How to Read a Health Study in 5 Minutes

Every week, a headline announces that something lengthens life, melts fat, or rewires the brain. Underneath almost every one sits a single study, and the study almost never says what the headline says. You don't need a science degree to catch the gap. You need five minutes and six checks. This is the exact routine I run before I write a word.

1. What kind of study is it?

Find the design before you read a single result. A randomized controlled trial, where people are assigned by chance to treatment or control, can support cause and effect. A cohort study, where researchers just watch, can only show associations. A study in mice or in a dish can show a mechanism, never a benefit in you. Look for: the words randomized, placebo, cohort, or the species. Red flag: a headline that says “proves” over a study that watched, or worse, over mice.

2. What was actually measured?

Separate the endpoint from the surrogate. An endpoint is the thing you care about: deaths, heart attacks, grades, infections. A surrogate is a number that stands in for it: a biomarker, a lab value, a scan. Surrogates are cheaper and faster, which is why marketing loves them. The Lp(a) trial I covered this week is the cleanest recent lesson: the biomarker fell beautifully, and the events did not. Red flag: a product sold on a surrogate, priced like an endpoint.

3. How big is the effect, in real units?

“Cuts risk by 50 percent” can mean from 2 in a million to 1 in a million. Always ask: 50 percent of what, over how long, in how many people? Prefer absolute numbers. When a paper reports standard deviations, translate: the AI tutor trial's 0.37 standard deviations sounds abstract until the authors convert it to about 4 points on a 100 point grade scale. Red flag: percentages with no base rate anywhere in the article.

4. Who was studied, and is it you?

Effects shrink and flip across populations. A drug tested in 8,323 patients who already had heart disease says little about a healthy 30 year old. A sleep finding in elite athletes says little about new parents. Look for: age, sex, health status, and the country of the sample, usually in the first table. Red flag: advice aimed at everyone, from a study of almost no one like them.

5. What is the paper's status, and who paid?

A peer reviewed paper, a preprint, a conference abstract and a company press release are four different levels of evidence wearing the same costume. Check the label before the claim. Then find the funding line: industry funding doesn't fake results, but it reliably chooses which questions get asked. Red flag: a press release with no data table, quoted as settled science.

6. What would change the answer?

Ask what the study can't rule out: short follow up, no replication yet, harms measured loosely or not at all. One trial is an opening argument, not a verdict. If a claim would change what you eat, buy or inject, wait for the second study. Red flag: urgency. Real evidence can wait a week; sales copy can't.

The 5 minute checklist

• Design: randomized trial, observational, or animal/in vitro?
• Endpoint: real outcome, or a surrogate number?
• Size: absolute effect, base rate, time window?
• People: who was studied, and how close are they to me?
• Status and sponsor: peer reviewed, preprint, or press release? Who paid?
• Limits: follow up length, replication, harms?

Run those six and you will read health news better than most of the people writing it. Every article on this site runs the same checks in public, sources at the bottom, so you can audit me with my own list.