Skip to content
All resources

Reading the evidence

Relative Risk vs Absolute Risk: How to Read the Numbers in a Trial

AR

By

· 10 min read

A summary of published research, with citations. It states no amount, schedule or route, and makes no claim about what this compound does in a person. Supplied for laboratory research use only.

On this page · 8 sections
  1. What are relative risk and absolute risk?
  2. A worked example: SURPASS-CVOT
  3. A worked example in the other direction: CAST
  4. What is a hazard ratio?
  5. How do you read a confidence interval?
  6. When the relative risk breaks down
  7. A checklist for any risk figure
  8. Frequently asked questions

Relative risk compares the rate of an event in one group with the rate in another, as a ratio. Absolute risk is the difference between those rates, in percentage points. A drug that cuts an event from 2% to 1% has halved the relative risk, a 50% reduction. It has lowered absolute risk by one percentage point, so 100 people would need treatment for one to avoid the event. Both numbers are true. Relative figures sound larger and travel further, so always ask for the absolute one.

A printed grid of a hundred small figures with a few highlighted, beside a pencil
Illustration: absolute risk counts people, not percentages of percentages.

What are relative risk and absolute risk?

Start with two groups and count events. The event rate in each group is the number of people with the event divided by the number in the group. Everything else is built from those two rates.

  • Relative risk (or risk ratio): the treated group's rate divided by the control group's rate. Below 1 means fewer events on treatment; above 1 means more.
  • Relative risk reduction: one minus the relative risk, usually expressed as a percentage. This is the "cuts risk by 30%" figure.
  • Absolute risk reduction: the control rate minus the treated rate, in percentage points. If treatment increases risk, the same subtraction gives an absolute risk increase.
  • Number needed to treat: one divided by the absolute risk reduction. It is the number of people who would need treating, for the duration of the trial, for one fewer to have the event. Laupacis, Sackett and Roberts proposed it in 1988 as a measure clinicians could actually use. For harms the same calculation gives a number needed to harm.

The same relative effect means very different things at different baseline risks. Halving a 40% risk prevents 20 events per 100 people; halving a 0.4% risk prevents one event per 500. The baseline risk is the half of the story that a relative figure leaves out.

A worked example: SURPASS-CVOT

SURPASS-CVOT randomised people with type 2 diabetes and cardiovascular disease to tirzepatide or dulaglutide, an active drug with established cardiovascular evidence rather than a placebo. The abstract gives everything needed:

  • Tirzepatide: 801 primary events among 6,586 people, 12.2%.
  • Dulaglutide: 862 events among 6,579 people, 13.1%.
MeasureCalculationResult
Relative risk12.2% ÷ 13.1%0.93
Relative risk reduction1 − 0.93about 7%
Absolute risk reduction13.1% − 12.2%0.9 percentage points
Number needed to treat1 ÷ 0.0094about 106, over the trial's follow-up
Hazard ratio (reported)From a time-to-event model0.92 (95.3% CI 0.83 to 1.01)

Three lessons sit in that table. First, the relative figure ("7% fewer events") and the absolute one ("nine fewer per thousand") describe the same result. Second, this is relative to dulaglutide, not to no treatment, so it says nothing directly about the effect against placebo. Third, the confidence interval runs from 0.83 to 1.01, crossing 1. The trial met its prespecified test of noninferiority, but superiority was not shown (P = 0.09 for superiority). Tirzepatide cardiovascular outcomes explains the noninferiority design in full.

A worked example in the other direction: CAST

Absolute figures matter just as much for harm. In the Cardiac Arrhythmia Suppression Trial, 755 patients received encainide or flecainide and 743 received matching placebo. Over a mean of 10 months, 43 on active drug died of arrhythmia against 16 on placebo.

  • Event rates: 5.7% against 2.2%.
  • Relative risk: about 2.6, so arrhythmic death was more than two and a half times as common.
  • Absolute risk increase: about 3.5 percentage points.
  • Number needed to harm: about 28. For roughly every 28 patients treated for those ten months, one additional arrhythmic death.

