Despite numerous studies published in the past two decades, the role of patent foramen ovale (PFO) as a risk factor of stroke remains a matter of debate. A significant correlation between PFO and cryptogenic stroke has repeatedly been shown. However, most studies included only small numbers of patients and thus did not allow to adjust the analysis for gender and age.

The aim of our study was to re-evaluate the association between right-to-left shunts (RLS) and stroke subtypes in a patient community large enough to allow multivariate analysis with special consideration of gender-related differences.

We found a significant association between RLS and cryptogenic stroke. This has frequently been reported in previous publications, which initially led to the consideration of PFO as an important risk factor in stroke. 456789 Also in accordance with previous studies, ES positive patients were younger and, as expected, less likely to have traditional risk factors. A higher prevalence of PFO in young subjects has also been reported in the general population.1011 The association of RLS and cryptogenic stroke might therefore be coincidence. When adjusting for age, there was no longer a significant correlation between RLS and cryptogenic stroke in our study, which reduced the suggested statistical association between cryptogenic stroke and PFO. This is in line with a recently published population-based study which also describes a much weaker association between PFO and cryptogenic stroke than has been reported earlier.1213

Unlike most previous studies, we stratified for gender: In female patients, who were significantly more likely to show RLS, the correlation between RLS and cryptogenic stroke did not reach significance and further decreased after adjusting for age. However, among male patients cryptogenic stroke still weakly correlated with RLS in multivariate analysis (CI 1.00–2.43). Conflicting with our results, neither the autoptic study of Hagen et al. nor the abovementioned population-based study showed the incidence of PFO to differ significantly between men and women.111214 Of both previous large multicenter studies only the PFO/ASA collaborative study reported a significant correlation of female gender and PFO.415 The reasons for any gender-related differences among stroke patients remain uncertain and require further exploration.

Our results furthermore question the common theory of paradoxical embolism: Presuming arterial embolism via PFO secondary to a venous embolic source, the amount of shunt volume would be expected to correlate with the risk of stroke. This thesis has support from prior TEE-based studies.1617 Anzola et al found that detection of more than 10 bubbles by c-TCD correlated with stroke recurrence.18 However, the number of patients included in these studies were small, and the criteria for grading the size and thresholds for distinguishing a large shunt from a small one were arbitrary. In contrast, in both previous large TEE-based multicenter studies, PFO size failed as a significant predictor of stroke recurrence.415 Further studies revealed that the amount of contrast shunting did not correlate with the size of the PFO whether measured by two-dimensional TEE or invasively by balloon sizing.19 Exploiting the method of c-TCD, we detected the shunted contrast medium directly in the target organ. Furthermore our method allowed us to discriminate between the presence of a shunt at rest versus RLS during Valsalva maneuver. Only in the first subgroup would the condition for paradoxical embolism be continuously satisfied. In the remaining patients, particular circumstances would be required that cause a transient right-to-left intracardiac shunt precisely at the moment an embolus passes the right atrium. This is less likely, and previous investigations did not reveal an association between Valsalva-provoking activities preceding stroke onset and the presence of PFO.52021 Hence, we expected a particularly high percentage of patients with RLS at rest among patients who otherwise had no identifiable causes of stroke. However, this was not the case;

Another possible explanation for stroke secondary to PFO but independent of paradoxical embolism is secondary cardiac arrhythmias.22 We did not find atrial fibrillation to be associated with RLS. Thus, our study does not provide support for cardiac arrhythmia as a relevant mechanism of stroke in PFO carriers. These findings are in line with previous studies.20

Other possible explanations include abnormalities of the endocardial surface of the septum or within the PFO that are a focus for thrombus formation.23 A substantial limitation is that c-TCD is not applicable in detecting distinctive features like an atrial aneurysm. A further unavoidable limitation of this method is that patients with severe stroke and very old patients are probably underrepresented, as the former may not be able to perform Valsalva maneuver and in the latter group it might be difficult to perform the transcranial Doppler examination because of insufficient "bone windows."

Despite these limitations, the large number of patients included and the use of c-TCD contribute new arguments, particularly by weakening the thesis of a significant correlation between PFO and cryptogenic stroke and assessing potential gender-related differences that should be taken into account in future studies.

The authors declare that they have no competing interests.