Skip to main content

Welcome to a Grasp lesson

Create a free account to keep studying and create your own study path.

Lesson 2

Converting Relative Risk Changes into Events per 1,000 People

Recommended resources

Hello. The previous lesson showed why a symptom or cardiovascular event occurring after treatment does not, on its own, establish what the treatment caused. A well-designed comparison helps estimate a treatment effect. This lesson asks the next question: how many more or fewer people would experience an event?

You will translate a relative risk change into expected events per 1,000 people, using a stated baseline risk and time period. That skill will help you read statin benefit and harm claims without mistaking a striking percentage for a large difference in the number of people affected.


The missing number in a relative-risk claim

A claim that a treatment “reduces risk by 25%” describes a relative change. It does not say whether the starting risk was 20 in 1,000 or 200 in 1,000. Yet 25% of those two numbers is very different.

Watch the following explanation of absolute risk, relative risk, and why the same relative reduction can yield different absolute benefits.

NCCMT - URE - What’s the Risk? Understanding Absolute and Relative Risk Reduction

Watch “What’s the Risk? Understanding Absolute and Relative Risk Reduction” from NCCMT | CCNMO for its step-by-step comparison of treatment and control groups.

Watch baseline risk to see how an event rate is obtained from a group, then absolute reduction and relative reduction to distinguish subtraction from division. Finish with different baselines; focus on how a fixed relative effect produces different numbers of prevented events.

Baseline risk is the expected chance of a specified event in a specified group without the treatment being evaluated, over a specified period. A baseline risk of 8% over ten years means that, among 1,000 comparable people, about 80 would be expected to experience the event during those ten years. It does not identify which 80 people.

The time period and event belong to the number. “An 8% risk” is incomplete; “an 8% risk of a cardiovascular event over ten years” is interpretable. A risk of a heart attack is not necessarily the same measure as a risk of either a heart attack or a stroke.



Turning the percentage into events per 1,000

The Cochrane Handbook distinguishes relative risk reductions from absolute risk reductions and recommends expressing absolute effects as natural frequencies. Read the indicated sections before working through the calculation below.

Chapter 15: Interpreting results and drawing conclusions

Read the Cochrane Handbook’s account of why a relative effect needs a baseline risk before it can become a number of events.

In Section 15.4.1, “Relative and absolute risk reductions,” read the distinction, beginning with the example comparing a relative reduction with a percentage-point reduction. In Section 15.4.3, “Expressing risk differences,” read natural frequencies from the opening paragraph through the explanation of why different risk groups and follow-up periods need separate absolute estimates. In Section 15.4.4.2, “Computing risk differences or NNT from a risk ratio,” read the worked example, from the need for an assumed baseline risk through the result of 24 fewer events per 1,000. The displayed formulas may appear as images; the calculation is written out below.

Here is a reusable method. Let be the number of people expected to have the event per 1,000 without treatment, and let be the relative risk reduction written as a decimal. For a 25% reduction, .

Suppose a hypothetical group has a ten-year baseline risk of 8% for a defined cardiovascular event. That is events per 1,000 people over ten years. If a suitable study estimates a 25% relative reduction for that event and period, the calculation is:

In percentages, risk changes from 8% to 6%. That is a 25% relative reduction but a 2-percentage-point absolute reduction—equivalently, 20 fewer people with the event per 1,000 over ten years. “25% fewer” does not mean “25 percentage points fewer.”

A study may instead report a risk ratio. A risk ratio of means the treated group’s event risk is estimated at times the comparison group’s risk: the same 25% relative reduction. Check the measure’s name before calculating; an odds ratio is not automatically a risk ratio.



Why starting risk changes the answer

Keep that hypothetical 25% relative reduction and ten-year period, but change the baseline risk:

Ten-year events without treatment Ten-year events with treatment Difference
80 per 1,000 60 per 1,000 20 fewer per 1,000
200 per 1,000 150 per 1,000 50 fewer per 1,000

The relative reduction is 25% in both rows. The absolute benefit is larger in the second because 25% of 200 is more than 25% of 80. Holding the relative effect constant here is a teaching assumption; whether it transfers well between real populations requires evidence.