Here the relative and absolute figures both alarm, because the baseline risk was not small. The story of why this trial was run at all is in surrogate endpoints explained.

What is a hazard ratio?

Most large outcome trials report a hazard ratio rather than a simple relative risk. A hazard ratio compares the rate at which events occur over time in the two groups, using a model that accounts for people being followed for different lengths of time and leaving the study at different points. In practice it is read like a relative risk: 0.80 means events occurred at about four-fifths the rate; 1.25 means a quarter faster.

It inherits the same weakness. A hazard ratio is relative and says nothing about absolute risk on its own. In the ILLUMINATE trial of torcetrapib, the hazard ratio for cardiovascular events was 1.25 and for death from any cause 1.58. To know how many people that represents, you need the event counts or rates, which a careful reader looks for in the results table rather than the abstract's headline.

How do you read a confidence interval?

A 95% confidence interval is the range of effect sizes reasonably compatible with the data, given the trial's assumptions. Two rules cover most cases:

  • For a ratio, the line of no effect is 1. An interval that includes 1 means the data are compatible with no difference. SURPASS-CVOT's 0.83 to 1.01 includes 1.
  • The width is information. A narrow interval means a precise estimate; a wide one means the trial could not pin the effect down.

In ACCORD, intensive glucose lowering produced a hazard ratio for death of 1.22, with an interval of 1.01 to 1.46: just clear of 1, and wide. Cheng and Chang's meta-analysis of GLP-1 drugs and hair loss reported a risk ratio of 3.25 with an interval of 1.44 to 7.36. That interval is compatible with risk being less than one and a half times as high, or more than seven times. A headline "three times the risk" hides most of that range. The trial-level hair-loss data are in GLP-1 hair loss, and how such pooled figures are built is in systematic review vs meta-analysis.

When the relative risk breaks down

When the control rate is zero. In TRANSCEND-T2D-1, a phase 3 trial of retatrutide in type 2 diabetes, discontinuation because of adverse events was 2% to 5% across retatrutide groups and 0% on placebo. A relative risk cannot be calculated when the denominator rate is zero, and any "times higher" figure would be meaningless. The absolute difference, two to five percentage points, is the honest summary. Retatrutide is not approved by Health Canada or any other regulator; its trial record is in retatrutide side effects.

When the outcome is continuous. Weight change is not a risk. In TRIUMPH-1, mean weight change was −23.7% and −25.0% in the two higher dose groups against −3.9% on placebo. The meaningful comparison is the placebo-adjusted difference the authors report, −19.8 and −21.0 percentage points. Dividing one percentage by another would produce a ratio that describes nothing.

When it is an odds ratio. Observational studies and meta-analyses often report odds ratios. When the outcome is rare, an odds ratio approximates a relative risk; when it is common, the odds ratio exaggerates it. The GLP-1 alopecia meta-analysis by Viquez Burboa and colleagues reported an odds ratio of 1.40 from health-record cohorts. It is not directly comparable with Cheng and Chang's risk ratio from trials.

When it is a relative change in a percentage. A rise from 2% to 3% is a 50% relative increase and a one-point absolute increase. Reports that say "50% higher" without the starting point are giving half the information.

A checklist for any risk figure

  • What are the event rates in each group, not just the ratio?
  • Is the comparison against placebo, no treatment, or another active drug?
  • Over what period? A number needed to treat is meaningless without its time frame.
  • Does the confidence interval include 1, and how wide is it?
  • Is the figure a risk ratio, an odds ratio or a hazard ratio?
  • Was this the trial's prespecified primary outcome, or one of many?

The last question is the subject of how to read a PubMed abstract, and the general limits of a result are in how to read a peptide study. Nothing here is medical, dosing, or therapeutic guidance.

Frequently asked questions

Why do news reports and press releases favour relative risk? Because it produces larger, more memorable numbers, and it does not depend on the baseline risk, which varies between populations. Both are reasonable motives, but a relative figure without the absolute one leaves a reader unable to judge how many people are affected. Ask for both, plus the time frame.

