T levels in archived serum were measured using liquid chromatography tandem mass spectrometry (LC-MS/MS). The LC-MS/MS system includes equipment for extraction of serum samples and automated sample loading, as well as a cohesive high-turbulence liquid chromatography system, and a Thermo-Finnigan Quattro

Tandem Mass Spectrometer (Thermo Fisher Scientific Inc., Waltham, GDC-0199 mouse MA, USA). This assay has a sensitivity of 2 ng/dL, and the interassay coefficient of variation ranged from 3.3 to 7.7%. Sex-hormone binding globulin (SHBG) was measured using radioimmunoassay (RIA). The SHBG assay uses a two-site directed immunofluorometric assay with a sensitivity of 2.5 nmol/L and is highly specific, with less than 0.1% cross-reactivity with known circulating proteins. Finally, FT was calculated Inhibitor Library from total T and SHBG measurements using the Vermeulen equation [17]. Additional variables of interest included race, body

mass index (BMI), the presence of hypertension, current smoking status and the use of lipid-lowering agents. These variables were collected from semi-annual interview data. Fasting specimens for high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, insulin and glucose to calculate homeostatic model assessment of insulin resistance (HOMA-IR) [18] were collected. In HIV-infected individuals, CD4 cell count and viral load were collected. Current and prior exposure to the major antiretroviral classes, i.e. protease inhibitors (PIs), nonnucleoside reverse transcriptase inhibitors (NNRTIs) and nucleoside reverse transcriptase inhibitors (NRTIs), was collected from semi-annual data. Current and prior exposure to individual Non-specific serine/threonine protein kinase antiretroviral agents was also collected. This was a cross-sectional analysis of data from a single visit. We compared demographic and clinical characteristics of participants with and without HIV infection, using χ2 tests to compare counts and prevalence measurements.

To compare continuous variables, we used the two-sample t-test or Wilcoxon rank sum test, depending on the distribution of the variables. We first constructed two multivariable logistic regression models using data for all participants, with CAC presence and carotid lesion presence as the two outcomes being modelled and log-transformed FT as the primary independent variable in each model. Other covariates included in the models were HIV status, age, BMI, race (Black vs. other), smoking, HDL cholesterol ≤40 mg/dL, LDL cholesterol ≥ 160 mg/dL, HOMA-IR, hypertension, lipid-lowering agent use and clinic site. We then modelled the relationship between carotid IMT and FT using a multivariable linear regression model. The dependent variable for the model was log-transformed IMT, and the primary independent variable was log-transformed FT. We included HIV status, age, BMI, race (Black vs.