Unusually High Incidences of Pseudomonas Bacteremias Within Topical Polymyxin-Based Decolonization Studies of Mechanically Ventilated Patients: Benchmarking the Literature

Abstract

Background: Topical polymyxin (PM)-based regimens to decolonize patients receiving prolonged mechanical ventilation (MV) have been widely studied. However, paradoxical bacteremia incidences remain unexplained. Methods: The literature was searched for studies of topical PM-based regimens used to decontaminate MV patients reporting incidences of overall and Pseudomonas bacteremia data. In addition, observational groups without any intervention and trials of various interventions other than topical PM (non-PM studies) served to provide external benchmarks and additional points of reference, respectively. The bacteremia incidences were extracted from the control and intervention (component) groups of these studies and compared with metaregression using generalized estimating equation methods. Results: The summary odds ratio derived from studies of topical PM-based interventions against overall bacteremia was 0.60 (95% confidence interval [CI], 0.53-0.69). Benchmark incidences per 100 MV patients for overall (mean, 8.9%; 95% CI, 6.9% to 10.9%) and Pseudomonas (mean, 0.7%; 95% CI, 0.5% to 1.1%) bacteremia were derived from 16 observational studies. By contrast, among 17 studies of topical PM, the mean incidences among control groups for overall (mean, 15.3%; 95% CI, 11.5% to 20.3%) and Pseudomonas (mean, 1.6%; 95% CI, 0.9% to 3.1%) bacteremia were both higher, whereas these incidences in the intervention groups for both topical PM and non-PM studies were in each case more similar to the respective benchmarks. These paradoxical incidences cannot readily be explained in metaregression models. Conclusions: Paradoxically, despite an apparent prevention effect of topical PM-based methods against bacteremia overall, the incidences of Pseudomonas bacteremia within the component groups of these studies are unusually high vs literature-derived benchmarks.