A practical framework for research question, design, outcomes and conflicts of interest.
Start with the protocol
Before the results, identify the research question, population or model, intervention, comparator and duration. For registered trials, compare published endpoints with the original registration. This reduces the risk of giving excessive weight to later analyses.
Read absolute, not only relative, quantities
Examine sample size, losses to follow-up, confidence intervals and handling of missing data. Statistical significance alone does not describe the size or certainty of an effect. Prespecified and exploratory analyses should be distinguished.
Funding and replication
Funding and author relationships do not automatically invalidate a study, but they are essential context. Look for an available protocol, statistical analysis plan and independent replication. A strong conclusion rests on the totality of evidence, not one striking result.