In this study, TTE detected a major CSE in only 1% of 1084 patients with ischaemic stroke or TIA of undetermined cause. Notably, 10 of 11 patients with a major CSE also had major ECG abnormalities that warranted TTE assessment anyway. Our findings suggest that the current strategy of routine echocardiography in patients with ischaemic stroke or TIA of undetermined cause might be replaced by selective echocardiography in patients with major ECG abnormalities only. Additionally, contrast echocardiography should be performed in patients eligible for PFO closure [15]. With this strategy, the number of TTEs would be reduced by 73%.

In accordance with Dutch guidelines [16] we considered PFO screening appropriate in patients < 60 years old, with a RoPE score of ≥ 6 points. A PFO was detected in 23% of those patients who underwent contrast-enhanced echocardiography, similar to the estimated prevalence in healthy individuals but lower than the 40–50% reported in patients with ischaemic stroke of undetermined cause [17]. This discrepancy might reflect dissimilar patient populations and differences in prior diagnostic analyses.

The prevalence of TTE-detected major CSE in this study is lower than detection rates reported by most previous studies (0.3–37%) [18,19,20,21,22,23,24,25,26]. Several aspects may play a role. Most previous studies were retrospective, and patients were often selectively included after referral for echocardiography. Also, some studies included patients with known CSE. Another factor that can explain differences is a change of insights into treatment implications. For instance, it has been shown that a change to anticoagulant therapy does not lower recurrent stroke risk in the presence of poor LV function or complex aortic atheromas [10,11,12]. Furthermore, although sometimes reported as CSE, for mitral valve prolapse, mitral annular calcification and spontaneous echo contrast no causal relationship with stroke has been demonstrated [3, 4]. Notably, a previous retrospective study by our group [25] found a proportion of major CSE that would have been similar to our present rate, if we then had classified LV dysfunction as minor CSE.

Previous studies showed that cardiac symptoms and ECG abnormalities were the strongest predictors for detecting CSE [23, 25]. Our results corroborate these findings: major ECG abnormalities were present in the vast majority of patients with a major CSE. Although having major ECG abnormalities was not a pre-defined outcome measure, we consider this a highly relevant finding that deserves more attention in future studies. However, in order to assess the effect of TTE screening on clinical outcomes in a randomised controlled trial, a very high number of participants is needed to show a significant difference in clinical endpoints.

Studies of cardiac CT in ischaemic stroke patients reported higher detection rates of 3–17% [27,28,29]. However, these populations included patients with known CSE [27, 28]. Timing was also different: cardiac CT was often performed in the acute stroke setting, whereas TTE was performed at a later time. The study in patients with ESUS was retrospective and included 48% of eligible patients [29]. Moreover, several CSE other than thrombi are better detected with TTE because of its dynamic nature. In conclusion, (selective) cardiac CT might be of additional value in patients with ischaemic stroke or TIA of undetermined cause, but this approach needs confirmation by prospective data.

This is the first large-scale, prospective multicentre study to examine the yield of TTE in a well-defined population. Major CSE were defined as pathologies for which a change of therapy reduces the risk of recurrent stroke.

This study has several limitations. First, not all patients with ischaemic stroke or TIA of undetermined cause underwent TTE; for most patients a reason was stated. Furthermore, not all patients had MRI-proven ischaemic stroke; yet, the inclusion of patients with clinically suspected ischaemic stroke and TIA ensures that results are applicable to everyday practice. In addition, not every patient had intracranial vasculature imaging. Also, TOE was not performed routinely. The superiority of TOE compared to TTE in ischaemic stroke patients lies mainly in detecting left atrial (LA) and left atrial appendage (LAA) thrombi, complex aortic atheromas and valve vegetations [4]. Yet, LA and LAA thrombi are extremely rare in patients in sinus rhythm [30], complex aortic atheromas do not change therapy [11, 12], and a stroke in a patient with (suspicion of) endocarditis is no stroke of ‘undetermined cause’.

As the most recent guidelines on PFO treatment were published after enrolment had started [15], contrast echocardiography was not always performed when it currently would be indicated. In addition, specific patient characteristics and preferences might have prevented intravenous contrast injections. Finally, the median period until TTE assessment was 30 days, reflecting logistic capabilities of the participating hospitals. Although the effect of timing on TTE results is unknown, we cannot exclude the possibility that a thrombus may have dissolved at the time of echocardiography. Nonetheless, we observed no significant difference in major CSE detection with TTE during the 1st week as compared to a later point in time.