e. doubling baseline CD4 count in those with baseline counts of 300–499, and >1000 cells/μL if baseline was ≥500 cells/μL. Demographics and HIV clinical and treatment history were documented at baseline. Thereafter, patients were seen every 4 months for the study duration, and information was captured on standardized case report forms (CRFs). Events were reported

using specific CRFs with supporting source documentation as soon as sites became aware of them. Criteria for a confirmed bacterial pneumonia event during follow-up included clinical, radiographic and microbiological evidence; a probable bacterial pneumonia event required clinical and radiographic evidence or diagnosis by doctor, physicians’ assistant or nurse practitioner without microbiological evidence. For a diagnosis of recurrent

bacterial pneumonia, both pneumonia episodes had to occur after enrolment and satisfy the criteria above with the additional requirements; GSK126 clinical trial i.e. the second pneumonia episode had onset of symptoms <365 days after the first episode and there was strong evidence that the first episode was resolved, such as an intervening clear chest this website x-ray or absence of symptoms after >1 month off antibacterials effective against pathogens commonly producing pneumonia. All endpoints, including the initial episode of bacterial pneumonia, were reviewed by the Endpoint Review Committee (ERC) blinded to treatment group against predetermined criteria as described above and designated as confirmed/probable or did not meet the criteria for an endpoint. CD4 cell count closest to the event and randomization arm were redacted prior to ERC review. Only bacterial pneumonia events designated by

the ERC as confirmed or probable were included in this analysis. Multivariate proportional hazards regression models were used to compare the treatment groups (IL-2 and control) and to summarize associations between baseline and time-updated factors and bacterial pneumonia – defined as the first episode of confirmed or probable bacterial pneumonia following randomization. The comparison of treatment groups was intention to treat. The proportional hazards assumption was examined by including an interaction Sirolimus mouse term between the treatment indicator and log-transformed failure time. Baseline predictors included age, gender, ethnicity, IDU, hepatitis B and/or C virus coinfection, nadir and baseline CD4 cell count, viral load (VL), prior ADI, prior recurrent bacterial pneumonia as an ADI, and PcP prophylaxis; time-dependent covariates updated during follow-up included proximal CD4 cell count, i.e. the CD4 cell count closest to the event, and VL, incident ADI, and time since rIL-2 receipt. Smoking and pneumococcal vaccination histories were not considered in the model as these data were not collected in ESPRIT. Statistical analyses were performed using sas software, version 9.1 (SAS Institute, Cary, NC, USA). P-values are two-sided.