Silent Strike: Stroke in Context of Endocarditis – Brain Imaging as a Catalyst for Diagnosis

Endocarditis is an inflammatory disorder affecting the innermost layer of the heart which usually involves valvular endocardium but may involve other regions of the endocardium as well. Its etiology can be either bacterial, resulting in infective endocarditis, or non-bacterial, also known as non-bacterial thrombotic endocarditis, the latter being commonly associated with malignancies. A daunting aspect of this disease is the polymorphic spectrum of complications, including neurovascular events, primarily ischemic stroke and cerebral abscess. Ischemic stroke is a consequence of reduced or arrested blood flow to a brain region, resulting in metabolic dysregulation, generated by various pathophysiological mechanisms [1,2,3].

Current practice for the diagnosis of infective endocarditis is based on the use of the Modified Duke Criteria [4]. Research has been conducted to explore the potential of brain imaging to enhance the pre-existing methods of diagnosing endocarditis in the context of ischemic stroke. This is due to the growing necessity to ensure timely and effective treatment for this condition [5,6]. This study aims to compare the clinical and imaging aspects of stroke caused by endocarditis with stroke induced by a different etiology (another source of cardioembolism or large-artery atherosclerosis) and to assess the usefulness of brain imaging as a tool for diagnosing endocarditis when Modified Duke Criteria are not met.

We conducted a nested case-control study in which we enrolled 42 patients admitted to the Neurology Department of the National Institute of Neurology and Neurovascular Diseases Bucharest over the course of 99 months (September 2014 – December 2022). The inclusion of the patients in the study group was based on coexisting diagnoses of endocarditis and ischemic stroke. Modified Duke Criteria were employed for diagnosing infective endocarditis [4]. Subsequently, we assembled an equally sized control group similar to the study group in terms of demographics and cardiovascular risk factors and including endocarditis-free patients with ischemic stroke due to cardioembolism (originating from sources other than valvular vegetations) or to large-artery atherosclerosis.

The following parameters were evaluated for both cohorts: demographics, risk factors, laboratory results, brain imaging (MRI + CT), and outcome.

Regarding patients’ outcome (evaluated at discharge), three categories were employed: improved (patients discharged in a better condition than at admission), worsened (patients transferred to other departments or hospitals due to various serious complications), and deceased.

Descriptive statistics included the calculation of the median (Q1, Q3) as the continuous numerical variables had an asymmetric distribution. Hemorrhagic transformation was classified according to the ECASS II system [9].

Shapiro-Wilk test was used to assess the normality of data distribution. Mann-Whitney test with continuity correction was used to compare continuous (numerical) variables.

Fisher’s exact test was employed to compare categorical parameters; the dependent categorical variable was endocarditis (either present or absent); the independent categorical variables were: age, gender, arterial hypertension, type 2 diabetes, dyslipidemia, chronic kidney disease, chronic heart failure, history of myocardial infarction, concomitant neoplasm, hemorrhagic transformation, presence of multiple ischemic lesions, involvement of multiple arterial territories, presence of watershed lesions, associated bleeding, deceased – for all these the two possible values being yes/present and no/absent.

When multiple comparisons were performed, the significance level (commonly set at 0.05) was lowered according to Bonferroni correction: the corrected significance level was 0.05 divided by the number of comparisons. All the statistical calculations were performed using the R language and environment for statistical computing and graphics (version 4.2.3.); the employed packages were dplyr, tidyverse, ggplot2, ggpubr, mosaic, mediation. For calculating sensitivity (Sn), specificity (Sp), positive likelihood ratio (LR+), negative likelihood ratio (LR−) with confidence intervals, we used the diagnostic test calculator from https://ebm-tools.knowledgetranslation.net/calculator/diagnostic/. To calculate the area under the receiver operating characteristic curve (AUROC) and the corresponding 95% confidence interval we have employed the ROC Calculator from https://www.statskingdom.com/roc-calculator.html and the 95% confidence interval calculator from https://riskcalc.org/ci/, respectively.

Out of the 42 patients in the endocarditis group, the definite diagnosis of endocarditis was established in 21 cases (50%), while in the remaining 21 (50%), the diagnosis was deemed possible, based on the Modified Duke Criteria.

From 11511 patients admitted during the same period with stroke not related to endocarditis, we chose another 42 patients as controls, matched with the cases by age, sex, cardiovascular diseases/risk factors (arterial hypertension, type 2 diabetes, dyslipidemia, chronic kidney disease, chronic heart failure, history of myocardial infarction), and concomitant neoplasm (see Table 1).

There was a statistically significant difference between the two groups when assessing the baseline laboratory parameters, with higher values in the endocarditis group for ESR (p=0.009) and neutrophil count (p<0.001), and lower values for hemoglobin level (p<0.001) as shown in table 2.

As regards the findings on cerebral imaging, there was a statistically significant difference in lesion patterns between the two groups, with multiple injuries (p<0.001), multi-territorial (including both posterior and anterior circulation) involvement (p<001), and watershed lesions (p=0.0072) being more common in endocarditis group. In terms of hemorrhagic transformations, there was no statistically significant difference between the two groups (p=0.35). Associated cerebral hemorrhage (intraparenchymal or subarachnoid) occurred only in the endocarditis group (in 4 patients). The diagnostic performance of these findings is illustrated in Table 3.

