Confidence interval
The range around a trial result that the data are consistent with. The point estimate gets the headline; the interval tells you how much to trust it.
A trial measures an effect in a sample of people and uses it to estimate the effect in everyone like them. The single best guess is the point estimate. The confidence interval (CI) is the range of values that are reasonably consistent with the data. By convention it is a 95 percent interval: if the same trial were repeated many times, 95 percent of the intervals calculated this way would contain the true value.
It is not quite "a 95 percent chance the true value is inside", though in practice it is read that way. What matters more is how to use it.
Three ways to read one
Width. A narrow interval means a precise estimate; a wide one means the true effect could be quite different from the headline. Width shrinks with more participants and more events.
Where it sits relative to no effect. For a difference, no effect is 0. For a ratio, such as a hazard ratio, it is 1. An interval that excludes that value corresponds to a statistically significant result at the matching level.
Both ends. The end nearest to no effect is the most modest effect consistent with the data. If that end is still clinically meaningful, the result is strong.
| Result | Point estimate | 95% CI | How to read it |
|---|---|---|---|
| STEP 1: semaglutide minus placebo, weight change at 68 weeks | -12.4 points | -13.4 to -11.5 | Narrow; even the smallest value is a large effect |
| SELECT: major cardiovascular events | HR 0.80 | 0.72 to 0.90 | Excludes 1; benefit between 10% and 28% relative reduction |
| SURMOUNT-1: tirzepatide 15 mg weight change | -20.9% | -21.8 to -19.9 | Narrow, from about 630 people per arm |
| SURPASS-CVOT: tirzepatide vs dulaglutide | HR 0.92 | 0.83 to 1.01 (95.3%) | Crosses 1; not shown to be better |
Why the interval often matters more than the p-value
A p-value says only whether a result clears a threshold. The interval shows the size and the uncertainty together. A trial with thousands of people can produce a tiny effect with a tiny p-value; a small trial can produce a large effect with an interval so wide it includes almost nothing. Phase 2 trials, with a few dozen people per arm, often have wide intervals, which is one reason phase 3 numbers come in lower.
Some intervals are not 95 percent. SURPASS-CVOT reported a 95.3 percent interval because of how its statistical testing was planned. The level is chosen in advance and should be stated beside the numbers.
Intervals on your own numbers
Lab tests and scales have their own error. A fasting glucose of 98 one month and 103 the next may be the same value measured twice. Reading a single lab change as a trend is the personal version of ignoring the interval. The lab report guide covers reference ranges and normal variation.
The bloodwork tracking feature plots each marker across draws, so one result can be seen against several, and against the dose timeline.
Related terms: hazard ratio, estimand, placebo-controlled trial, clinical trial phases. Back to the glossary.
Questions people ask
What does a 95 percent confidence interval mean?
It is the range of values consistent with the data. If the trial were repeated many times, 95 percent of intervals calculated the same way would contain the true effect.
What was the confidence interval in SELECT?
The hazard ratio was 0.80 with a 95 percent confidence interval of 0.72 to 0.90, which excludes 1.
What does it mean if the interval crosses 1?
For a ratio, 1 means no difference. An interval that includes 1 means the trial did not show a difference with the usual level of certainty.
Why are phase 2 confidence intervals wide?
Small groups give imprecise estimates. Width shrinks as the number of participants and events grows.
Sources
- Once-weekly semaglutide in adults with overweight or obesity (STEP 1), New England Journal of Medicine, 2021
- Semaglutide and cardiovascular outcomes in obesity without diabetes (SELECT), New England Journal of Medicine, 2023
- Tirzepatide once weekly for the treatment of obesity (SURMOUNT-1), New England Journal of Medicine, 2022
- Cardiovascular outcomes with tirzepatide versus dulaglutide in type 2 diabetes (SURPASS-CVOT), New England Journal of Medicine, 2025
- ICH E9(R1): addendum on estimands and sensitivity analysis in clinical trials, International Council for Harmonisation, adopted 20 November 2019