The same principle applies to increases in risk. In the infographic below, a 50% relative increase changes 2 in 10 to 3 in 10, but changes 4 in 10 to 6 in 10. The added numbers differ even though the relative increase is identical. These are illustrations of risk arithmetic, not estimates of statin effects.

Infographic comparing absolute and relative risk: its middle examples show that the same 50% relative risk increase adds one event when baseline risk is 2 in 10, but two events when baseline risk is 4 in 10.

This is why a headline giving only a relative change leaves out information needed to judge its scale. It is also why figures from different periods cannot be placed side by side as though they were directly comparable. A ten-year baseline risk must not be combined uncritically with a relative effect measured for a different follow-up period, and a ten-year risk cannot simply be divided by ten to obtain a one-year risk.



Reading a statin benefit estimate

A patient decision aid from NICE shows what absolute frequencies look like in a statin decision. Its figures are modelled estimates, presented for different starting risks of heart disease or stroke over ten years; they are not a guarantee of what will happen to any individual.

CG181 Patient decision aid on should I take a statin?

Read the NICE patient decision aid to see how starting cardiovascular risk and estimated statin benefit are expressed as numbers of people, rather than a relative claim alone.

On page 2, under “What is my risk of heart disease or stroke?”, read the risk explanation, from the description of a ten-year risk estimate through the 20-in-100 example. On page 4, under “How much will a statin reduce my risk?”, read the benefit explanation, including who has an event despite treatment. Then compare the diagrams beginning with a 10% risk on page 8 and with a 20% risk on page 9. Focus on the numbers expected to have a heart disease or stroke outcome, with and without a statin, over the same ten years.

At the aid’s 10% ten-year starting risk, about 10 of 100 people would have heart disease or a stroke without a statin, compared with about 6 of 100 with one. Expressed per 1,000, that is approximately 100 versus 60, or 40 fewer people with an event over ten years.

At its 20% starting risk, the corresponding figures are about 20 versus 13 of 100: approximately 200 versus 130, or 70 fewer per 1,000 over ten years. The aid rounds its displayed counts and models a particular treatment effect, so these two examples need not imply an identical relative reduction. Their central lesson is that the starting risk matters when estimating absolute benefit.

Keep the outcome label attached to these figures. “Fewer people with heart disease or stroke” does not mean “the same number of deaths prevented.” A later lesson will separate cardiovascular events from deaths when comparing benefit estimates.



Using the same method for an increased risk

The arithmetic for a possible harm runs in the other direction. If people per 1,000 would experience a specified outcome without treatment and the relative risk increases by a fraction , then:

For example, imagine a hypothetical outcome affecting 10 per 1,000 people over five years, with a 50% relative increase. Since , treatment would be associated with 15 rather than 10 events per 1,000: 5 additional events per 1,000 over five years. This is not an estimate for any particular statin side effect. It illustrates the calculation you will use when examining actual evidence about harms later in the course.

Before translating any claim, pin down four things: the outcome, the comparison group’s baseline risk, the relative change, and the time period. Then report the expected events in each group and their difference, all per 1,000 people over that same period. The result is a group estimate, not a prediction of which individual will benefit or be harmed.



Key takeaways

A relative percentage needs a baseline risk before it can tell you how many events may be added or prevented. For a relative reduction, multiply baseline events per 1,000 by the reduction expressed as a decimal; for a relative increase, use the same multiplication to find additional events. Always name the outcome and period, and distinguish a relative percent change from an absolute percentage-point change.

These calculations turn an effect estimate into a more useful frequency, but they do not tell you how certain that estimate is. The next lesson examines confidence intervals, including why an inconclusive result does not rule out an effect—especially when the outcome is rare.