Hidden Heart Infections: Landmark Analysis Finds Endocarditis in a Quarter of Enterococcal Bloodstream Cases

One of the most stubborn puzzles in infectious disease medicine has just been cracked open. For decades, clinicians have argued over how often a common gut bacterium, Enterococcus faecalis, silently attacks the heart after slipping into the bloodstream. Published studies have reported rates of infective endocarditis in E. faecalis bloodstream infection ranging from a seemingly reassuring 1 percent to an alarming 26 percent — a more than twentyfold spread that has left physicians unsure whether the infection is rare or dangerously underdiagnosed. Now, the first systematic review and meta-analysis of the question, published in eClinicalMedicine, concludes that the discrepancy is largely an illusion created by how hard doctors have looked, and that the true burden of heart infection is far higher than routine practice figures suggest.

The research team, led by Giovanni Mori and Nicola Susca, searched PubMed, Embase, Scopus, and Web of Science from their inception through August 2026, screening 1,274 unique records and ultimately pooling data from 15 observational studies spanning 1995 to 2024. Together, these studies covered 11,157 episodes of microbiologically confirmed E. faecalis bloodstream infection, including 1,483 cases of infective endocarditis. The analysis was conducted under the PRISMA 2020 reporting framework and prospectively registered in PROSPERO, and the authors report no external funding or competing interests.

The headline number from the raw pooling — a combined prevalence of 11.8 percent, with a 95 percent confidence interval of 8.1 to 16.8 percent — is, paradoxically, the least meaningful figure in the paper. Between-study heterogeneity was extreme, at an I² of 97.3 percent, meaning the individual studies disagreed with one another to an almost total degree. In most meta-analyses, such heterogeneity would render the pooled estimate clinically uninterpretable. Here, however, the investigators suspected the disagreement had a specific, measurable cause: the intensity of echocardiographic ascertainment, that is, the proportion of patients in each cohort who actually received cardiac ultrasound imaging.

Echocardiography occupies a peculiar dual role in this field. It is the diagnostic gold standard for infective endocarditis, capable of revealing the vegetations — masses of bacteria, fibrin, and inflammatory cells — that anchor themselves to heart valves. Yet it is also the principal source of detection bias in prevalence estimates, because a definite diagnosis under the Duke criteria is rarely possible without imaging evidence. Studies that scanned every patient found high rates of endocarditis; studies that scanned selectively, or relied on administrative coding, found low ones. Whether the low figures reflected genuinely less disease or simply missed disease had never been formally tested until now.

The meta-regression analysis delivered a striking answer. When the proportion of patients undergoing transoesophageal echocardiography (TEE) was used as the statistical moderator, a quadratic model explained up to 91.5 percent of the between-study variance — an extraordinary figure for study-level epidemiology. The fitted curve rose steeply with increasing imaging coverage and then flattened, projecting a plateau prevalence of 30.4 percent (95 percent CI 23.1 to 38.9) at a TEE coverage of about 58 percent. An analogous model using transthoracic echocardiography (TTE) coverage explained 87.4 percent of the variance and projected a plateau of 29.4 percent. In plain terms: once doctors image nearly everyone, roughly three in ten patients with E. faecalis in their blood turn out to have endocarditis.

Two independent lines of evidence converge on this conclusion. The two cohorts in the review that applied systematic echocardiography to all patients — a Danish prospective study and a multinational prospective cohort — reported closely concordant prevalences of 26.2 and 23.0 percent, with zero heterogeneity between them. The regression-derived plateau of approximately 30 percent matches these direct observations almost exactly. The relationship also held under both the older modified Duke 2000 criteria and the newer 2023 Duke–ISCVID criteria, indicating a structural feature of the literature rather than an artefact of more inclusive definitions. Notably, in the multinational EfFAECT cohort, applying the 2023 criteria rather than the 2000 criteria reclassified definite endocarditis from 13.3 percent to 23.0 percent in the very same patients with the very same imaging — showing that the criterion effect and the ascertainment effect are separable.

