Relative Risk vs Absolute Risk: Why Both Numbers Matter

TLDR

Relative risk describes the size of a difference compared with another group. Absolute risk shows how often an outcome actually occurs. A claim of “50% lower risk” could mean a decline from 20 in 100 people to 10 in 100, or from 2 in 10,000 to 1 in 10,000. The relative reduction is identical, but the absolute benefit is dramatically different. To interpret a health claim, ask for both numbers, the baseline risk, the outcome, the comparison group, the population, and the timeframe.

The central lesson in relative risk vs absolute risk is not that one measure is honest and the other is misleading. Both are legitimate, but they answer different questions. Relative risk helps show the strength of a comparison. Absolute risk reveals the practical scale of the difference.

Whenever a headline says a treatment “cuts risk by 50%” or an exposure “doubles risk,” important context is missing. Risk of what? Compared with what? Among whom? Over what period? And how many people experienced the outcome in each group? Until those questions are answered, the percentage is incomplete.

What is absolute risk?

Absolute risk is the probability that an outcome will occur in a defined population over a specified period. It may be expressed as a percentage, a decimal, or a natural frequency such as “2 out of 100 people over five years.” The National Cancer Institute defines it in terms of the probability of developing a disease in a particular population during a stated period.

The timeframe and population are essential. A 10-year risk is not interchangeable with a one-year risk, and a rate measured among older adults with an existing condition may not describe younger adults without that condition. An absolute risk estimate describes what happened or is expected in a defined group; it is not automatically a personalized forecast.

Natural frequencies can make the scale easier to see. A Cochrane review found that formats such as “10 out of 1,000” were better understood than percentages in the diagnostic and screening comparisons it examined, although those comparisons did not establish effects on actual health behavior.

What is relative risk?

Relative risk, also called the risk ratio, compares the probability of an outcome in one group with the probability in another. It is calculated by dividing the risk in the treatment or exposed group by the risk in the comparison group.

A relative risk of 1 means the observed risks are equal.
A relative risk below 1 means the outcome was less common in the treatment or exposed group.
A relative risk above 1 means the outcome was more common in that group.
A relative risk of 0.50 means the first group had half the risk of the comparison group. It does not mean that everyone had a 50% chance of the outcome.

Relative risk is especially useful for comparing effects across groups or studies. Its limitation is that it does not, by itself, show whether the outcome was common or rare. That is why guidance on interpreting clinical evidence supports presenting relative and absolute effects together. The Cochrane Handbook’s guidance on interpreting results emphasizes clear reporting of important beneficial and adverse outcomes, uncertainty, and applicability.

A simple relative risk vs absolute risk example

Consider a hypothetical five-year study of 200 people. In the comparison group, 2 of every 100 people experience the outcome. In the intervention group, 1 of every 100 experiences it.

Measure	Calculation	Result
Comparison-group risk	2 ÷ 100	2%
Intervention-group risk	1 ÷ 100	1%
Absolute risk reduction	2% − 1%	1 percentage point
Relative risk	1% ÷ 2%	0.50
Relative risk reduction	1 − 0.50	50%
Number needed to treat	1 ÷ 0.01	100 over five years

Every result in the table describes the same hypothetical data. “A 50% relative reduction” sounds substantial, while “1 fewer event for every 100 people treated over five years” provides the practical scale. Neither statement is mathematically wrong. Together, they give a more complete picture.

An NIH health-risk explainer uses this same 2-in-100 versus 1-in-100 pattern to illustrate how a 50% relative reduction can correspond to an absolute reduction of one event per 100 people and an NNT of 100.

Why baseline risk changes the meaning

Baseline risk is the event rate in the relevant comparison group. Because an absolute difference depends on that starting risk, the same relative effect can produce very different practical outcomes.

Hypothetical scenario	Without intervention	With intervention	Relative reduction	Absolute reduction
Rare outcome	2 in 10,000	1 in 10,000	50%	1 fewer event per 10,000
Common outcome	20 in 100	10 in 100	50%	10 fewer events per 100

In both scenarios, the relative risk is 0.50 and the relative risk reduction is 50%. But the first scenario changes the event count by 1 per 10,000, while the second changes it by 10 per 100. Baseline risk determines how a relative effect translates into an absolute difference.

This also explains why the same treatment may offer different absolute benefits to different populations. If one group begins with a higher probability of the outcome, an equivalent relative reduction usually produces a larger absolute reduction. Whether the same relative effect truly applies across populations is a separate clinical question that depends on the available evidence.

ARR, RRR, NNT, and NNH in plain English

Absolute risk reduction

Absolute risk reduction, or ARR, is the comparison-group risk minus the intervention-group risk when an outcome becomes less frequent. If risk falls from 8% to 6%, the ARR is 2 percentage points, or 0.02 as a decimal.

Relative risk reduction

Relative risk reduction, or RRR, describes the reduction relative to the starting risk. In the 8%-to-6% example, the relative risk is 6% ÷ 8% = 0.75. The RRR is 1 − 0.75 = 0.25, or 25%.

Number needed to treat

Number needed to treat, or NNT, estimates how many people would need the intervention for one additional person to avoid the defined outcome during the stated period. It is calculated as 1 divided by ARR when ARR is written as a decimal. An ARR of 0.02 produces an NNT of 50.

