Patent Foramen Ovale—Assessment and Treatment


Simon Ray, MB, MRCP, Department of Cardiology, University Hospitals of South Manchester, Manchester Academic Health Sciences Centre, Manchester, UK.
Tel.: +44 161 2912402; Fax: +44 161 2915634;


A Patent Foramen Ovale (PFO) is detectable in 20–25% of the population. Some, but not all, case control studies have found an increased incidence of PFO in patients with cryptogenic stroke. Prospective cohort studies have failed to convincingly demonstrate a link between PFO and first stroke, and evidence linking PFO to recurrent stroke is far from compelling. The rate of recurrent stroke in medically treated patients is low, but the development of devices for PFO closure has lead to enthusiasm in some quarters to pursue a strategy of device closure. Nonrandomized studies have suggested a lower risk of recurrent events with device closure but the data are heterogeneous, and potentially prone to bias. Device implantation is associated with a risk of major adverse events of between 1.5% and 2.3%, and there is a significant rate of failure to close shunts. The results of randomized trials of device closure are keenly awaited. Migraine with aura has been linked with PFO. A recent metanalysis suggested an association, but the one prospective population study did not. The well-publicized and controversial MIST Trial is the only randomized trial of device closure in migraineurs yet published, and failed to demonstrate a convincing benefit from device closure. Other conditions such as platypnea-orthodeoxia syndrome and prevention of decompression sickness in divers, may justify device closure. Evidence for a role of PFO in the etiology of cryptogenic stroke and migraine is contradictory. It is possible that some patients might benefit from PFO closure but there is scant evidence of sufficient quality to justify routine PFO closure in either group. It is essential that ongoing randomized trials of device closure are completed.

Prevalence and Detection

A patent foramen ovale (PFO) is a slit or tunnel like passage in the interatrial septum (IAS) formed by failure of postnatal fusion of the septum primum and septum secundum. Persistent PFO occurs in around 20–25% of the adult population, the exact frequency dependent on the method of detection. An autopsy based study suggested that the prevalence of PFO decreased with age implying either spontaneous closure or, intriguingly, a significant early death rate in those with PFOs [1]. However recent evidence from a large surgical series suggests that this is not the case and prevalence does not relate to age [2].

Transthoracic echocardiography (TTE), transoesophageal echocardiography (TOE), and transcranial Doppler (TCD) can all be used to detect PFO [3–7]. TOE has traditionally been regarded as the gold standard but has some significant disadvantages. Whilst in some individuals a PFO is obvious on 2D echo or color flow on TOE, injection of bubble contrast is needed in many to clearly demonstrate a right to left shunt (see Figure 1). Patient sedation, positioning in the left decubitus position and the inability to perform a complete Valsalva may make it more difficult to visualize a shunt with bubbles on TOE than on TTE where the patient is semiupright and able to perform a good Valsalva. On the other hand suboptimal image quality means that some shunts may be missed on TTE. Femoral rather than antecubital contrast injection increases the detection of PFO as the inferior vena caval bloodstream is directed onto the atrial septum [8,9]. In practice it makes sense to perform both contrast TTE and TOE. Importantly, TOE may detect otherwise unsuspected causes of stroke, including aortic atheroma, left atrial appendage thrombus, or spontaneous left atrial contrast. If available TCD is probably the best screening technique as it allows an optimal Valsalva without deterioration of image quality but cannot distinguish between PFO, ASD and less common pulmonary arteriovenous malformations. If positive TCD needs to be followed by TOE to confirm the site of the shunt, and the presence of associated features such as atrial septal aneurysm (ASA) and a Eustachian valve (see Figure 2).

Figure 1.

Image A (top left): 2-Dimensional TOE image demonstrating a PFO with clear separation of the thin septum primum from the septum secundum. Image B (top right): 3-Dimensional TOE image viewed from the left atrial aspect. The tunnel-like nature of the defect can be appreciated in this view. Image C (bottom left): Bubble contrast TOE. A mixture of air, saline and blood has been used to create bubble contrast. Bubbles can be visualized entering the left atrium through a PFO.

Figure 2.

In this patient with PFO an Atrial Septal Aneurysm (ASA) is present. ASA is most commonly defined as > 10mm excursion of the atrial septum from the midline.

In our institution, we investigate in a step-wise fashion (see Figure 3). If bubble contrast TTE is negative from ante-cubital injection after provocation manoeuvres, and there is a high index of suspicion, we proceed to femoral injection. TOE is performed in any patient with inadequate transthoracic images or a demonstrated shunt. Furthermore, we have a low threshold in young patients with cryptogenic stroke to perform a TOE to exclude the abnormalities outlined above.

