98) among those who started in the recent period, compared with the early period. HIF inhibitor Patients who had a previous history of injecting drug use (IDU) (of whom 65% had been enrolled in the early period) had almost a threefold increased risk of discontinuation because of poor adherence compared with those who were infected with HIV by heterosexual intercourse (ARH

2.85; 95% CI 1.89–4.30, P<.0001). Female gender (ARH 1.42, 95% CI 1.07–1.89 vs. male gender; P=0.01) and a higher CD4 cell count at baseline (ARH 1.08, 95% CI 1.02–1.14 per 100 cells/μL higher; P=0.002) were independently associated with a higher risk of discontinuation because of poor adherence. We observed a tendency towards a higher rate of discontinuation because of poor adherence in patients younger than 30 years compared with those aged 30–45 years (ARH 1.34, 95% CI 0.97–1.84, P=0.07) (Table 3). The results of the model were similar when we analysed separately patients

who received coformulated boosted PI (72% of those who started a boosted PI) and those Buparlisib cell line who received ritonavir and another PI as a separate drug (data not shown). The Kaplan–Meier estimates of discontinuation by 1 year because of immunovirological and clinical failure were about 60% lower in patients who started HAART recently (3.4%; 95% CI 1.9–4.9%) and in the intermediate period (2.4%; 95% CI 1.3–3.4%) compared with those who started in 1997–1999 (5.5%; 95% CI 4.3–6.6%) (log rank P=0.0013) (Fig. 1). In the multivariable model, we observed a significant decline

in the incidence of discontinuation because of failure in patients who started HAART in 2000–2002 (ARH 0.46, 95% CI 0.26–0.82, P=0.008 vs. 1997–1999) and, surprisingly, comparable rates of discontinuation because of failure between the early and recent periods (ARH 0.81, 95% CI 0.40–1.63, P=0.57). The number of CD4 T cells at HAART initiation showed an independent association with this outcome (ARH 0.88, 95% CI 0.80–0.97, per 100 cells/μL higher; P=0.01) (Table 3). Fifty-seven patients out of 101 (56%) who discontinued because of virological failure had viral load >500 copies/mL at the time of switching, five of 11 (45%) who discontinued because of immunological failure had an increase in CD4 cell count of <10% from the pre-therapy value, and six of 14 (43%) categorized as having Thalidomide clinical failure had an AIDS-defining illness at the time of discontinuation. In order to validate the accuracy of the reason for discontinuation given by the clinicians, the analysis with the endpoint immunovirological and clinical failure was repeated only with those patients, and provided results that were very similar to those of the main analysis (Table 4). A significant declining trend with calendar period of HAART initiation was observed in the viral load at switch of patients who discontinued because of virological failure [1.69 log10 copies/mL (95% CI 1.69–2.75 log10 copies/mL) in patients who started HAART in the recent period, 2.37 log10 copies/mL (95% CI 2.00–4.