Skip to main content
v2026.11,986 entries · CC-BY 4.0

How to Interpret a Hazard Ratio (and Why It Is Not a Risk Ratio)

A hazard ratio from a Cox proportional-hazards model is an instantaneous rate ratio, not a risk ratio. This guide walks through a worked example, a plain-language lookup table, and the proportional-hazards assumption check that determines whether a single HR is a valid summary at all.

Written and maintained by CASRAI Editorial Board

Published · Last updated

On this page

A hazard ratio (HR) is one of the most consistently misread numbers in the clinical and epidemiological literature. It gets reported alongside a 95% confidence interval from a Cox proportional-hazards model, and then gets talked about as if it were a risk ratio — “patients on the new drug were 30% more likely to die.” That sentence is usually wrong, or at least imprecise in a way that changes what a reader takes away. A hazard ratio is a ratio of instantaneous event rates, summarized across the follow-up period under a proportionality assumption — it is not a ratio of the probability that an event ever happens to a given patient. This guide works through what a reported HR is actually telling you, gives a lookup table for translating a value into a plain-language statement, and covers the assumption check that determines whether a single HR is even a valid summary of your data in the first place.

What Does a Hazard Ratio Mean?

A hazard ratio (HR) compares how fast an event, such as death or disease progression, is happening in one group versus another at any given moment of follow-up. HR = 1 means no difference, HR < 1 means a lower event rate in the first group, and HR > 1 a higher one. Read it together with its 95% confidence interval: if the interval includes 1, the result is not statistically significant.

A Worked Example First

Suppose a randomized trial compares a new maintenance therapy against standard care for time to disease progression, and the Cox proportional-hazards model produces this line in the results table:

Comparison Hazard Ratio 95% CI p-value
New therapy vs. standard care 0.68 0.51 – 0.90 0.007

The sentence a statistician would write for the results section: “The new therapy was associated with a 32% reduction in the instantaneous hazard of disease progression relative to standard care (HR 0.68, 95% CI 0.51–0.90, p = 0.007), consistent with the proportional-hazards assumption over the observed follow-up period.” Notice what is present and what is deliberately absent. The HR describes a rate reduction, not “32% of patients were spared progression,” and not a claim about how much longer any individual patient will remain progression-free — that is a separate question answered by median survival times or restricted mean survival time, not by the HR itself. The confidence interval excluding 1.0 is what licenses calling the reduction statistically significant at the conventional alpha = 0.05 threshold; the point estimate of 0.68 alone does not.

What a Hazard Ratio Actually Measures

The hazard function, h(t), is the instantaneous rate at which an event occurs at time t, given that a subject has survived event-free up to that point. A Cox proportional-hazards model assumes the hazard in one group is a constant multiple of the hazard in a comparison group at every point in time — that constant multiple is the hazard ratio. An HR of 1.0 means the two groups have the same instantaneous event rate throughout follow-up. An HR of 2.0 means one group’s instantaneous rate is twice the other’s at any given moment, not that twice as many people in that group eventually experienced the event. For a rate ratio built from person-time rather than a continuously updated hazard, see incidence rate ratio.

Because the HR is estimated from the full follow-up period and from everyone still at risk at each event time, it is a summary measure, not a snapshot at one moment. That is precisely why the proportional-hazards assumption (the ratio staying roughly constant over time) matters so much: if it does not hold, a single reported HR is averaging over a relationship that is actually changing shape, and can misrepresent both the early and late parts of the follow-up.

Why an HR Is Not a Risk Ratio (or an Odds Ratio)

A risk ratio (relative risk) compares the cumulative probability of an event by a fixed point in time between two groups — a simple proportion. A hazard ratio compares instantaneous rates among those still event-free, re-evaluated continuously across follow-up. When an event is rare and follow-up is short, the two converge and an HR can be read loosely as an approximate risk ratio. As cumulative incidence rises, they diverge, and the HR is typically the more extreme number of the two (further from 1 than the corresponding risk ratio) — not because the treatment effect is more extreme, but because of how the two statistics are constructed. For the related distinction between odds and risk, see odds ratio vs. risk ratio.

