Relative and absolute risk in a trial: Why both numbers change the meaning of a headline

By MedClinRes.org Editorial Team. This article has not been peer-reviewed or clinically reviewed, and it is general education, not medical advice.

Short answer: a relative risk compares two groups as a ratio, while an absolute risk is the actual chance of an outcome in one group over a stated period. One trial result can be described as a big relative change or a small absolute one, and both descriptions can be accurate. To judge what a headline means, a reader needs both numbers, the time frame, and the interval around each. The examples below use invented numbers to show how far apart the two can sit.

What each number measures

In a randomized trial, treatments are assigned to participants by chance rather than by choice. The National Institutes of Health explains that this is done to avoid bias in who gets which treatment. Researchers then count how many people in each group had the outcome during follow-up. Those counts produce four numbers:

Absolute risk: the number of people with the outcome divided by the number followed in that group. It only makes sense with the time period attached. Twenty events among 1,000 people followed for one year is a one-year risk of 2.0%.
Relative risk (risk ratio): the risk in the intervention group divided by the risk in the comparison group. A value of 1.0 means the groups had the same risk. A value of 0.8 means the intervention group's risk was 20% lower. That 20% is the relative risk reduction, which is simply 1 minus the relative risk.
Absolute risk reduction (risk difference): the comparison group's risk minus the intervention group's risk. When risks are percentages, the answer is in percentage points, the plain difference between two percentages.
Number needed to treat (NNT): 1 divided by the absolute risk reduction. It is the average number of people who need the intervention for one additional person to avoid the outcome during the study period.

A relative figure compresses two risks into a single ratio. That makes it compact, but it also hides the underlying absolute risks, and readers tend to overestimate an effect when it is presented in relative terms. Papers sometimes report a hazard ratio or an incidence rate ratio instead of a risk ratio. These are also relative measures, and the same caution applies.

A worked comparison: same ratio, different meaning

Imagine two invented trials of a hypothetical intervention. Each randomizes 1,000 people to the intervention and 1,000 to a comparison group and follows everyone for one year. Both produce the same headline: “cuts risk in half.” None of these numbers come from a real study.

Trial A: the outcome is uncommon

Comparison group: 20 of 1,000 had the outcome (2.0%).
Intervention group: 10 of 1,000 had the outcome (1.0%).
Relative risk: 0.50, a 50% relative reduction.
Absolute risk reduction: 1.0 percentage point.
NNT: 100 people treated for one year for one fewer outcome.

Trial B: the outcome is common

Comparison group: 200 of 1,000 had the outcome (20%).
Intervention group: 100 of 1,000 had the outcome (10%).
Relative risk: 0.50, a 50% relative reduction.
Absolute risk reduction: 10 percentage points.
NNT: 10 people treated for one year for one fewer outcome.

The relative risk is identical, yet the absolute difference is ten times larger in Trial B. A reader who saw only “50%” could not tell the two trials apart.

The intervals add a second layer. I calculated approximate 95% intervals from the invented counts using standard large-sample formulas, so treat them as illustrations, especially for Trial A, where there are very few events. In Trial A, the relative risk interval runs from about 0.24 to 1.06, and the absolute risk reduction interval runs from about -0.1 to 2.1 percentage points. Both include the value that means “no difference” (1.0 for a ratio, 0 for a difference), so Trial A's result is also compatible with no effect at all. Trial B's intervals are much tighter: about 0.40 to 0.63 for the relative risk, about 7 to 13 percentage points for the absolute reduction, and an NNT of roughly 8 to 15. The same headline can sit on top of a result that is imprecise or one that is precise, and only the interval shows which.

Where the two numbers diverge

Baseline risk. A relative figure ignores how common the outcome was to begin with. The absolute figure depends entirely on it. Noordzij and colleagues give an example of a claim that a treatment halves mortality when it moves death rates from 0.002% to 0.001%, an improvement whose clinical relevance they say may be questioned.
Rare outcomes. The same authors note that relative risks can become extremely large when the outcome is rare in the comparison group, which can make an effect look more important than its absolute size supports.
Time frame. A risk only applies to the period it covers. A one-year risk and a five-year risk from the same trial are different numbers, and so are their NNTs.
Precision. A relative result is conventionally called statistically significant when its 95% interval excludes 1.0, and an absolute difference when its interval excludes 0. Each measure needs its own interval.

