Enterococcus faecalis bacteraemia and infective endocarditis: a systematic review and meta-analysis

Rates of infective endocarditis (IE) in Enterococcus faecalis bloodstream infection (EF-BSI) vary more than 20-fold across published cohorts. We aimed to estimate the true IE prevalence, quantify the role of echocardiographic ascertainment in driving this variation, and assess the diagnostic performance of IE prediction scores.

We conducted a systematic review and meta-analysis (PubMed, Embase, Scopus, and Web of Science searched from inception to 16 August 2026; PROSPERO registration CRD420251240242) of observational studies reporting IE prevalence in confirmed EF-BSI episodes. Risk of bias was assessed with the Quality in Prognostic Studies (QUIPS) tool. Pooled prevalence was estimated with a generalised linear mixed model (logit scale). Meta-regression with echocardiographic coverage as the moderator was used to derive an ascertainment-adjusted prevalence. Diagnostic performance of the NOVA, DENOVA, and MEFIER scores was summarised by random-effects pooling.

Fifteen studies (11,157 EF-BSI episodes) were included. Crude pooled IE prevalence was 11.8% (95% confidence interval [CI] 8.1-16.8; I² = 97.3%). In studies with systematic echocardiography (k = 2), pooled prevalence was 24.2% (21.5-27.2; I² = 0.0%). Meta-regression explained up to 91.5% of between-study variance; estimated plateau IE prevalence was approximately 30%. Pooled DENOVA sensitivity was 88.7% (78.9-94.3%) and specificity 84.0% (77.8-88.7%); prospective validation yielded sensitivity 79.2% (negative predictive value [NPV] 93.0%). NOVA had high sensitivity (95.6%) but low specificity (30.7%). Certainty of evidence was low (Grading of Recommendations Assessment, Development and Evaluation [GRADE]).

Once ascertainment is accounted for, approximately one quarter of EF-BSI episodes are complicated by IE. These findings support a shift from selective to routine echocardiography, with transoesophageal echocardiography (TEE) as the definitive modality. Existing scores can prioritise and sequence imaging but do not identify a subgroup in whom it may be safely omitted.

Cassa di Risparmio di Trento e Rovereto (GRAM+ TN project).