Reviewer #4: The authors are still failing to take the problem of heterogeneity seriously. The latest response was "well, yes, the results are all over the place, but most results are pretty good". First off, it is not at all clear to this reviewer that a negative predictive value of 85%, the threshold used by the authors, is acceptable. Would most patients be happy to be told that they only have a 15% risk of high-grade cancer, so not to worry about it? That is approximately the risk for a PSA of 10. How many urologists would argue against giving a biopsy to a man with a PSA of 10?Second, the point about heterogeneity is absolutely not to find out where the majority of results lie. For instance, if this was a meta-analysis of complication risk after abortion, and there was gross heterogeneity, with most clinics having very low rates and a few having very high rates, I'm pretty sure the conclusion wouldn't be only that "abortion is a safe procedure". More likely, there would be something about there being some unsafe clinics, and then an analysis to determine the characteristics of those clinics (e.g. solo practice) or a call for further research to determine why some clinics have poor outcomes. The current paper reads as if the authors came up with the conclusion first and then went through the motions of reporting forest plots and I2 statistics and so on. The conclusion that "Multiparametric MRI of the prostate is an accurate test for ruling-out clinically significant prostate cancer" is simply not an appropriate reflection of the data presented by the authors.Formatted: Font: Bold Commented [AL1]: I know we'd already done it but we don't need to point that out to AV. Hopefully he'll think we've now spotted the wisdom of his words and made the change requested. Don't want to give him any reason to dig his heals in.

Background and objectivesPleural infection is a major cause of morbidity and mortality among adults. Identification of the offending organism is key to appropriate antimicrobial therapy. It is not known whether the microbiological pattern of pleural infection is variable temporally or geographically. This systematic review aimed to investigate available literature to understand the worldwide pattern of microbiology and the factors that might affect such pattern.Data sources and eligibility criteriaOvid MEDLINE and Embase were searched between 2000 and 2018 for publications that reported on the microbiology of pleural infection in adults. Both observational and interventional studies were included. Studies were excluded if the main focus of the report was paediatric population, tuberculous empyema or post-operative empyema.Study appraisal and synthesis methodsStudies of ≥20 patients with clear reporting of microbial isolates were included. The numbers of isolates of each specific organism/group were collated from the included studies. Besides the overall presentation of data, subgroup analyses by geographical distribution, infection setting (community versus hospital) and time of the report were performed.ResultsFrom 20 980 reports returned by the initial search, 75 articles reporting on 10 241 patients were included in the data synthesis. The most common organism reported worldwide was Staphylococcus aureus. Geographically, pneumococci and viridans streptococci were the most commonly reported isolates from tropical and temperate regions, respectively. The microbiological pattern was considerably different between community- and hospital-acquired infections, where more Gram-negative and drug-resistant isolates were reported in the hospital-acquired infections. The main limitations of this systematic review were the heterogeneity in the method of reporting of certain bacteria and the predominance of reports from Europe and South East Asia.ConclusionsIn pleural infection, the geographical location and the setting of infection have considerable bearing on the expected causative organisms. This should be reflected in the choice of empirical antimicrobial treatment.

Early surgery is advisable for radiologically suspected IPNB as it is frequently invasive. The pathobiology of IPNB demonstrates geographic variation. Pancreaticobiliary IPNB expresses MUC1 and is more frequently associated with invasive disease than other IPNB subtypes.