You are reviewing a recent lipid panel on John, a 50-year-old man who has been following an exercise and diet program since you discovered a high cholesterol level at his wellness physical six months ago. John's total cholesterol has gone from 315 to 280, his HDL from 40 to 45 and his LDL from 205 to 185. John's wife read an ad in a magazine about a cholesterol-lowering medication that will reduce the risk of heart attack by 30 percent, so he asks you about taking it. What will you tell him?

To start, you will certainly evaluate John's other heart attack risk factors — smoking status, hypertension, family history and diabetes — but it will also help if you can explain to John and his wife what this 30 percent risk reduction actually means. To do that, you need to understand how risks are calculated.

Three kinds of risk

The concept of risk has not been well taught in medical education, but the growing interest in evidence-based medicine is starting to change this. There are essentially three ways of reporting risk in a study, and they are all derived from the same statistics:

• Relative risk reduction (RRR) refers to the percentage decrease in risk achieved by the group receiving the intervention vs. the group that did not receive the intervention (the control group).

• Absolute risk reduction (ARR) refers to the actual difference in risk between the treated and the control group.

• Number needed to treat (NNT) refers to the number of people who need to be treated to prevent one undesirable outcome.

To apply reported risks clinically, you also need to know 1) the population studied (e.g., men with or without known coronary artery disease) to see if the results apply to your patient, 2) what outcome was measured (e.g., coronary artery deaths, nonfatal MIs) to know if the study will matter to your patient and 3) how long the study patients were treated.

For example, in a study relevant to John's situation, Scottish men without coronary heart disease but with elevated cholesterol (total over 252; LDL over 170) were studied.¹ Half of the men were given a statin for five years, and half were not. At the end of five years, the researchers counted deaths from coronary disease and non-fatal MIs and found that for every 100 people followed for five years, 7.9 of the control and 5.5 of the treated patients suffered one of the two events being studied. Let's calculate the risks below:

What the study shows, then, is that treating white men who have no known coronary disease with a particular statin medication for five years reduced their chance of having a non-fatal heart attack or coronary death by 30 percent. Put another way, the study shows that you would need to treat 100 men for five years to prevent 2.4 of them from having an MI or coronary death. Finally, you can also say that 42 people with similar cholesterol profiles would need to be treated for five years to prevent one nonfatal MI or coronary death. All these risks are derived from the same data.

Beyond relative risk reduction

Almost all reports in the popular media, and many in the medical literature, present risk results as relative risk reductions rather than absolute risk reductions or number needed to treat. Why? Most likely, it has to do with the perceived impact on readers; that is, relative risk reductions often make data seem more impressive than they actually are. Lest one think that only the public can be misled in this way, a study done by Naylor et al showed a similar effect on primary care physicians' interpretation of risk data.²

Relative risk reductions can be misleading in another way. Consider a relatively rare condition such as the rate of thromboembolic events in reproductive-age women (1/100,000 women per year). Now consider the rate in women on oral contraceptives (7/100,000 women per year). With these numbers, one can calculate that stopping oral contraceptives would lead to a relative risk reduction in thromboembolic events of about 86 percent. However, the absolute risk reduction is only 6/100,000 (or 0.00006), which translates to an NNT of 1/0.00006 (or 16,667). In this case, despite the high relative risk reduction, almost 17,000 women would need to stop taking oral contraceptives for one year to prevent one thromboembolic event.

The point to remember is that the underlying rate of events in the control group has a lot to do with the number of people who need to be treated to prevent one event. This information is not reflected in the relative risk reduction but will be reflected in the absolute risk reduction and the NNT. In fact, many experts recommend presenting risks to patients in the form of an NNT. For example, in John's case, you might say, “We would need to treat 42 people with your level of cholesterol for five years to prevent one nonfatal MI or one death from heart disease.”

You and your patient could then have an informed discussion of the benefits and risks of the medical therapy and whether other factors might change the risks. For example, if John already had known coronary artery disease, the NNT would be lower than 42.

In studies of treatment of hypercholesterolemia in patients with known coronary artery disease (secondary prevention), NNTs have varied from 12 to 33.³ In other words, 12 to 33 patients with known coronary artery disease need to be treated for five years to prevent one fatal or nonfatal heart attack. For patients without coronary artery disease, studies have shown 42 to 55 patients need to be treated for five years to prevent one event. In contrast, a very effective treatment such as triple-drug therapy to eradicate H. pylori has an NNT in the 2 to 4 range.

Clearly, the lower the NNT, the better. As the NNT rises, factors such as patient and physician values, benefits and risks, and costs need to be taken into consideration when deciding if a particular therapy is “worth it.” Over the past few years, more studies of treatment are reporting the NNT. If the NNT is not reported, it can be helpful to calculate the NNT yourself and explain it to your patient. There certainly will be times when patients will decide to forgo a particular intervention when they understand the potential benefits and risks in this way.