On this page
A hazard ratio is a measure of relative risk that compares the instantaneous rate of an event between two groups in a clinical trial. A hazard ratio of 1 means equal risk; below 1 means lower risk in the treatment group; above 1 means higher risk. It quantifies how much faster or slower an event occurs at any given moment, not the total number of events that will occur.
- A hazard ratio compares instantaneous risk of an event between two groups at any moment in time, not cumulative risk over the entire study period.
- A hazard ratio of 0.7 does not mean 30% fewer total events will occur; it means the instantaneous risk is 30% lower at any given point.
- The confidence interval around a hazard ratio tells you the range of plausible values and whether the treatment effect is statistically distinguishable from no effect.
- A hazard ratio assumes the risk difference between groups remains constant over time, which may not hold true if one treatment's advantage grows or shrinks as follow-up continues.
- The same hazard ratio can represent vastly different real-world outcomes depending on baseline event rates, so hazard ratios must be paired with absolute risk or survival curves to interpret clinical meaning.
Clinical trials report survival and disease-free survival outcomes using a metric called the hazard ratio. It appears in nearly every oncology, cardiology, and infectious disease trial, often with a confidence interval attached. Readers encounter it constantly but frequently misinterpret what it means. A hazard ratio of 0.65 does not mean 35% fewer patients will experience the event. Understanding what a hazard ratio actually measures, and what it does not, is essential to reading trial results accurately and avoiding the most common misreadings that lead researchers and decision-makers astray.
What a hazard ratio actually is
A hazard ratio is a measure of relative risk that compares two treatments in the same study. It quantifies how much faster or slower an event occurs in one group compared to another. The key word is instantaneous: a hazard ratio describes the risk of an event at any given moment in time, not the total number of events that will accumulate over the full study period.
Think of it as a snapshot of risk. If you freeze time at any point during a trial, a hazard ratio tells you how the odds of experiencing the event compare between the treatment group and the control group at that exact moment. This is different from asking how many total events occurred by the end of the study, which is a cumulative measure. The hazard ratio answers a narrower, more specific question: at this instant, is one group experiencing the event faster than the other?
How hazard ratios are calculated
Hazard ratios emerge from survival analysis, a statistical method designed to handle data where not all participants experience the event by the end of the study and not all participants are followed for the same length of time. In a typical trial, some patients may drop out, move away, or simply not yet have experienced the outcome when the trial ends. Survival analysis accounts for this incomplete information.
The calculation compares the event rate between groups using methods that weight each observation by how long that patient was at risk. Suppose a treatment group had 15 events among 100 patients followed for an average of 2 years, while a control group had 20 events among 100 patients followed for the same period. The hazard ratio would reflect not just the count difference but the rate at which events occurred in each group. The exact mathematical approach depends on the survival model used, but the principle is consistent: divide the hazard (instantaneous event rate) in one group by the hazard in the other group to get the ratio.
Reading a hazard ratio: a worked example
Suppose a cancer trial compares a new chemotherapy to standard treatment, measuring time to death.
Now suppose the control group's death rate is 10 events per 100 patient-years of follow-up. That is what the hazard ratio describes: the relative speed at which events occur in each group, measured per unit of time at risk.
But here is what it does not tell you: the actual number of deaths depends on how long patients are followed, how many start in each group, and whether the hazard ratio stays constant over time. Two trials with identical hazard ratios can have very different numbers of total events if one follows patients longer or starts with more participants.
Why hazard ratios matter in clinical trials
Hazard ratios are the standard way to report survival outcomes because they handle the messy reality of clinical data. Patients enter trials at different times. Some are lost to follow-up. Others withdraw. Some never experience the event before the trial ends. Simply counting total events and dividing by group size would ignore all this complexity and produce a misleading comparison.
Survival analysis methods, which produce hazard ratios, solve this problem by accounting for how long each patient was actually at risk. A patient followed for 5 years contributes more information than a patient followed for 6 months. A hazard ratio incorporates this weighting automatically. It also allows researchers to report a single summary statistic that captures the treatment effect across the entire follow-up period, even if that period is different for different patients. This makes hazard ratios far more useful than simpler measures for comparing treatments when follow-up times vary.
What a hazard ratio does not tell you
The most common misinterpretation of hazard ratios stems from confusing instantaneous risk with cumulative risk. This section clarifies what hazard ratios do and do not convey.
Hazard ratio versus absolute risk reduction
The absolute number of events prevented depends entirely on how many events would have occurred in the control group to begin with.
