This retrospective study used a hospital-based prospective cohort of consecutively enrolled patients with ischemic stroke within 7 days of symptoms onset. Routine evaluations performed were as follows: standard blood tests including D-dimer; brain magnetic resonance imaging (MRI) and/or computed tomography (CT) and cerebral angiography (magnetic resonance angiography, CT angiography, or conventional angiography); Holter monitoring or continuous electrocardiography monitoring during admission in the stroke unit; and transesophageal echocardiography (TEE) and/or transthoracic echocardiography (TTE). TEE was not performed in patients with decreased consciousness, aphasia, impending brain herniation, poor systemic conditions, severe swallowing difficulty, tracheal intubation, or when informed consent could not be obtained [17].

The study hospital is a tertiary university hospital with a dedicated cancer center where patients with cancer routinely undergo extensive evaluations for the presence of metastasis. When cancer or NBTE is diagnosed during admission, patients are referred to oncologists for further evaluation and treatment.

This study included consecutive patients with active cancer who were registered between January 2010 and December 2018 and who underwent echocardiography. Active cancer was defined as cancer that was diagnosed within 6 months, required chemotherapy or surgical treatment within 6 months, or was recurrent, metastatic, or inoperable [18]. Patients with brain malignancy or hematologic malignancy were excluded. The cohort and this study were approved by the Institutional Review Board of the Yonsei University Health System. The need for an informed consent was waived due to the retrospective nature of the study.

Because the study participants had active cancer, patients were categorized into those with NBTE, those with cryptogenic etiology, and those with determined etiology [11]. NBTE was determined based on formal reports by cardiologists who diagnosed vegetation based on echocardiographic examinations. Five patients with NBTE and co-existing atrial fibrillation were classified into the NBTE group. Twenty-one patients with a patent foramen ovale without venous thrombosis and other determinable etiologies were classified into cryptogenic etiology group.

All patients were regularly followed up after discharge by neurologists and clinical research assistants in the outpatient clinic using face-to-face interviews or by a telephonic interview using a structured questionnaire [19]. We reviewed the medical records to obtain information on mortality and cause of death. The cause of death was categorized as stroke-related, cancer-related, or others/unknown [20]. Information on the occurrence of any new stroke events (cerebral infarction or intracerebral hemorrhage [ICH]) after the initial cerebral infarction was also obtained from medical records. The diagnosis of stroke was based on brain MRI and/or CT results.

Variables were expressed as mean±standard deviation, median (interquartile range), or number (percentage), as appropriate. Baseline characteristics among the groups were compared using one-way analysis of variance or the Kruskal-Wallis test for continuous variables and the chi-square test or Fisher’s exact test for categorical variables. Post hoc analysis was performed using Bonferroni’s correction or Dunn’s test. Kaplan-Meier estimates and log-rank tests were performed (adjusted by Bonferroni’s correction) for mortality, occurrence of stroke, and composite outcomes of mortality and stroke. To determine whether the presence of NBTE was associated with composite events of mortality and stroke, we performed multivariate Cox proportional hazard regression analyses. As all patients with NBTE had metastasis and 90% of patients with NBTE had multiple territory infarctions, we constructed two different multivariate regression models to avoid multicollinearity. Model 1 included the presence of NBTE and Model 2 included metastasis and multiple territory infarctions. All statistical analyses were performed using R statistical software version 3.5.1 (http://www.R-project.org). A P<0.05 was considered statistically significant.

Of the 245 patients, 20 patients (8.2%) had NBTE, 96 (39.2%) had a cryptogenic etiology, and 129 (52.7%) had determined etiologies. In patients with determined etiologies, those with cardioembolism were 51, those with large artery atherosclerosis were 46, and those with two or more causes were 19 (Supplementary Table 2). Of 150 patients (61.2%) with metastasis, NBTE was diagnosed in 20 patients (13.3%). Compared with patients with determined etiologies, those with NBTE were younger, had more frequent multiple vascular territory involvement, lower platelet counts, lower fibrinogen levels, and higher D-dimer levels (Table 1).

All 20 patients with NBTE had metastasis, and NBTE was not detected in patients without metastasis. The presence of metastasis was significantly higher in patients with NBTE (100%) than in those with stroke with a cryptogenic etiology (69.8%; P=0.014 after Bonferroni’s correction) or those with a determined etiology (48.8%; P<0.001 after Bonferroni’s correction) (Table 1). The cancer type was not significantly different among the three groups (Table 1).

All patients were followed up for 6 months after the initial stroke event. During the follow-up, 127 patients (51.8%) developed stroke events and/or died (110 [44.9%] died and 55 [22.4%] patients experienced stroke events). The median survival of patients with NBTE was 44.5 days, and event-free survival of composite outcomes of mortality and stroke was 36 days. During the 6-month follow-up, 90% (18/20) patients with NBTE died or developed stroke, while only 38.8% patients with a determined etiology died or developed stroke (50/129, P<0.001) (Table 2).

Compared to event-free patients, those with events (mortality or stroke) were younger and had lower body mass indices and higher initial NIHSS scores. They also had lower hemoglobin levels, platelet counts, and fibrinogen levels, and higher white blood cell counts and D-dimer levels (Table 3). Metastasis and multiple vascular territory involvements were more common in patients with mortality or stroke events. Cancer type was also different between patients with or without events (Table 3).

In the Kaplan-Meier survival analysis, the 6-month mortality was different among the three groups (P<0.001). Patients with NBTE showed lower 6-month survival probability, lower stroke-free probability, and lower event-free probability than those with cryptogenic or determined etiologies (Figure 1). Mortality, stroke events, and composite outcomes were more common in patients with a cryptogenic etiology than in those with a determined etiology (Figure 1).

During the follow-up, more patients with NBTE died (16/20 [80.0%]) compared to those with a cryptogenic etiology (52/96 [54.2%]) and a determined etiology (32/129 [32.6%]) (P<0.001). Causes of death were cancer-related in 83 (75.5%), stroke-related in 21 (19.1%), and other/unknown in six (5.5%) patients. The cause of death did not significantly differ among the three groups (P=0.307) (Table 2).

During the 6-month follow-up, stroke events occurred in 55/245 (22.4%) of patients; 53 (21.6%) with ischemic stroke and two (0.8%) with ICH. Stroke was diagnosed using diffusion-weighted imaging in 49 patients (89.1%) and by CT in five patients (9.1%). The type of imaging was uncertain in one patient who was diagnosed at another hospital (1.8%). Stroke events occurred most frequently in patients with NBTE (10/20, 50%), followed by those with a cryptogenic etiology (24/96, 25.0%) and those with a determined etiology (21/129, 16.3%) (Table 2). Mortality was more frequent in patients with stroke events than in those without (69.1% [38/55] vs. 37.9% [72/190], P<0.001).

In the multivariate Cox proportional hazard analysis, female sex and a lower body mass index were associated with composite outcomes of mortality and stroke both in Model 1 and Model 2. The composite outcomes were independently associated with the presence of NBTE (hazard ratio [HR], 1.941; 95% confidence interval [CI], 1.052 to 3.690) (Table 4, Model 1). The presence of metastasis (HR, 2.870; 95% CI, 1.648 to 4.996) and the involvement of multiple vascular territories (HR, 2.524; 95% CI, 1.570 to 4.058) were also independently associated with the composite outcomes (Table 4, Model 2).