Observational studies often make clinical practice recommendations: an empirical evaluation of authors' attitudes

Although observational studies provide useful descriptive and correlative information, their role in the evaluation of medical interventions remains contentious. There has been no systematic evaluation of authors' attitudes toward their own nonrandomized studies and how often they recommend specific medical practices.

We reviewed all original articles of nonrandomized studies published in 2010 in New England Journal of Medicine, Lancet, Journal of the American Medical Association, and Annals of Internal Medicine. We classified articles based on whether authors recommend a medical practice and whether they state that a randomized trial is needed to support their recommendation. We also examined the types of logical extrapolations used by authors who did advance recommendations.

Of the 631 original articles published in 2010, 298 (47%) articles were eligible observational studies. In 167 (56%) of 298 studies, authors recommended a medical practice based on their results. Only 24 (14%) of 167 studies stated that a randomized controlled trial (RCT) should be done to validate the recommendation, whereas the other 143 articles made a total of 149 logical extrapolations to recommend specific medical practices. Recommendations without a call for a randomized trial were most common in studies of modifiable factors (59%), but they were also common in studies reporting incidence or prevalence (51%), studies examining novel tests (41%), and association studies of nonmodifiable factors (32%).

The authors of observational studies often extrapolate their results to make recommendations concerning a medical practice, typically without first calling for a RCT.