For Scientific Readers

#Start with the question and intended use

Scientific findings are easier to interpret when the research question is clear. Identify the population, the measurement or intervention, the comparison and the outcome. Then ask what decision the evidence is meant to support. An exploratory finding and a result intended to guide care require different kinds of support.

Begin with Formulating a Research Question, then read Protocols and Prespecified Analysis. Sampling and Representativeness explains who the results may describe. Sample Size and Statistical Power covers the assumptions behind study size, including why a large study can still give a misleading answer.

#Separate measurement from clinical evidence

A method can measure its target consistently without identifying a clinical condition accurately. It can also identify a condition without improving care. These are separate questions, so evidence for one should not be treated as proof of the others. The setting and intended use matter at every stage.

Reference Standards and Comparators describes what a method is evaluated against. Analytical Validation addresses measurement performance. Clinical Validation considers relationships with clinical states or outcomes. Clinical Utility examines whether using the method leads to useful effects, taking account of harms as well as benefits.

#Assess transferability and transparency

Results may change across settings, populations, equipment and time. External and Prospective Evaluation explains two ways to test performance beyond an initial analysis. Neither label alone establishes that a study is independent, unbiased or relevant to every future use. Read the design details rather than relying on the label.

Reproducibility and Provenance covers traceable inputs and repeatable workflows. Reporting Guidelines explains how structured descriptions help readers assess a study. Evidence Extraction Method presents a proposed approach to recording findings and limitations. Together, these notes support critical reading; they do not replace appraisal of the underlying evidence.

#Common misunderstandings

A precise measurement is not necessarily an accurate one: repeated results can agree closely while remaining systematically biased. Likewise, agreement with a reference method does not, by itself, establish that a measurement predicts health outcomes or improves clinical decisions. These are different questions requiring different evidence.

Statistical significance is another frequent source of confusion. It does not establish that an effect is large, clinically meaningful, or free from bias. A non-significant result also does not prove that no effect exists; uncertainty and study size matter.

Validation is best understood as evidence for a specified purpose, population and setting, rather than a permanent seal of approval. Performance may change when methods, users or conditions differ. Finally, a detailed report is not necessarily a strong study. Transparent reporting makes strengths and limitations easier to assess, but cannot repair weaknesses in recruitment, measurement, comparison groups or analysis after the fact.

#Questions worth asking a clinician

• How does the chosen measurement capture what the research question asks, and is its performance adequate for the intended use?
• What evidence separately supports measurement performance, clinical validity, and whether using the result improves clinical decisions or patient outcomes?
• Beyond the study label, how do participant selection, comparison groups, and follow-up affect the conclusions this design can support?
• Was validation performed in an independent population relevant to the intended use, and how were missing data, confounding, and uncertainty handled?
• Where can readers find the protocol, prespecified analyses, deviations, and complete results needed to assess selective reporting and reproduce the analysis?