Introduction

Infective endocarditis (IE) is a potentially life-threatening syndrome with an in-hospital mortality rate of approximately 20% despite optimal medical and surgical management [1–3]. Echocardiography is an essential tool in the diagnosis of IE, but also in the identification of heart conditions that predispose to its development [4–7]. The designation of a predisposing heart condition is nonspecific for left-sided native valve endocarditis (LS-NVE) and may include a myriad of congenital and acquired conditions such as congenital abnormalities, rheumatic heart disease, hypertrophic cardiomyopathy, bicuspid aortic valve, or mitral valve prolapse [8–13].

Considering the: (i) nonspecific classification of predisposing cardiac conditions associated with the development of LS-NVE; (ii) fact that 40% of patients with NVE do not have a structural abnormality or valve damage prior to the development of IE [14]; and (iii) evolution of technical advances in echocardiography over recent years, it seems appropriate to evaluate echocardiography as a potential tool in assessing the likelihood of the subsequent development of LS-NVE. Thus, the primary objective of the current investigation was to determine whether transthoracic echocardiography (TTE) measurements were associated with the subsequent development of LS-NVE in patients with no predisposing cardiac conditions.

Materials and methods

A case-control study design was conducted. Cases and controls were adult (≥ 18 years) patients from Mayo Clinic sites (Rochester, Florida, Arizona, and the Health Systems) from 2008 to 2020. To determine the cases, institutional databases of patients with an initial diagnosis of LS-NVE were queried. Medical charts were reviewed, and IE cases were identified according to the modified Duke criteria [10], and categorized into either possible or definite LS-NVE. Only patients who had a TTE performed 6 months to 3 years before the development of IE were considered for study inclusion, and every individual was only included once. Controls were selected from a large retrospective cohort of adult patients with at least one TTE report over a similar period to that of the cases. Patients with prosthetic valves, cardiovascular implantable electronic devices (CIEDs), intracardiac devices, previous IE, injection drug use, and predisposing heart conditions (congenital abnormalities, rheumatic heart disease, hypertrophic cardiomyopathy, bicuspid aortic valve, and mitral valve prolapse) were excluded among cases and controls.

Case and controls were matched using age, sex, Charlson comorbidity index (CCI), and date of echocardiogram. Our matching algorithm attempted to sample all LS-NVE cases and 3 times as many controls as cases, matching them exactly on sex and CCI category, and closely on age (±3 years) and timing of echocardiography (±1 year).

Details characterizing the degree of valve abnormality and other echocardiographic measurements were then collected for both matched cases and controls. Clinical characteristics, medical history, procedures, and comorbid conditions (at the time of the patients’ echocardiography) were also collected from the electronic health record. The CCI was calculated as described previously [15]. This study was approved by the Mayo Clinic Institutional Review Board and included only patients with institutional research authorization.

Results

Overall, a total of 579 patients were included in the analysis: 148 LS-NVE cases [124 definite (83.8%) and 24 possible] and 431 matched controls (Figure 1). The median age was 63.8 (IQR: 53.6–74.6) years, and 329 (56.8%) patients were males (Table 1). The patients’ race was predominantly white in both cases (91.8%) and controls (92.2%). LS-NVE cases had a higher prevalence of diabetes mellitus (46.6% vs. 30.4%) and chronic kidney disease (46.6% vs. 28.1%) when compared with controls (P < 0.001 for both comparisons). The median CCI score was 3 (IQR: 0–6) for cases and 2 (IQR: 0–6) for controls.

Display full size

Figure 1. 

Selection flow chart demonstrating cases and controls

Echocardiographic variables that were routinely (i.e., in > 80% of patients) measured in cases and controls are presented in Table 2. In these unadjusted analyses, anatomical echocardiographic measurements were different in patients with LS-NVE than those without IE, as evidenced by higher IVSDT, LVPWDT, and LV mass index (P ≤ 0.01 for all comparisons). LS-NVE cases also exhibited different hemodynamics compared with controls based on higher LVOT velocity, LV cardiac index, mitral E and A wave, E:e’ ratio, and AVSP velocity (P ≤ 0.002 for all comparisons).

