percentages below):
- The ABSOLUTE difference in mortality rates is 11% minus 7% = 4%. So far so good. We agree there was a 4% difference in mortality rates.
- The RELATIVE difference, however, is ~37%. How? Well, decreasing from 11 to 7 represents a 37% drop. As an example, a 50% drop in mortality would have decreased 11 to 5.5 (half of 11). We can even show this using a formula: (11 minus 7) divided by 11 = ~0.37 * 100 = 37%. If you need to re-read this section that's okay, I'll wait.
Almost ALL drug studies tout the relative difference between treatments, which can be misleading. For this reason, it is IMPERATIVE that you understand which type of risk reduction a study is referencing. Is it absolute (e.g. 4%)? Or is it relative (e.g. 37%)?
Let's dive a little deeper for a second and discuss the medical term Number Needed to Treat (NNT). This is often used within the medical community to understand how many people need to take a drug in order to have one person benefit. For example, an NNT of 10 simply means that 10 people need to take a drug for 1 person to show the benefit. The other 9 people will have no benefit.
How do we calculate NNT? It's pretty simple. Divide 100 by the absolute risk reduction (e.g. 4%). Using our example from above, the absolute risk reduction was 4%. Therefore, the NNT would be 100 divided by 4 = 25. So 25 people will need to take this new drug in order for 1 person to show any benefit.
This "breakthrough" drug, touting a 37% reduction in mortality, actually has a 4% ABSOLUTE risk reduction in mortality and requires 25 people to be treated in order to save 1 life. Interesting.
Now that we're equipped with all this knowledge…let's explore a more common drug class. Cholesterol lowering medications (aka "Statins").
(Note: Like before, I will be over-simplifying the following information.)
In high risk patients (e.g. high cholesterol), statins reduce the risk of having a heart attack by about 35%. This is roughly the accepted number used within the medical community.
Do you think 35% represents the absolute OR relative risk reduction?
If you said 'relative', then pat yourself on the back. That's right. So what is the absolute risk reduction? How about the Number Needed to Treat (NNT)? Good questions!
First we need the raw numbers. In high risk patients (e.g. high cholesterol), the following is generally true:
- Placebo: 9% heart attack rate
- Statin: 6% heart attack rate
This is more like it. We're used to dealing with numbers like this. Just like before, let's calculate absolute vs relative risk reduction:
- The ABSOLUTE difference between heart attack rates is 9% minus 6% = 3%.
- The RELATIVE difference is (9% minus 6%) divided by 9% = 0.33 * 100 = ~35%
Ah, so that's where the 35% reduced risk of heart attacks comes from!
How about the Number Needed to Treat (NNT)? Well, just like before we divide 100 by the absolute risk reduction (e.g. 3%). So 100 divided by 3 = ~33.
Therefore, 33 people would need to take a statin in order for one person to prevent a heart attack.
(Note: The actual heart attack risk reduction and NNT for statins varies depending on the study in question. This is a general example.)
I am not advocating that you alter drug therapy based on these over-simplifications. However, I AM advocating that you gain a better understanding of what these numbers mean. Hopefully this will help foster more productive conversations between you and your Doctor (or patient).
The next time you hear "This drug reduced mortality by 40%!", you'll know the right questions to ask in order to determine the absolute truth.