Impact of septic cerebral embolism on prognosis and therapeutic strategies of infective endocarditis: a retrospective study in a surgical centre

We have built a local registry of patients affected by non-device-related IE, where we recorded the 551 incident cases admitted to our department from January 2013 to November 2021. Device-related IE was defined as an infection on cardiac devices other than valve prosthesis. Data for analysis were retrieved from electronic hospital charts and were fully anonymized. The local Ethics Committee (Regional Ethics Committee of Tuscany for Experimental Medicine, Section: AREA VASTA CENTRO, n 12113_oss) approved the study and, in accordance with Italian laws for observational studies, granted a waiver of informed consent from study participants. We followed the current international IE guidelines for diagnostic work-up and treatment strategies and all methods were performed in accordance with them [4]. Brain CT or MR were performed in 352 (64%) patients, either for neurologic symptoms (89) or for the screening of embolism in asymptomatic patients (263). In patients with cerebral embolism (126), we reported the site of embolism, the presence of CT-detected hemorrhagic lesions, the Glasgow Coma Scale (GCS) and the Barthel index [6] determined immediately after the onset of the neurologic event. For the purpose of analysis, we categorized the GCS into two groups (15 or ≤ 14) and the Barthel index into three groups, indicating a substantially preserved overall functional independence (≥ 65), as opposed to a moderate (40–64) or severe (0–39) disability. Renal failure was defined as GFR < 60 mL/min/1.73 m² (Mild GFR 45–59; Moderate GFR 30–44; Severe GFR 15–29). Multivariable analysis of all-cause in-hospital and long-term mortality was adjusted for the treatment received: surgery, medical therapy for the absence of surgical indication or exclusion from surgery for prohibitive clinical conditions. Age, gender, history of drug abuse, the microbiologic agent involved, left ventricular ejection fraction, type of valve (native or prosthetic) affected, double valve infection, the paravalvular extension of infection, severe valvular dysfunction, vegetation length, EUROSCORE II, CT-detected haemorrhagic complication of cerebral ischemic lesion, as well as the GCS and the Barthel index, were all entered into an initial multivariable Cox proportional hazards model.

We calculated the follow-up duration from the time of IE diagnosis. A structured phone interview updated the follow-up of all patients to March 2022.

Identification of predictors of cerebral embolism and mortality in patients with neurologic complications of IE were the primary study endpoints.

We used the chi-square and the Mann–Whitney or Kruskal–Wallis tests to compare respectively proportions and continuous variables with normal or non-normal distribution. We performed univariable and multivariable analyses using logistic regression and general linear models. We used the Kaplan–Meier method to estimate the univariate survival analysis and the Cox regression to identify the multivariable associations with mortality and estimate their hazard ratio with 95% confidence interval. All tests were 2-sided, and statistical significance was defined as a p-value < 0.05. We performed the analyses with SPSS 23.0 and R 3.6.3.

Of 551 patients with IE in the registry, 126 (23%) had a neurologic complication. The median follow-up was 3.4 years (95% CI 3.2–3.6). The main demographic, clinical, echocardiographic and microbiologic characteristics of the cohort by the presence or absence of cerebral embolism are reported in Table 1.

Cerebral embolism was significantly more frequent in patients with large vegetations (p = 0.004), mitral valve infection (p = 0.001), and Staphylococcus aureus infection (p = 0.025). Patients with neurologic complications also had a significantly higher left ventricular ejection fraction and a lower proportion of severe valvular dysfunction. Such a difference is the probable consequence of a referral to a surgical centre since, beyond embolic complications, surgery for IE is indicated by heart failure or severe valvular dysfunction [4]. At multivariable analysis, only vegetation length was an independent predictor of cerebral embolism (HR 1.057, 95% CI 1.025–1.091, p = 0.001), with a best predictive threshold of 10 mm length at ROC analysis (AUC 0.54, p = 0.001).

Of 551 patients with IE, 431 (78%) underwent surgery, 60 (11%) received only medical therapy because of the absence of surgical indication, and 60 (11%) were excluded from surgery, despite surgical indication, because of prohibitive general conditions.

Compared to those without, patients with neurologic complications were more often excluded from surgery despite an indication for it (Table 2). As shown in Table 3, severe disability (i.e. Barthel index score class 3) was the main reason for denying surgery. When eligible, patients with or without neurologic complications were operated on within two weeks of IE diagnosis in similar proportions (70% vs 72%) and with similar survival rates for each treatment group (i.e. surgery; medical therapy for the absence of surgical indication; excluded from surgery because of prohibitive general conditions; Fig. 1) and for patients undergoing early surgery (Fig. 2).

Hemorrhagic transformation (p = 0.018), GCS ≤ 14 (p = 0.001) or Barthel index score < 40 (i.e. severe disability; p = 0.001) at presentation were univariate predictors of mortality in IE patients with cerebral embolism. Of these, only the Barthel index latter was kept in the final multivariable model (HR 2.4, 95% CI 1.47–3.95, p = 0.001), together with known adverse prognostic factors of infective endocarditis such as age (HR 1.036 for each increasing year, 95% CI 1.024–1.050), double valve infection (HR 2.27, 95% CI 1.59–3.23), EUROSCORE II (HR 1.015 per unitary increase, 95% CI 1.009–1.021), and exclusion from surgery (HR 4.82, 95% CI 2.42–9.58) [7]. Figure 3 shows the survival curves of patients with cerebral embolism stratified by levels of disability, and Fig. 4 shows the same curves by levels of disability and treatment, highlighting the independent prognostic impact of disability itself.