Odds Ratio

The odds ratio (OR) is a measure of association that compares the odds of an event occurring in one group to the odds of it occurring in another group. It is widely used in case control studies and logistic regression analyses to quantify the strength and direction of the relationship between an exposure and an outcome.

The odds ratio is one of the most frequently reported effect measures in epidemiological and clinical research, particularly in studies of disease risk. It is the natural measure of association for case control studies, where the sampling design precludes direct calculation of risk or incidence rates. In prospective studies and cross sectional analyses, odds ratios are produced by logistic regression, one of the most commonly applied multivariable methods. Understanding odds ratios is therefore essential for interpreting a vast body of published research, for conducting meta analyses that pool results across studies, and for making evidence based decisions in clinical practice and public health.

The odds of an event are defined as the probability that the event occurs divided by the probability that it does not occur. If 20 of 100 exposed individuals develop a disease, the odds of disease in the exposed group are 20/80, or 0.25. If 10 of 100 unexposed individuals develop the disease, the odds in the unexposed group are 10/90, or approximately 0.11. The odds ratio is 0.25/0.111 = 2.25, indicating that the odds of disease are 2.25 times as high in the exposed group as in the unexposed group. The risk ratio for the same data, 20% against 10%, is 2.0, closer to 1 than the odds ratio.

An OR of 1.0 indicates no association between the exposure and the outcome. An OR greater than 1.0 indicates a positive association (higher odds of the outcome among the exposed), and an OR less than 1.0 indicates a negative association (lower odds of the outcome among the exposed). Confidence intervals around the OR are used to assess statistical precision; if the 95% confidence interval includes 1.0, the association is not statistically significant at the 0.05 level.

When the outcome is rare (typically less than 10% in all groups), the OR approximates the relative risk (RR), making interpretation more intuitive. However, when the outcome is common, the OR exaggerates the apparent strength of association compared to the RR. This distinction is important because many readers interpret ORs as if they were RRs, leading to overestimation of risk when events are frequent. For common outcomes, researchers should consider reporting risk ratios or prevalence ratios alongside or instead of odds ratios.

Adjusted odds ratios from multivariable logistic regression adjust for measured confounding variables by holding the other variables in the model constant while estimating the independent association between a specific exposure and the outcome. This ability to adjust for multiple confounders simultaneously is one reason logistic regression and its associated ORs are so widely used.

Report odds ratios when conducting case control studies, when using logistic regression for binary outcomes, or when pooling results across studies in a meta analysis. When the outcome is common and a cohort or cross sectional design is used, consider supplementing or replacing the OR with a risk ratio or prevalence ratio for clearer interpretation. Always report ORs with confidence intervals and the sample sizes for each comparison group.

A common mistake is interpreting odds ratios as relative risks, particularly when the outcome is common. An OR of 2.5 for a disease that affects 40% of the unexposed group does not mean the disease is 2.5 times as likely in the exposed group; the relative risk is about 1.6 (2.5 ÷ (1 − 0.4 + 0.4 × 2.5)). Another shortcoming is reporting only odds ratios from prospective cohort data when risk ratios could be computed directly and would be easier to interpret. Failing to report confidence intervals or to assess statistical significance is a further shortcoming. Finally, in meta analyses, combining ORs and RRs without conversion can produce misleading pooled estimates, especially when the outcome is common.

A case control study examines the association between regular use of nonsteroidal anti inflammatory drugs and the risk of colorectal cancer. The study enrolls 500 patients diagnosed with colorectal cancer (cases) and 500 controls without cancer drawn from the same source population. Among cases, 120 (24%) report regular use of the drug class, compared to 180 (36%) of controls. The odds of use among cases are 120/380 = 0.316, and among controls 180/320 = 0.563. The OR is 0.316/0.563 = 0.56 (95% CI: 0.43 to 0.74), suggesting that regular use is associated with 44% lower odds of colorectal cancer.