Hazard Ratio

What it means

A hazard ratio (HR) is the ratio of hazards (instantaneous event rates) between two groups. From Cox proportional-hazards regression:

• HR = 1.0: no difference.
• HR > 1.0: more events in the comparison group (e.g. HR 1.5 = 50% higher hazard).
• HR < 1.0: fewer events (e.g. HR 0.7 = 30% lower hazard).

What it doesn’t tell you

• The absolute event rate in either group.
• The timing of when the effect manifests.
• Whether the proportional-hazards assumption holds (effect constant over time).

Worked example

A trial reports HR 0.75 (95% CI 0.65–0.86) for stroke with the treatment. That means:

• The treatment group’s instantaneous stroke rate is 75% of the control group’s.
• The 95% confidence interval excludes 1.0 (so it’s statistically significant).
• But you still need the absolute event rate to know how many people must be treated to prevent one stroke.

Common pitfalls

• "Risk halved" sounds dramatic; if baseline risk is 1%, halving it takes the absolute risk to 0.5% — small in absolute terms.
• HRs from very small subgroups have wide CIs; treat with caution.
• HR ≠ odds ratio ≠ relative risk in all situations; in long-follow-up trials they diverge.

Related entries

NNT, ARR, RRR, Mendelian randomization, Bradford-Hill criteria.