The predictive value of a discharge diagnosis of PE in administrative databases has previously been reported to be 80–90%, and somewhat lower for deep venous thrombosis [42–45]. Up to 10–20% of VTE cases listed in Scandinavian hospital discharge registries therefore may be misclassified [42], and this lack of specificity may have biased our results. However, as we used the same source of data to ascertain VTE for all study subjects, we presume that any potential misclassification

was nondifferential and buy BGJ398 therefore did not influence our estimates of relative risk. HIV-infected patients usually have frequent hospital contacts, so we cannot exclude the possibility that, because they are monitored more closely than individuals in the general population, they may be more prone to be diagnosed with VTE. We used previously developed models to

stratify the results by provoked vs. unprovoked VTE [34,35]. The specificity of classifying VTE as provoked/unprovoked has been described as high, given the validity of the cancer diagnosis and surgical procedure models used find more to define provoked VTE [46]. Although our results were adjusted for several risk factors for VTE, we did not have access to information on all the classic risk factors for a hypercoagulable state, including use of oral contraceptives, postmenopausal hormone replacement, immobility as a result of acute medical illness and family history of VTE. We did adjust the risk of VTE for obesity, based on a discharge diagnosis of this condition, but the validity of this diagnosis selleck products seems questionable. HAART, particularly treatment with protease inhibitors (PIs), has previously been posited as a risk factor for VTE [13,16]. This risk has been ascribed to a PI-induced abnormality in platelets or endothelium [13]. However, the association between HAART and risk of thrombosis may arise

from mutual associations with other risk factors, such as advanced stage of disease [12]. Of note, three studies have found no association between HAART and VTE [14,17,18]. Our data showed that HAART nearly doubled the risk of overall VTE in non-IDU HIV-infected patients. In contrast, risk of VTE did not increase after HAART initiation in the IDU group. It is probable that IDU patients receiving HAART are less affected by their drug abuse and thereby at decreased risk of VTE. It has been suggested that alterations in several thrombophiliac components correlate with HIV-induced immunodeficiency and thereby with a low CD4 cell count [16,25–27]. The association between free protein S deficiency and CD4 cell count has been observed most consistently, but the clinical significance of this association remains controversial [47]. The increased risk of VTE in sick HIV-infected patients with low CD4 cell counts also might stem from immobilization, as suggested by Saif [16]. Ahonkai and Saif et al.