Most nutrition headlines describe a study far more confidently than the study describes itself. The gap between the two is where careers in misinformation are built — and where careful practitioners earn their credibility.
Start with the design, not the conclusion
Before reading the abstract's final sentence, find the design. A cross-sectional survey cannot establish that a food causes an outcome, no matter how large the sample. A randomised trial can, but only for the population, dose and duration actually studied.
Ask who was studied
A trial in 24 young male athletes tells you very little about a 68-year-old with chronic kidney disease. Generalisability is not a technicality; it decides whether the finding belongs in your consultation at all.
Look for the effect size, not just the p-value
"Statistically significant" answers whether an effect is likely to be real. It does not answer whether the effect is large enough to matter. A 0.4 kg difference over twelve months is real and clinically irrelevant at the same time.
Check what was measured
Self-reported dietary intake carries substantial measurement error. Surrogate endpoints — a biomarker moving in the right direction — are not the same as the outcome patients care about.
Read the funding and conflicts
Industry funding does not invalidate a study, but it belongs in your weighting, especially where the outcome measured is unusually favourable to a single product.
A working checklist
- What was the design, and what can it support?
- Who was studied, and for how long?
- How large was the effect in absolute terms?
- Was the outcome meaningful or a surrogate?
- Who funded it, and were conflicts declared?
- Does it agree with the wider body of evidence?
If a study fails several of these, it is not evidence you should be changing practice on — it is a hypothesis someone else needs to test properly.
#critical appraisal #evidence #research methods