Relative Risk vs Odds Ratio: Differences and Examples

Relative risk (RR) compares the probability of an outcome between two groups, whereas the odds ratio (OR) compares their odds. RR is usually easier to interpret and is preferred when risks can be measured directly. OR is essential in conventional case-control studies and is commonly produced by logistic regression, but it can exaggerate the apparent size of an association when the outcome is common.

Relative Risk vs Odds Ratio at a Glance

Start With a 2×2 Table

Both measures can be calculated from a table that cross-classifies exposure and outcome:

What Is Relative Risk?

Relative risk, also called the risk ratio, is the risk of an outcome in the exposed or treated group divided by the risk in the comparison group.

Risk in exposed group = a ÷ (a + b)

Risk in unexposed group = c ÷ (c + d)

Relative risk = [a ÷ (a + b)] ÷ [c ÷ (c + d)]

How to Interpret Relative Risk

• RR = 1: the observed risk is the same in both groups.
• RR > 1: the outcome is more likely in the exposed group.
• RR < 1: the outcome is less likely in the exposed group.

For example, an RR of 2 means that the outcome risk in one group is twice the risk in the comparison group. An RR of 0.70 means that the observed risk is 30% lower on a relative scale because 1 − 0.70 = 0.30.

What Is an Odds Ratio?

Odds are calculated as the probability that an event occurs divided by the probability that it does not occur. In a 2×2 table, the odds ratio compares the odds of the outcome between the two groups.

Odds in exposed group = a ÷ b

Odds in unexposed group = c ÷ d

Odds ratio = (a ÷ b) ÷ (c ÷ d) = ad ÷ bc

How to Interpret an Odds Ratio

• OR = 1: the observed odds are the same in both groups.
• OR > 1: the odds are higher in the exposed group.
• OR < 1: the odds are lower in the exposed group.

An OR of 2 means that the odds are twice as high. It does not necessarily mean that the probability, or risk, is twice as high.

Why Risk and Odds Are Not the Same

Risk uses the entire group as the denominator. Odds compare people with the outcome against people without it.

Suppose 20 of 100 people experience an outcome:

• Risk = 20 ÷ 100 = 0.20, or 20%
• Odds = 20 ÷ 80 = 0.25

At low probabilities, risk and odds are numerically close. As the probability increases, they separate. A risk of 50% corresponds to odds of 1, while a risk of 80% corresponds to odds of 4.

Worked Example 1: A Common Outcome

Consider a study in which 40 of 100 exposed participants and 20 of 100 unexposed participants develop an outcome.

Relative Risk Calculation

Risk in exposed group = 40 ÷ 100 = 0.40

Risk in unexposed group = 20 ÷ 100 = 0.20

RR = 0.40 ÷ 0.20 = 2.00

The exposed group has twice the risk of the outcome.

Odds Ratio Calculation

Odds in exposed group = 40 ÷ 60 = 0.667

Odds in unexposed group = 20 ÷ 80 = 0.25

OR = 0.667 ÷ 0.25 = 2.67

The odds are approximately 2.67 times as high. Reporting this OR as “2.67 times the risk” would be incorrect because the directly calculated RR is 2.00.

Worked Example 2: A Rare Outcome

Now suppose the outcome occurs in 2 of 1,000 exposed participants and 1 of 1,000 unexposed participants.

Relative Risk

RR = (2 ÷ 1,000) ÷ (1 ÷ 1,000) = 2.00

Odds Ratio

OR = (2 × 999) ÷ (998 × 1) ≈ 2.00

Because the event is rare in both groups, the OR closely approximates the RR. This is known as the rare-outcome approximation. “Rare” is not a rigid universal threshold, so researchers should examine the actual baseline risk rather than relying on a single cut-off.

When Should You Use Relative Risk?

Relative risk is appropriate when the study design allows the investigator to calculate outcome risk in each group. Common examples include:

• Prospective or retrospective cohort studies with defined denominators
• Randomised controlled trials
• Cross-sectional analyses when the measure is explicitly treated as a prevalence ratio rather than an incidence risk ratio

RR is often preferable for communication because clinicians and patients can understand a probability comparison more directly than an odds comparison.

When Should You Use an Odds Ratio?

The odds ratio is commonly used in:

• Traditional case-control studies: participants are selected according to outcome status, so the sampled proportion of cases generally does not represent population risk.
• Logistic regression: the model naturally estimates adjusted odds ratios for binary outcomes.
• Matched case-control studies: specialised analyses estimate odds ratios while accounting for matching.

Some specialised case-control sampling designs can estimate rate ratios or risk ratios under particular assumptions. Nevertheless, the conventional unmatched case-control analysis is most commonly expressed with an OR.

Why Odds Ratios Can Appear More Dramatic

When an association is above 1 and the outcome is common, the OR is usually farther from 1 than the corresponding RR. For protective associations below 1, the OR may likewise appear more strongly protective.