Multiple lesions were proven to be the most sensitive tool among these findings, with Sn = 0.786, Sp = 0.857, LR+ = 5.497, LR− = 0.25. On the other hand, the involvement of multiple arterial territories had the highest specificity and positive likelihood ratio for endocarditis-related stroke (Sn = 0.738, Sp = 0.929, LR+ = 10.394, LR− = 0.282). Watershed lesions and hemorrhagic transformation have also shown high specificity, but with lower sensitivity and high negative likelihood ratio (Sn = 0.429, Sp = 0.857, LR+ = 3, LR− = 0.66 for watershed lesions, Sn = 0.109, Sp = 0.905, LR+ = 2, LR− = 0.895, for hemorrhagic transformation).

Ischemia extent (evaluated using ASPECTS and pc-ASPECTS scores [7,8]) was larger in the endocarditis group, with larger infarctions both in the posterior circulation (p<0.001), and the anterior circulation (p=0.011), with both models having a predictive power, but with a higher AUROC for pc-ASPECTS, as shown in Figure 1.

Figure 1.

The area under the receiver operating characteristic curve (AUROC) was used to evaluate the performance of ASPECTS and pc-ASPECTS in predicting endocarditis-related stroke.

Concerning the outcome, there were no fatalities in the control group, all patients being improved at discharge; by contrast, in the endocarditis group there were 16 deaths, while other 14 patients needed to be transferred due to their worsened condition (Table 4). The median in-hospital survival time in the endocarditis group was 12 days.

This study aimed to identify the role of brain imaging in the context of endocarditis related stroke. Given the lack of specificity of the wide array of neurological symptoms engendered by endocarditis induced ischemic stroke [11,12], one may not suspect at first the underlying condition. The Modified Duke Criteria alone cannot ensure a definitive diagnosis in all cases [6], thus prompting clinicians to develop further methods [13,14] to uncover this pathology. This was also visible in our study, where only half of the patients from the study group met the criteria for the definite diagnosis. Consequently, in recent years, there has been increasingly more research focusing on the use of brain imaging in such cases [13,14,15,16,17].

Laboratory results usually point to systemic involvement, as highlighted by our study too, consistent with previous literature [2]: elevated neutrophil count and ESR and decreased hemoglobin level on admission were more common in endocarditis associated stroke.

Early cerebral imaging, whether it is CT or MRI, has proven to be of great usefulness in the context of ischemic stroke, particularly when it occurs as a complication of endocarditis. Our study suggests that some patterns of injury (including the involvement of multiple territories, the coexistence of multiple lesions, and a greater extent of the ischemia-afflicted cerebral territory) plead for endocarditis as the source of the cerebral ischemia-generating emboli. Our study also pointed out an association between watershed ischemia and endocarditis, in agreement with previous studies [18], despite the control group having a significant prevalence of major artery atherosclerosis, which could result in hemodynamically significant stenoses known to be associated with watershed lesions. We also attempted to assess the impact of susceptibility weighted imaging MRI (as was done in some recent studies [10]), but unfortunately our study was limited by the reduced access to such technology. Further studies comparing only cardioembolic ischemia with endocarditis-related ischemia, consistently evaluated by MRI with sequences that have a high sensitivity for vascular events may help advance our knowledge.

Better results in terms of management and outcome of patients suffering from endocarditis-related stroke may be achieved if the Modified Duke Criteria and cerebral imaging are employed timely [5] in cases of stroke patients presenting with fever and altered lab results, until blood culture results arrive, as empirical antibiotic therapy might be ineffective and insufficient [19]. The relative reluctance to perform thrombolysis (endocarditis being regarded as a contraindication, albeit long-debated) and endovascular treatment in the context of endocarditis [9,20,21] did not interfere with our results, as none of the patients in the study group underwent thrombolytic therapy or mechanical thrombectomy, although etiology was unknown at presentation.

The mortality rate was much higher in the study group, however the real survival rate may be significantly lower, as we were unable to further monitor the patients after being transferred to other departments (e.g. Infectious Diseases department).

While our study has managed to prove its hypothesis, it is imperative to acknowledge its limitations, such as its retrospective character, the long time span over which data was collected, during which the standard of diagnosis and care changed, the small size of our study cohort, the monodisciplinary character of our unit, resulting in limited access to various investigation methods [22] (such as transesophageal echocardiography), the scarcity of MRI examinations, and the need to transfer many of our patients to other units (which prevented follow-up imaging of the ischemic lesions and hemorrhagic transformations, as well as outcome and lab results monitoring). However, CT scan, although one of the simplest and most accessible cerebral imaging methods, has proven to be reliable in uncovering key elements suggesting endocarditis as the etiology of ischemic stroke.

To this day, endocarditis remains a dreaded clinical entity with a challenging diagnosis process, despite the recent revision of the Duke Criteria. Nonetheless, both clinical and laboratory findings are useful in suggesting the diagnosis, as our study has proven.

This study found that early brain imaging is a valuable and reliable instrument that aids the clinician in raising the suspicion of endocarditis in patients presenting with stroke and no apparent underlying condition. Considering the wide array of possible cerebral imaging techniques, further studies with larger and more homogenous samples are required to determine if extensive brain imaging investigations are warranted as cost-effective tools able to provide useful clues for diagnosis and treatment.