Figure 3.

Protocol for investigation of possible PFO in our institution. Bubble contrast TTE is used first line, with femoral injection if required. TOE is used to define anatomy if a shunt is confirmed, or as a first line modality if TTE images are deemed inadequate.

PFO and Stroke/TIA

Association of PFO and Cryptogenic Stroke

Cryptogenic strokes account for 25% of ischaemic strokes overall and up to 50% in younger patients [10]. A number of case control studies have demonstrated an increased incidence of PFO in younger patients with cryptogenic stroke as compared with normal controls. Equally there are some good quality studies, which have failed to show an association [11,12]. A meta-analysis of 566 patient and 458 controls showed an odds ratio of 3.1 (95% CI 2.29–4.21) for the presence of a PFO in cryptogenic stroke in individuals < 55 years of age, increasing to 6.0 (95% CI 3.72–9.68) when compared with patients with a known cause of stroke [13]. The prevalence of PFO in younger stroke patients was 40% as compared with 17% in controls. A more recent study has demonstrated a similar association in older patients, OR 2.92 (95% CI 1.70–5.01) [14], and both found an increased risk with the presence of an ASA in association with a PFO. Whilst on balance these findings seem to demonstrate an association between PFO and cryptogenic stroke, they do not necessarily demonstrate causality. There are a number of published reports of thrombus identified stuck in transit through a PFO in subjects with stroke or peripheral embolism where there is a self-evident causal relationship and the obvious assumption is that the association between PFO and ischaemic stroke is on the basis of paradoxical embolisation. Against this is the low frequency of identifiable DVT in patients with cryptogenic stroke. However deep vein thrombosis (DVT) may not be clinically evident, especially if present in the pelvic veins [15] and a thrombus does not have to be large to cause a devastating stroke.

Other mechanisms of embolic stroke in association with a PFO are possible and a recent paper by Rigatelli et al. investigated the possibility that PFO might be associated with abnormalities of atrial function [16]. In a series of 100 patients with recurrent stroke referred for PFO closure they demonstrated abnormalities in atrial function similar to those in patients with chronic atrial fibrillation (AF) and these were more pronounced in those with moderate to large ASAs who also had a greater frequency of coagulation abnormalities and more frequent appearance of spontaneous left atrial contrast. Intriguingly the atrial function parameters normalized after device closure. Other authors have found using event-loop recording in patients with PFO and cryptogenic stroke, that the incidence of AF is significant at approximately 8–15%, and unaffected by the presence of a closure device [17,18].

It is possible that in some individuals the PFO is a bystander or a marker for some other embolic risk. Alsheikh-Ali et al. used a Bayesian analysis to examine the likelihood that a PFO in a subject with a cryptogenic stroke was an incidental finding [19]. Overall the probability that a PFO was an incidental finding was about one third, falling to one fifth in younger patients (typically < 55 years) and rising to about half in older patients (typically ≥ 55 years). Corresponding figures for PFO plus ASA were 11%, 9%, and 26%.

PFO and First Stroke—Prospective Cohort Studies

There is limited prospective data on the occurrence of first stroke in subjects with a PFO. The Northern Manhattan Study (NOMAS) study [20] followed 1100 subjects aged over 39 years of whom 164 (15%) had a PFO identified by contrast transthoracic echo. Over a mean follow up of about 6.5 years the hazard ratio for occurrence of a first stroke was 1.64 (95% CI 0.87–3.09) with a PFO and 1.25 (95% CI 0.17–9.29) with an associated ASA. A second study (SPARC) in 550 individuals over 45 years found a hazard ratio of 1.46 (95% CI 0.74–2.88) with a PFO [21]. In the NOMAS study the rate of PFO detection using TTE alone was 15%, as compared with 24% in SPARC using TOE. Although there was a trend towards a greater incidence of first stroke with a PFO in both studies, this did not reach statistical significance. Interestingly both studies also suggested an increased risk of stroke in the small number of individuals with ASA alone.

It is likely that the discrepancy between the results of case control studies and these prospective cohort studies can be explained at least in part by ascertainment bias in the former as there is evidence that a PFO is searched for more diligently in an individual with a cryptogenic stroke as opposed to a control or a patient in whom there is another established cause for stroke [22].