Two statistical properties are involved, and they are separate. Hazard ratios are non-collapsible, meaning an adjusted HR can differ from the crude one even without confounding. And period-specific hazard ratios carry a built-in selection bias from the depletion of susceptible individuals: as higher-risk members of one group experience the event and leave the risk set, the people still at risk in that group are on average less susceptible. The second point is the subject of Miguel Hernán’s note “The Hazards of Hazard Ratios” (Epidemiology, 2010), which states that endowing an HR with a causal interpretation is risky because the HR may change over time and has a built-in selection bias. It is worth reading directly if the HR in your own analysis is being used to support a causal claim rather than reported as a descriptive association. The practical takeaway for interpretation: never convert a reported HR into a “times more likely” or “% more likely” statement about outcomes without stating explicitly that this refers to the instantaneous hazard, not cumulative risk.

A Hazard Ratio Is Not a Time Effect

A hazard ratio of 2 is often read as “twice as fast” or “half the time.” Spruance and colleagues, writing in Antimicrobial Agents and Chemotherapy, show why that reading fails: the HR is a unitless ratio of rates, and the benefit in days or months also depends on the shape of the underlying time-to-event distribution. Their comparison of the hazard ratio with the median ratio (placebo median time divided by treatment median time) in three antiviral trials shows that the two statistics can move in either direction relative to each other.

  • Valaciclovir 500 mg for recurrent genital herpes episodes: HR 1.9, but median ratio 1.5 (5.9 days versus 4.0 days). The HR looks larger than the time benefit.
  • Famciclovir for time to resolution of postherpetic neuralgia: HR 1.9 and median ratio 2.0. The two roughly agree.
  • Acyclovir for the duration of herpes zoster pain: HR 1.5, but median ratio 3.1 (62 days versus 20 days). The HR understates the time benefit.

The data table below the guide lists these values beside the period-specific figures discussed in the next section. The practical rule: when every participant will eventually experience the outcome (healing, pain resolution), report median or percentile times alongside the HR, because the HR alone cannot tell a reader how many days or weeks were saved.

Reading an HR as the Probability of Healing First

Spruance et al. give one correct plain-language reading of an HR for a time-to-event outcome: the HR equals the odds that a treated patient experiences the event before a comparable control patient. With P as the probability of the treated patient being first, HR = P / (1 − P), so P = HR / (1 + HR). An HR of 2 therefore corresponds to a 67% chance of the treated patient’s event occurring first, and an HR of 3 to 75%. For an HR of 1.9, P = 1.9 / 2.9 = 0.66 to two decimals (the paper rounds the same case to 0.67). Because natural variation between patients is large, even a clearly significant HR does not guarantee that any one treated patient fares better than any one control patient. This is a statement about ordering of events, not about how much sooner they happen.

Hazard ratio P = HR / (1 + HR) Reading
1.0 0.50 Event equally likely to occur first in either group
1.5 0.60 60% chance the event occurs first in the numerator group
2.0 0.67 About two chances in three
3.0 0.75 Three chances in four

When the HR Changes With Follow-Up: the Women’s Health Initiative Example

The proportional-hazards assumption is not a technicality. Hernán (Epidemiology, 2010) uses the Women’s Health Initiative randomized trial of combined estrogen-plus-progestin hormone therapy, which followed more than 16,000 women for an average of 5.2 years and compared the risk of coronary heart disease. The headline result was a hazard ratio of 1.24. The period-specific HRs in the same trial were 1.81, 1.34, 1.27, 1.25, 1.45 and 0.70 for years 1, 2, 3, 4, 5 and 6 or more. A single value of 1.24 therefore averages over a ratio that was well above 1 early and below 1 after year 5.

Hernán makes two points that matter for interpretation. First, the conclusion can depend on how long the study ran: the average HR would have been about 1.8 had the study stopped after 1 year, 1.7 after 2 years and 1.2 after 5 years. Second, period-specific HRs carry a built-in selection bias. The hazard in year t is computed only among people who reached year t free of the outcome, and if susceptible people are depleted faster in the higher-risk arm, the remaining people in that arm are on average less susceptible. This depletion of susceptibles can produce an HR below 1.0 after year 5 even if the treatment has no truly preventive effect in anyone. This is why a late-period HR should not be read as evidence that a harmful exposure becomes protective.