Applying it to a headline about a trial

Suppose a news story says a trial found that “Intervention Y cut the risk of Event Z by 50%.” Before deciding what that means, work through these steps:

Find the counts. Look in the abstract or results for how many people were randomized to each group and how many had the outcome in each. The CONSORT group's guidance, as Noordzij and colleagues quote, says event rates, or denominators, should be reported so readers can see how the ratios were calculated.
Confirm the outcome and the time frame. Check that the outcome in the headline is the one the trial was built to measure. Our guides to primary and secondary outcomes and to finding the primary outcome and time frame in a trial abstract show where to look.
Calculate or locate the absolute difference. Subtract the intervention group's risk from the comparison group's risk. Divide 1 by that figure to get the NNT.
Check the intervals. Look for a 95% interval for both the relative and absolute results. Our guide to confidence intervals explains what the range does and does not say.
Compare the enrolled group with the reader. A trial describes the people it enrolled. Our guide to inclusion and exclusion criteria explains how to check who was included and who was excluded.

If the story gives only the “50%,” you cannot yet tell whether the trial looks like Trial A or Trial B. That is not a reason to dismiss the result. It is a reason to treat the claim as incomplete until the counts, the time frame and the intervals are in view.

What reporting guidance asks for

According to Noordzij and colleagues, the CONSORT 2010 reporting guideline for randomized trials recommends presenting both absolute and relative effect sizes for binary outcomes, and the authors themselves recommend giving a 95% interval for each. They also cite a review of randomized medication trials published in six high-impact general medicine journals between June 2008 and September 2010. Of 157 trials with positive, statistically significant findings, 69 (44%) reported only relative measures in the abstract. Those figures are more than a decade old and may not reflect current practice, but they show why you should check the absolute numbers yourself.

Limits of this comparison

The numbers are invented. Trials A and B are teaching examples. They say nothing about any real intervention, product or condition.
Trial results describe groups. An NNT or risk reduction is an average across enrolled participants. It is not a forecast for any one person. Individual absolute risk estimates come from prediction models built for that purpose, and a trial headline is not one of them.
Benefit figures say nothing about harms. A risk reduction describes one outcome. For the other side of the ledger, see our guide to adverse event tables in clinical trials.
Neither measure repairs a weak study. Both are calculated from the trial's data, so they are only as sound as the design, follow-up and reporting behind them.
NNT is tied to its context. It applies to a specific outcome, population and time period, and it changes when any of them changes. This point is this article's own reading of how NNT is defined, not a statement from a source.

If a trial result bears on your own treatment, screening or medication decisions, discuss it with a physician, pharmacist or other qualified clinician who knows your history. See also this site's medical disclaimer.

What you can do next

Whenever a headline gives a percentage change, ask what the risk was to begin with, over what period, and how precisely it was measured. For a fill-in-the-blanks version of this check, use our reader worksheet on absolute, relative and baseline risk. For related reading habits, see correlation and causation in health headlines and the full How to Read Health Research section.

Sources

Sources checked October 9, 2026.

Noordzij M, van Diepen M, Caskey FC, Jager KJ. Relative risk versus absolute risk: one cannot be interpreted without the other. Nephrology Dialysis Transplantation. 2017;32(suppl_2):ii13-ii18. doi:10.1093/ndt/gfw465. Used for the definitions, the NNT formula, the 0.002% to 0.001% example, the CONSORT 17b recommendation and the 69-of-157 figure (the last two as reported by the authors, not checked against the original CONSORT paper or the original review).
National Institutes of Health. NIH Clinical Research Trials and You: The Basics, page last reviewed April 24, 2025. Used only for the description of randomization.