Absolute Risk Reduction: Your Secret Weapon in Literature Evaluation
If you read this post, you'll be 43% less likely to get a C or below on your next journal club.
Of course. That's the relative risk (RR) we're reporting. But that's fine, right? That's all anyone ever reports. If you peruse through any given journal article, that's what you'll find. If you read some story about a new wonder drug you'll find RR there too. RR even shows up in news stories about that popular diet all of the celebrities are talking about.
So great. RR is the bees knees then, right? Maybe not...
A few weeks ago we wrote a two part series on How to Outsmart Big Pharma (Part I) (Part II). We talked about study design and marketing tactics used by drug companies to make their drugs seem like the second coming of penicillin.
We got some emails asking us to dig a little deeper. It seems a few of you want to fine tune your BS detectors.
Ask and ye shall receive. Today we're going to give you one of the best tools to separate the wheat from the chaff. A way to quickly determine if the magical results in that new study are all hype or actually legit.
We're going to teach you about absolute risk (AR). And, more importantly, we'll teach you how to calculate the...
Absolute Risk Reduction (ARR)
Allow me to present our friendly mascot for today's lesson, the ARR Pirate:
Look at him. So vigilant. Ready at a moment's notice to introduce you to the business end of his fist. Think of this little guy whenever you read a new study or journal article. Let him be your conscience and guiding light.
So what is Absolute Risk Reduction?
Time for some vocabulary. The absolute risk is the total amount of risk of a given 'thing' occurring after all risk factors and confounding variables are summed up. For example you could sum up your lifetime risk of having and atherosclerotic event based on the incidence and prevalence of your demographic.
Relative risk is different. It's the risk of a given 'thing' in comparison (ie. relative) to something else. For example, your risk of developing a DVT if you're a smoker compared to if you weren't a smoker.
When we talk about relative risk reduction (RRR) and absolute risk reduction (ARR), we're talking about an intervention. We're reducing the absolute and relative risk by giving some treatment. The absolute risk reduction is the total (ie. the absolute) reduction in risk that results by choosing a given treatment. This number is often very different from the relative risk reduction.
With RRR, the reduction of risk is compared to some other group. You could compare the effectiveness of Entresto in reducing mortality from heart failure to an existing treatment like enalapril. Or you could compare Paxil CR to no intervention (placebo) at reducing symptoms of depression.
Great. So what's the point?
You can have a lot of fun with relative risk.
For example, I could choose to no longer fly in planes. By doing so, my relative risk of dying in a fiery plane crash would drop faster than the ratings of Batman v Superman. Maybe the RRR would be somewhere around 99% (because a fiery plane could crash into me while I'm doing yard work or something). But what was my absolute risk of dying in a fiery plane crash in the first place? 0.000009%. A 99% reduction from 1 in 11 million isn't doing me a whole lot of good, is it?
Let's keep going with our new heart failure medication Entresto. You have a study that shows a 19.4% RRR of cardiovascular death in heart failure patients using Entresto compared to enalapril. You've possibly seen news stories boldly claiming that Entresto reduces your risk of death by 20% (because rounding). The ARR is not mentioned in such stories. But in this particular case, it's 3.2%.
That's nothing to sneeze at (this study was actually stopped early due to the benefit seen in the Entresto group). But you don't need me to tell you that 3.2% <<< 20%.
We'll dig more into this study further down the page....
Another place relative risk likes to poke it's nose into is in dietary studies. You may have recently heard about the WHO naming red meat as a probable carcinogen. If you read through the preceding links, you'll read about how each serving of red meat is associated with a 17% increased risk of colorectal cancer.
Holy shizz. 17%!? Why hasn't the FDA banned all red meat consumption?
Because that 17% is relative risk.
What would the ARR Pirate do? (#WWARRPD)?
He'd go look at data from the CDC. That data would show that if you were a 60 year old man living in 'Merica, you'd have a 4.04% chance of developing colorectal cancer....sometime in the next 30 years. If you were a 60 year old woman, your risk is 3.46%. These are the highest risk groups as the risk of colorectal cancer increases with age. If you're 30 years old (male or female) and living the US, your risk of getting colorectal cancer in the next 30 years is less than 1%.