PFO and Recurrent Stroke

There are numerous studies on the influence of PFO on recurrent events after an index cryptogenic stroke. There is considerable heterogeneity in the patients included and major variability in the results. Several have reported particular characteristics of a PFO associated with recurrent paradoxical embolism [23–25]. Mas et al. demonstrated a greater incidence of recurrent stroke in patients with ASA in association with PFO but not PFO alone; Goel found that PFOs in patients with recurrent cryptogenic stroke were larger with bigger shunts, longer tunnels, and more likely to be associated with ASA. In contrast some studies have shown no link with either the presence of a PFO or septal morphology [26,27], while others have suggested a link between abnormalities of coagulation, PFO, and cryptogenic stroke [28].

A recent metanalysis [29] examined data on 15 studies published between 1994 and 2007. There was marked heterogeneity in the rates of recurrent events in the individual studies. Thirteen of the 15 allowed treatment at the discretion of the usual physician. Treatment was generally antiplatelet or anticoagulant drugs but 28% of patients underwent PFO closure during follow up. The pooled risk for recurrent TIA was 4.0/100 patients years (95% CI 3.0–5.1) and for recurrent stroke 1.6 (95% CI 1.1–2.1). There was no increase in the risk of either TIA or stroke with a PFO. The relative risk of recurrent TIA with versus without a PFO was 1.1(95% CI 0.8–1.5) and for ischaemic stroke 0.8(95% CI 0.5–1.3).

A study published too late to be included in the metanalysis [27] used TCD to look for right to left shunting in 486 patients with cryptogenic stroke. Overall around 60% of patients had a shunt and in around 40% this was classified as massive. Treatment was left to the referring physician and the risk of recurrent stroke after a mean of about 2 years was 5.8%. There was no difference in the risk in those patients with a massive shunt, nonmassive shunt or no shunt, irrespective of age, antithrombotic treatment, and the presence of an ASA. There is therefore evidence from a number of prospective studies that the presence of a PFO does not increase the risk of stroke recurrence. There are a number of possible explanations for this apparent discrepancy [29]. First, it is possible that medical treatment is highly effective, thus obscuring the additional risk of a PFO. Second, it is possible that the risk of a PFO is balanced by some other undetected risk in patients without a PFO and third it is possible that there might be an ascertainment bias in the case control studies with PFO being more rigorously sought in patients known to have suffered a cryptogenic embolic event.

Treatment to Prevent Recurrent Cerebral Embolism

Medical Treatment

As described in the preceding section the rate of recurrent events in medically treated patients with cryptogenic stroke is actually relatively low [29]. A recent metanalysis included 8 studies with sufficient data to compare treatment effects and found that recurrent events were about half as likely on warfarin as compared with antiplatelet treatment (OR 0.5 95% CI 0.4.0.) [29]. However the one randomized study of aspirin versus warfarin in cryptogenic stroke patients failed to demonstrate a difference in outcomes (22).

PFO Closure

Primary surgical PFO closure has now more or less died out with the development of percutaneous closure devices (see Figure 4). Patients undergoing cardiac surgery for other indications may have an incidental PFO discovered at the time of surgery and a recent study suggests that closure of incidental PFO in these circumstances may actually increase postoperative stroke risk [30]. There a number of percutaneous devices available and new technologies such as closure by suture or radiofrequency energy are under development. Randomized controlled trials are underway to test the efficacy of device closure of PFO compared with medical therapy but as yet none has reported. The availability of a relatively simple means to close a PFO has lead to great enthusiasm amongst interventional cardiologists and the lack of clear evidence to an understandable temptation on behalf of some physicians, patients, and relatives to push for device closure especially in a younger patient with perceived high-risk features. Three systematic reviews [31,32] of nonrandomized data have concluded that the rate of recurrent stroke with medical treatment is greater than that with device closure, the most recent [33] suggesting an annualized risk of recurrent TIA or CVA with device closure of 1.3% as compared with 5.2% with medical treatment. Analyzed studies were highly heterogenous in entry criteria, duration of follow up and definition of end points and these data cannot be regarded as a reliable guide to treatment. Another potential confounding issue is the prolonged duration between index event and device closure in some studies that could potentially bias in favor of device closure [34].

Figure 4.

Image A (left): 2-Dimensional TOE image in the bi-caval view with an Amplatzer PFO occluder device in situ. Note the smaller left atrial disc seen uppermost in this image. Image B (right): 3-Dimensional TOE image in the same patient, viewed from a left atrial aspect. This image helps confirm adequate coverage of the defect, and the absence of impingement on adjacent structures.