Show Absolute Numbers Beside the HR

Hernán’s recommended remedy is to report survival (or cumulative-incidence) curves, which show absolute risk, alongside or instead of a single HR. His WHI illustration: the average HR of about 1.8 in year 1 corresponds to a one-year risk of about 0.49% in the treatment group and 0.28% in the placebo group, not 49% versus 28%. The relative figure sounds large while the absolute difference is small. The same reasoning is why the guidance above says to pair a large HR on a rare outcome with absolute event counts. Cumulative measures such as survival curves or a series of average HRs summarize follow-up more meaningfully than a single averaged number. Hernán notes that hazards can cross during follow-up because of depletion of susceptibles even when the survival curves never cross.

Report What it answers What it does not answer
Hazard ratio with 95% CI How the instantaneous event rate compares between groups, under proportional hazards How many people are affected, or how much sooner events occur
Median or percentile time ratio How much sooner the typical participant reaches the event Behaviour of the tails; undefined if fewer than half the group has an event
Survival or cumulative-incidence curves Absolute risk at each time point A single summary number for quick comparison

Translating an HR Into a Plain-Language Statement

The following is a reading guide, not a set of fixed cutoffs — unlike some effect-size measures, there is no widely adopted convention (comparable to, say, Cohen’s d thresholds) for calling a hazard ratio “small” or “large.” What counts as a meaningful HR depends entirely on the outcome, the baseline event rate, and the clinical or research context.

Hazard Ratio Plain-language reading
1.00 No difference in instantaneous event rate between groups.
0.90 – 0.99 Slightly lower hazard in the comparison group; check whether the CI excludes 1.0 before treating this as a real effect.
0.50 – 0.89 Meaningfully lower instantaneous hazard (roughly an 11–50% relative reduction in event rate at any given moment).
< 0.50 Hazard reduced by more than half at any given moment — a large effect if the CI is precise and the proportional-hazards assumption holds.
1.01 – 1.10 Slightly higher hazard; again, check the CI before treating this as a real effect.
1.10 – 2.00 Meaningfully higher instantaneous hazard (10–100% relative increase in event rate at any given moment).
> 2.00 Hazard more than doubled at any given moment — report alongside absolute event counts, since a large HR on a rare outcome can still mean a small absolute difference.

Every row in that table is a statement about instantaneous rate, and every row should be paired with the confidence interval before it is described as an effect at all, not just a point estimate.

Reading the Confidence Interval Together With the HR

A hazard ratio without its confidence interval is close to unreportable. The 95% CI reflects the precision of the estimate: a narrow interval close to the point estimate (e.g., HR 0.68, 95% CI 0.60–0.77) supports a confident reading; a wide interval (e.g., HR 0.68, 95% CI 0.32–1.45) means the data are consistent with anything from a large protective effect to a modestly harmful one, and the point estimate alone overstates what is actually known. If the interval crosses 1.0, the result is not statistically significant at conventional thresholds regardless of how far the point estimate itself sits from 1.0 — see CASRAI’s guide to confidence intervals for how the interval is constructed and what it does and does not claim about repeated sampling.

The Assumption That Makes a Single HR Valid: Proportional Hazards

Reporting one HR for the entire follow-up period is only defensible if the hazard ratio is genuinely stable over time — the proportional-hazards assumption. Before reporting a Cox model’s HR as a single summary number, check it:

  • Kaplan–Meier curves for the groups should not cross. Crossing curves are a visual red flag that the direction of effect itself may be reversing over time, which a single HR cannot represent.
  • Log(−log(survival)) plots against log(time) should be roughly parallel across groups; converging, diverging, or crossing lines indicate non-proportionality.
  • Schoenfeld residuals, tested formally (e.g., via cox.zph() in R), should show no significant trend over time for a covariate. A significant test result is direct statistical evidence against the proportionality assumption for that covariate.

