Wednesday, January 22, 2014

Putting "Relative Risk" in perspective - Eades

Absolute risk versus relative risk: Why you need to know the difference

Six years ago, the airwaves were alive with Lipitor ads. Lipitor, a statin and the largest selling drug in the world at that time, was being challenged by other less expensive statins that had gone off patent, so Pfizer, the manufacturer, was blanketing the media in an effort to keep sales humming.

Pfizer had a series of ads featuring Dr. Robert Jarvik, one of the developers of the first artificial heart. These ads were a first in that a real doctor had never appeared in an ad touting a drug. As I posted at the time, Pfizer and Dr. Jarvik got into some trouble because they all played a little fast and loose with the truth about Dr. Jarvik’s credentials.

Jarvik Lipitor AD 36 Absolute risk versus relative risk: Why you need to know the difference

Aside from the misrepresentation by both Pfizer and Jarvik, one of the claims many of these ads made was that by taking Lipitor you could reduce your risk of heart disease by 36 percent. Sounds pretty good. I would like to reduce my risk of heart disease by 36 percent as did millions of people who went on the drug.

These ads weren’t technically misleading, but they didn’t tell the whole truth, because the 36 percent reduction in heart attack risk was what’s called a relative risk reduction.

Relative Risk
Relative risk is always stated as a percentage. Let’s say we do a study in which we randomize 200 subjects in two groups of 100. One group (the study group) takes a drug, and the other (the control group) takes a sugar pill. We keep the subjects in the two groups on their pill regimen and wait for, say, ten years to see what happens. After ten years, we find that 90 people in the group taking the sugar pill died while only 60 people taking the drug died.

We can then do the following calculation. 90 − 60 = 30. 30 divided by 90 = 0.33. Converting 0.33 to a percentage = 33 percent. Which is the relative risk. If this were an actual study, you could say people taking the drug reduced their risk of dying by 33 percent.

Would this be important? Absolutely. In this case as least. Why? Because 150 people out of 200 died. This means most of the people in the study died, so a 33 percent reduction in risk is huge. I would be all over this drug in a heartbeat.

So what’s the problem, then, with the 36 percent reduction in risk found in the Lipitor study? And why shouldn’t people be begging to go on Lipitor?

Because they need to know the absolute risk before the relative risk becomes important.

Absolute Risk
The absolute risk is simply the total risk for whatever is being studied. In our made up example above, the study end point we were looking for was death. In that example, 150 out of 200 died. So if you were a subject in that study, your absolute risk of dying would be 150 divided by 200 or 75 percent, which is very high.

If your absolute risk of crashing during a commercial airline flight were 75 percent, you would never fly. But if you absolutely had to fly, and one airline offered a 33 percent reduction in risk of crashing (relative risk), you would be stupid not to fly that airline.

To make sense out of relative risk, you’ve got to know the absolute risk.

Let’s go back to our Lipitor claim.

The 36 percent relative risk reduction figure came from a study published in Drugs titled Prevention of coronary and stroke events with atorvastatin in hypertensive patients who have average or lower-than-average cholesterol concentrations, in the Anglo-Scandinavian Cardiac Outcomes Trial–Lipid Lowering Arm (ASCOT-LLA): a multicentre randomised controlled trial.

This study was a subset of a larger study in which about 20,000 subjects with high blood pressure and three other cardiovascular risk factors were randomized into two groups, the subjects in each of which got one of two blood pressure lowering medications. From this larger group, a little over 10,000 subjects were found who had cholesterol levels at or below 250 mg/dl (6.5 mmol/l). In addition to the blood pressure medications, half of this group got 10 mg of Lipitor (atorvastatin) per day while the other half got a placebo along with their blood pressure medication.

In the Lipitor/placebo arm of this study, the endpoint was defined as a non-fatal heart attack or death from heart disease.

After about 3.3 years, this arm of the study was discontinued because there were a significant number of either heart attacks of deaths from heart disease in the placebo group, and the researchers felt it was unethical to continue the study.

When the data were examined, it turned out that the group taking Lipitor experienced a 36 percent decrease in relative risk for heart disease. Thus the barrage of ads for Lipitor that followed.
Even websites aimed at doctors, using info from the Lipitor package insert, showed graphics designed to make any doctor grab for the Lipitor pre-printed prescription pad. This graph is 100 percent accurate, but, as we shall see, hugely misleading.

Lipitor vs placebo graph Absolute risk versus relative risk: Why you need to know the difference
Figure 1 below is a graphic showing what that 36 percent relative risk looks like.
Control vs Lipitor vs 36percent Absolute risk versus relative risk: Why you need to know the difference
Figure 1. Relative risk of 36 percent with a large absolute risk.
Pretty impressive, isn’t it? Makes you wonder why anyone wouldn’t want to take Lipitor. Problem is, Figure 1, which I made up, shows a relative risk differential of 36 percent, but it also shows a large absolute risk.

