Evidence Literacy

How to read scientific studies

A paper is an argument supported by methods and data. Read beyond the headline and abstract to see whether the methods answer the question and whether the conclusion matches the results.

1 · Question
2 · Methods
3 · Results
4 · Interpretation

1. Identify the research question

Write the question in plain language. Identify the population, exposure or intervention, comparison and outcome. A clear question makes it easier to decide whether the chosen design is suitable.

2. Identify the study design

Randomised controlled trial

Participants are allocated by chance to comparison groups. Randomisation aims to reduce systematic differences between groups, although conduct, missing data and reporting can still introduce bias.

Cohort study

Groups are followed or reconstructed according to an exposure. Cohort studies can examine associations and rates but may be affected by confounding and selection.

Case-control study

People with an outcome are compared with people without it to examine previous exposures. Selection of controls and accurate exposure measurement are important.

Cross-sectional study

Exposure and outcome are measured around the same time. It can describe prevalence and associations, but the time order may be unclear.

Systematic review and meta-analysis

A systematic review uses pre-specified methods to find and assess relevant studies. A meta-analysis statistically combines compatible results. Its reliability depends on the search, included evidence, risk-of-bias assessment and suitability of the synthesis.

3. Read the methods before the conclusion

4. Read the results in absolute as well as relative terms

Find the number of participants and events in each group. Relative measures can make a difference appear large when the absolute difference is small, while absolute figures alone may obscure proportional differences. Both can be useful when presented with the underlying time period and population.

Ask: What is the effect estimate? What does its confidence interval include? Is the estimate precise? Is the outcome meaningful to the question being asked?

5. Separate statistical significance from importance

A p-value addresses a statistical question under a model; it does not measure the probability that the claim is true, the size of an effect, its practical importance or the absence of bias. Read it alongside the effect size, confidence interval, absolute numbers and study limitations.

6. Check whether the conclusion overreaches

7. Put one study into the wider evidence base

Compare the paper with other relevant studies, systematic reviews, regulatory assessments and guidance. Differences may reflect populations, interventions, outcomes, study quality or chance. A new paper can add information without overturning the wider evidence by itself.

8. Check funding and conflicts without using them as shortcuts

Read funding and conflict-of-interest declarations and consider how study design, analysis or reporting might be affected. A disclosed interest does not automatically invalidate a study, and absence of a declared interest does not establish that the methods are sound. Judge the evidence and its safeguards directly.

Methodology sources

Cochrane Handbook — bias and conflicts of interest

Cochrane Handbook — assessing risk of bias in randomised trials

Cochrane Handbook — bias due to missing evidence

ANTIVAX.co.uk — Medical Evidence Review Policy

Educational scope

This guide provides general research-literacy information. It does not independently appraise every study and does not replace personal medical advice from a suitably qualified healthcare professional.