# Odds ratio

Source: https://onco.cc/terms/odds-ratio/  
OnCo record `odds-ratio` (Term). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

An odds ratio compares events with non-events in two groups. Odds use a different denominator from probability, so doubling the odds does not generally double the chance.

## Summary

An odds ratio divides the odds of an event in one group by the odds in a reference group. Probability uses all assessed outcomes in the denominator; odds compare events with non-events. The two scales can differ substantially when an event is common. An odds ratio does not supply an absolute risk or a causal treatment effect by itself. Its interpretation depends on outcome coding, the comparison, follow-up and model adjustment; a case-control sample does not reveal population prevalence from its case fraction.

## Fields

- Kind: Term
- Last checked: 2026-11-10
- Also known as: odds ratio; OR

## Notes

- Probability counts events among everyone; odds count events relative to non-events. In a fictional bag with 20 red and 80 blue counters, the probability of red is 20/100 = 20%, but the odds are 20/80 = 0.25. For probability p, odds are p/(1 - p). An odds ratio divides one group's odds by the reference group's odds for a defined event. A value above one means higher odds in the numerator group; whether that is favourable depends on whether the event is response or harm. Specify the event, comparator and horizon. When events are rare in both groups the odds ratio can approximate the risk ratio; when they are common it can differ substantially and cannot be applied directly as a multiplier to a percentage.
- Logistic regression commonly reports odds ratios. An estimate per one unit of a predictor differs from one per ten units, so its increment and outcome coding matter. Report the adjustment set and confidence interval alongside the estimate. An adjusted odds ratio can differ from the unadjusted value even without confounding, a property called noncollapsibility. A change after adjustment alone therefore does not prove confounding, and adjustment alone does not establish causation.
- In a fictional comparison with complete response assessment, 40 of 100 people respond in group A and 20 of 100 in group B. A's response probability uses all 100 people, whereas its response odds use the 60 nonresponders. The response risk ratio is 0.40/0.20 = 2; the odds ratio is (40/60)/(20/80), about 2.67; and the absolute response difference is 20 percentage points. The same counts support three different measures. “2.67 times the odds” does not mean “2.67 times the chance”, and the study design determines whether a causal claim is justified.
- Case-control studies deliberately sample cases and controls, so the case fraction in that sample is not population prevalence and cannot simply supply absolute risk. Zero counts can make a raw odds ratio zero, infinite or undefined; estimation method and uncertainty matter. Reversing the comparison takes the reciprocal. For another fictional example, 20 adverse events among 100 people versus 10 among 100 gives a risk ratio of 2 but an odds ratio of (20/80)/(10/90) = 2.25. A response odds ratio is also different from a survival hazard ratio. An assay-based association with response does not quantify the benefit of acting on a biomarker; its model, endpoint and treatment comparison need their own evidence. All counts here are invented.

## Sources

- CDC: risk ratios and odds ratios: https://archive.cdc.gov/www_cdc_gov/csels/dsepd/ss1978/lesson3/section5.html
- Statistics notes. The odds ratio.: https://doi.org/10.1136/bmj.320.7247.1468
- Confounding and Collapsibility in Causal Inference: https://doi.org/10.1214/ss/1009211805
- Hernan and Robins: Causal Inference: What If: https://miguelhernan.org/whatifbook

## Connected records

- terms: [Confidence interval](https://onco.cc/terms/confidence-interval/), [Hazard ratio (HR)](https://onco.cc/terms/hazard-ratio/), [Kaplan-Meier curve, censoring and proportional hazards](https://onco.cc/terms/kaplan-meier-curve/)
- fronts: [Drug Discovery Platforms](https://onco.cc/fronts/drug-discovery/)

---
JSON: https://onco.cc/api/v1/entities/odds-ratio.json