Periods off cART with a duration of >90 days were omitted from the primary analysis. A new cART regimen was defined as a regimen created from an existing Entinostat ic50 regimen by the addition of one or more new antiretrovirals, or by the replacement of one or more antiretrovirals in the existing regimen with one or more new antiretrovirals. NeurocART status was assigned

to those regimens with a CPE rank of 8 or more, with the CPE rank calculated using the 2010 rankings process [17]. CD4 cell counts and viral loads were taken as the latest measurement from up to 90 days prior to regimen commencement. HIV viral load measurements of ≤400 copies/mL were defined as undetectable because more sensitive assays were not uniformly available for all observations. Coinfection with hepatitis

B virus (HBV) or hepatitis C virus (HCV) was defined as the detection of HBV surface antigen or HCV antibody, respectively. A secondary composite endpoint of AIDS or mortality within 90 days of cessation of treatment was also investigated. Follow-up was calculated from the start date of each new cART regimen (or the date of cohort enrolment if later), until cessation of that cART regimen. Loss to follow-up was defined as no clinic visit in the 12 months prior to 31 March 2009 (cohort censoring date). Patients lost to follow-up were censored at their last clinic visit. We used an intention-to-continue treatment approach and ignored any changes to, or interruptions or termination of, treatment after baseline. For each new cART regimen we created a new set of baseline covariates and assessed Dabrafenib solubility dmso the risk of death on that cART regimen adjusted for those baseline covariates. Variables updated at change in cART regimen were neurocART status, Cisplatin supplier CD4 count (<50, 50–99, 100–199, 200–349 and ≥350 cells/μL, or missing), HIV viral load (≤400 or >400 HIV-1 RNA copies/mL,

or missing), prior AIDS-defining illness (ADI), cART regimen count (first, second, third, fourth or more), months of prior neurocART exposure (never, or 1–9, 10–18 or >18 months), and months of prior cART (not neurocART) exposure (never, or 1–18 or >18 months). Additional variables examined were age (<30, 30–39, 40–49 or ≥50 years), sex, mode of HIV exposure [men who have sex with men (MSM), heterosexual, injecting drug use (IDU), other or missing], HCV coinfection, HBV coinfection, and neurocART type prior to entry (naïve, cART and not neurocART, or neurocART). We also analysed the incidence of HAD. As there is some evidence that progressive multifocal leucoencephalopathy (PML) may respond better to neurocART than non-neurocART [20], PML data were also analysed. We did not have data on patients’ CD4 cell count nadirs. An administrative censoring date of 31 March 2009 was used. Univariate Cox proportional hazards models were developed for all variables.