Looking at Figure 2 below, which shows the actual absolute risk of experiencing a heart attack or dying from heart disease in this study, you can see that the 36 percent relative risk reduction is accurate. But does it really make you want to stampede to the pharmacy to pick up your Lipitor? Remember, these subjects all had high blood pressure and two other risk factors for heart disease, yet their absolute risk is pretty low. Would you want to take a medicine that could give you muscle aches and pains along with muscle wasting, short term memory loss and possibly fatal liver or kidney damage based on the absolute risk shown in Figure 2?

LIpitor vs Control MI Absolute risk versus relative risk: Why you need to know the difference
Figure 2.  Relative risk of 36 percent with actual absolute risk in the Lipitor study. 
If you compare Figure 1 to Figure 2 above, both of which have the same relative risk, you can readily see that the absolute risk is extremely important. If the absolute risk is high, as it is in Figure 1 at the top, then the relative risk becomes important.

If, as in Figure 2, the real Lipitor vs placebo graphic, the absolute risk is small, then not so much so. Both Figure 1 and Figure 2 show the same relative risk, but not anywhere close to the same absolute risk. Which is why you always want to know the absolute risk before you make a decision on anything based on relative risk. Because absolute risk is usually pretty low in drug studies, the pharmaceutical industry typically uses the relative risk number to sell their medicines.

In the case of statins, most people go on statins because what they really fear is dropping dead of a heart attack, and they feel the statins are protective.

What happened in this study? There were two more deaths in the placebo group than in the Lipitor group, but that number wasn’t statistically significant. Figure 3 below is a graphic showing the absolute risk of having a fatal heart attack in the Lipitor trial.

Control vs Lipitor vs Fatal MI Absolute risk versus relative risk: Why you need to know the difference
Figure 3. Relative and absolute risk of fatal heart attack in the Lipitor study.
Not a huge absolute risk.

Let’s look at another example that is much more dramatic than the Lipitor study above.

I pulled a paper from the New England Journal of Medicine (NEJM) looking at coronary artery calcification (CAC) scores and risk for having a heart attack.

Approximately 20 percent of the plaque in coronary arteries is composed of calcium, which shows up on X-rays. But since the heart is constantly in motion, it’s difficult to see the calcium in standard X-rays of the chest. Specialized CT machines, however, can take extremely fast photos of the heart and actually see the calcium in the coronary arteries. A computer program then converts this calcium into a score, which is simply a number. If you have a zero score, you have no calcium accumulation, which means you probably don’t have any plaque to speak of. You could have some soft plaque that hasn’t yet calcified, but a zero CAC score is definitely a good one.

The NEJM study looked at CAC scores of 6,722 subjects and broke them out into one of four groups. Zero, CAC score of 1-100, CAC score of 101-300, and CAC score of greater than 300. These groups were followed for about 4 years on average for the main endpoint of the study, which was a major coronary event, defined as a heart attack or death from coronary heart disease.

You can see the relative and absolute risk in the graphics below. Upper left is zero CAC score. Upper right, 1-100 CAC score. The lower left is 101-300 CAC score, and the lower right is greater than a 300 CAC score.

CAC 0 vs CAC 1 100 vs MI Absolute risk versus relative risk: Why you need to know the difference
CAC 100 300 vs CAC300 vs MI Absolute risk versus relative risk: Why you need to know the differenceFigure 4. Relative and absolute risks for various CAC scores.  Zero upper left. 1-100 CAC upper right.
101-300 CAC lower left.  Greater than 300 lower right.
As you can see, the risk for a major coronary event is negligible with a CAC score of zero. And the risk increases as the CAC scores go up. The subjects in this study who had CAC scores above 300 experienced 19 times more major coronary events than did those with zero CAC scores. Which would mean a 300+ CAC score carries a relative risk of 1900 percent* as compared to a zero, which is humongous. But looking at the absolute risk tells a different story.

I’ve seen patients with 300+ CAC scores come unglued. Granted, it’s not a score you want to see, but it doesn’t mean you’re going to die the next day. A lot of people feel that way, though. Which is why knowing the absolute risk is important before you decompensate over a high relative risk for this or any condition.

If you would like to play around with these absolute and relative risk graphics, you can do so at this site.  Keep the site bookmarked, so the next time you come across a relative risk, look up the actual risk and graph it.  Only then will you know the true risk involved.

*I used a relative risk score based on the raw data from the study.  If you look at the actual study (linked above), you see different relative risks than what I have.  The ones in the study are lower because the authors extrapolate the data out longer and take into account when the events happened.  So, if you happen to have a high CAC score, your risk is lower than what I show above, which was strictly for illustrative purposes.
Read the complete article here.

20140721 Added this link to good video on relative risk. Well worth the watch.

No comments:

Post a Comment

I appreciate appropriate comments but reserve the right to publish those with credible, verifiable, significant information to contribute to the topic at hand. I will not post comments with commercial content nor those containing personal attacks. Thank You.