This is a retrospective, case-control, single center study. Data were collected from the electronic medical records of patients admitted between October 2019 and December 2022. The study population included consecutive individuals aged 18–65 years, who presented with either stroke or transient ischemic attack (TIA) associated to a PFO diagnosed by TEE. Patients with other potential stroke or TIA etiologies, as suggested by 24-h ECG monitoring, cervical and cerebral vascular imaging, transthoracic and transesophageal echocardiography, positive work-up suggesting vasculitis, systemic diseases with possible central nervous system (CNS) involvement, including antiphospholipid syndrome, and those with lacunar strokes confirmed by magnetic resonance imaging (MRI) studies were excluded from the study. The control group was selected from consecutive patients referred to or hospitalized during the same time period in our department who underwent brain MRI for vertigo or headache, with no history of stroke/TIA, inflammatory CNS disorders, or other systemic disorders with possible CNS involvement. For vertigo, we included only patients having persistent symptoms for at least 24 h, in order to exclude possible TIAs presenting as vertigo. They were subsequently invited to participate in the study and undergo contrast-enhanced transcranial Doppler (c-TCD) and transesophageal echocardiography (TEE). Patient selection flowchart is presented in Figure 1; a detailed description of the inclusion and exclusion criteria used in the study is presented in the Supplementary material. All patients were required to have undergone cerebral MRI within the preceding month or during their hospitalization, with a quality that met the technical specifications required by the semi-automated software Quantib ND™ (Quantib B.V., Rotterdam, Netherlands), for quantifying WM hyperintensities volume. To avoid omitting lesion visualization in patients with PFO-related stroke/TIA, only MRIs that were performed more than 24 h after symptom onset were included. All patients had MRIs performed within 3 weeks of symptom onset in order to ensure that DWI and ADC sequences show potential subacute or early chronic lesions suggestive of stroke. All patients underwent TEE and contrast transcranial Doppler. Comprehensive morphological TEE characteristics of the PFO, such as lengths and heights, were available for a subset of individuals; severity of right-to-left shunt (RLS) was documented for all participants. Two groups were defined: patients with a history of PFO-related stroke/TIA, and patients with PFO and no history of stroke/TIA. Demographic data, cardiovascular and cerebrovascular risk factors, including hypertension, diabetes, smoking status, chronic kidney disease, and dyslipidemia were collected.

The study was approved by the Local Ethic Committee. All patients signed an informed consent. Research was performed in accordance with the Declaration of Helsinki and GDPR standards.

Transesophageal echocardiography was conducted using an E95 echo machine from General Electric (GE) Healthcare, Horten, Norway, equipped with a dedicated TEE probe. A single examiner (SMB) conducted all examinations. Interatrial septum was analyzed using dedicated views, including long-axis, short-axis, and bicaval views. These views were used to assess the presence of a PFO, and to identify any color shunt at this level. Length of the PFO was measured as the longest distance of the overlap between the ostium primum and ostium secundum, while the height of the PFO was defined as the maximum distance between these two structures. Measurements of PFOs lengths and heights were available in a limited number of patients due to the retrospective nature of data collection. All patients underwent a “bubble test” for the assessment of RLS, using intravenous administration of agitated saline solution, at rest and after intense Valsalva maneuver; RLS was assessed by the presence of microbubbles passing through the PFO. Additionally, presence of an interatrial septal aneurysm was assessed, defined as an excursion of more than 10 mm of the interatrial septum. Other anatomical features, such as a prominent Eustachian valve or a prominent Chiari’s network, were also assessed.

Contrast transcranial Doppler was performed using a Digital DWL® TCD device equipped with DWL® routine software (Doppler-Box™, Compumedics Germany GmbH, Singen, Germany). A single examiner (RȘB) conducted all examinations. Standard protocol involved insonating both middle cerebral arteries (MCAs) at a depth of 55 mm, using 2 MHz pulsed wave (PW) monitoring probes affixed to a DiaMon® headset (Compumedics Germany GmbH, Singen, Germany). A 23-gauge cannula was inserted into one of the antecubital veins, and a blood sample of 1 mL was withdrawn from the patient. The blood sample was then mixed with 1 mL of air and 8 mL of 0.9% saline solution. The resulting mixture was agitated 10 times using two syringes and a three-way stopcock before being injected into the intravenous cannula. Subsequently, the patient was instructed to perform two Valsalva maneuvers. The total count of microbubbles passing through both MCAs during the entire test (during resting and during the two Valsalva maneuvers) was recorded, and the severity of the RLS was assessed using Spencer logarithmic scale (SLS), as follows: grade 0, no microembolic signals (MES); grade 1, 1–10 MES; grade 2, 11–30 MES; grade 3, 31–100 MES; grade 4, 101–300 MES; and grade 5, over 300 MES (curtain pattern) (5).

In order to be compatible with the image processing software and thus accepted in the study, MRI studies had to be performed on either 1.5 or 3 Tesla MRI machines, with a non-contrast enhanced 3D T2-FLAIR acquisition with an inversion time of 2,250 ms, an echo time of 140 ms, a voxel size ranging between 1 and 3, and a slice thickness of 1.5–3 mm; patients with examinations utilizing filters such as GE filter SCIC and GE filter E were excluded from the study due to incompatibility with the software. MRI images were assessed using the semi-automated software Quantib ND™ to measure WM lesional volume. Quantib ND™ was selected for its advanced artificial intelligence (AI) algorithms, which facilitate accurate and efficient quantification of brain lesions. The process involved uploading the MRI scans into the software, where Quantib ND’s AI algorithms semi-automatically identified and measured the WM lesions. Any necessary manual adjustments were made to ensure accuracy, following which the software provided quantitative data on lesional volume. This data were then extracted for statistical analysis. For each of the 3D FLAIR acquisitions, each of the T2 weighted hyperintensity, surrounded by normal isointense WM, was marked (Figure 2). Lesions located in the basal ganglia and brainstem were excluded from measurement, as they were ascribed to small vessel disease. Measurement protocol for patients with stroke comprised two phases: initially, all cerebral lesions were automatically identified and measured, including the volume of the stroke; subsequently, the volume of the stroke was excluded, and the remaining volumes were recorded. Two examiners, a trained neurologist (RȘB) and a radiologist (ANM), conducted all examinations.