NNT must stay connected to its outcome, comparator, population, and timeframe. “NNT of 50” is incomplete; “NNT of 50 to prevent one defined event over three years compared with the control” is meaningful. NNT also should not be calculated from a relative-risk claim alone because the calculation requires an absolute event rate.

Absolute risk increase and number needed to harm

If an unwanted event becomes more common, the absolute risk increase, or ARI, is the difference between the groups. Number needed to harm, or NNH, is 1 divided by the absolute increase expressed as a decimal. For example, if an adverse outcome rises from 2% to 4%, the ARI is 2 percentage points and the NNH is 50 over the study period.

Benefits and harms are easiest to compare when they use the same population, timeframe, denominator, and format. Trial readers should also distinguish an event that happened after treatment from one shown to have been caused by it; our guide to adverse events versus side effects explains why that distinction matters.

Five questions to ask about a risk claim

What is the outcome? “Risk” could refer to a symptom, diagnosis, hospitalization, laboratory threshold, or death. These outcomes do not carry equal practical importance.
Who was studied? Look at age, baseline health, disease severity, treatment setting, and other characteristics that may affect applicability.
What is the comparison? Determine whether the study compared an intervention with placebo, usual care, another treatment, or no exposure.
Over what timeframe? A difference over six weeks cannot be interpreted as though it were measured over five years.
What happened in absolute numbers? Ask for the event count in each group, ideally as the same denominator, along with confidence intervals, harms, and the certainty of the evidence.

CDC risk-communication guidance makes a similar point: saying an exposure “doubles the risk” does not reveal the underlying magnitude. A change from 1 in 100 to 2 in 100 communicates both the increase and the starting level more clearly.

Common mistakes when interpreting risk

Confusing percentages with percentage points

A decline from 10% to 5% is a reduction of 5 percentage points, not 5%. Relative to the original 10%, it is a 50% reduction. Percentage points describe subtraction between rates; percent change describes that difference relative to the starting rate.

Confusing relative risk with relative risk reduction

A relative risk of 0.70 means the first group’s observed risk was 70% of the comparison group’s risk. The corresponding relative risk reduction is 30%. The two numbers are related, but they are not interchangeable.

Treating odds ratios as risk ratios

Odds and risks are calculated differently. An odds ratio can approximate a risk ratio when outcomes are rare under appropriate conditions, but the measures can diverge when outcomes are common. The National Cancer Institute distinguishes odds ratios from relative risk measures, so an odds ratio should not simply be described as a relative risk.

Assuming association proves causation

An observed relative risk may describe an association rather than a causal effect. Randomization, study design, confounding, measurement quality, missing data, and consistency with other evidence all influence whether a causal interpretation is justified. A lower event rate in an observational study does not, by itself, prove that the exposure caused the difference.

Equating statistical significance with practical importance

A statistically significant result can represent a small absolute difference, while an important possible effect may remain uncertain in a small study. Confidence intervals help show the range of effects compatible with the data. Cochrane guidance recommends reporting confidence intervals rather than relying only on statistical-significance labels.

How to translate a health headline

When you encounter a headline such as “Treatment lowers risk by 40%,” translate it into a short worksheet:

Outcome: What event became less common?
Baseline: How many people in the comparison group experienced it?
Intervention: How many people in the treatment group experienced it?
Absolute difference: How many fewer events occurred per 100 or 1,000 people?
Time: Over how many months or years?
Uncertainty: What does the confidence interval include?
Tradeoff: What important harms, burdens, or costs occurred?

If a report gives only the relative change, look for a study table, abstract, prescribing document, or evidence summary containing the underlying event counts. Without a baseline event rate, you cannot reliably recover the absolute difference or calculate NNT.

Frequently asked questions

Why can a 50% relative reduction be a small benefit?

Because the reduction is measured relative to the starting risk. Cutting a risk from 2 in 10,000 to 1 in 10,000 is a 50% relative reduction but only one fewer event per 10,000 people. Cutting it from 20 in 100 to 10 in 100 is also a 50% relative reduction, with a much larger absolute difference.

What does it mean when a headline says risk doubled?

It means the reported relative risk is about 2, but the practical size depends on the original rate. A rise from 1 in 1,000 to 2 in 1,000 and a rise from 10 in 100 to 20 in 100 both represent doubled risk, despite very different absolute increases.

Can NNT be calculated from relative risk alone?

No. NNT requires an absolute risk reduction, which requires knowing the baseline event rate and the event rate under the intervention. It also needs a defined outcome and timeframe.

Which measure should I trust more?

Do not choose one and ignore the other. Relative effects describe proportional differences, while absolute effects show practical magnitude. Interpret both alongside the study design, confidence interval, population, timeframe, outcome importance, and harms.

The useful next step

The next time a health claim presents a large percentage, ask to see the underlying event counts. Translate the result into events per 100, 1,000, or 10,000 people over a stated period. Then examine uncertainty, harms, and whether the study population resembles the group to which the claim is being applied.

Relative risk can show that a comparison is meaningful. Absolute risk shows how large that difference may be in practice. Using both turns an attention-grabbing percentage into information that is much easier to evaluate and discuss.