The position of van Troostenburg de Bruyn and Dormandy on the prospective/retrospective distinction of a study seems to be that timing of subject identification is the deciding factor for which label to use. We defined our study as prospective because exposure and covariates were recorded before fractures.1 This is a proper definition that is more relevant to the common concern with retrospective studies of potential bias caused by the outcome event influencing ascertainment of the exposure.
Dormuth CR, Bassett K, Carleton B. Risk of Thiazolidinedione-Associated Fracture Should Be Appropriately Assessed—Reply. Arch Intern Med. 2010;170(2):209-210. doi:10.1001/archinternmed.2009.488