By MedClinRes.org Clinical Research Team. This article has not been peer-reviewed or clinically reviewed, and it is general education, not medical advice.
Short answer: a confidence interval shows how precisely a study pinned down its estimate. It marks a range of effects that are compatible with that study's result. In this article's reading (an editorial interpretation, not a statement from the sources), that range describes the studied group and the uncertainty in the summary number. It is not a forecast of what will happen to you or to any one person.
What the range adds to the number
A study result usually arrives as a single figure, the estimate. The interval beside it addresses a second question: how much uncertainty surrounds that figure. According to the Cochrane Handbook chapter on analyzing data, the interval conveys the precision of a summary estimate, while the P value conveys how strongly the data argue against there being no effect. The two answer different questions, so neither replaces the other.
The Handbook also describes the forest plot, the standard chart for combined results. Each study appears as a block at its estimate, with a horizontal line through it. That line is the confidence interval, most often drawn at the 95% level, and it shows the range of effects compatible with that study's result. The size of the block reflects the weight the study carries. Larger studies have smaller standard errors and receive more weight, and heavily weighted studies usually have narrower intervals.
Most of that chapter concerns meta-analysis, the statistical combination of results from separate studies. Where a point below applies only to combined results, this article says so.
A result card you can fill in
When you read a study, separate the four parts below. If you cannot fill one in from the paper, treat that gap as information too.
- Estimate: the single summary figure, and which measure it is (for example, a ratio or a difference between groups).
- Interval: the lower and upper limits and the confidence level, usually 95%. Note whether the interval includes the value that means “no difference” for that measure. The paper should say which value that is.
- Study population: who was enrolled, in what setting, for how long, and how many people.
- Unanswered question: what the study did not test, such as other populations, longer follow-up, or outcomes it did not measure.
Four common ways intervals get over-read
1. Treating the range as a personal prediction
The Handbook notes that absolute measures of effect are often easier for clinicians to interpret than relative ones, but are less likely to generalize. Converting a relative result into an absolute one requires assuming a baseline risk for the comparison population. Taken together, these points suggest why an interval calculated across a study group should not be read as a personal prediction. That conclusion is this article's interpretation.
2. Confusing the interval with the spread between studies
For a pooled random-effects result, the Handbook explains that the interval describes uncertainty in the location of the average effect. It does not describe how much the studies disagree, and a tight interval can coexist with large variation between studies. Prediction intervals show that variation, and the Handbook encourages them when there are a reasonable number of studies (it gives about five or more as an example).
3. Assuming a narrow interval means a trustworthy result
Precision and validity are different things. The Handbook states that statistical synthesis does not guarantee valid results, and that meta-analyses can mislead seriously when study design, within-study bias, or variation between studies is not carefully considered. It also says a fixed-effect analysis run on studies that genuinely differ can produce an interval that is too narrow and even meaningless.
4. Reading “not significant” as “no effect,” or skipping the population
In its discussion of subgroup comparisons, the Handbook cautions that a non-significant result may simply reflect too little information, not a smaller or absent effect. By extension, a wide interval is best read as a study that could not settle the question (this article's interpretation). Population matters too. A field-trials chapter on NCBI Bookshelf notes that results from trials in high-income countries may not apply directly to low- and middle-income countries, for reasons such as differing prevalence of other infections or nutritional deficiencies. This article extends that caution to any mismatch between a study group and a reader.
Limits of this article
- It draws on one methods handbook chapter, written for people conducting systematic reviews, and one trials-methods textbook chapter. It is not a full statistics course.
- The Handbook describes intervals mainly for meta-analyses. Applying the same reading habits to single studies is editorial guidance.
- Different measures, study designs, and analysis methods change how you should read an interval. The paper's own methods section governs.
- One Handbook caution deserves a special mention: for study-level comparisons, the pattern across trial averages can differ from, or even run opposite to, the pattern within trials. This limits what study-level results can say about individual differences.
Who should seek professional advice?
If a study result bears on your own treatment, screening, or medication decisions, discuss it with a physician, pharmacist, or other qualified clinician who knows your history. They can weigh the result against your circumstances, which no interval can do. For what this publication does and does not provide, see the About page and the Editorial Team and Review Process page.
What you can do next
- Find the full text or registry entry, not just the headline number, and locate the interval.
- Fill in the four-part result card above.
- Ask whether the interval comes from one study or a pooled analysis, and whether the paper reports how much the studies differed.
- Compare the enrolled population with your own situation before drawing any conclusion.
- Bring your questions, along with the paper, to a clinician.
Sources
- Deeks JJ, Higgins JPT, Altman DG, McKenzie JE, Veroniki AA, editors. Chapter 10: Analysing data and undertaking meta-analyses (last updated November 2024). In: Cochrane Handbook for Systematic Reviews of Interventions, version 6.5. Cochrane, 2024. cochrane.org
- Ross DA. Types of intervention and their development. In: Smith PG, Morrow RH, Ross DA, editors. Field Trials of Health Interventions: A Toolbox, 3rd edition. Oxford: OUP; 2015. NCBI Bookshelf (cited only for the point about applying trial results across populations).