The aim of the study was to evaluate the cardiovascular risk factors associated with subclinical carotid atherosclerosis in antiretroviral therapy-naïve HIV-infected patients. The HERMES (HIV Exposure and Risk of Metabolic Syndrome) study enrolled therapy-naïve patients attending hospitals in the Italian coordination group for the study of allergies and HIV infection (CISAI [Coordinamento Italiano per lo Studio Allergia e Infezione da HIV]) in 2007. It was designed to identify metabolic syndrome (MS) and cardiovascular risk factors. The present analysis is a nested cross-sectional study with a subset of patients examined by carotid ultrasonography. Consecutive antiretroviral therapy-naïve HIV patients attending the facilities involved in the CISAI were included. Their 10-year probability of cardiovascular events was calculated using the Framingham Risk Score (FRS) and three other cardiovascular algorithms (the Global Framingham Risk Score - GFRS, 'Progetto Cuore' and 'SCORE'). Vascular age was estimated using a new model derived from GFRS and was compared with chronological age. The diagnosis of MS was based on the National Cholesterol Education Programme and International Diabetes Federation (IDF) definitions. Subclinical atherosclerosis was determined as ultrasound carotid intima-media thickness >0.9 mm. Out of 140 patients enrolled in the HERMES study by the four centres participating in the nested study, a total of 72 (51.4%) subjects, with no overt cardiovascular disease, were examined using carotid ultrasonography. The median age was 40 years, 79.2% men. The vascular age was 7.6 years higher than the chronological age. The factors associated with subclinical atherosclerosis were age (P < 0.0001), vascular age (P = 0.0002), body mass index (P = 0.003), waist circumference (P = 0.0002), MS (IDF definition, P = 0.004) and all the cardiovascular (CV) models (FRS, P = 0.01, GFRS, P = 0.002, Progetto Cuore, P = 0.018, SCORE, P = 0.03). Independent of other significant factors, waist circumference was significantly associated with pathological results (P = 0.007). The GFRS (area under the receiver-operating characteristic curves, 0.78; P < 0.001) had slightly better predictive accuracy than the other three CV models (FRS, areas under the curve [AUC] = 0.71, P = 0.003; Progetto Cuore, AUC = 0.74, P = 0.0005; SCORE, AUC = 0.77, P < 0.0001); 55% of patients at intermediate risk (6-20%) had subclinical carotid lesions. Subclinical carotid lesions had a highly significant direct association with all the CV risk predictors. The GFRS and vascular age were highly predictive. We recommend a carotid ultrasonographic examination at least among HIV patients with GFRS > or =6% or with an elevated waist circumference.