When the assumption fails, options include reporting time-varying hazard ratios (splitting follow-up into intervals with a separate HR per interval), adding a time interaction term to the model, or switching to a method that does not require the assumption (e.g., restricted mean survival time, which summarizes an absolute time difference rather than a rate ratio). Reporting a single HR without checking this assumption — and without saying whether it was checked — is one of the most common and consequential gaps in published survival analyses.

Other Conditions Worth Checking

  • Non-informative censoring. The Cox model assumes that being censored (lost to follow-up, study ends) is unrelated to a subject’s underlying risk of the event. If sicker patients are disproportionately lost to follow-up, the HR can be biased in either direction.
  • Enough events, not just enough subjects. Precision in a Cox model is driven by the number of events observed, not the number of subjects enrolled. A trial with many enrolled patients but few events will still produce a wide, imprecise confidence interval around the HR.
  • Competing risks. If subjects can experience a different, competing event that removes them from being able to experience the event of interest (e.g., death from another cause before disease progression), a standard Cox HR can be inflated or biased; a subdistribution hazard model (Fine–Gray) or cause-specific hazard framing may be more appropriate, and the choice should be stated explicitly.
  • Multiple covariates. An adjusted HR from a multivariable Cox model answers a different question than an unadjusted (crude) HR — state which one is being reported and what was adjusted for. See CASRAI’s confounding variable entry for what adjustment is and is not able to control for.

Adjusted vs. Crude Hazard Ratio

A crude (unadjusted) hazard ratio comes from a model containing only the exposure or treatment of interest. An adjusted hazard ratio (aHR, sometimes written “adjusted HR”) comes from a multivariable Cox model that also includes covariates such as age, sex or disease stage, so the HR describes the exposure effect while those covariates are held at the same values. In a randomized trial the crude and adjusted values are usually close, because randomization already balances the groups on average. In an observational study the gap between them shows how much of the crude association was due to measured confounding.

Version What is in the model How to word it
Crude HR Exposure only “Unadjusted hazard ratio for X versus Y was 1.80 (95% CI 1.40–2.31).”
Adjusted HR Exposure plus named covariates “After adjustment for age, sex and stage, the hazard ratio was 1.35 (95% CI 1.04–1.75).”

Always list the covariates, because an adjusted HR is only as informative as the adjustment set. Adjustment cannot remove bias from confounders that were not measured. Also, an adjusted HR can differ from the crude one even without confounding, because hazard ratios are non-collapsible.

Hazard Rate vs. Hazard Ratio

The two terms are easy to confuse. A hazard rate is the instantaneous event rate in one group at a point in time (the h(t) defined above), and its unit is events per unit of time. A hazard ratio divides one group’s hazard rate by another’s, so it has no unit. Hernán defines the HR as “the hazard in the exposed groups divided by the hazard in the unexposed groups” and notes that hazards can roughly be read as incidence rates, so the HR approximates an incidence rate ratio (Hernán, Epidemiology 2010). The same article explains why one averaged HR can mislead when the true ratio changes over follow-up, and argues that adjusted survival curves are more informative than HRs because they show absolute risks. For choosing between Kaplan-Meier, Cox and parametric approaches, see choosing between Kaplan-Meier, Cox and parametric models; when proportional hazards fails, restricted mean survival time gives an absolute alternative.

Frequently Asked Questions

Is a hazard ratio of 1.5 the same as “50% more likely to die”?

No. It means the instantaneous rate of the event at any given moment during follow-up is 1.5 times higher in one group than the other, under the proportional-hazards assumption. It is not a statement that 50% more people in that group experienced the event by the end of the study — that is a cumulative-risk claim, and a risk ratio (not a hazard ratio) is the correct statistic for it. To express that cumulative difference in absolute terms, see absolute risk reduction and number needed to treat.

Can a hazard ratio be negative?

No. A hazard ratio is a ratio of two non-negative rates, so it is always greater than or equal to zero. Values below 1.0 indicate a lower hazard in the numerator group; values above 1.0 indicate a higher hazard; a value of exactly 1.0 indicates no difference.

