Brief Definition: The Risk Ratio (RR), also known as relative risk, is a measure for comparing event probabilities between two groups, typically an intervention group and a control group. It is calculated as the quotient of the risks: risk in the intervention group divided by risk in the control group.
Interpretation of the Risk Ratio
An RR of 1 indicates no difference between the groups. Values less than 1 suggest a lower risk in the intervention group (potential benefit), while values greater than 1 suggest a higher risk (potential harm). The interpretation must always consider the endpoint under investigation, e.g., efficacy endpoints, safety endpoints, or combined endpoints.
It is important that the RR is a relative measure. A seemingly strong relative risk can mean a small absolute risk difference for rare events. Therefore, absolute measures such as Risk Difference or Number Needed to Treat are often reported in clinical reports as supplementary information.
For safety endpoints, it is also crucial whether the event is definition-driven (e.g., “serious adverse event”) and how completely events were recorded. Different reporting and follow-up processes, for example, in various countries or centers, can distort the RR, even if the therapeutic effects were identical.
Calculation and Data Basis
For a binary event: Risk = Number of events divided by the number of persons in the group. The RR is the quotient of these two risks. In randomized studies, this metric is often well interpretable because randomization reduces confounding and increases group comparability.
In observational studies, the RR can also be used but must be carefully evaluated for bias and confounding. Adjustments (e.g., regression, propensity scores), defined time windows, and sensitivity analyses are important here. The choice of the analysis population (intention-to-treat, per-protocol, or exposed-as-treated) also influences the result.
Another practical point is whether an event is counted only once per person or if recurrent events occur. For recurrent events, other models are often more suitable; the RR is then only of limited interpretability if it merely reflects the proportion of individuals with at least one event.
RR, Odds Ratio, and Hazard Ratio: Demarcation
The RR differs from the Odds Ratio (OR). The OR is based on odds (event vs. non-event) and can deviate significantly from RR for common events, which complicates communication. In case-control studies, the OR is often used because the risk itself cannot be directly determined.
The Hazard Ratio (HR) originates from time-to-event analyses (e.g., Kaplan-Meier and Cox models). It describes the ratio of hazard rates over time and is not identical to RR, even if both are often read as “relative effects.” With censoring and variable observation time, the HR is usually more appropriate than the RR.
For communication, it is helpful to translate the metric into understandable statements, e.g., “the event risk is reduced by 20%” for an RR of 0.8. However, it should always be supplemented with information on the baseline risk and the period over which the event was recorded.
Confidence Intervals, Significance, and Clinical Relevance
An RR should practically never be reported without a Confidence Interval (CI). A 95% CI that includes 1 typically argues against a statistically significant difference. Nevertheless, statistical significance is not equivalent to clinical relevance: a small but statistically significant effect may be clinically irrelevant, while a clinically relevant effect may remain statistically uncertain with a small sample size.
For regulatory dossiers, a clear, consistent presentation is important, including endpoint definition, handling of missing data, multiple testing, and subgroup analyses. A solid study protocol and a detailed Statistical Analysis Plan contribute to this. For safety signals in pharmacovigilance, additional contextual information (exposure, background incidences, underreporting) is required.
Practical Application in Clinical Development and Pharmacovigilance
In drug development, RR is frequently used for safety endpoints and for certain binary efficacy endpoints, such as responder rates or defined events (e.g., exacerbations). In pharmacovigilance and Post-Authorisation Safety Studies, it can be used to compare risks between exposed and unexposed groups, although interpretation heavily depends on data quality, confounding control, and defined exposure duration.
In EU contexts, EU Regulation 536/2014 is relevant for clinical trials, while quality and documentation requirements are framed, among others, by the principles of Good Clinical Practice according to ICH E6(R3). For medical devices, safety and performance endpoints under MDR 2017/745 may also require relative risk measures, e.g., when evaluating clinical data from clinical evaluation or Post-Market Clinical Follow-up.
For Clinical Study Reports, it is also important to consistently define RR (numerator/denominator), clarify the time basis, and design the presentation (tables, forest plots) so that readers can interpret the effect without misinterpretation. Overinterpretation should be avoided in subgroups; interaction tests and pre-defined hypotheses are crucial here.
FAQ
When is an RR preferable to an Odds Ratio?
If the risk can be directly determined (e.g., in randomized trials or cohort studies) and events are not extremely rare, the RR is often easier to understand than the Odds Ratio.
Can I infer the absolute risk reduction from an RR?
Only if the absolute risks are known. The RR alone contains no information about the baseline risk in the control group; for absolute risk reduction, both group risks are needed.
How do I correctly report RR in a study report?
Typically with a point estimate, 95% confidence interval, p-value (if applicable), and a clear definition of the endpoint, analysis population, and sensitivity analyses.
Regulatory References (Selection): EU Regulation (EU) No. 536/2014; Regulation (EU) 2017/745 (MDR); ICH E6(R3) Good Clinical Practice.