Evidence Literacy

How to interpret evidence

A practical framework for asking what a source can show, what it cannot show and how confidently its findings should be interpreted.

Core principle: the strength of a conclusion should match the strength, relevance and consistency of the evidence behind it.

Start with the question

Evidence is easier to judge when the question is specific. Is the source asking whether an intervention works, whether an event is associated with an exposure, how often something occurs, or whether a possible safety signal needs investigation? Different questions require different study designs.

Six checks for any evidence claim

1. Identify the sourceIs it a primary study, systematic review, regulatory assessment, guidance document, spontaneous report, case report or personal account?

2. Check the designWas there a comparison group? Were participants selected or allocated in a way that reduces bias? Was follow-up adequate?

3. Look at the populationWho was studied, how many people were included and how closely do they match the population discussed in the claim?

4. Read the result preciselyCheck the outcome measured, effect size, absolute numbers, confidence interval and period of follow-up—not only whether a result was called significant.

5. Consider bias and confoundingCould selection, measurement, missing data, reporting choices or differences between groups offer another explanation?

6. Compare the wider evidenceDoes the finding agree with other well-conducted studies and reviews, or is it an isolated result?

Evidence hierarchy is a guide, not an automatic verdict

ANTIVAX.co.uk generally gives greater weight to systematic reviews and well-conducted meta-analyses, large well-designed studies, regulatory assessments and public-health guidance. Individual studies, spontaneous-report data, case reports and anecdotes can still matter, particularly for identifying questions or possible signals, but they answer different questions and carry different limitations.

Important: a study’s place in a hierarchy does not remove the need for critical appraisal. A poorly conducted review can be less informative than a strong, directly relevant study.

Association, signal and causation are different

Association

Two things occur together more or less often than expected. This does not by itself show that one caused the other.

Safety signal

Information suggesting a possible new or changed risk that warrants investigation. A signal is a reason to investigate, not a confirmed conclusion.

Causation

The evidence supports that one factor contributed to an outcome. Assessment may consider timing, comparison groups, alternative explanations, biological plausibility, consistency and findings from multiple sources.

Read uncertainty as part of the result

All estimates have uncertainty. Confidence intervals show a range of values compatible with the data under the statistical model. Wide intervals usually indicate less precision. A precise estimate can still be biased, and a statistically significant result is not automatically important in practical or clinical terms.

Avoid false certainty: “no statistically significant difference” does not necessarily prove that two things are equivalent or that no effect exists.

Apply the conclusion only as far as the evidence allows

Check whether the article’s conclusion matches its methods and results. Be cautious when a source moves from association to causation, from a subgroup to everyone, from a surrogate outcome to a health outcome, or from one study to a settled general conclusion.

Methodology sources

Cochrane Handbook — considering bias and conflicts of interest

WHO — causality assessment of an adverse event following immunisation

ANTIVAX.co.uk — Medical Evidence Review Policy

Educational scope

This guide provides general evidence-literacy information. It does not assess an individual’s medical circumstances or replace advice from a suitably qualified healthcare professional.