The clinical implications are immediate. An ascertainment-adjusted prevalence of 24 to 30 percent places E. faecalis bloodstream infection in the same territory as, or above, Staphylococcus aureus bacteraemia, for which a formal decision analysis estimated that echocardiography becomes the survival-optimal strategy once the probability of occult endocarditis exceeds roughly 1.1 percent. The E. faecalis figure exceeds that benchmark by more than an order of magnitude. Current European Society of Cardiology guidelines assign echocardiography in E. faecalis bacteraemia only a Class IIa recommendation, meaning it should be considered. The new data, the authors argue, support moving from a selective to a routine recommendation — with the important caveat that TTE alone is insufficient, since the prospective Danish cohort found TTE missed approximately half of vegetations compared with TEE.

The review also delivered the first pooled diagnostic accuracy estimates for the clinical prediction scores designed to triage which patients need imaging. The DENOVA score, at its standard threshold, achieved a pooled sensitivity of 88.7 percent and specificity of 84.0 percent across seven studies, but its performance degraded under scrutiny: in the largest prospective validation to date, an international cohort of 543 patients undergoing protocol echocardiography, DENOVA missed 26 of 125 endocarditis cases, yielding a sensitivity of just 79.2 percent and a negative predictive value of 93.0 percent. The NOVA score combined near-perfect sensitivity of 95.6 percent with dismal specificity of 30.7 percent, classifying so many patients as high-risk that it spares almost no one an scan. The MEFIER score, assessed in only two studies, showed a pooled sensitivity of 54.5 percent, disqualifying it as a screening tool. Critically, the authors calculate that even a negative DENOVA result still leaves a 7 to 10 percent probability of endocarditis at the derived prevalence — far above any defensible threshold for omitting imaging.

The analysis also consolidated the clinical predictors of endocarditis in this population. The most robust factors, each with zero heterogeneity across studies, were the presence of a prosthetic heart valve (odds ratio 4.78) and community-acquired infection (odds ratio 3.11). Previous endocarditis and an unknown portal of entry were also positively associated, though with wider uncertainty. Single-cohort analyses additionally linked a heart murmur on auscultation, symptom duration of at least seven days, embolic events, a higher number of positive blood cultures, immunosuppression, monomicrobial bacteraemia, and prior E. faecalis bacteraemia to elevated risk, while nosocomial acquisition and a urinary source were associated with lower risk. These predictors overlap heavily with the components of NOVA and DENOVA, but the authors caution that odds ratios cannot be converted into individual post-test probabilities and cannot identify a subgroup in whom imaging can safely be skipped.

Limitations temper the conclusions. The systematic-imaging subgroup comprised only two cohorts, and the regression plateau is anchored by few high-coverage studies; coverage also cannot capture imaging quality or operator expertise, and it correlates with prospective design and contemporary criteria, so imaging cannot be fully isolated from these covariates. None of the selective-echo studies compared outcomes between imaged and unimaged patients, and the frequency of endocarditis among unimaged patients is largely unobserved rather than merely unreported. Individual patient data were unavailable, sex-disaggregated analyses could not be pooled, and no cost-effectiveness analysis or randomised comparison of imaging strategies exists for this infection. Formal certainty of evidence under the GRADE framework ranged from low to moderate. Even so, the triangulated conclusion stands: once ascertainment is accounted for, infective endocarditis complicates roughly one quarter of E. faecalis bloodstream infections. The authors call for prospective derivation of a new, more sensitive prediction rule calibrated to the Duke–ISCVID 2023 criteria — one that might incorporate prosthetic valves, prior endocarditis, intracardiac devices, and blood-culture time-to-positivity — but until such a rule is validated, their message to clinicians is unambiguous: in enterococcal bacteraemia, what the heart has not been shown, the heart may still be hiding.

Subject of Research: Prevalence of infective endocarditis in Enterococcus faecalis bloodstream infection and the role of echocardiographic ascertainment

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

Article References: Mori, G., Castelli, A., Cordin, A., Lorenzin, G., Bonmassari, R., Onorati, F., Maino, A., Lanzafame, M., Racanelli, V., & Susca, N. (2026). Enterococcus faecalis bacteraemia and infective endocarditis: a systematic review and meta-analysis. eClinicalMedicine, 101, Article 104241. https://doi.org/10.1016/j.eclinm.2026.104241

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104241

Keywords: Enterococcus faecalis, infective endocarditis, bacteraemia, echocardiography, meta-analysis, systematic review, Duke-ISCVID criteria, DENOVA score, NOVA score, bloodstream infection, transoesophageal echocardiography, prevalence