Normality of data distribution for continuous variables was assessed using the Shapiro–Wilk test. Continuous variables with a normal distribution were presented as mean ± standard deviation, while those with a non-normal distribution were reported as median (interquartile range). Categorical variables were presented as rates (percentages). To compare differences between groups, independent t-test was used for continuous variables, while χ² or Fisher’s exact test was used for categorical variables. To determine differences in WM lesions volume between groups, the Mann–Whitney U test was conducted. Additionally, a Quade nonparametric ANCOVA test was employed to assess differences in WM lesions volumes between groups, while controlling for the following variables: age, gender, hypertension, glomerular filtration rate, LDL-cholesterol, and triglycerides blood levels. To assess the relationship between morphological characteristics (height and length) of the PFO and WM lesional volume, Spearman’s rank-order correlation test was conducted. Missing values were left in for the analysis. This decision was based on the nature of the missing data, which was determined to be missing at random (MAR), and our assessment that their inclusion would not significantly bias the results. Inter-observer reliability between the two estimations of WM lesion volume, using the semi-automated software Quantib™, was determined using the Intraclass Correlation Coefficient (ICC, two-way mixed, absolute agreement). A p value of less than 0.05 was considered statistically significant. Statistical analysis was performed using IBM SPSS Statistics for Windows, Version 29.0. (Armonk, NY: IBM Corp).

In order to understand the sensitivity of our study, a post hoc power analysis was conducted, considering the observed effect size, the sample sizes of the two groups, and the alpha level. The effect size (Cohen’s d) was derived from the mean difference in the total volume of WM lesions (without the stroke volume in cases of patients with a history of stroke) between patients with and without a history of stroke/TIA, divided by the pooled standard deviation of both groups. The post-hoc power analysis revealed that the study achieved a power of 88.31%. This suggests that, given the observed effect size, our study had an 88.31% chance of detecting a true effect of this magnitude, assuming it exists. Additionally, in a retrospective evaluation to address the adequacy of our sample size, we estimated the required number of participants per group to achieve a power level of 80% with an alpha level of 0.05, based on the observed effect size (Cohen’s d) of 0.802 from our study. This estimation suggested that approximately 26 participants per group would be necessary to adequately power a study to detect the differences in the total volume of WM lesions between patients with and without a history of stroke/TIA. Statistical power calculations were performed using G*Power software (Version 3.1.9.7, Heinrich Heine University Düsseldorf, Germany). (6)

After reviewing the medical records from October 2019 to December 2022, a total of 461 patients aged 18–65 years, diagnosed with stroke and screened for PFO were initially identified in the neurology department. Among these, 146 patients were confirmed to have PFO. After applying the exclusion criteria, 45 patients were found to have other potential causes of stroke or had insufficient etiological work-up and were excluded from the study. An additional 60 patients were excluded because they either did not undergo cerebral MRI or had MRIs that were incompatible with the Quantib™ software specifications. Consequently, the final sample comprised 41 patients diagnosed with PFO-attributable stroke.

For the control group, 442 patients who were evaluated in the neurology department for headache and/or vertigo were identified from medical records and invited for PFO screening. Of these, 118 patients received a confirmed PFO diagnosis. Subsequently, 23 patients were excluded due to diagnoses of other conditions that could result in cerebral lesions, and one patient declined participation. Furthermore, 68 patients were excluded for not undergoing a cerebral MRI or having an MRI incompatible with the Quantib™ software. Ultimately, 26 patients, with a PFO diagnosis but no history of stroke, were included in the control group (Figure 1).

Demographic characteristics of the groups are reported in Table 1. There were no statistically significant differences, apart from gender.

When comparing the overall burden of WM lesions, which includes the stroke volume, the lesional volumes appear similar between the two groups. However, upon excluding the stroke volume from the analysis, patients with PFO with no history of stroke displayed significantly higher volumes of WM lesions (0.27 cm³, IQR 0.03–0.60) compared to patients with PFO-related stroke/TIA (0.08 cm³, IQR 0.02–0.18), p = 0.019 (Figure 3). This statistically significant difference persisted even after controlling for cardiovascular risk factors (F = 7.27, p = 0.010). The inter-observer agreement in the assessment of WM lesions volume in patients with PFO was excellent among the two observers (ICC = 0.980).

Patent foramen ovales morphological features are presented in Table 1. There was a tendency of patients with PFO-related stroke / TIA to have larger PFOs and greater shunts than patients with no history of stroke/TIA (Table 1). Meanwhile, presence of interatrial septal aneurysms, Chiari’s network, or of a prominent Eustachian valve were not different between groups (Table 1), and they did not contribute to the burden of cerebral WM lesional volume (Table 2). There was no significant correlation of cerebral WM lesions volume (excluding stroke volume) and the PFOs’ lengths [ρ(45) = −0.139, p = 0.351], heights [ρ(45) = 0.005, p = 0.972] or shunt size [ρ(65) = −0.115, p = 0.372] (Figure 4).