Introduction

Right-to-left shunt through a patent foramen ovale (PFO) is more common in patients with cryptogenic stroke [1]. Although long-term observational studies suggested a reduced risk of stroke after PFO closure with endovascular devices, randomized trials failed to demonstrate a clear benefit [2]. However, the evaluation for a right-to-left shunt remains an important part of the cryptogenic stroke workup, particularly in younger patients or those with thrombophilia, who are both at higher lifetime cumulative risk of recurrent stroke [2]. Although cardiac catheterization and MRI may diagnose the anatomic pathway of the right-to-left shunt, the dynamic nature of transthoracic and transesophageal echocardiogram (TTE and TEE) bubble study with agitated saline or contrast has made them, respectively, the screening test and gold standard to evaluate for right-to-left shunt [3]. Studies comparing TTE or TEE to transcranial Doppler (TCD) ultrasound bubble study, which costs less than either TTE or TEE [4], have shown good agreement between the studies [5], apart from the 15–25 % of patients that are positive on TCD and negative on TTE or TEE for right-to-left shunt [6, 7].

There are several explanations for this phenomenon, including the need for direct visualization of bubbles on TTE or TEE compared to the more easily detected acoustic bubbles on TCD and the superior ability of patients to Valsalva during TCD when patients are not sedated, which is standard practice for a TEE. Prior studies have not fully examined the contribution of different stroke subtypes, medical comorbidities, or the presence of a delayed right-to-left shunt, also known as “late bubbles,” which is suggestive of an extracardiac shunt [8]. To further explore these questions, we conducted a retrospective review of acute ischemic stroke patients tested for right-to-left shunt with both modalities.

Methods

With approval of our institutional review board, we queried the records of the TCD lab at a major urban teaching hospital for TCD bubble studies with the indication of acute ischemic stroke from 2011 to 2013. One hundred fifty-six patients met these criteria, and after chart review, we confirmed that 109/156 also had either a TTE or TEE bubble study. These 109 patients constituted the study cohort. Studies were considered positive for right-to-left shunt if the presence of the bubbles was detected in the MCA via TCD or left atrium via TTE or TEE, either at rest or with Valsalva. We recorded 12 comorbid medical conditions and determined stroke subtype according to the TOAST classification [1 = large-artery atherosclerosis (LAA), 2 = cardioembolism (CE), 3 = small-vessel occlusion (SVO), 4 = stroke of other determined etiology (SDE), 5 = stroke of undetermined etiology (cryptogenic)] [9]. The severity of the right-to-left shunt on TCD was quantified by the Spencer Grade [0 = 0 microembolic signals (MES), 1 = 1–10 MES, 2 = 11–30 MES, 3 = 31–100 MES, 4 = 101–300 MES, 5 = >300 MES or shower effect] [10] and an extracardiac pathway was defined, per our institutional criteria, as late bubbles appearing >15 cardiac cycles from injection, which is comparable to previously described methods [11]. Statistical significance for intergroup differences was assessed by Pearson χ 2 or Fisher exact test for categorical variables and by Student t or Mann–Whitney U test for continuous variables. A chi-square goodness-of-fit test was used to compare the TOAST subtypes to a typical distribution of TOAST subtypes. No correction for multiple comparisons was used since this is exploratory data.

Results

Baseline demographics are shown in Table 1. All 109 patients had a TCD, 94/109 had a TTE, and 36/109 had a TEE, of which all also had a TTE. Results from these studies are shown in Fig. 1. In the 27 patients with right-to-left shunt diagnosed by TCD but not by TTE or TEE (TCD+Echo−), all 27 patients had a TTE and 10 had a TEE. The mean Spencer Grade of the right-to-left shunt on TCD during Valsalva of these 27 patients was low at 1.76 (SD 0.9) with a range of 1–4.

Table 1 Baseline demographics of the entire cohort, the TCD+Echo− cohort, and the remaining patients who were not TCD+Echo−
Full size table
Fig. 1
figure 1

TCD and TTE/TEE bubble study results for 109 patients with recent ischemic stroke

Full size image

TOAST classification for the 27 TCD+Echo− patients was 11 % LAA, 7 % CE, 4 % SVO, 33 % SDE, and 44 % cryptogenic, which is significantly weighted towards the SDE subgroup (33 %) compared to a typical TOAST distribution (p < 0.001) [12]. When we divided the cohort between the 27 TCD+Echo− studies and the 82 remaining studies, there was no statistically significant association with age, ethnicity, gender, or comorbid medical conditions, with the exception of active malignancy, defined as stage IV cancer or patients receiving treatment for cancer, which was present in 6/109 and more likely to be associated with TCD+Echo− (4/27 in the TCD+Echo− cohort and 2/82 in the remainder, p = 0.032). Twenty-nine out of 109 had late bubbles on TCD, which was strongly associated with TCD+Echo− (p = 0.001) (Table 2).