Suppose a trial follows two groups of 1,000 patients each for 5 years. The absolute risk reduction depends on the baseline event rate in the control group and the hazard ratio between groups.
Now suppose a different trial has a much lower baseline event rate. The absolute risk reduction would be correspondingly smaller. Same hazard ratio, vastly different clinical impact. This is why hazard ratios must always be paired with information about baseline risk to understand real-world meaning.
Proportional hazards assumption
Hazard ratios rest on a hidden assumption: that the risk difference between groups stays constant over time. In technical terms, this is called the proportional hazards assumption. If the treatment effect is strong early but fades, or weak early but grows, the proportional hazards assumption is violated.
When this assumption holds, a single hazard ratio accurately summarizes the treatment effect across the entire follow-up period. When it does not hold, the hazard ratio becomes misleading. It may overstate or understate the true benefit depending on when in the trial the effect is strongest. Researchers can test this assumption using statistical methods, and careful readers of trial results should look for mention of this check. If the hazard ratio is reported without discussion of proportional hazards, the assumption was likely tested and held; if it was violated, that violation should be disclosed.
Confidence intervals and statistical significance
A hazard ratio is a point estimate: a single number calculated from the trial data. But that number has uncertainty around it. The confidence interval quantifies that uncertainty. Most clinical trials report a 95% confidence interval, written as 95% CI.
Suppose a trial reports a hazard ratio of 0.80 with a 95% CI of 0.65 to 0.98. This means the researchers are 95% confident that the true hazard ratio lies somewhere in that range. A narrow confidence interval suggests more precision and less uncertainty than a wide one. A wide interval means the data are consistent with a range of possible treatment effects.
The confidence interval also reveals statistical significance. If the confidence interval includes the value 1, then the treatment effect is not statistically significantly different from no effect. A hazard ratio of 1 implies that the event is equally probable in both groups. For example, a hazard ratio of 0.95 with a 95% CI of 0.70 to 1.30 includes 1, so despite the point estimate suggesting a small benefit, the data do not rule out no effect or even harm. Conversely, a hazard ratio of 0.80 with a 95% CI of 0.65 to 0.98 excludes 1, indicating a statistically significant difference between groups.
How hazard ratios connect to other trial outcomes
Hazard ratios do not exist in isolation. They are part of a constellation of metrics used to report survival data. Understanding how they relate to adjacent concepts helps readers interpret trial results more completely.
Median survival time is the time at which half the patients in a group have experienced the event. If a trial reports that the control group had a median survival of 12 months and the treatment group had a median survival of 18 months, that is a concrete, easy-to-understand outcome. The hazard ratio connects to median survival: a lower hazard ratio typically corresponds to a longer median survival time, but the relationship is not perfectly linear.
Kaplan-Meier curves are plots that show the proportion of patients still event-free over time in each group. These curves visualize what the hazard ratio summarizes numerically. A curve that drops more steeply indicates a higher hazard; a flatter curve indicates a lower hazard. The hazard ratio is the ratio of the slopes, in a sense, though the actual calculation is more sophisticated.
Number needed to treat (NNT) is the number of patients who must receive the treatment for one additional patient to avoid the event compared to control. NNT is calculated from absolute risk reduction, not directly from the hazard ratio. But readers often want to know NNT because it answers the practical question: how many patients do I need to treat to prevent one bad outcome? To calculate NNT, you need both the hazard ratio and the baseline event rate in the control group. This is why hazard ratios alone are insufficient for clinical decision-making; they must be combined with absolute risk information.
Do this automatically
Calculating hazard ratios from raw trial data requires statisticians to manually extract event times, censor observations, fit survival models, and interpret confidence intervals, a process that compounds in complexity across multiple studies and treatment comparisons. Software can automate the extraction of participant timelines, the computation of instantaneous risk ratios, and the generation of hazard ratio estimates with their associated uncertainty bounds, reducing the arithmetic burden and enabling faster analysis of clinical outcomes. Healthcare research runs this kind of analysis end to end and cites every figure back to its original source, across as many subjects as you need, with Mimir attaching a citation to every figure. It will not predict a trial's outcome or advise on anyone's treatment. Those decisions stay with clinicians and regulators, and deliberately so. Every plan includes it, including the free one.
Start your first report and run it on something you already know well.
This article is general information about healthcare and life-sciences topics. It is not medical advice, diagnosis, or treatment guidance, and it is not a substitute for a qualified professional. Any treatment decision should be discussed with a clinician.