What sample size or number of events does a Cox model need?

There is no single universal minimum, but the relevant quantity is the number of events, not enrolled subjects (see sample size for survival analysis) — a widely used rule of thumb in applied survival analysis is roughly 10 events per covariate included in the model, though this is a heuristic rather than a formal requirement, and simulation-based power calculations are preferable when feasible.

Why do published hazard ratios sometimes look more extreme than the corresponding risk ratio in the same paper?

A hazard ratio and a risk ratio answer different questions: the HR compares instantaneous rates among people still event-free, while the risk ratio compares cumulative proportions. When events are common the two diverge, and the HR is typically the one further from 1.0, as described above. The difference reflects how the statistics are constructed, not a modelling error.

What is the difference between a hazard ratio and a median survival ratio?

The HR compares event rates; the median ratio divides the control group’s median time to event by the treatment group’s. Spruance et al. show the two can disagree widely: an HR of 1.5 alongside a median ratio of 3.1 for acyclovir in herpes zoster pain, and an HR of 1.9 alongside a median ratio of 1.5 for valaciclovir in genital herpes episodes. Report both when the outcome is one that everyone eventually experiences.

Does a hazard ratio of 2 mean the treatment works twice as fast?

No. It means the odds that a treated patient has the event before a comparable control patient are 2 to 1, a probability of about 67% (P = HR / (1 + HR)). It says nothing about how many days or months sooner the event occurs.

Can a hazard ratio change over time?

Yes. In the Women’s Health Initiative combined hormone therapy trial the reported period-specific hazard ratios for coronary heart disease ran from 1.81 in year 1 to 0.70 from year 6 onward, against an overall HR of 1.24 (Hernán 2010). When the ratio changes, report period-specific or time-varying HRs, survival curves, or restricted mean survival time rather than one averaged number.

Is an HR of 0.7 a 30% reduction in the risk of death?

It is a 30% lower instantaneous rate of death at any given moment, under proportional hazards. Whether cumulative risk is 30% lower depends on the baseline risk and follow-up length, so state “30% lower hazard” and give the absolute risks from the survival curves.

Sources

Hazard ratios are one output among several from a broader family of inferential statistics used to summarize and report research results. For adjacent concepts:

Data: Published hazard ratios beside time-scale and follow-up context: antiviral trials and the Women's Health Initiative

Compiled from two peer-reviewed methodology papers: Spruance et al. 2004 (hazard ratio versus median ratio in three antiviral trials) and Hernán 2010 (period-specific and overall hazard ratios for coronary heart disease in the Women's Health Initiative combined hormone therapy trial). Shows that an HR can differ from time-based effects and from itself across follow-up windows.

Period covered: 2004/2010.

Published hazard ratios beside time-scale and follow-up context: antiviral trials and the Women's Health Initiative
ExampleWindow or settingHazard ratioMedian ratio (placebo/treatment)Source
Valaciclovir 500 mg, recurrent genital herpes episodesWhole trial1.91.5Spruance 2004
Famciclovir, postherpetic neuralgia resolutionWhole trial1.92Spruance 2004
Acyclovir, herpes zoster pain durationWhole trial1.53.1Spruance 2004
WHI combined hormone therapy, coronary heart diseaseYear 11.81not reportedHernán 2010
WHI combined hormone therapy, coronary heart diseaseYear 21.34not reportedHernán 2010
WHI combined hormone therapy, coronary heart diseaseYear 31.27not reportedHernán 2010
WHI combined hormone therapy, coronary heart diseaseYear 41.25not reportedHernán 2010
WHI combined hormone therapy, coronary heart diseaseYear 51.45not reportedHernán 2010
WHI combined hormone therapy, coronary heart diseaseYear 6 or later0.7not reportedHernán 2010
WHI combined hormone therapy, coronary heart diseaseOverall (mean follow-up 5.2 years)1.24not reportedHernán 2010

Sources

Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about How to Interpret a Hazard Ratio (and Why It Is Not a Risk Ratio)

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →