METAANALYSIS OF RANDOMIZED CONTROLLED TRIALS OF SELECTIVE DECONTAMINATION OF THE DIGESTIVE-TRACT

Objective-To determine the clinical benefits of selective decontaminat ion of the digestive tract in patients treated in intensive care units . Design-Meta-analysis of 22 randomised trials that compared different combinations of oral non-absorbable antibiotics, with or without a sy stemic component, with no treatment in controls. Subjects-4142 patient s seen in general and specialised intensive care units around the worl d. 2047 received some form of antibiotic treatment, the remainder no p rophylaxis. Data analysis-Each trial was reviewed through direct conta ct with study investigators. Data collected were: the randomisation pr ocedure, number of patients, number excluded from the analysis, and nu mbers of respiratory tract infections and deaths. Data were combined a ccording to an intention to treat analysis with the Mantel-Haenszel-Pe to method. Main outcome measures-Respiratory tract infections and tota l mortality. Results-Selective decontamination of the digestive tract significantly reduced respiratory tract infections (odds ratio 0.37; 9 5% confidence interval 0.31 to 0.43). The value of the common odds rat io for total mortality (0.90; 0.79 to 1.04) suggested at best a modera te treatment effect, reaching statistical significance only when the s ubgroup of trials of topical and systemic treatment combined was consi dered separately (odds ratio 0.80; 0.67 to 0.97). No firm conclusions could be drawn owing to large variations in patient mix and severity w ithin and between trials. Conclusions-The findings strongly indicate t hat selective decontamination significantly reduces infection related morbidity in patients receiving intensive care. They also highlight wh y definite conclusions about the effect of prophylaxis on mortality ca nnot be drawn despite the large number of trials available. Based on t he most favourable results obtained by pooling data from trials in whi ch combined topical and systemic treatment was used it may be estimate d that 6 (range 5-9) and 23 (13-139) patients would need to be treated to prevent one respiratory tract infection and one death respectively .