On bias.

We performed a search in Medline to assess the quality of clinical journals in orthopedics from the point of view of study design. 3 levels of quality were chosen: prospective studies, random allocation or double-blind methods and randomized controlled trials; all entries were Medical Subject Headings (MeSH). Out of 25,538 articles indexed in Medline since 1966-1993 in the 8 most cited general orthopedic journals, 994 were indexed as prospective studies, while 138 were indexed as randomized, controlled trials. In recent years the number of well-designed articles has increased, as has the percentage. In a random check of 208 articles, approximately half were of a clinical type where these issues can be addressed. The agreement between the manual check and Medline indexing was good, but not perfect. It was concluded that the retrospective study representing clinical production control accounts for the vast majority of all published clinical articles in orthopedics. In recent years, a sharp increase in controlled trials had, however, occurred.