Objectives: To determine the incidence, mode of presentation, and outcome of pulmonary embolism (PE) in patients with HIV infection. Methods: Retrospective review of clinical case records, imaging department database, and necropsy reports of patients admitted to the specialist HIV/AIDS unit at UCL hospitals from April 1993 to August 1997 in order to identify those with a diagnosis of PE. Results: During the study period there were 3792 admissions of whom 10 (0.26%) had PE. All patients with PE presented with fever, seven were dyspnoeic, and seven had cough: all were thought initially to have respiratory infection. Only five patients had pleural pain. All 10 patients had abnormal baseline chest radiographs. The diagnosis in six was made by computed tomograph (CT) pulmonary angiography, in two was made by ventilation perfusion (V/Q) scanning, in one by both techniques, and in one at necropsy. CT angiography in addition to identifying thrombus also showed concomitant lung parenchymal abnormalities in all but one patient. Nine patients with PE had one or more risk factors for venous thromboembolism as did 34/40 case matched controls (odds ratio = 1.67; 95% confidence interval = 0.18-15.5). All patients diagnosed in life were anticoagulated and five survived. Conclusions: PE was uncommon in this HIV infected population. The diagnosis should be considered in patients with respiratory infection which does not respond to antibiotics. Identifiable risk factors for venous thromboembolism appear to be unhelpful in increasing clinical index of suspicion for PE. As baseline chest radiographs are frequently abnormal, the diagnostic utility of V/Q scanning may be reduced and CT pulmonary angiography is the imaging modality of first choice.