Relative risk vs absolute risk comes down to a ratio versus a gap. Relative figures compare groups, while absolute figures show how many people are actually affected. Check the baseline risk before judging any headline percentage.
Health stories often report only one of these numbers. This guide shows how to read both, and it supports the wider coverage on medical news today.
Absolute and relative risk defined
Absolute risk is the number of new cases that occur during a specified time period, and it is often written as a percent. A tutorial on risk measures in PubMed Central uses this definition. Another teaching source describes absolute risk, also called incidence, as the actual probability of an event occurring, as set out in a biostatistics lesson.
Relative risk is built from two absolute risks. It divides the risk of an event in one group by the risk in a second group. The result says how the groups compare, not how common the event is in either one. That gap is why a story can sound dramatic while the change for any single person stays small.
Quick reference for the formulas
| Measure | Formula | Planning use |
|---|---|---|
| Absolute risk | New cases in a specified period among the people at risk, often shown as a percent | Shows how likely the event is for a person in that group |
| Relative risk (RR) | Risk in one group divided by risk in a second group | Shows how groups compare; says nothing about how common the event is |
| Absolute risk reduction (ARR) | Risk in the control group minus risk in the treatment group | Gives the real-world size of a benefit |
| Number needed to treat (NNT) | 1 divided by the risk difference | Turns the gap into a count of people treated |
These formulas come from the risk-measures tutorial and from OpenAnesthesia’s explainer.
Worked example from event rates to NNT
The same pair of event rates produces every measure in the table. The PubMed Central tutorial gives an illustrative example of revision risk, with 15% in one group and 6% in the other. Follow the stages in order.
- Write down each group’s risk. The example compares 15% with 6%.
- Subtract for the absolute figure. The risk difference is the difference between the absolute risks, 9% in this example.
Only the second tells you how many people are affected.
The same logic with a reduction
OpenAnesthesia gives a nerve block example. If the block reduces the incidence of chronic postsurgical pain from 15% to 5%, the ARR is 10 percentage points and the NNT is 10. The ARR formula there is the risk in the control group minus the risk in the treatment group.
The same explainer shows how a relative reduction is read. In an anesthetic nausea example comparing 10% with 20%, the RR is 0.5, which indicates a 50% reduction in risk. This relative reduction is the number a headline would quote. Read it next to the two underlying percentages, because the percentages show how many people are involved. Details are in the OpenAnesthesia explainer.
Why relative figures dominate headlines
A large relative risk reduction can correspond to a tiny absolute change. That is the central warning in a biostatistics lesson on risk measures, which advises readers to be wary of interventions promoted with large relative risk reduction figures.
OpenAnesthesia makes a related point: in populations with low event rates, RR may overstate the clinical significance of findings. A relative figure hides the starting level of risk, so the same percentage can describe a common problem or a rare one. Headlines gain brevity by dropping that context, and readers lose the ability to judge size.
When a story quotes only a relative figure, treat it as incomplete. Look for the event rate in each group, then work out or find the absolute difference. If the article gives neither, go to the trial report or the press release’s data table.
What number needed to treat tells you
Number needed to treat converts the absolute difference into a count of people. The tutorial formula is 1/RD, and in its example the result is 11 patients treated to prevent a single revision. In the nerve block example the NNT is 10.
Use NNT as a reading of the absolute effect. A small NNT means a bigger absolute benefit, and a large NNT means that many people must be treated for one to benefit. It rests on the same inputs as ARR, so it inherits the same limits. ARR assumes the baseline risk does not change between groups, so an NNT from one population is not automatically right for a person whose starting risk is very different.
Baseline risk and the odds ratio trap
Baseline risk is the event rate in the control group, and it sets the ceiling on how much any treatment can change. A low baseline means a big relative cut still leaves little absolute benefit. Before accepting a claim, find that starting figure.
Baseline also decides whether a different statistic can be trusted. According to the PubMed Central tutorial, odds approach risk when the risk is less than 10%, and the odds ratio can estimate the risk ratio when the probability of an event is 10% or less. The same source cautions readers to beware how they interpret the odds ratio when the risk of an outcome is higher than 10%. If a story reports an odds ratio for a common outcome, do not read it as a plain risk comparison.
Study design matters too. OpenAnesthesia notes that RR is unsuitable for case-control studies. Observational findings also raise questions about cause, which is covered in correlation vs causation in medical studies.
Finding raw event counts in a trial report
Absolute risk is easiest to verify from the counts of events and participants in each arm. The tutorial gives an example: C5 palsy occurred in 15.7% of patients (16/102) within 1 week following laminoplasty for cervical myelopathy.
Risk is not the only way to express events. The tutorial also shows a rate, which divides events by person-time: 4/22.5 person-years, or 17.8 per 100 person-years in its example. A rate and a risk are different quantities, so do not compare one with the other.
The trial stage also shapes how much weight a result deserves, as explained in clinical trial phases explained.
A checklist for headline numbers
- Find the baseline. Look for the event rate in the control group, since a relative figure means little without it.
- Get both groups’ risks. With the two absolute risks you can rebuild RR, ARR and NNT yourself.
- Prefer the absolute difference. It states how many more or fewer people are affected, and a large relative reduction can sit on a tiny absolute change.
- Convert to NNT. Divide 1 by the risk difference to see how many people are treated per event avoided.
- Check the statistic type. An odds ratio approximates risk only when risk is about 10% or less.
- Match the population. ARR assumes baseline risk is unchanged, so the result may not carry over to people with a very different starting risk.
Frequently asked questions
What is the difference between relative and absolute risk?
Absolute risk is the chance of an event in a group over a period. Relative risk divides one group’s risk by another’s, so it shows comparison rather than size.
What does number needed to treat tell me?
It is 1 divided by the risk difference. In the tutorial example it means 11 patients treated to prevent a single revision.
Why do headlines favor relative figures?
Relative reductions can look large even when the absolute change is tiny, and they omit the starting risk. Teaching sources advise caution when a treatment is promoted using a large relative reduction alone.
What baseline risk should I check before judging a claim?
Check the control group’s event rate. If it is above 10%, be careful with odds ratios, because they may no longer approximate risk.
