Evidence Policy

Start with the source

We link to primary research or original documentation wherever possible. Reviews and expert explanations can provide context. If only an abstract, press release, conference presentation, or company statement is available, that limitation should be clear. Preprints and other findings that have not undergone peer review should be identified.

Match the claim to the study

Cell experiments, animal studies, observational research, clinical trials, and evidence syntheses answer different questions. We identify the relevant population, intervention or exposure, comparison, and measured outcome when these are needed to understand a finding.

An association does not by itself establish causation. A change in a biomarker does not by itself demonstrate a benefit in how people feel, function, or survive. Findings from animals or cells should not be described as proven human benefits.

Quality requires context

Study labels alone do not establish reliability. Relevant considerations include sample size, controls, randomization, blinding, follow-up, missing data, effect size, uncertainty, funding, and consistency with other evidence. Systematic reviews and meta-analyses also depend on the quality and comparability of the studies they include.

For technology and engineering stories, distinguish a demonstration, simulation, prototype, independent evaluation, and a product available in practice. Company performance claims should be attributed and distinguished from independently verified findings.

Explain uncertainty

Coverage should distinguish what was observed, how it is interpreted, and what would need to be tested next. A promising result can be worth reporting without implying that it is ready for clinical use or that a proposed benefit is certain. New evidence can change an earlier assessment.

• Editorial Policy
• Medical Disclaimer