Relative rate of an event occurring in one group versus another over time
The hazard ratio compares the instantaneous rate of an event—such as redislocation or graft failure—between two groups at any point during follow-up, as estimated from a time-to-event model like Cox regression. An HR of 1 means the event rate is equal between groups; values above or below 1 indicate a higher or lower rate in the comparison group relative to the reference.
Unlike a simple proportion at a single time point, the HR uses the full follow-up period and accounts for varying observation times and censoring (patients lost to follow-up or not yet reaching the event by study end). This matters when appraising a study because it protects against bias from unequal follow-up length between groups, but only if the underlying proportional hazards assumption holds — if the hazard ratio actually changes over time, a single summary HR can be misleading.
It appears in survival analyses reported alongside Kaplan-Meier curves, typically from a Cox proportional hazards regression, with a 95% confidence interval and often a p-value. Appraisal should check whether the proportional hazards assumption was tested (e.g., Schoenfeld residuals or log-log plots), whether the model was adjusted for relevant confounders, and whether the number of events (not just sample size) was adequate to estimate the HR precisely.
With few events, the confidence interval around the HR can be wide even when the point estimate looks striking, so a single value like 'HR 2.5' should not be read as precise or clinically definitive without checking the interval. The HR is also a relative measure — it says nothing about absolute event rates or numbers needed to treat, and it can be misinterpreted as a risk ratio or applied over any time horizon when it strictly describes the instantaneous rate ratio over the observed follow-up.
This guide was auto-drafted and is pending editorial review.