This difference does not make the OR mathematically wrong. The error occurs when odds are described as risks or probabilities. Authors should label the measure accurately and, where possible, provide absolute risks alongside the relative estimate.

Converting an Odds Ratio to an Approximate Relative Risk

If the baseline risk in the unexposed or control group is known, an OR can be converted to a corresponding RR using:

RR = OR ÷ [(1 − P₀) + (P₀ × OR)]

Here, P₀ is the baseline risk in the comparison group.

For an OR of 2.67 and a baseline risk of 0.20:

RR = 2.67 ÷ [(1 − 0.20) + (0.20 × 2.67)] ≈ 2.00

This conversion depends on a valid baseline risk. In a conventional case-control sample, the percentage of controls or cases in the dataset should not automatically be treated as the population baseline risk.

Adjusted RR and Adjusted OR

Crude measures come directly from a 2×2 table. Adjusted measures account for variables such as age, sex, disease severity, or other potential confounders.

• Logistic regression generally gives adjusted odds ratios.
• Log-binomial regression can estimate adjusted risk ratios directly.
• Modified Poisson regression with robust standard errors is frequently used to estimate adjusted risk ratios for binary outcomes.

An adjusted OR should still be interpreted as an odds ratio. Adjustment does not turn it into a risk ratio.

Confidence Intervals and Statistical Interpretation

RR and OR are usually reported with a confidence interval, commonly a 95% confidence interval. A narrower interval indicates greater precision than a wider interval under the same analytical framework.

Because the null value for both measures is 1, an interval that includes 1 is compatible with no association at the stated confidence level. However, clinical importance should not be judged from statistical significance alone. The estimate, interval width, baseline risk, study design, bias, confounding, and absolute effect should all be considered.

Relative Measures Should Be Paired With Absolute Effects

Neither RR nor OR shows how common the outcome is by itself. The same RR can correspond to very different absolute changes:

• An increase from 1% to 2% gives RR = 2, with an absolute increase of 1 percentage point.
• An increase from 20% to 40% also gives RR = 2, with an absolute increase of 20 percentage points.

For treatment studies, useful companion measures may include the risk difference, absolute risk reduction, number needed to treat, or number needed to harm.

Common Mistakes to Avoid

Calling an Odds Ratio a Relative Risk

Do not say “twice the risk” when the reported measure is an OR of 2 unless the RR has also been calculated or a justified rare-outcome approximation applies.

Assuming Every Case-Control Study Gives a Population Risk

In a conventional case-control study, the number of cases and controls is set by the sampling design. These sampled proportions usually cannot be used to calculate population incidence directly.

Ignoring the Outcome Frequency

The OR and RR can differ materially when the outcome is common. Always inspect the baseline risk when comparing or communicating these measures.

Reporting Only a Ratio

A ratio without group-specific event counts, risks, confidence intervals, or absolute differences may be technically correct but clinically difficult to interpret.

Practical Reporting Example

A clear results statement might read:

“The outcome occurred in 40% of exposed participants and 20% of unexposed participants, corresponding to a relative risk of 2.00. The odds ratio was 2.67. The absolute risk difference was 20 percentage points.”

This wording separates probability, odds, relative effect, and absolute effect rather than treating them as interchangeable.

Key Takeaway

The main difference in relative risk vs odds ratio is the quantity being compared: RR compares probabilities, while OR compares odds. Use RR when risks can be measured directly and OR when required by the design or model. When outcomes are common, avoid interpreting an OR as though it were an RR, and report absolute risks whenever the data allow.

Frequently Asked Questions

Is an odds ratio the same as relative risk?

No. Relative risk compares probabilities, while the odds ratio compares odds. They may be numerically similar for rare outcomes but are not the same measure.

Why is the odds ratio used in case-control studies?

In a conventional case-control study, investigators select participants based on outcome status. The sampled case proportion therefore usually cannot estimate population risk directly, but exposure odds can be compared between cases and controls.

When does an odds ratio approximate relative risk?

The approximation is closest when the outcome is uncommon in both groups. As outcome frequency rises, the two measures can differ substantially.

Can an odds ratio be converted to relative risk?

Yes, if a valid baseline risk is available. The conversion is RR = OR ÷ [(1 − P0) + (P0 × OR)], where P0 is the risk in the comparison group.

Which measure is easier for patients to understand?

Relative risk is usually easier because it compares probabilities directly. Absolute risks or risk differences should also be provided to show the size of the real-world change.

Does an RR or OR prove causation?

No. Both quantify an association. Causal interpretation additionally depends on study design, temporality, randomisation or confounding control, measurement quality, bias, consistency, and biological or clinical context.

Medical disclaimer: This article is for education and research interpretation only. It does not replace consultation with a qualified clinician, epidemiologist, or biostatistician for decisions about individual care or study analysis.