Understanding Relative Risk
Relative risk (RR) is a comparison of disease incidence between two populations. The exposed group has undergone some intervention or experience a particular factor; the control group has not. If RR equals 1, both groups face identical risk. An RR greater than 1 indicates elevated risk in the exposed group; less than 1 suggests protective effect.
Consider a hypothetical study of 200 smokers and 200 non-smokers tracked for lung cancer incidence over ten years. If 16 smokers develop cancer versus 2 non-smokers, the relative risk would be 8—meaning smokers are eight times more likely to develop the disease under the study conditions.
Relative risk differs from odds ratio, which compares odds rather than probabilities. For rare outcomes, the two metrics converge; for common ones, they diverge substantially. RR is more intuitive for cohort studies and prospective research designs.
Relative Risk and Confidence Interval Formulas
The fundamental calculation divides the risk in the exposed group by the risk in the control group. The confidence interval accounts for sampling variability and indicates the range where the true population parameter likely resides.
Risk (Exposed) = a ÷ (a + b)
Risk (Control) = c ÷ (c + d)
Relative Risk = Risk (Exposed) ÷ Risk (Control)
Standard Error = √(1/a + 1/c − 1/(a+b) − 1/(c+d))
Lower CI = exp(ln(RR) − z × SE)
Upper CI = exp(ln(RR) + z × SE)
a— Count of disease cases in the exposed groupb— Count of non-cases in the exposed groupc— Count of disease cases in the control groupd— Count of non-cases in the control groupz— Z-score corresponding to your chosen confidence level (1.96 for 95%)SE— Standard error of the log relative risk
Interpreting Your Results
A relative risk of 2.0 means the exposed group is twice as likely to experience the outcome. An RR of 0.5 means they are half as likely. The confidence interval reflects precision: narrow intervals suggest more reliable estimates from larger samples, while wide intervals indicate uncertainty and call for caution in drawing conclusions.
Statistical significance occurs when the 95% confidence interval excludes 1.0. If your lower bound is 1.1 and upper bound is 3.5, the result is statistically significant at the 0.05 level—you can reject the null hypothesis of no difference. Conversely, an interval spanning 0.8 to 1.3 includes 1.0, so the effect is not statistically significant.
Always examine effect size alongside significance. A relative risk of 1.05 may be statistically significant in a trial of 50,000 participants yet clinically trivial. Context, study design, and biological plausibility matter equally.
Common Pitfalls and Caveats
Relative risk is powerful but demands careful interpretation.
- Confounding bias distorts results — If exposed and control groups differ systematically in ways unrelated to the exposure, associations may be spurious. Age, sex, socioeconomic status, and comorbidities can all confound findings. Randomised trials minimise this; observational studies require statistical adjustment or matching.
- Rare outcomes inflate uncertainty — When disease counts are very small (fewer than 5 cases per cell), standard error estimation becomes unstable and confidence intervals widen dramatically. Consider combining rare categories or using alternative approaches like Fisher's exact test.
- Study design shapes interpretation — Cohort studies yield true relative risk directly. Case–control studies produce odds ratios that approximate RR only for rare outcomes. Cross-sectional studies cannot establish causality. Always match your question to the appropriate design.
- Publication and selection bias skew literature estimates — Studies showing large effects are more likely published; null results languish in file drawers. A meta-analysis of relative risks without accounting for this bias will overestimate true population effects. Funnel plots help detect such asymmetry.
Practical Applications in Research
Pharmaceutical companies use relative risk when comparing drug efficacy between treatment and placebo arms in randomised controlled trials. A cholesterol medication reducing myocardial infarction risk by 30% has an RR of 0.70. Environmental epidemiologists quantify occupational hazards: welders exposed to fumes might have an RR of 3.2 for respiratory disease relative to office workers.
Public health authorities rely on relative risk to set intervention priorities. If vaccination reduces hospitalisation by 80% (RR = 0.20), that high-impact prevention strategy justifies resource allocation. Clinicians counselling patients benefit from translating RR into absolute risk reduction, which speaks more directly to individual decision-making.