Table 2 Medical comorbidities and presence of late bubbles in the TCD+Echo− subgroup of patients compared to the remainder of the cohort
Full size table

Discussion

In our cohort, 25 % of patients were TCD+Echo− with only 2 % of patients TCD−Echo+, a finding that has been previously described [6, 7]. From a technical perspective, this is plausible given that TCD allows for highly sensitive acoustic detection of bubbles, whereas TTE and TEE rely on visualization of bubbles with a limited field of view of the left atrium that does not encompass all four pulmonary veins. The duration of the echocardiographer’s observation following injection of contrast could be too short to allow detection of late bubbles, while during TCD, the recording continues uninterrupted throughout the study’s multiple contrast injections. In addition, patients undergoing TEE are not able to properly Valsalva due to sedation and the presence of an ultrasound probe in their pharynx and esophagus, which is not an issue during TCD. These data reinforce the superiority of TCD as a screening test for right-to-left shunt, based on its accuracy, low-cost, and non-invasive nature. A recent meta-analysis of prospective studies comparing TCD to TEE for diagnosis of right-to-left shunt, which included 1968 patients from 27 studies, came to a similar conclusion regarding the superiority of TCD as a screening test [13].

Active malignancy and late bubbles were more common in the 27 patients who were TCD+Echo−. Active malignancy has been reported before in patients who were TCD+Echo−, but a statistically significant association is novel [6]. Pulmonary arteriovenous malformation (AVM), the most common extracardiac right-to-left shunt that would cause late bubbles, has a high risk of stroke, as does malignancy [14, 15]. Both of these stroke mechanisms are compatible with the low mean Spencer Grade of 1.76 seen in TCD+Echo− patients. Active malignancy patients form more venous thrombi and thus have more opportunity for stroke through a small right-to-left shunt [14]. Another explanation for stroke with a low Spencer Grade is a pulmonary AVM, which results in bubbles dissolving due to a longer transit pathway, and may be very difficult to detect visually with an echocardiogram, but would be more apparent acoustically during a TCD.

We also report that the TOAST classification of our cohort differs from a typical distribution. This was driven by an increase in the proportion of SDE, which typically comprises less than 10 % of all strokes, but was 33 % of the TCD+Echo− cohort [12]. In part, this reflects the selection bias of complicated stroke patients referred for TCD at our institution, but it was also accompanied by a normal percentage of cryptogenic stroke. Yet, despite the high number of patients with late bubbles, there were no patients with a diagnosis of pulmonary AVM. In the 29/156 patients with late bubbles on TCD, the cause of stroke in 48 % was SDE. Apart from a single case, these patients did not undergo screening for pulmonary AVM, suggesting pulmonary AVMs may have been missed in a significant number of patients.

Our study has several limitations. The most important is that it was a retrospective analysis with potential selection bias. We also collected data on multiple variables that we theorized would explain the TCD+Echo− phenomenon but did not have a definitive a priori hypothesis concerning causality. However, we consider our results hypothesis generating and recognize that a prospective study of a diverse sample of stroke patients with extensive and standardized diagnostic workups would ultimately be more informative. Such a study could further explore the relationship that we describe between active malignancy, extracardiac right-to-left shunt, and TCD+Echo−.

Summary

As other authors have reported, TCD outperforms both TEE and TTE in the detection of right-to-left shunt and has a negligible false-negative rate when compared to echocardiography. Based on this data and other recent studies, TCD should be considered the screening test of choice for evaluating right-to-left shunt in ischemic stroke patients. TTE and TEE still have a role in defining cardiac anatomy, but TCD outperforms them for right-to-left shunt detection, is non-invasive, and costs less than either a TTE or TEE. Our cohort suggests that patients with active malignancy and extracardiac right-to-left shunt are particularly prone to being missed by echocardiography. These conditions require additional testing to diagnose and are often amenable to curative treatment. In the appropriate clinical scenario, patients who are either TCD+Echo− or have late bubbles on TCD should undergo additional screening directed at discovering malignancy or pulmonary AVM.