91 (95% CI 0.83–1.00); P = 0.040]. The proportional hazards assumption was not violated. Overall, 495 patients were diagnosed with TB in the first year after ART initiation during 5728 person-years of follow-up. The overall incidence rate was 8.6/100 PYAR (95% CI 7.9–9.4 PYAR). The incidence

MK-1775 ic50 rate fell from 8.2/100 PYAR (95% CI 7.0–9.5 PYAR) in 2005 to 6.5/100 PYAR (95% CI 5.3–8.1 PYAR) in 2007, and then rose in 2008 (9.5/100 PYAR (95% CI 7.6–11.9 PYAR)] and 2009 [15.6/100 PYAR (95% CI 12.4–19.7 PYAR)]; the log-rank test for equality of survivor functions was P = 0.003. The cumulative incidence of TB in the first year after ART initiation is depicted in Figure 3. The proportional hazards assumption of the multivariable Cox BMS-777607 proportional hazards model was violated. We therefore stratified our analysis for the years 2005, 2006 and 2007 versus 2008 and 2009, based on the difference in TB incidence. The two models showed the same covariates to be significantly associated with the outcome with similar HR estimates, and visual inspection of the curves also showed a great similarity between the two periods. A multivariable

Cox proportional hazards model was therefore run on all data and showed lower baseline CD4 cell count and male sex at ART initiation to be significantly associated with TB incidence in the first year (Table 3). Patients initiating ART later were more likely to be diagnosed with TB in the first year of ART initiation

[HR per year of later ART initiation, 1.13 (95% CI 1.05–1.21); P = 0.001]. This is one of the first studies to relate decreasing mortality rates in ART initiators to changing patient characteristics and improved TB case finding Teicoplanin after the rapid scale-up of ART in East Africa. In our large urban HIV-infected cohort, baseline CD4 cell counts increased significantly over time, which was associated with a decrease in mortality. A later year of ART initiation was independently associated with improved survival. Our findings show that major programmatic changes are possible in resource-limited settings and that these are associated with a measurable effect on all-cause mortality. There are some published data on changing CD4 cell counts over time since the roll-out of large-scale antiretroviral therapy in the developing world. The ART-LINC of IeDEA (ART in Lower Income Countries collaboration of the International Databases to Evaluate AIDS) cohort study, reporting on 17 ART programmes and 36 715 patients initiating ART in 12 countries in sub-Saharan Africa, South America and Asia, showed increasing median baseline CD4 counts between 2001 and 2006, with the lowest CD4 counts for the sub-Saharan African region (60 cells/μL in 2001 increasing to 122 cells/μL in 2006) [19]. In studies specifically looking at sub-Saharan Africa, most data originate from South Africa, where this trend has also been noted [20, 21].