In the August 10/24 issue of the Archives, Dormuth et al1 reported an increased fracture rate in provincially insured residents of British Columbia, Canada, exposed to thiazolidinediones between January 1998 and December 2007. We wish to comment on several statements in the article.
Dormuth et al1 described their study as a prospective cohort study, when in fact it was a retrospective database analysis. It is well recognized that prospective studies offer a higher probability of detecting true associations than retrospective studies because they allow greater control to be exercised over confounding factors and data collection. We do not mean to question the validity of the results of their analysis but to suggest that the data should be interpreted in the context of the retrospective study design used.