Is a hazard ratio the same as a relative risk? Not exactly. A relative risk compares the proportion with an event by the end of follow-up; a hazard ratio compares the rate of events over time and handles people followed for different lengths. When events are uncommon and follow-up is similar, the two are usually close. Both are relative measures and need event rates beside them.

What is a good number needed to treat? There is no universal threshold. It depends on how serious the outcome is, how long treatment lasts, what it costs and what harms come with it. An NNT of 100 over several years may be worthwhile for preventing death, and an NNT of 5 may not be worthwhile for a trivial symptom with serious side effects.

If the confidence interval crosses 1, did the drug have no effect? Not necessarily. It means the trial could not rule out no effect at its chosen level of confidence. The true effect could be a modest benefit, nothing, or a modest harm. "No significant difference" is a statement about the data, not proof of equivalence, which requires a trial designed to test it.

Selected Research References

  1. Laupacis A, Sackett DL, Roberts RS. An assessment of clinically useful measures of the consequences of treatment. New England Journal of Medicine. 1988;318(26):1728-1733. pubmed.ncbi.nlm.nih.gov/3374545
  2. Nicholls SJ, Pavo I, Bhatt DL, et al.; SURPASS-CVOT Investigators. Cardiovascular Outcomes with Tirzepatide versus Dulaglutide in Type 2 Diabetes. New England Journal of Medicine. 2025;393(24):2409-2420. pubmed.ncbi.nlm.nih.gov/41406444
  3. Echt DS, Liebson PR, Mitchell LB, et al. Mortality and morbidity in patients receiving encainide, flecainide, or placebo. The Cardiac Arrhythmia Suppression Trial. New England Journal of Medicine. 1991;324(12):781-788. pubmed.ncbi.nlm.nih.gov/1900101
  4. Barter PJ, Caulfield M, Eriksson M, et al.; ILLUMINATE Investigators. Effects of torcetrapib in patients at high risk for coronary events. New England Journal of Medicine. 2007;357(21):2109-2122. pubmed.ncbi.nlm.nih.gov/17984165
  5. Gerstein HC, Miller ME, Byington RP, et al.; Action to Control Cardiovascular Risk in Diabetes Study Group. Effects of intensive glucose lowering in type 2 diabetes. New England Journal of Medicine. 2008;358(24):2545-2559. pubmed.ncbi.nlm.nih.gov/18539917
  6. Cheng PL, Chang HC. Glucagon-like peptide-1 receptor agonists and hair loss: A systematic review and meta-analysis. Diabetes Research and Clinical Practice. 2026;237:113333. pubmed.ncbi.nlm.nih.gov/42155605
  7. Viquez Burboa GU, Flores Guillén MÁR, Duardo González VS. GLP-1 Receptor Agonists and Alopecia: A Systematic Review and Meta-Analysis of Incidence, Risk, Subtypes, and Mechanisms. Skin Appendage Disorders. 2026. pubmed.ncbi.nlm.nih.gov/42621629
  8. Bajaj HS, Welch M, Shah P, et al. Efficacy and safety of retatrutide, a GIP, GLP-1, and glucagon receptor agonist, in people with type 2 diabetes and inadequate glycaemic control with diet and exercise (TRANSCEND-T2D-1): a double-blind, randomised, phase 3 trial. The Lancet. 2026;407(10546):2402-2413. pubmed.ncbi.nlm.nih.gov/42250575
  9. Jastreboff AM, Kaplan LM, Davies MJ, et al.; TRIUMPH-1 Investigators. Retatrutide, a Triple Hormone Receptor Agonist, for Treatment of Obesity. New England Journal of Medicine. 2026. pubmed.ncbi.nlm.nih.gov/42814954

All products sold by NovaEvum are for research and laboratory use only. They are not intended for human or animal consumption, diagnosis, treatment, cure, or prevention of any disease.