All studies were reviewed regardless of effect measure, study design and publication language. Studies on combinations of polysaccharide and conjugate vaccines were excluded. We searched online databases (PubMed, EMBASE and the Cochrane Library) using the following search line: HIV AND vaccine AND (pneumococcal OR pneumonia OR pneumoniae). Reference lists of studies found in the initial search were examined for studies that had not been previously identified. The outcome of interest see more was the vaccine effectiveness in preventing any of three pre-specified clinical

endpoints: all-cause pneumonia, all-pneumococcal disease and IPD. Thus, all studies that reported at least one of these endpoints for HIV-infected individuals who had or had not received PPV-23 were included. The search was conducted independently by two reviewers (RHP and OSS) and data were extracted in duplicate from 1 June 2009 to 1 March 2010. Varying definitions were accepted for the diagnosis of pneumonia (e.g. physician-diagnosed or X-ray-confirmed). For infection with S. pneumoniae and IPD there had to be a positive culture of S. pneumoniae, and for IPD the

specimen had to originate from a normally sterile site (any sterile site was accepted). Risk estimates from each study were recorded, along with other important study details (including study design, setting, statistical models used to control for potential confounding factors, baseline characteristics of study participants and study limitations). Selumetinib molecular weight For each study, the extent of confounders controlled for was tabulated and risk estimates were stratified according to clinical endpoints. Plots of vaccine effectiveness

were produced for all clinical endpoints. Further, we stratified study subjects as controls vs. cases in case–control studies, and as vaccinated vs. unvaccinated Tacrolimus (FK506) in other study types and tabulated major risk factors for pneumococcal infection (treatment, race, smoking and CD4 cell count) to look for trends in the risk estimates. In one instance, the same study was described in two publications [14,35]. Both publications were examined in order to retrieve maximal information. In another instance, the risk estimate was published without a confidence interval, which had to be calculated from the P-value [36]. We included one randomized trial and 15 observational studies in our review. The randomized trial had data on the three outcomes of interest. Seven observational studies reported data on all-cause pneumonia, six on S. pneumoniae infection and six on IPD. This randomized, double-blind, placebo-controlled trial took place in Uganda from 1995 to 1998 and initial results were published in 2000 [14]. A subsequent report included an additional 3 years of follow-up and was published in 2004 [35]. The trial was conducted among 1323 Ugandan adults not receiving HAART. Participants were randomized 1:1 to immunization with a single dose of PPV-23 or placebo inoculation (buffered sodium phosphate).