Effects of confounders and intermediates on the association of bacteraemia caused by extended-spectrum -lactamase-producing Enterobacteriaceae and patient outcome: a meta-analysis

Bacteraemia caused by Enterobacteriaceae (EB) producing extended-spectrum -lactamase (ESBL) has been associated with higher mortality compared with non-ESBL-producing (ESBL) EB bacteraemia in observational studies. We conducted a systematic review and meta-analysis of these studies to assess how adjusting for confounding in multivariate analyses affects the pooled estimate, and whether multivariate analyses that include intermediates in the causal pathway of outcome (sepsis severity and inadequate empirical therapy) have lower estimates of attributable mortality. PubMed search on 23 November 2010 followed by manually searching reference lists of included studies. Cohort studies published in English with separate mortality rates for ESBL and ESBL EB bacteraemia. Random-effects pooling of unadjusted and adjusted ORs followed by subgroup analyses to explore effects of adjustment procedures on adjusted ORs. The pooled OR for the unadjusted mortality associated with ESBL production was 2.35 (95 CI 1.902.91, I(2)42, 32 studies). The pooled adjusted OR was 1.52 (95 CI 1.152.01, I(2)32, 15 studies). Adjustment for more intermediates was associated with decreasing ORs. The pooled OR for the analyses adjusting for inadequate empirical therapy was 1.37 (95 CI 1.041.82). ESBL production in EB bacteraemia is associated with a higher mortality compared with bacteraemia with ESBL EB, although the estimate of this association is affected by adjustment procedures. Adjustment for inadequate empirical therapy leads to a reduction in ORs, indicating that higher mortality is likely to be mediated through this phenomenon.