Risk of VTE among users of oral contraceptives.

contraceptives We have recently reviewed two studies,1 a cohort study conducted in Denmark,2 and a case-control study conducted in The Netherlands,3 in which it was claimed that the risk of venous thromboembolism (VTE) among users of oral contraceptives (OCs) containing desogestrel, gestodene, drospirenone and cyproterone is greater than among users of levonorgestrelcontaining OCs. We concluded that in both studies the comparisons among the progestogens were not valid due to methodological limitations. The Danish study linked prescription data recorded in one national registry to hospital discharge diagnoses of VTE recorded in another registry. The investigators stated that in an earlier validation study 10% of the diagnoses documented between 1994 and 1998 “were uncertain”. In the study under review they acknowledged that they relied on the “final discharge diagnoses as reported”, and that they were unable to “evaluate the validity of each included diagnosis of [VTE]”.2 Since publication of our review new information has come to light that bears on the validity of the registry-recorded diagnoses. In a cohort study that included 27 178 men and 29 876 women aged 50–64 years, Severinsen and her colleagues examined the medical records of 1100 cases of registry-recorded VTE.4 The diagnosis was incorrect in 25% of cases diagnosed in hospital wards, and in 69% of cases diagnosed in emergency departments; the latter cases constituted 41% of the total. Incorrect diagnoses were more commonly recorded among women than among men. A stratified analysis did not show an impact of age on diagnostic precision. It is difficult to reconcile the findings of Severinsen et al. with the assumption that the diagnosis was uncertain in about 10% of the cases of VTE,2 even though that estimate was made among women of fertile age. Based on the wording used by the authors it can be assumed that the VTE incidence rates among the compared OCs were based on all VTE diagnoses – including VTE diagnosed in emergency departments. If so, Severinsen’s results suggest that the diagnosis was not only uncertain, but in at least 40% of the cases it was wrong. If the analysis was based only on hospital ward cases, the diagnosis was incorrect in about 29% of the female patients. Relative to levonorgestrel the relative risks for the compared OCs were small (<2), and the major diagnostic imprecision suggested by Severinsen’s data would be sufficient to nullify the findings. It obliges Lidegaard to verify the diagnoses in his study.