Issues in reporting epidemiological studies: No data are given to support generalisationBMJ 2005; 330 doi: https://doi.org/10.1136/bmj.330.7483.146-b (Published 13 January 2005) Cite this as: BMJ 2005;330:146
- David F Blackburn, assistant professor ()
EDITOR—The article by Pocock et al outlined deficiencies in epidemiological research.1 Undoubtedly, the deficiencies they reported will be generalised to all observational studies, even though they specifically excluded papers examining treatments or patients with diseases. Even the accompanying editorial revisits the hormone replacement therapy debacle.2 Interestingly, a paper about hormone replacement therapy would not have been eligible for this review as studies using pharmacological treatments were excluded. However, the authors dedicate only two paragraphs to the methods section so this fact may not be obvious at first glance, or not at all if the abstract is reviewed in isolation.
Before this study is used as another reason to discredit all observational research, its limitations should be appreciated. The authors say that clinical epidemiology in people with disease and studies in pharmacoepidemiology raise similar issues,1 but they provide no data to support this generalisation. On the basis of their inclusion and exclusion criteria, they would have examined studies such as those linking coffee and diabetes.3 I am neither surprised nor concerned that papers such as this may contain some residual confounding, and I think it is a reasonable expectation that healthcare professionals interpret these reports appropriately. Often all that is required is to read beyond the abstract.
Competing interests None declared.
Log in using your username and password
Log in through your institution
Sign up for a free trial