How to Read a Sleep Study in the News

Sleep is one of the most heavily covered health topics and one of the worst reported, because the findings are usually small, preliminary and interesting — a combination that produces confident headlines from uncertain data.

Six questions, in the order worth asking them.

1. How many people were in it?

The first thing to look for and usually the hardest to find in the article.

Sleep research runs small. The BMJ trial that found didgeridoo playing reduced snoring and daytime sleepiness had 25 participants. The systematic review most often cited on mouth taping covered ten studies and 213 patients in total. The frequently quoted finding linking heavy snoring to carotid artery changes came from 110 people.

None of those is disqualifying. All of them mean the finding is a signal worth following rather than a settled fact — and headlines rarely make that distinction. The didgeridoo trial, reported properly.

2. What is the confidence interval?

The single most useful number and the one almost never reported.

An odds ratio of 10.5 sounds dramatic. An odds ratio of 10.5 with a confidence interval running from 2.1 to 51.8 tells you the study could not narrow the effect down to within an order of magnitude. Both describe the same carotid atherosclerosis finding in snorers. Only one is honest about the precision. How we report that one.

Wide intervals are normal in small studies. They are a reason for interest rather than confidence.

3. Is the outcome the thing you care about?

Watch for surrogate outcomes — something measurable standing in for something that matters.

A device that improves a sleep-tracker score has not been shown to improve your health. A supplement that shifts a blood marker has not been shown to make you feel better. An intervention that reduces snoring loudness has not been shown to treat sleep apnea, which is about airway closure rather than noise.

The question to ask: did they measure the outcome I actually want, or something correlated with it?

4. Cross-sectional, or over time?

Cross-sectional studies photograph a population at one moment and find associations. They cannot establish that one thing caused another, and they cannot rule out that both were caused by something else.

Longitudinal studies follow people over time, which is better. Randomised controlled trials assign the intervention deliberately, which is better still and much rarer in this field.

If a headline says "linked to", it is almost certainly the first kind. That is not worthless — it is where hypotheses come from — but "linked to" and "causes" are doing very different work.

5. Who paid for it, and what was tested?

Industry funding does not invalidate research, and it is worth knowing about.

More useful is being precise about what was tested. A biocompatibility test on an adhesive is not a test of the finished product it goes into. A study of a drug at one dose says nothing about another. Data commissioned by a component's manufacturer is not the same as data commissioned independently.

Those distinctions get flattened in reporting, and they are frequently the whole story. How we apply that to lab claims.

6. Does the expert quoted have anything to do with the study?

A common structure: a modest finding, then a quote from someone unconnected to the research offering a much more expansive interpretation. The quote is often what becomes the headline.

Where to read it instead

The practical answer is to go one layer closer to the source than the aggregators.

Publications with a dedicated health desk generally report the caveats that content farms strip out — Health Newspapers runs a sleep section covering research and policy, and its "What the Study Actually Says" franchise is doing precisely the job this article describes. Reading a study's own abstract is better again, and PubMed is free.

What you are looking for in either case is the sample size, the interval, and what was actually measured. If a write-up gives you none of those, it is not reporting a study, it is repeating a press release.

The version that matters for this site

Almost everything sold for snoring sits on thin evidence, including things we recommend. Mouth tape and nasal strips are tier 2 here — plausible mechanism, limited data — and we say so every time rather than borrowing the certainty of better-evidenced interventions. Our evidence standards.

The two claims on this site with genuinely strong evidence behind them are that positional therapy helps positional snorers, and that mandibular advancement devices help tongue-base snorers. Everything else is weaker than the marketing implies, ours included. Every remedy ranked.