Mohammad A Mansournia, Maryam Nazemipour
Even with sophisticated analytic methods, causal inference from observational studies is a risky business, and so their results should be critically appraised.
BACKGROUND: Observational studies are often used for assessing the causal effect of an exposure on the outcome though they suffer from several validity threats such as confounding, selection bias, measurement bias, over-adjustment bias, model misspecification/failure bias, and cognitive bias.
METHODS: We illustrate how to appraise causality studies through a recently published paper assessing the effect of tattoo on malignant lymphoma.
RESULTS: We argue how the study results should be cautiously interpreted in terms of bias in the design and analysis as well as the magnitude of the effect/impact measure values compatible with data. Also we show how a better understanding of the study results can be achieved by calculating some useful indices.
CONCLUSIONS: Even with sophisticated analytic methods, causal inference from observational studies is a risky business, and so their results should be critically appraised.