There are other reasons why clinical decisions on device closure are not straightforward. The rate of major complications of percutaneous closure is low at around 1.5–2.3% but these include death, major haemorrhage, emergency surgery, tamponade, and fatal pulmonary embolism [31,33]. High volume units probably have a lower rate of complication [35,36]. New onset or worsening of existing valvular regurgitation has been reported after device deployment [37], although a recent study using cardiac magnetic resonance quantification showed no significant effect [38]. The incidence of AF appears to be increased following device closure of PFO [39], and may be associated with use of larger devices [40]. This is a cause for concern, as device closure may expose patients to a new and potentially more important risk factor for cerebrovascular events. Another important consideration is that device implantation is not necessarily the same as defect closure and the prevalence of residual shunting depends on how vigorously it is sought. Rigorous follow up using TCD at a mean of a year postprocedure reveals incomplete closure in around 20% and large residual shunts in around 14%, irrespective of the device used [41]. Some of these may have represented pulmonary arteriovenous malformations rather than a persistent PFO but this in itself would be important, implying a potential source of paradoxical embolism not dealt with by PFO closure. Although most PFOs that close completely or near completely do so within a month a significant minority do so between 6 and 12 months post procedure [41].

Current Guidelines for the prevention of recurrent cerebral embolism in the context of a PFO are summarized in Table 1. It should be noted that whilst both the AHA/ASA and European guidelines specifically mention device closure as a possible treatment this recommendation is based on evidence of the lowest possible category, relying on case studies and expert opinion rather than randomized trials.

Table 1.  International guidance for prevention of recurrent cerebral embolism in the context of PFO
American Academy of Neurologists [42]PFO is not associated with increased risk of recurrent stroke (level A) possible increase in risk with atrial septal aneurysm age <55. Justifies investigation of PFO for risk stratification (Level C)Insufficient evidence to determine superiority of warfarin over antiplatelet drugs insufficient evidence to determine the role of device closure
American Academy of Chest Physicians [43]No clear advantage of warfarin over antiplateletsAntiplatelet treatment recommended
American Heart Association/American Stroke Association [44] Antiplatelet therapy reasonable Warfarin reasonable for high-risk patients with other indications such as hypercoagulable state or venous thrombosis Insufficient evidence to recommend device closure for a first stroke. PFO closure may be considered for recurrent cryptogenic stroke on optimal medical treatment
European Stroke Organisation [45] Patients with cardioembolic stroke unrelated to AF should receive warfarin if the risk of recurrence is high Device closure should be considered in patients with cryptogenic stroke and high-risk PFO

The lack of clarity surrounding the role of device closure in the secondary prevention of cryptogenic stroke was reflected in a recent advisory statement from the American Heart Association/American Stroke Association/American College of Cardiology Foundation that strongly advocated referral of suitable patients for inclusion in one of several ongoing randomized trials [46].

PFO and Migraine

A link between PFO and migraine was suspected when individuals who had a PFO (or ASD) closed for other reasons noted an improvement in the frequency and severity of migraine headaches [47]. A recent metanalysis [48] of data from 2636 subjects from 18 highly heterogenous studies concluded that there was low grade of evidence to support an overall increased frequency of migraine in individuals with PFO with a summary odds ratio of 5.13 (95% CI 4.67–5.59), and for migraine with aura an odds ration of 3.21(95% CI 2.38–4.70) but no association for migraine without aura. Similarly there was low to moderate grade evidence to support an overall increased frequency of PFO in migraineurs with a summary odds ratio of 2.54 (95% CI 2.01–3.08). The weakness of the evidence means that these findings must be interpreted with caution. A recent large case control study failed to show any association between PFO and migraine diagnosed by specialists [49]. The one prospective population study [50] also failed to show any association between self-reported migraine and the presence of a PFO, despite both having a prevalence of around 15% in the population.

It has been variously proposed that particulate matter or vasoactive substances passing through the shunt might trigger the onset of a migraine but there is little if any hard evidence to support this [51,52]. Indeed if it were the case, migraine in patients with a PFO would be expected to present early in life and follow a pattern of frequent episodes, usually with aura and be provoked by a Valsalva manoeuvre. Septal defects are clearly not the sole cause of migraine as 50% of patients with migraine with aura do not have a PFO and not all individuals with a PFO suffer from migraine. Anzola et al. studied 460 patients with migraine with aura and investigated them systematically for a right to left shunt with TCD [52]. Patients with associated shunts did not exhibit the above features but were slightly younger, had more sensory symptoms of aura and a greater proportion of first-degree relatives with migraine. This last observation raises the possibility of the coinheritance of migraine and PFO, at least in some lineages, as suggested by Wilmshurst [53].

