convert hazard ratio to odds ratio

Convert Hazard Ratio to Odds Ratio: Simple Formula and Step-by-Step Guide

By Hannah Jones
AD

Converting a hazard ratio (HR) to an odds ratio (OR) is a common request in evidence synthesis, clinical decision modeling, and observational research. The two measures appear similar because both compare the likelihood of an event between groups, but they answer different questions. An HR comes from time-to-event analysis and describes the instantaneous relative hazard over follow-up, while an OR comes from logistic regression and describes the relative odds of a binary outcome at a fixed time point.

There is no universal, assumption-free formula that turns every HR into an OR. The conversion depends on the baseline event risk in the control group, the length and completeness of follow-up, and whether the proportional hazards assumption holds. This article explains when a conversion is defensible, how to apply the most widely used exponential survival formula, and which pitfalls can make an HR-to-OR conversion misleading.

Understanding Hazard Ratios and Odds Ratios

A hazard ratio is estimated from survival models such as the Cox proportional hazards model. It represents the ratio of the hazard rate in one group to the hazard rate in another group at any given time, assuming that this ratio stays constant over follow-up. Because hazard is an instantaneous risk, an HR does not directly translate into a simple probability of experiencing the event by a certain date.

AD
a diagram showing the steps to get rids from other people and how they can help them
a diagram showing the steps to get rids from other people and how they can help them
Odds and Odds Ratio
Odds and Odds Ratio
Calculating Number Needed to Treat/Harm (NNT/H) with Odds Ratio | R-bloggers
Calculating Number Needed to Treat/Harm (NNT/H) with Odds Ratio | R-bloggers

An odds ratio is estimated from logistic regression or a contingency table for a binary outcome. It compares the odds of the event in one group with the odds in another group, where odds equal the probability of the event divided by the probability of no event. Unlike an HR, an OR is tied to the cumulative event probability observed over a specific period, which means it changes if the follow-up window changes.

Why HR and OR Are Not Interchangeable

The fundamental difference is that HR operates on the hazard scale, while OR operates on the odds scale. A constant HR can produce very different ORs depending on how common the outcome is in the control group. If the event is rare, the OR and HR will often be numerically close, but when the event is common, the OR can be substantially farther from 1 than the HR.

When can odds ratios mislead?
When can odds ratios mislead?
the words hazard in front of an image
the words hazard in front of an image
How to Calculate Odds Ratio.
How to Calculate Odds Ratio.

For example, an HR of 2.0 might correspond to an OR near 2.05 when the control event risk is 5%, but the same HR can correspond to an OR of 3.0 when the control event risk is 50%. This occurs because the odds ratio compounds the baseline risk, while the hazard ratio does not. Any conversion therefore requires an explicit baseline risk and a defined follow-up period.

When Conversion Is Appropriate

Conversion from HR to OR is most defensible when several conditions are met: the proportional hazards assumption is reasonable, survival times follow an exponential distribution, the control event risk is available for the same follow-up period, there are no major competing risks, and censoring is not strongly informative. Under these conditions, the HR can be mapped to a cumulative event probability and then to an OR.

AD

Conversion is not appropriate when hazards are clearly non-proportional, when the HR changes meaningfully over time, when the baseline risk comes from a different population or a different follow-up length, or when competing events are common. In those situations, the conversion becomes a rough modeling exercise rather than a statistical transformation, and the resulting OR may not reflect the actual data.

printable worksheet for rounding numbers
printable worksheet for rounding numbers
Risk Management And Best Risk Reward Ratio For Trading
Risk Management And Best Risk Reward Ratio For Trading
the graph shows that there are two different types of rato explained
the graph shows that there are two different types of rato explained
Example 8.29: Risk ratios and odds ratios | R-bloggers
Example 8.29: Risk ratios and odds ratios | R-bloggers
The Concepts Behind Logistic Regression | Towards Data Science
The Concepts Behind Logistic Regression | Towards Data Science
Ratio Worksheets
Ratio Worksheets

How to Convert a Hazard Ratio to an Odds Ratio

The standard approach begins by using the HR to estimate the treatment-group event probability from the control-group event probability. If p0 is the control-group event risk over the study period, and the HR is assumed constant, then under an exponential survival model the treatment-group event risk p1 can be written as p1 = 1 - (1 - p0)^HR. Once p0 and p1 are known, the odds ratio follows directly from the odds in each group.

The resulting formula is OR = [p1 / (1 - p1)] / [p0 / (1 - p0)], which simplifies to OR = [1 - (1 - p0)^HR] × (1 - p0)^(1 - HR) / p0. This equation is exact only under the exponential survival and proportional hazards assumptions. It should be treated as an approximation when those assumptions are uncertain, which is common in real-world meta-analyses.

AD

The Exponential Survival Formula in Practice

To apply the formula, first confirm that the HR and baseline risk refer to the same follow-up period and the same outcome definition. Then compute the treatment-group risk using p1 = 1 - (1 - p0)^HR. Next, calculate the odds in the control group as p0 / (1 - p0) and the odds in the treatment group as p1 / (1 - p1). Finally, divide the treatment odds by the control odds to obtain the OR. Confidence intervals can be approximated by converting the lower and upper bounds of the HR separately, because the formula is monotonic.

a balance scale with the words, reward and reward on top of it in front of a
a balance scale with the words, reward and reward on top of it in front of a
printable worksheet for rounding numbers
printable worksheet for rounding numbers
Ratio Riesgo-Beneficio
Ratio Riesgo-Beneficio
Welcome to the world of ratios 📊
Welcome to the world of ratios 📊

As a worked example, suppose a trial reports HR = 0.75 for a treatment versus control, and the control-group event risk over five years is 20%. Then p1 = 1 - (1 - 0.20)^0.75 = 1 - 0.8^0.75 ≈ 0.154. The OR is (0.154 / 0.846) / (0.20 / 0.80) ≈ 0.182 / 0.25 = 0.728. So the HR of 0.75 converts to an OR of approximately 0.73 under the exponential model. A second example with HR = 2.0 and p0 = 10% gives p1 = 1 - 0.9^2 = 0.19 and an OR of approximately 2.11.

Approximations, Limitations, and Best Practices

A useful rule of thumb is that when the outcome is rare, with a control event risk below about 10%, the OR is often close to the HR. In that setting, many analysts treat the two as approximately interchangeable for meta-analysis. However, as the event becomes more common, the OR moves farther from 1 than the HR, and ignoring the baseline risk can substantially distort the pooled result. Avoid the shortcut of treating an HR as if it were a risk ratio, because that leads to a different and often inappropriate formula.

AD

Major limitations include non-proportional hazards, informative censoring, competing risks, and baseline risks that do not match the population of interest. Censoring can also make the observed event risk lower than the true cumulative risk, which biases the conversion. Best practices are to report the conversion formula, state the source of the baseline risk, convert confidence limits on the same scale, and run sensitivity analyses using plausible ranges of p0. Whenever possible, analyze time-to-event data directly or pool HRs with HRs rather than forcing all studies onto an OR scale.

A careful HR-to-OR conversion can be useful when a common effect measure is required for a specific analysis, but it cannot erase the underlying differences between time-to-event and binary outcomes. Use the exponential formula as a transparent tool, test how sensitive your conclusions are to realistic baseline risks, and let the clinical question decide whether an OR is truly the right target. When the assumptions are shaky, the most honest approach may be to present the HR and OR side by side rather than pretending they are interchangeable.

AD