Stepwise forward selection was used to select independent predictors of the event occurrence, with 0.50 and 0.15 as P-values for entry into the model and being retained in the DAPT model, respectively. Known recorded risk factors for the SNA events were forced into the model [25]. Thus, smoking status, diabetes mellitus and hyperlipidaemia were forced into the cardiovascular events model;

hyperlipidaemia, HBV and HBC coinfections and alcohol abuse were forced into the model for terminal liver conditions; and smoking status was forced into the non-AIDS malignancies model. All of the former factors were forced into the model that estimated risk for SNA as a composite outcome. In addition, the indicator of ever received antiretroviral treatment was always forced into the models because all the variables associated with antiretroviral treatment were defined as interactions; i.e. 0 or missing if never treated. The following variables were considered as potential predictors: race, mode of transmission, HIV infection history, immunological factors and exposure to antiretroviral treatment. Although age and gender

are known to be associated with most non-AIDS events, they were not included in the models see more because they were used as matching variables. As of February 2008, 6007 patients had been included in the LATINA retrospective cohort, with a mean of 3.2 years and a median of 2.5 years of follow-up. Of the 6007 patients, 30% were women and 21% had a history of AIDS-defining conditions before the baseline visit. The incidence of AIDS events was 4.7 per 100 person-years

of follow-up. A total of 130 patients had an SNA event (94 confirmed and 36 probable) and were defined as cases, with an incidence rate of 8.6 events per 1000 person-years (95% CI 7.2, 10.0). Twenty-eight of these patients (21%) were female. Forty patients (30.7%) had a cardiovascular condition [11 had an MI (five confirmed), 13 had cardiovascular disease requiring an invasive procedure and 16 had a stroke (nine confirmed); incidence of cardiovascular events: 2.2 events per 1000 person-years (95% CI 1.5, 2.9)]; 54 patients (41.5%) had liver failure/cirrhosis (34 confirmed) [incidence: 2.9 events per 1000 person-years (95% CI 2.1, 3.7)]; 35 patients (27%) had a non-AIDS-defining malignancy (34 confirmed) Astemizole [incidence 1.9 events per 1000 person-years (95% CI 1.2, 2.5)] and two (1.5%) had terminal renal insufficiency (both confirmed). One patient experienced simultaneously a liver failure and a cardiovascular disease. The median time of follow-up until the index date for cases and controls was 1.42 and 2.45 years, respectively (P=0.12; univariate conditional logistic regression). Table 1 compares the general characteristics of all cases and controls. The frequency of injecting drug use was significantly higher in the cases (P=0.001), as were the frequencies of histories of some traditional risk factors such as HCV coinfection (P<0.