The metanalysis discussed above also included six retrospective studies of 194 patients undergoing device closure of PFO and although there was a trend to a reduction in attacks with device closure the quality of the evidence was insufficient to draw any conclusions [48]. A recent nonrandomized study of migraineurs with moderate to severe symptoms and brain MR lesions found a significant reduction in both the frequency and severity of attacks with device closure compared with controls [54]. The well publicized and controversial MIST Trial is the only randomized trial of device closure in migraineurs yet published and failed to demonstrate a clear benefit from device closure in individuals with particularly severe migraine [55, 56]. The rate of complications in the device arm was 6.8%. Several other trials are currently in progress but there is at the moment no evidence to support device closure of PFO in migraineurs outwith the context of a clinical trial.

Decompression Illness

Decompression illness in scuba divers occurs on ascent from deeper waters. Gas bubbles develop by expansion of preexisting gas nuclei found at normal atmospheric pressure in joints, the spine, sweat glands, and skin pores. Tissue bubbles may gain access to the capillary or lymphatic bed and enter the venous circulatory system. The lungs filter small gas volumes, but large ones may cause pulmonary barotrauma. Venous gas bubbles may be transferred to the systemic circulation via intra-atrial defects, or via pulmonary arterio-venous malformations.

Torti et al. investigated 230 scuba divers with contrast TOE to diagnose PFO [57]. PFO was identified in 23%. The presence of a PFO was associated with a 4.8 to 12.9-fold increased risk of a significant decompression event. In addition, the risk of a decompression event was positively associated with defect size. Interestingly, divers with a PFO have been shown to have twice the number of ischaemic brain lesions than those without, although diving even in the absence of PFO also increases the incidence of brain lesions [58]. Device closure should be considered in divers with otherwise unexplained decompression who wish to return to diving with the standard caveats regarding the potential for complications and incomplete shunt eradication.

Platypnoea–Orthodeoxia Syndrome

The platypnea–orthodeoxia syndrome is characterized by subjective dyspnoea and objective hypoxemia on assuming an upright posture. The syndrome can be caused by severe lung disease, but also through venous to arterial shunting at atrial level. In the case of PFO, the upright posture may increase shunting via the redirection of inferior vena cava (IVC) inflow towards the IAS. The syndrome has classically been described in association with right pneumonectomy [59]. Here postoperative cardiac position is shifted to predispose to increased shunting from the IVC across the atrial septum. Redirection of IVC flow to towards the IAS has also been reported in an enlarged or “horizontalized” aortic root [60]. Here counter-clockwise cardiac rotation may distort the position of the atrial septum [61]. The presence of a prominent Eustachian valve may also contribute to the degree of shunting [61]. Device closure is an effective treatment.

Other Types of Embolism and PFO

A number of case reports have linked the presence of a PFO to myocardial infarction in the absence of another identifiable cause [62–64]. The YAMIS study [65] investigated 101 adults under the age of 40 years who had suffered a myocardial infarction for evidence of a venous to arterial circulation shunt (v-aCS). The investigators found no significant difference in the frequency of v-aCS between this group and the control, reflecting the findings of a previous smaller study [66]. A systematic causal link between myocardial infarction in the young and PFO has not been demonstrated although this does not exclude paradoxical embolism as a cause of MI in some individuals.

Systemic embolism in many forms has been linked with PFO. These include sudden onset deafness [67], monocular blindness [68–70], hemianopia [70] or peripheral embolism [71]. In general the reports and studies lack quality, hence these scenarios do not represent a rationale for PFO closure.


Recent work suggests significant clinical overlap between Alzheimer's disease (AD) and vascular dementia (VaD), and vascular pathology has increasingly been linked with AD [72]. In a case control study, spontaneous cerebral emboli were detected in 40% of patients with AD and 37% of those with VaD compared with 15% and 14% of their controls. The resulting odds ratios (adjusted for vascular risk factors) were 2.7 for AD and 5.4 for VaD [73]. In this study, the odds ratio for the presence of a venous to arterial shunt using TCD was 1.57 in AD and 1.67 in VaD, neither of which reach significance. A recent study found a significant venous-to arterial shunt indicative of PFO, was associated with more severe deep white matter hyperintensities in AD [74]. These hyperintensities are associated with a more rapid cognitive decline in AD, and are postulated to represent recurrent microemboli. Further work is required define the role of right to left shunts and traditional vascular risk factors in the etiology of dementia.


Evidence for a role of PFO in the etiology of cryptogenic stroke and migraine is confused and contradictory. It is certainly possible that some patients with cryptogenic stroke and some with migraine with aura might benefit from PFO closure. However, despite enthusiastic advocacy by some clinicians based on case studies and individual experience there is little if any existing evidence of sufficient quality to justify routine PFO closure in either group. It is essential that ongoing randomized trials of device closure are completed.

Conflict of Interest

The authors have no conflict of interest.