Reading nutrition research

Write the research question in your own words. Identify the population, the exposure or intervention, the comparison and the outcome. Keep the time period visible: a short-term response and a longer-term health outcome are different questions.

Next, identify how the comparison was made. Were participants assigned to an intervention, followed over time or assessed at one point? Consider whether the methods can address the stated question. Cochrane's guidance on including non-randomised studies explains why design features matter when evaluating intervention effects.

Different instruments capture different aspects of diet. A 24-hour recall asks about a recent day; a food-frequency questionnaire asks about usual consumption over a specified period. The choice needs to fit the research question, and both approaches involve measurement error.

The National Cancer Institute's Dietary Assessment Primer discusses these trade-offs and approaches to reducing their effects. When reading a paper, note how often diet was measured, how the tool was used and whether the authors considered measurement error. A single mention of a “validated questionnaire” is a reason to find the validation source, not to stop examining the method.

In a non-randomised comparison, groups may differ in ways that also affect the outcome. For example, an observed difference between dietary groups may accompany differences in smoking, income or other behaviours. The question is how the study addressed plausible alternative explanations, not simply whether the analysis contains many variables.

Cochrane's chapter on risk of bias in non-randomised intervention studies discusses confounding and other sources of bias. Its approach concerns a defined study question and result; it should not be reduced to a single generic quality label for every type of nutrition research.

Locate the numerical result and its uncertainty. Check which comparison it refers to and whether it is the outcome described in the methods. Distinguish absolute quantities from relative changes. For an illustration, a hypothetical change from 2 to 1 events per 100 people is both a reduction of 1 event per 100 and a relative reduction of 50%; those descriptions communicate different information. This is an arithmetic example, not a result from a nutrition trial.

Then compare the authors' conclusion with the result. Does the discussion concern the population actually studied? Does it distinguish a measured association from an intervention effect? Does it explain limitations that could change the interpretation?

For each paper, record five points: the question; the main finding in its original units; two strengths of the methods; the main uncertainty; and what further evidence would help. Include the DOI or original link so that another reader can check your account.

For a journal club, give each participant one task: explain the question, examine the diet measure, assess the comparison or trace a claim back to the results. Finish by agreeing on one conclusion the paper supports and one it does not resolve.

Use reporting guidance to identify missing information, and our resource library to find the original sources.