statistics
Odds Ratio vs Relative Risk: What They Mean and When to Use Each
Relative risk (RR) is the risk of the outcome in the exposed group divided by the risk in the unexposed group. Odds ratio (OR) is the odds of the outcome (or exposure) in one group divided by the odds in the other
Resora Research Team · 14 Nov 2026 · 4 min read
Relative risk (RR) is the risk of the outcome in the exposed group divided by the risk in the unexposed group. Odds ratio (OR) is the odds of the outcome (or exposure) in one group divided by the odds in the other. Both equal 1 when there is no association. Use RR in cohort studies and randomised trials, where risk can be calculated directly. Use OR in case–control studies, where risk cannot be calculated, and in logistic regression. When the outcome is rare (under about 10%), OR approximates RR. When the outcome is common, OR overstates the effect, so never read it as "times more likely". Alongside either measure, report the absolute difference (risk difference or NNT), since a large relative effect on a rare outcome may be clinically small.
Start with the 2×2 table
Both measures come from the same table:
| Disease present | Disease absent | Total | |
|---|---|---|---|
| Exposed | a | b | a + b |
| Unexposed | c | d | c + d |
Risk = the proportion who develop the outcome, = events ÷ everyone at risk. Odds = events ÷ non-events.
RR = \frac{a/(a+b)}{c/(c+d)} \qquad\qquad OR = \frac{a/b}{c/d} = \frac{ad}{bc}Worked example
A cohort study follows 100 smokers and 100 non-smokers. Chronic cough develops in 40 smokers and 20 non-smokers.
| Cough | No cough | Total | |
|---|---|---|---|
| Smokers | 40 | 60 | 100 |
| Non-smokers | 20 | 80 | 100 |
| Measure | Calculation | Result |
|---|---|---|
| Risk in smokers | 40 / 100 | 0.40 (40%) |
| Risk in non-smokers | 20 / 100 | 0.20 (20%) |
| Relative risk | 0.40 / 0.20 | 2.0 (95% CI 1.26 to 3.17) |
| Odds in smokers | 40 / 60 | 0.67 |
| Odds in non-smokers | 20 / 80 | 0.25 |
| Odds ratio | 0.67 / 0.25 = (40 × 80) / (60 × 20) | 2.67 (95% CI 1.42 to 5.02) |
Interpretation. Smokers had twice the risk of chronic cough (RR 2.0). The OR of 2.67 is larger because the outcome is common (30% overall), so the two measures drift apart.
How to read the value
| Value | Meaning |
|---|---|
| 1 | No association |
| Above 1 | Exposure linked to higher risk or odds |
| Below 1 | Exposure linked to lower risk or odds (protective) |
An RR of 0.6 means a 40% reduction in risk. Always check that the 95% CI excludes 1.
Which one should you use?
| Study design | Measure | Why |
|---|---|---|
| Randomised trial | RR (and risk difference, NNT) | Risk can be calculated directly |
| Cohort study | RR (or hazard ratio if time matters) | Participants are followed from exposure, so incidence is known |
| Case–control study | OR only | Investigators chose how many cases and controls to recruit, so risk cannot be calculated |
| Cross-sectional study | Prevalence ratio or OR | Prevalence ratio is easier to interpret; OR is common because of logistic regression |
| Logistic regression (any design) | Adjusted OR | The model's natural output |
Why a case–control study can only give an OR
In a case–control study the investigator fixes the number of cases and controls, so the proportion with disease in each exposure group depends on the sampling, not on real risk. The OR, however, stays the same whether you calculate it by exposure among cases and controls or by disease among exposed and unexposed. That mathematical property is why it works.
When OR overstates RR
The OR approximates the RR only when the outcome is rare.
| Outcome frequency | Exposed | Unexposed | RR | OR |
|---|---|---|---|---|
| Rare | 2 of 100 | 1 of 100 | 2.0 | 2.02 |
| Common | 60 of 100 | 30 of 100 | 2.0 | 3.5 |
In the common-outcome row, the risk doubled, but an OR of 3.5 would suggest the effect is much bigger. This is why a result such as "OR = 3.5" must not be described as "3.5 times more likely" unless the outcome is rare.
When the outcome is common and you want a ratio you can interpret as risk, use log-binomial or Poisson regression with robust variance to get prevalence ratios or relative risks directly.
Beyond ratios: absolute measures
Ratios hide how many people are actually affected. Always add an absolute measure.
| Measure | Formula | Smoker example |
|---|---|---|
| Risk difference (attributable risk) | Risk exposed − risk unexposed | 0.40 − 0.20 = 0.20, or 20 extra cases per 100 smokers |
| Attributable fraction among exposed | (RR − 1) / RR | (2 − 1) / 2 = 50% of cough in smokers is due to smoking |
| Number needed to harm / treat (NNH / NNT) | 1 / risk difference | 1 / 0.20 = 5: one extra case per 5 smokers |
Why it matters. A treatment that halves the risk of an event (RR 0.5) is impressive when risk falls from 20% to 10% (NNT = 10) and trivial when it falls from 0.002% to 0.001% (NNT = 100,000). The RR is the same in both.
Common mistakes
- Calculating RR from a case–control study. It is not valid.
- Saying "OR times more likely" when the outcome is common.
- Forgetting that the OR is relative to the reference group. Always state it.
- Reporting the ratio without the 95% CI or absolute risk.
- Mixing up OR with hazard ratio (time-to-event) and prevalence ratio (cross-sectional).
- Treating an OR below 1 as "no effect". It means a protective association; check the CI.
- Reversing the direction: an OR of 0.5 for exposure A is an OR of 2 when the reference group is switched.
Keywords
odds ratio vs relative risk, odds ratio, relative risk, risk ratio, OR vs RR, how to calculate odds ratio, odds ratio interpretation, 2x2 table, case control odds ratio, cohort relative risk, adjusted odds ratio, logistic regression odds ratio, number needed to treat, attributable risk, epidemiology measures of association
References
- Park K. Park's Textbook of Preventive and Social Medicine — chapter on epidemiology.
- Gordis L. Epidemiology. 5th ed. Elsevier; 2014.
- Davies HTO, Crombie IK, Tavakoli M. When can odds ratios mislead? BMJ. 1998;316:989–991.
- Zhang J, Yu KF. What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA. 1998;280:1690–1691.
- Altman DG. Practical Statistics for Medical Research. Chapman & Hall/CRC; 1991.
Related services
- Research Topic Selection & RefinementRefinement of a research topic based on feasibility, relevance, available literature and study setting. Final topic approval remains with the researcher/institution.
- Research Question, Objectives & HypothesisSupport to formulate a focused research question, primary/secondary objectives and testable hypothesis where applicable.
- Literature Search StrategySearch-term development, database search strategy and inclusion/exclusion framework for a literature search.
- Preliminary Literature ReviewStructured summary of relevant literature to help identify context and research gaps; scope and number of papers agreed in advance.
- Synopsis Preparation SupportStructured synopsis support based on the researcher's own study idea, institutional format and supplied references; includes one revision.
- Research Protocol Development SupportSupport with protocol structure, rationale, objectives, study design, methodology and analysis plan. Does not include fabricated content or guaranteed ethics approval.