What is the difference between an odds ratio and a risk ratio?

A risk ratio divides the probability of an outcome in one group by the probability in another. An odds ratio divides the odds, the chance of the event over the chance of no event, between the groups. They agree when the outcome is rare and diverge as it becomes common.

When should I use an odds ratio instead of a risk ratio?

Use an odds ratio for case-control studies, where participants are sampled by outcome so true risks cannot be measured, and when reporting logistic regression results. Use a risk ratio for randomised trials and cohort studies, where it is easier to interpret as plain probability.

Why do the odds ratio and risk ratio differ for common outcomes?

When an event is common, the odds inflate faster than the probability, so the odds ratio moves further from 1 than the risk ratio. For rare events, probability and odds are nearly equal, so the two measures are almost identical.

How do you convert an odds ratio to a risk ratio?

You can convert an odds ratio to a risk ratio when you know the baseline risk in the comparison group, using RR = OR / (1 minus baseline risk times (1 minus OR)). The two are nearly equal when the outcome is rare, so the conversion matters most for common outcomes, where reading an odds ratio as if it were a risk ratio overstates the effect.

Can you mix odds ratios and risk ratios in one meta-analysis?

Not directly. Every study must contribute the same effect measure before pooling, so you decide on the risk ratio or the odds ratio in the protocol and convert the others to match. Pooling a mix of measures without converting them produces a summary estimate that does not correspond to any real quantity.

Odds Ratio vs Risk Ratio: Which to Use

Q: Why is the odds ratio described as non-collapsible?

Non-collapsibility means the odds ratio averaged over subgroups does not equal the odds ratio in the combined data, so an adjusted odds ratio can differ from the crude one even when the adjustment variable is not a confounder. The risk ratio does not behave this way, which makes odds ratios from logistic regression harder to compare across models with different covariates.

The difference between odds ratio vs risk ratio is what each one divides. A risk ratio compares the probability of an outcome between two groups, the chance of the event happening in the treated group divided by the chance in the control group. An odds ratio compares the odds, the chance of the event divided by the chance of no event, between the same two groups. They answer related but distinct questions, and they only agree numerically when the outcome is rare.

How probability and odds pull apart

Probability and odds describe the same data differently. If an event happens to 1 in 5 people, the probability is 0.2 and the odds are 0.2 divided by 0.8, which is 0.25. When events are uncommon, probability and odds are nearly identical, so the risk ratio and odds ratio track each other closely. As the event becomes common, odds inflate faster than probability, and the two measures diverge sharply. This is the heart of why the same study can report a risk ratio of 1.5 and an odds ratio of 2.0 from one dataset. You can compute both from a two-by-two table with our odds ratio and risk ratio calculator.

A worked two-by-two example

Take a trial where, of 100 treated patients, 40 have the event, and of 100 controls, 60 have the event. The risk in the treated arm is 40 divided by 100, which is 0.40; in the control arm it is 0.60. The risk ratio is 0.40 divided by 0.60, which is 0.67, a one-third relative reduction. The odds in the treated arm are 40 divided by 60, which is 0.67; in the control arm they are 60 divided by 40, which is 1.5. The odds ratio is 0.67 divided by 1.5, which is 0.44. Same data, yet the odds ratio (0.44) looks like a much larger effect than the risk ratio (0.67), because the outcome here is common. A reader who treats that odds ratio as a risk would conclude the treatment more than halves the chance of the event, when it actually cuts it by a third.

Why the odds ratio is non-collapsible

A more technical reason to prefer the risk ratio for communication is non-collapsibility. The risk ratio averaged over subgroups equals the risk ratio in the combined data, but the odds ratio does not: an adjusted odds ratio can differ from the crude one even when the adjustment variable is not a confounder, purely because of how odds behave. That property makes odds ratios from logistic regression harder to compare across models with different covariate sets, and it is one more reason to fix your measure in advance and report it consistently across a synthesis.

Which measure to choose

When the risk ratio is preferable

For most clinical questions the risk ratio is easier to interpret because it speaks in plain probability: a risk ratio of 0.5 means the treated group had half the chance of the event. It is the natural choice for randomised trials and cohort studies, where you can actually observe the underlying risk in each group. Its limitation is that it depends on the baseline risk, so the same relative effect can look very different across populations with different baseline rates.

When the odds ratio is appropriate

The odds ratio is the default in case-control studies, where participants are sampled by outcome rather than exposure, so true risks cannot be measured directly but odds ratios can. It is also the quantity produced by logistic regression, which is why adjusted analyses so often report it. The drawback is interpretability: people routinely read an odds ratio as if it were a risk ratio, which overstates the effect when the outcome is common.

Why it matters in a meta-analysis

In a synthesis the choice is not just interpretive; it is structural. Every study must contribute the same effect measurebefore anything can be pooled, so you decide on risk ratio or odds ratio in the protocol, convert where needed, and stay consistent. The full menu of binary and continuous measures is set out in effect sizes in meta-analysis, and the pooling mechanics are covered in how to do a meta-analysis. Whichever you choose appears on the forest plot with the line of no effect at 1.

A note on absolute effects

Both ratios are relative measures, and a large relative effect can be trivial in absolute terms when the baseline risk is tiny. That is why guideline panels translate a risk ratio into an absolute risk reduction and a number needed to treat, which patients and clinicians can actually act on. Reporting the ratio without the absolute context is one of the most common ways a true result gets oversold.

Common mistakes when reporting the two

A handful of errors do most of the damage. The first is reading an odds ratio as a risk ratio for a common outcome, which overstates the effect, as the worked example showed. The second is mixing measures across a pool: combining studies that report odds ratios with others reporting risk ratios without converting them to a single metric produces a meaningless summary, which is why the choice is a structural decision made in the review protocol. The third is quoting either ratio without its baseline, so a large relative effect on a tiny absolute risk reads as more important than it is, the gap that a number needed to treat calculation is built to close. The fourth is converting between the two with the rare-disease approximation when the outcome is common, which only holds when events are uncommon.

Getting it right

Choose the measure that matches your study design and your audience, fix it in the protocol, never read an odds ratio as a risk when the outcome is common, and pair the relative effect with an absolute one. Confirm the pooled choice does not change your conclusion with a robustness check that swaps the effect measure. Do that, and the odds ratio and risk ratio stop being a source of confusion and become two complementary lenses on the same evidence. If you want the measure chosen and reported correctly for your data, our meta-analysis service handles it end to end.