ABSTRACT
Epidemiological studies are rarely as convincing as clinical trials in attributing a causal mechanism. In a well designed and well performed clinical trial, a significant treatment effect can, with some confidence, be said to be caused by the different treatments patients received. With an epidemiological study, attributing causality is far more problematic and would normally be claimed only after a range of studies carried out in a variety of settings had all found, say, the same relationship between some risk factor of interest and some particular disease. Investigators tempted to ascribe causality on the basis of a single, perhaps relatively small epidemiological study should think again.
