How to read a study

Anatomy of a digest entry

Each study in the Research Digest follows the same pattern. Here it is with an entry from the October 20, 2026 issue.

1. The grey line under the heading says the design, who was studied and for how long. Example: Randomized within-subject trial · 31 healthy untrained young adults · 12 weeks. Read this first: it tells you who the result is about. A study in untrained young adults says little about experienced lifters.
2. The bold sentence is the main finding in one line.
3. The body gives the numbers the abstract reports, and the biggest limitation.
4. "In practice" says what the finding supports, and for whom. It is never a training prescription.
5. Trust level (Low, Moderate or High) is one judgment from the design, the number of people, the length, who was studied and how consistent the results were. It is assigned when each issue is written and reviewed author-assigned. It is not a formal grading system, and the reason is always written next to it.
6. "Read the study" links to the paper. Digest summaries are based on the published abstract only, so the full paper can add detail or change the picture.

Study designs, roughly from weaker to stronger for cause and effect

• Observational study. Researchers watch what people already do and look for links. Useful for ideas; it can't show that one thing caused another, because other differences between people get mixed in.
• Randomized trial. People are randomly put into groups that get different treatments. Randomizing spreads the other differences evenly, so a difference in results is more likely to come from the treatment.
• Within-person randomized trial. Each person does both conditions, for example one leg trained one way and the other leg another way. Each person is their own comparison, which cuts out a lot of noise, but it only works for things that can be split that way.
• Systematic review and meta-analysis. Researchers gather every study on a question and pool the results. Usually the strongest single source, but only as good as the studies it pools: five small, weak trials pooled together are still five small, weak trials.

Confidence interval reader

Most results come with a 95% confidence interval: a low and a high end around the main estimate. Roughly, it is the range of true effects the data are compatible with. The key question is whether it crosses the "no effect" line. Pre-filled with the time-restricted eating result from the October 20 issue.

Effect size reader

Many studies report a standardized mean difference (SMD), sometimes called Cohen's d or Hedges' g: the gap between two groups measured in standard deviations, so results on different scales can be compared. The labels "small, medium, large" you'll see for values near 0.2, 0.5 and 0.8 are rough conventions, not rules. Pre-filled with the soreness result (−0.81) from the cold-water study in the October 20 issue.

What the digest can't tell you

• It reads abstracts, not full papers, so details like dropouts, exact methods and side results may be missing.
• Most lifting studies are small, short and in specific groups (often young, untrained or recreationally trained people). One study rarely settles a question.
• An average result doesn't tell you how any one person will respond.