A multivariable conditional logistic regression model was developed to determine the association of anatomical and hemodynamic measurements with the risk of LS-NVE. The selection of variables to be included in the model was based on a priori knowledge and data reduction methods blinded to the case-control group, resulting in 10 variables selected for inclusion (Table 3). Ranked by importance according to their adjusted χ² values from the model (Figure 2), LVOT velocity (P = 0.017), LVEF (P = 0.018), and E:e’ (P = 0.050) were significantly associated with LS-NVE. When re-expressed in terms of grouped variables, the hemodynamic-related variables had a stronger contribution to the model (corrected χ² = 31.7; P < 0.001) than the anatomical landmark-related variables (corrected χ² = 11.1; P = 0.086). A partial effects plot for the continuous variables in the final model is presented in Figure 3.

Display full size

Figure 2. 

The relative importance of individual and grouped echocardiographic variables. LVOT: left ventricular outflow tract; LVEF: left ventricular ejection fraction; MV: mitral valve; LVESD: left ventricular end-systolic dimension; IVSDT: interventricular septum diastolic thickness; LVPWDT: left ventricular posterior wall diastolic thickness; AVSP: aortic valve systolic peak; LV: left ventricular; df: degrees of freedom

Display full size

Figure 3. 

Partial effects plot for continuous variables. IVSDT: interventricular septum diastolic thickness; LVESD: left ventricular end-systolic dimension; LVPWDT: left ventricular posterior wall diastolic thickness; LVEF: left ventricular ejection fraction; LVOT: left ventricular outflow tract; LV: left ventricular; MV: mitral valve; AVSP: aortic valve systolic peak

The bias-corrected C statistic of the model was 0.67, indicating modest discrimination in predicting LS-NVE for an individual patient. Using the bootstrap method to internally validate and calibrate the model, the estimated slope shrinkage factor of 0.79 suggested minor overfitting. The bias-corrected calibration curve presented in Figure 4 demonstrates that the model is generally well calibrated across the risk range, except for higher risk profiles where the model tends to underestimate the probability of IE. Model-based predictions of LS-NVE (in the matched sample), as estimated by the mathematical formula in Figure S1, are illustrated for ten hypothetical patients in Table S1.

Display full size

Figure 4. 

Bias-corrected calibration curve

A secondary analysis in which terms for diabetes mellitus and chronic kidney disease were added to the original model yielded similar findings concerning echocardiographic variables, except for E:e’ which was no longer significant (Figures S2 and S3).

Discussion

This is the first study to evaluate multiple specific TTE measurements and determine whether they were associated with the subsequent development of LS-NVE in patients with no known cardiac predisposing conditions. Our primary finding was that increased peak LVOT velocity and LVEF were hemodynamic echocardiographic measurements associated with the subsequent development of LS-NVE. It is conceivable that increased LV pressures and velocities could be operative in IE predisposition by causing substantial turbulence that could create a substrate for endocardial injury as highlighted in an experimental model highlighting a “damage-induced” pathway [12]. Nevertheless, it is important to mention that there is limited clinical application for these results, future larger-scale studies should be performed to assess the relevance of our findings.

It has been estimated that more than 800,000 individuals are at a high risk of IE in the United States alone [21]. Nevertheless, there is not a clear NVE diagnostic scheme for patients with ‘formally’ structural normal heart valve conditions. Thus, implementing our analysis, which included an in-depth characterization of cardiac disease with the use of quantitative echocardiographic measurements, but also recognizing clinical parameters, may provide the most useful foundation for a future strategy in identifying patients who may have an increased risk of IE.

We did not detect an association between the degree of mitral and aortic regurgitation and subsequent development of IE in our multivariable model. Previous work has shown that valvular dysfunction is associated with IE, however, in contrast to our study, they included patients with predisposing cardiac conditions. Namely, the degree of mitral regurgitation could be a risk factor associated with the development of IE in patients with mitral valve prolapse [22]. Of note, the inclusion of stenotic lesions of the mitral and aortic valves in our multivariable model was not possible due to their limited prevalence in our study cohort.

Diabetes mellitus and chronic kidney disease were identified as predisposing comorbidities for the development of LS-NVE in the current investigation. Diabetes mellitus is well recognized as an independent risk factor associated with IE; these patients can have an impaired immune response, delayed healing, reduced tissue vascularization, and endothelial dysfunction [23]. Chronic kidney disease patients, particularly those with end-stage renal disease receiving hemodialysis, are at a significantly higher risk of developing IE when compared to the general population due to a higher prevalence of bacteremia related to vascular access and because of uremia-related impaired immune response [24, 25]. In addition, both diabetes mellitus and chronic kidney disease have a higher rate of S. aureus colonization than the general population, and hence, an increased risk of S. aureus bacteremia, which consequently increases the risk of IE [26].