That's absolute risk at work. 4.04% and 3.46% are sum totals looking at the entire US population. Of course there are populations that have a higher risk within those percentages. People with family histories of colon cancer. Certain demographic groups such as African Americans. The risk of colon cancer is higher than 4% in these groups. And it's lower than 4% in other groups.
But on the whole, you've got about a 4% chance of getting colon cancer if you're 60 and plan to live for another 30 years. In terms of "things that tend to kill 60 to 90 year olds in the US," that barely registers as a blip on the radar.
To be clear. I'm not trivializing colorectal cancer. It sucks. It's a bastard to treat and it's quite deadly. But there is a very big difference between 17% for each additional hamburger and a 4% chance by the time you're 90 years old.
And that's what we're driving at with absolute risk here. It lends perspective to the equation.
How to Calculate Absolute Risk Reduction
RRR is not a bad guy. He's just misunderstood. At the end of the day, RRR and ARR are just different ways to measure the size of an intervention. They're a way to help you determine the clinical usefulness of a drug. To help you decide if new treatment X will be beneficial to your patient.
But there's a reason we're writing this article. We have to teach you to "calculate" ARR because it's almost never reported. Why? Because RRR is much more impressive looking. Reducing risk by 20% is much more sexy than reducing risk by 3.2%.
And sexy makes for better headlines. It also makes for happier Gilead stockholders.
But in the medicine world, sexy also tends to mean "cripplingly expensive." And sexy doesn't always mean phenomenal clinical benefit.
So how do we calculate ARR?
Let's go back to our Entresto study. And let's zero in on that cardiovascular death rate part. You'll see that in the control group (enalapril), 16.5% of the patients died from cardiovascular causes. In the experimental group (Entresto), 13.3% died from cardiovascular causes.
ARR is the control event rate (CER) minus the experimental event rate (EER).
CER - EER = ARR
0.165 - 0.133 = 0.032
Or as we mentioned above, 3.2%. If you're doing this with a study (or test question) that doesn't give you the percentages you'll have to manually calculate them from the study. So if 800 patients in a control group of 1000 people had an event, your CER is 80% (or 0.80).
What about the relative risk reduction? The formula there is also pretty simple.
(CER - EER) / CER = RRR
(0.165 - 0.133) / 0.165 = 0.1939
Or (also as we mentioned above), 19.4%.
You'll notice that we're converting percentages to their decimal forms to carry out these calculations. This is a best practice that will prevent you from making any decimal point errors. It's also necessary to work in decimal form in order to calculate the Number Needed to Treat (NNT).
No discussion of ARR is complete without mentioning the NNT. The NNT is just another way of expressing the ARR. It's the number of patients you have to treat in order to prevent one 'event.' An event in this case would be cardiovascular death from heart failure.
NNT is the inverse of ARR.
1 / ARR = NNT
Using our Entresto numbers...
1 / 0.032 = 31.25
The NNT would be 32 in this case. Why 32? Because the NNT is always a whole number (an integer). You can't treat 0.25 of a person. So whenever your NNT ends in a decimal point, round up to the next whole number.
To apply this in "real terms," you'd need to give Entresto to 32 patients before you prevented one cardiovascular death.
How to Apply All of This
You can use ARR and NNT (and even the RRR) to determine if an intervention is worth it for your patient. As you've seen here, you can very easily calculate ARR and NNT. When you're weighing treatment options for patients, calculate ARR for a few different options. Look at what the drug is going to cost the patient. Look at the dosing schedule and determine if adherence will be an issue.
Then you can bring this information to the patient and the medical team and make a collaborative choice that is best for the patient. When everybody is involved (especially the patient), everybody wins.
Oh. Also, you can apply all of this info on your upcoming biostats exam. And the NAPLEX. You're gonna need to know this stuff.
In closing. When you're in doubt. Just think of what the ARR Pirate (and his judgmental glare) would do.